the labor supply of u.s. agricultural workers hill; us...the data for this paper come from the...
TRANSCRIPT
The Labor Supply of US Agricultural Workers
Alexandra E Hill
Abstract
Recent increases in immigration enforcement activities and observed changes in the
profile of US agricultural workers are expected to have large effects on the agricultural
industry This paper documents emerging trends in the demographic characteristics
of US crop workers estimates the implications of these trends for future industry
labor supply and investigates the viability of several employer actions for increasing
the hours and weeks of work that farmworkers are willing to provide Using a nationally
representative survey of employed farmworkers I show evidence that workers are aging
are more likely to have children and are more likely to be female I show that there has
not been a change in the legal status composition of the workforce however trends in
the other demographic characteristics vary significantly over legal status I examine the
specific implications of this for the future labor supply of agricultural workers I find
that the current trends in workforce aging suggest that the future labor force will be
willing to work more hours per week and weeks per year in agriculture However the
trends in gender and family composition move in the opposite direction suggesting that
the future labor force will provide fewer labor-hours and labor-weeks to agricultural
employers Finally I examine the potential of specific employer policies to increase
hours and weeks of work I find that increasing wages and offering health care coverage
have no significant effect on labor supply but that offering a bonus increases both hours
and weeks of work
1
1 INTRODUCTION
1 Introduction
The profile of US agricultural workers is changing and little is understood about the
implications for their employers Farmworkers are aging increasingly female and are
more likely to have children (Hernandez et al 2016) In addition tightening immi-
gration enforcement is expected to reduce the number of undocumented workers in the
labor force (Kostandini et al 2014 Zahniser et al 2012) Previous work has linked
heterogeneity across these demographics with differential behavior in the labor market
These studies suggest that the changing profile of farmworkers will decrease the aggre-
gate labor supply of US agricultural workers This could have large negative effects
on US agricultural production and profits
This paper shows how the changing profile of agricultural workers will affect labor
supply on the intensive margin ie the labor supply of those employed in agriculture I
compare the labor supply of US agricultural workers across four dimensions age gen-
der family composition and legal status The labor supply estimates reflect a workerrsquos
short-run choice of the number of hours to work each week and the number of weeks
worked each year I combine these estimates with an analysis of trends in the work-
force composition to predict future labor supply I then explore several avenues through
which employers can increase labor supply In particular I examine how wages health
care coverage and bonuses affect hours and weeks of work for employed agricultural
workers
Recent estimates of labor supply for US agricultural workers are limited (Emerson
amp Roka 2002 Pena 2010 Taylor amp Thilmany 1993) Significant changes in US im-
migration policy and the profile of workers have resulted in major structural changes
in the agricultural labor market that are not captured in existing estimates (Boucher
et al 2012 Martin amp Calvin 2010 Taylor 2010) Importantly literature studying
other industries and the US labor market more broadly find significant differences in
worker labor supply across many of the demographic dimensions in which the agricul-
tural workforce is changing
A large literature focuses on the labor supply of immigrants and the effects of
legal status on various labor market outcomes Generally this literature finds that
legalization is associated with increases of 5 to 15 in the workersrsquo real wage rate and
that these effects are largest for more skilled workers (Kaushal 2006 Kossoudji amp Cobb-
Clark 2002 Lofstrom et al 2013 Rivera-Batiz 1999) Recent work by Borjas (2017)
is the first to examine labor supply elasticities based on immigrant status He finds a
significant difference in the elasticity of labor supply between US workers of different
immigration statuses He finds that undocumented men have a wage elasticity close to
zero while the elasticity for US native men is closer to 04 Here I add to this finding
2
1 INTRODUCTION
and examine the labor supply differential across the legal statuses of US agricultural
workers Many of these workers are foreign born (76) and undocumented (50)1
This combined with the flexible nature of agricultural employment (ie limited use of
formal employment contracts) makes agricultural workers a useful group for examining
differential labor market behavior across legal statuses Further while Borjas and others
impute worker legal statuses to categorize workers as lsquolikely undocumentedrsquo I rely on
workersrsquo self-reported legal statuses To the best of my knowledge Pena (2010) provides
the only evidence on differences in the labor supply of agricultural workers across self-
reported legal statuses Using the same data source as this paper Pena finds that
foreign born workers are more likely to work in piece rate pay positions and that this
is associated with higher levels of poverty
Another body of literature examines differences in labor supply across gender and
family composition (marital and parental status) This literature shows that the labor
supply of women is lower than that of men that married women are more responsive
to changes in wages (ie have higher wage-elasticities) than single women single men
or married men and that children lead to a reduction in labor supply for females but
not for males (Angrist amp Evans 1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 Mc-
Clelland amp Mok 2012 Meyer amp Rosenbaum 2001) Because of these well-documented
differences in labor market behavior the common practice is to separately examine the
labor supply of married and single males and females Here I include all workers in the
same regression and control for these characteristics This allows me to compare labor
supply between these groups and ultimately enables me to show how future changes
across these demographics will affect industry labor supply
There is a small but growing literature on the labor market effects of workforce aging
Because workers of different ages behave heterogeneously in the workforce the norm
in the literature is to focus on working-age adults (those ages 16 - 65) and control for
age cohort so that labor supply differences are estimated from within groups of similar-
aged workers (Blundell amp Macurdy 1999) More recently due to the rising average
age of workers in the US and many other countries age has become an explanatory
variable of interest rather than a necessary control Evidence in the United States
suggests that productivity and labor force participation are significantly decreasing in
age (Fallick amp Pingle 2010 2006 Maestas et al 2016 Sheiner 2014) In this paper I
examine the relationship between worker age and labor supply among those employed
in US agriculture To the best of my knowledge this is the first empirical evidence on
the intensive margin effects of an aging workforce
This paper builds on existing literature in three important ways This paper is
1Estimates are from my own analysis of the NAWS data for the most recent survey round (2015-2016)
3
1 INTRODUCTION
the first comprehensive study of how the changing demographic profile of agricultural
workers will affect industry labor supply I find that native born citizens work fewer
hours per week and fewer weeks per year than foreign born legal or undocumented
workers Mid-aged workers (25 - 44 year olds) work more hours per week than younger
or older workers (16 - 24 and ge 45 respectively) but older workers work more weeks
each year Parents work significantly more weeks per year than non-parents but work a
similar number of hours Males work significantly more hours and weeks than females
In light of ongoing trends in these demographic characteristics my results imply that
the way in which the agricultural workforce is aging will cause the hours and weeks of
labor provided by employed farmworkers to increase while the changes in gender and
family composition will cause labor supply to decrease
Second this paper provides the first empirical evidence on multidimensional differ-
ences in labor supply across legal statuses and key demographic characteristics Existing
evidence has focused on separately identifying effects of legal status gender age mar-
ital status and parental status on labor supply This is the first study to examine
how these interact and provides important insights into heterogeneity in labor supply
across worker legal statuses Among my sample of US agricultural workers I find
surprisingly little heterogeneity across legal statuses and within demographic groups
This suggests that while on average natives foreign born legals and undocumented
workers have different labor market behavior they behave similarly within age gender
marital status and parental status groups Two notable exceptions to this are single
women with children who are legal immigrants work 65 more hours per week than na-
tives and single men with children who are both legal and undocumented immigrants
work 55 more weeks per year than natives
Finally this paper provides empirical evidence on how specific employer actions can
influence the labor supply among those employed in agriculture I examine whether
employers can increase hours or weeks of work by offering higher wages health benefits
or pay bonuses I show evidence from naive OLS regressions that these employer policies
are significantly and positively correlated with labor supply outcomes However causal
evidence from an IV approach indicates that little of this correlation is causal Among
these policies I find that bonuses are the only ones that causally affect labor-hours and
labor weeks I find that offering a bonus causes the average worker to increase weekly
hours of labor by ten percent and to increase annual weeks working in agriculture by
65 weeks I also find evidence that bonuses paid to workers for staying through the
end of the season increases weeks of work (by 35 weeks per year) but not hours
The paper proceeds as follows In the next section I introduce the data used for this
study and provide summary statistics on worker demographics employment employer
4
2 DATA
behavior and variation in state minimum wages In section 3 I introduce the data and
show recent trends in worker demographics and employment that are relevant to this
study In section 4 I introducte my empirical methodology and show estimates of labor
supply across demographic characteristics In section 5 I discuss employer options for
increasing intensive margin labor supply Section 6 summarizes the findings discusses
policy implications and suggests directions for future research
2 Data
The data for this paper come from the National Agricultural Workers Survey (NAWS)
The NAWS is an employment-based repeated cross-sectional survey administered an-
nually by the US Department of Labor The survey began in 1989 and data are
currently available through the 2016 survey round The survey is randomized and pro-
vides a nationally representative sample of workers employed in US crop production
The survey includes questions on household and worker demographics income legal
status wages hours worked weeks worked and various characteristics of a workerrsquos
current job Because NAWS provides detailed information on worker demographics and
employment it is the best available data for this analysis2
In my analysis I drop the first four years of the survey as several relevant variables
were not collected until the 1993 survey round I remove workers from the sample
who fall under the lsquosupervisorrsquo category and those who are salaried I keep low-skilled
and semi-skilled workers who are paid hourly piece rate or a combination of hourly
and piece rate I focus on these workers to minimize demand-side drivers of short-run
labor supply A major concern with examining labor supply comes from conflating
the supply decisions of a worker with the labor demands from the employer In most
industries workers have little control over the number of hours they work each weekmdash
service industry workers are put on a schedule office workers typically work either full-
or part-time (40 or 20 hours per week) and contract workers work until they complete
their task In these industries labor supply determinants can be interpreted as affecting
the workerrsquos willingness to agree to the contract In agriculture most non-managerial
jobs give workers considerable flexibility in their labor provision because most work
without contracts and without consequence for missing work days
Workers in the NAWS are first asked where they were born and foreign born workers
are later asked for their specific legal status Based on these legal statuses workers are
categorized as citizens green card holders undocumented or other work authorized I
2For this analysis I use the NAWS restricted access data so I can include controls and instruments for
potentially endogenous variables at the state rather than region level
5
2 DATA
exclude workers who are categorized as lsquoother authorizedrsquo The number of respondents
in this category is low (750) making it challenging to examine their behavior separately
from other groups of foreign born workers Rather than grouping them together with
foreign born workers I remove them from the sample I additionally divide citizens into
native and foreign born and group together foreign born citizens and green card holders
as foreign born legals3 The final legal status groups that I use are native citizens
foreign born legals and undocumented
A limitation to this study is that I do not examine the demographics or labor
provision of H2A Visa Holders (this is the work visa for temporary agricultural work)
These workers are not captured in the NAWS (the lsquoother authorizedrsquo category does not
include H2A workers) or any other nationally representative survey but are becoming
increasingly important for US agriculture As such my findings are only relevant for
currently employed workers who are not visiting on the agricultural visa
In my analysis I use information on worker demographic characteristics that in-
cludes age gender marital status parental status nativity and legal status I also use
information on employment that includes the payment type crop task wage employer-
offered benefits and weekly hours worked for the current farm job as well as annual
weeks worked in agriculture I exclude workers with missing values for these variables
from my analysis I summarize these variables in Tables 1 and 2
Table 1 shows the mean and standard deviation of the metrics of labor supply job
characteristics and employer provisions The first column shows information for the
entire sample and the remaining three columns divide the sample based on worker legal
status The two labor supply variables I use are hours worked per week and weeks
worked per year Hours per week are the number of hours spent working in the week
prior to the interview for the current farm employer This does not contain farmwork
for other employers or non-farmwork Weeks per year are the number of weeks spent
working in farmwork during the previous year Respondents are told to count all weeks
where they worked at least one day This includes weeks worked for fall farm employers
but does not include for non-farm jobs I remove workers from the sample who worked
zero weeks in the prior year (these are generally recent arrivals) The average worker
in the sample works 37 weeks per year and 44 hours per week This varies across legal
statuses with natives working the least (42 hours per week and 35 weeks per year) and
foreign born legals working the most (45 hours a week and 39 weeks per year)
3Foreign born legals are comprised of 88 green card holders and 12 foreign born citizens
6
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
1 INTRODUCTION
1 Introduction
The profile of US agricultural workers is changing and little is understood about the
implications for their employers Farmworkers are aging increasingly female and are
more likely to have children (Hernandez et al 2016) In addition tightening immi-
gration enforcement is expected to reduce the number of undocumented workers in the
labor force (Kostandini et al 2014 Zahniser et al 2012) Previous work has linked
heterogeneity across these demographics with differential behavior in the labor market
These studies suggest that the changing profile of farmworkers will decrease the aggre-
gate labor supply of US agricultural workers This could have large negative effects
on US agricultural production and profits
This paper shows how the changing profile of agricultural workers will affect labor
supply on the intensive margin ie the labor supply of those employed in agriculture I
compare the labor supply of US agricultural workers across four dimensions age gen-
der family composition and legal status The labor supply estimates reflect a workerrsquos
short-run choice of the number of hours to work each week and the number of weeks
worked each year I combine these estimates with an analysis of trends in the work-
force composition to predict future labor supply I then explore several avenues through
which employers can increase labor supply In particular I examine how wages health
care coverage and bonuses affect hours and weeks of work for employed agricultural
workers
Recent estimates of labor supply for US agricultural workers are limited (Emerson
amp Roka 2002 Pena 2010 Taylor amp Thilmany 1993) Significant changes in US im-
migration policy and the profile of workers have resulted in major structural changes
in the agricultural labor market that are not captured in existing estimates (Boucher
et al 2012 Martin amp Calvin 2010 Taylor 2010) Importantly literature studying
other industries and the US labor market more broadly find significant differences in
worker labor supply across many of the demographic dimensions in which the agricul-
tural workforce is changing
A large literature focuses on the labor supply of immigrants and the effects of
legal status on various labor market outcomes Generally this literature finds that
legalization is associated with increases of 5 to 15 in the workersrsquo real wage rate and
that these effects are largest for more skilled workers (Kaushal 2006 Kossoudji amp Cobb-
Clark 2002 Lofstrom et al 2013 Rivera-Batiz 1999) Recent work by Borjas (2017)
is the first to examine labor supply elasticities based on immigrant status He finds a
significant difference in the elasticity of labor supply between US workers of different
immigration statuses He finds that undocumented men have a wage elasticity close to
zero while the elasticity for US native men is closer to 04 Here I add to this finding
2
1 INTRODUCTION
and examine the labor supply differential across the legal statuses of US agricultural
workers Many of these workers are foreign born (76) and undocumented (50)1
This combined with the flexible nature of agricultural employment (ie limited use of
formal employment contracts) makes agricultural workers a useful group for examining
differential labor market behavior across legal statuses Further while Borjas and others
impute worker legal statuses to categorize workers as lsquolikely undocumentedrsquo I rely on
workersrsquo self-reported legal statuses To the best of my knowledge Pena (2010) provides
the only evidence on differences in the labor supply of agricultural workers across self-
reported legal statuses Using the same data source as this paper Pena finds that
foreign born workers are more likely to work in piece rate pay positions and that this
is associated with higher levels of poverty
Another body of literature examines differences in labor supply across gender and
family composition (marital and parental status) This literature shows that the labor
supply of women is lower than that of men that married women are more responsive
to changes in wages (ie have higher wage-elasticities) than single women single men
or married men and that children lead to a reduction in labor supply for females but
not for males (Angrist amp Evans 1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 Mc-
Clelland amp Mok 2012 Meyer amp Rosenbaum 2001) Because of these well-documented
differences in labor market behavior the common practice is to separately examine the
labor supply of married and single males and females Here I include all workers in the
same regression and control for these characteristics This allows me to compare labor
supply between these groups and ultimately enables me to show how future changes
across these demographics will affect industry labor supply
There is a small but growing literature on the labor market effects of workforce aging
Because workers of different ages behave heterogeneously in the workforce the norm
in the literature is to focus on working-age adults (those ages 16 - 65) and control for
age cohort so that labor supply differences are estimated from within groups of similar-
aged workers (Blundell amp Macurdy 1999) More recently due to the rising average
age of workers in the US and many other countries age has become an explanatory
variable of interest rather than a necessary control Evidence in the United States
suggests that productivity and labor force participation are significantly decreasing in
age (Fallick amp Pingle 2010 2006 Maestas et al 2016 Sheiner 2014) In this paper I
examine the relationship between worker age and labor supply among those employed
in US agriculture To the best of my knowledge this is the first empirical evidence on
the intensive margin effects of an aging workforce
This paper builds on existing literature in three important ways This paper is
1Estimates are from my own analysis of the NAWS data for the most recent survey round (2015-2016)
3
1 INTRODUCTION
the first comprehensive study of how the changing demographic profile of agricultural
workers will affect industry labor supply I find that native born citizens work fewer
hours per week and fewer weeks per year than foreign born legal or undocumented
workers Mid-aged workers (25 - 44 year olds) work more hours per week than younger
or older workers (16 - 24 and ge 45 respectively) but older workers work more weeks
each year Parents work significantly more weeks per year than non-parents but work a
similar number of hours Males work significantly more hours and weeks than females
In light of ongoing trends in these demographic characteristics my results imply that
the way in which the agricultural workforce is aging will cause the hours and weeks of
labor provided by employed farmworkers to increase while the changes in gender and
family composition will cause labor supply to decrease
Second this paper provides the first empirical evidence on multidimensional differ-
ences in labor supply across legal statuses and key demographic characteristics Existing
evidence has focused on separately identifying effects of legal status gender age mar-
ital status and parental status on labor supply This is the first study to examine
how these interact and provides important insights into heterogeneity in labor supply
across worker legal statuses Among my sample of US agricultural workers I find
surprisingly little heterogeneity across legal statuses and within demographic groups
This suggests that while on average natives foreign born legals and undocumented
workers have different labor market behavior they behave similarly within age gender
marital status and parental status groups Two notable exceptions to this are single
women with children who are legal immigrants work 65 more hours per week than na-
tives and single men with children who are both legal and undocumented immigrants
work 55 more weeks per year than natives
Finally this paper provides empirical evidence on how specific employer actions can
influence the labor supply among those employed in agriculture I examine whether
employers can increase hours or weeks of work by offering higher wages health benefits
or pay bonuses I show evidence from naive OLS regressions that these employer policies
are significantly and positively correlated with labor supply outcomes However causal
evidence from an IV approach indicates that little of this correlation is causal Among
these policies I find that bonuses are the only ones that causally affect labor-hours and
labor weeks I find that offering a bonus causes the average worker to increase weekly
hours of labor by ten percent and to increase annual weeks working in agriculture by
65 weeks I also find evidence that bonuses paid to workers for staying through the
end of the season increases weeks of work (by 35 weeks per year) but not hours
The paper proceeds as follows In the next section I introduce the data used for this
study and provide summary statistics on worker demographics employment employer
4
2 DATA
behavior and variation in state minimum wages In section 3 I introduce the data and
show recent trends in worker demographics and employment that are relevant to this
study In section 4 I introducte my empirical methodology and show estimates of labor
supply across demographic characteristics In section 5 I discuss employer options for
increasing intensive margin labor supply Section 6 summarizes the findings discusses
policy implications and suggests directions for future research
2 Data
The data for this paper come from the National Agricultural Workers Survey (NAWS)
The NAWS is an employment-based repeated cross-sectional survey administered an-
nually by the US Department of Labor The survey began in 1989 and data are
currently available through the 2016 survey round The survey is randomized and pro-
vides a nationally representative sample of workers employed in US crop production
The survey includes questions on household and worker demographics income legal
status wages hours worked weeks worked and various characteristics of a workerrsquos
current job Because NAWS provides detailed information on worker demographics and
employment it is the best available data for this analysis2
In my analysis I drop the first four years of the survey as several relevant variables
were not collected until the 1993 survey round I remove workers from the sample
who fall under the lsquosupervisorrsquo category and those who are salaried I keep low-skilled
and semi-skilled workers who are paid hourly piece rate or a combination of hourly
and piece rate I focus on these workers to minimize demand-side drivers of short-run
labor supply A major concern with examining labor supply comes from conflating
the supply decisions of a worker with the labor demands from the employer In most
industries workers have little control over the number of hours they work each weekmdash
service industry workers are put on a schedule office workers typically work either full-
or part-time (40 or 20 hours per week) and contract workers work until they complete
their task In these industries labor supply determinants can be interpreted as affecting
the workerrsquos willingness to agree to the contract In agriculture most non-managerial
jobs give workers considerable flexibility in their labor provision because most work
without contracts and without consequence for missing work days
Workers in the NAWS are first asked where they were born and foreign born workers
are later asked for their specific legal status Based on these legal statuses workers are
categorized as citizens green card holders undocumented or other work authorized I
2For this analysis I use the NAWS restricted access data so I can include controls and instruments for
potentially endogenous variables at the state rather than region level
5
2 DATA
exclude workers who are categorized as lsquoother authorizedrsquo The number of respondents
in this category is low (750) making it challenging to examine their behavior separately
from other groups of foreign born workers Rather than grouping them together with
foreign born workers I remove them from the sample I additionally divide citizens into
native and foreign born and group together foreign born citizens and green card holders
as foreign born legals3 The final legal status groups that I use are native citizens
foreign born legals and undocumented
A limitation to this study is that I do not examine the demographics or labor
provision of H2A Visa Holders (this is the work visa for temporary agricultural work)
These workers are not captured in the NAWS (the lsquoother authorizedrsquo category does not
include H2A workers) or any other nationally representative survey but are becoming
increasingly important for US agriculture As such my findings are only relevant for
currently employed workers who are not visiting on the agricultural visa
In my analysis I use information on worker demographic characteristics that in-
cludes age gender marital status parental status nativity and legal status I also use
information on employment that includes the payment type crop task wage employer-
offered benefits and weekly hours worked for the current farm job as well as annual
weeks worked in agriculture I exclude workers with missing values for these variables
from my analysis I summarize these variables in Tables 1 and 2
Table 1 shows the mean and standard deviation of the metrics of labor supply job
characteristics and employer provisions The first column shows information for the
entire sample and the remaining three columns divide the sample based on worker legal
status The two labor supply variables I use are hours worked per week and weeks
worked per year Hours per week are the number of hours spent working in the week
prior to the interview for the current farm employer This does not contain farmwork
for other employers or non-farmwork Weeks per year are the number of weeks spent
working in farmwork during the previous year Respondents are told to count all weeks
where they worked at least one day This includes weeks worked for fall farm employers
but does not include for non-farm jobs I remove workers from the sample who worked
zero weeks in the prior year (these are generally recent arrivals) The average worker
in the sample works 37 weeks per year and 44 hours per week This varies across legal
statuses with natives working the least (42 hours per week and 35 weeks per year) and
foreign born legals working the most (45 hours a week and 39 weeks per year)
3Foreign born legals are comprised of 88 green card holders and 12 foreign born citizens
6
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
1 INTRODUCTION
and examine the labor supply differential across the legal statuses of US agricultural
workers Many of these workers are foreign born (76) and undocumented (50)1
This combined with the flexible nature of agricultural employment (ie limited use of
formal employment contracts) makes agricultural workers a useful group for examining
differential labor market behavior across legal statuses Further while Borjas and others
impute worker legal statuses to categorize workers as lsquolikely undocumentedrsquo I rely on
workersrsquo self-reported legal statuses To the best of my knowledge Pena (2010) provides
the only evidence on differences in the labor supply of agricultural workers across self-
reported legal statuses Using the same data source as this paper Pena finds that
foreign born workers are more likely to work in piece rate pay positions and that this
is associated with higher levels of poverty
Another body of literature examines differences in labor supply across gender and
family composition (marital and parental status) This literature shows that the labor
supply of women is lower than that of men that married women are more responsive
to changes in wages (ie have higher wage-elasticities) than single women single men
or married men and that children lead to a reduction in labor supply for females but
not for males (Angrist amp Evans 1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 Mc-
Clelland amp Mok 2012 Meyer amp Rosenbaum 2001) Because of these well-documented
differences in labor market behavior the common practice is to separately examine the
labor supply of married and single males and females Here I include all workers in the
same regression and control for these characteristics This allows me to compare labor
supply between these groups and ultimately enables me to show how future changes
across these demographics will affect industry labor supply
There is a small but growing literature on the labor market effects of workforce aging
Because workers of different ages behave heterogeneously in the workforce the norm
in the literature is to focus on working-age adults (those ages 16 - 65) and control for
age cohort so that labor supply differences are estimated from within groups of similar-
aged workers (Blundell amp Macurdy 1999) More recently due to the rising average
age of workers in the US and many other countries age has become an explanatory
variable of interest rather than a necessary control Evidence in the United States
suggests that productivity and labor force participation are significantly decreasing in
age (Fallick amp Pingle 2010 2006 Maestas et al 2016 Sheiner 2014) In this paper I
examine the relationship between worker age and labor supply among those employed
in US agriculture To the best of my knowledge this is the first empirical evidence on
the intensive margin effects of an aging workforce
This paper builds on existing literature in three important ways This paper is
1Estimates are from my own analysis of the NAWS data for the most recent survey round (2015-2016)
3
1 INTRODUCTION
the first comprehensive study of how the changing demographic profile of agricultural
workers will affect industry labor supply I find that native born citizens work fewer
hours per week and fewer weeks per year than foreign born legal or undocumented
workers Mid-aged workers (25 - 44 year olds) work more hours per week than younger
or older workers (16 - 24 and ge 45 respectively) but older workers work more weeks
each year Parents work significantly more weeks per year than non-parents but work a
similar number of hours Males work significantly more hours and weeks than females
In light of ongoing trends in these demographic characteristics my results imply that
the way in which the agricultural workforce is aging will cause the hours and weeks of
labor provided by employed farmworkers to increase while the changes in gender and
family composition will cause labor supply to decrease
Second this paper provides the first empirical evidence on multidimensional differ-
ences in labor supply across legal statuses and key demographic characteristics Existing
evidence has focused on separately identifying effects of legal status gender age mar-
ital status and parental status on labor supply This is the first study to examine
how these interact and provides important insights into heterogeneity in labor supply
across worker legal statuses Among my sample of US agricultural workers I find
surprisingly little heterogeneity across legal statuses and within demographic groups
This suggests that while on average natives foreign born legals and undocumented
workers have different labor market behavior they behave similarly within age gender
marital status and parental status groups Two notable exceptions to this are single
women with children who are legal immigrants work 65 more hours per week than na-
tives and single men with children who are both legal and undocumented immigrants
work 55 more weeks per year than natives
Finally this paper provides empirical evidence on how specific employer actions can
influence the labor supply among those employed in agriculture I examine whether
employers can increase hours or weeks of work by offering higher wages health benefits
or pay bonuses I show evidence from naive OLS regressions that these employer policies
are significantly and positively correlated with labor supply outcomes However causal
evidence from an IV approach indicates that little of this correlation is causal Among
these policies I find that bonuses are the only ones that causally affect labor-hours and
labor weeks I find that offering a bonus causes the average worker to increase weekly
hours of labor by ten percent and to increase annual weeks working in agriculture by
65 weeks I also find evidence that bonuses paid to workers for staying through the
end of the season increases weeks of work (by 35 weeks per year) but not hours
The paper proceeds as follows In the next section I introduce the data used for this
study and provide summary statistics on worker demographics employment employer
4
2 DATA
behavior and variation in state minimum wages In section 3 I introduce the data and
show recent trends in worker demographics and employment that are relevant to this
study In section 4 I introducte my empirical methodology and show estimates of labor
supply across demographic characteristics In section 5 I discuss employer options for
increasing intensive margin labor supply Section 6 summarizes the findings discusses
policy implications and suggests directions for future research
2 Data
The data for this paper come from the National Agricultural Workers Survey (NAWS)
The NAWS is an employment-based repeated cross-sectional survey administered an-
nually by the US Department of Labor The survey began in 1989 and data are
currently available through the 2016 survey round The survey is randomized and pro-
vides a nationally representative sample of workers employed in US crop production
The survey includes questions on household and worker demographics income legal
status wages hours worked weeks worked and various characteristics of a workerrsquos
current job Because NAWS provides detailed information on worker demographics and
employment it is the best available data for this analysis2
In my analysis I drop the first four years of the survey as several relevant variables
were not collected until the 1993 survey round I remove workers from the sample
who fall under the lsquosupervisorrsquo category and those who are salaried I keep low-skilled
and semi-skilled workers who are paid hourly piece rate or a combination of hourly
and piece rate I focus on these workers to minimize demand-side drivers of short-run
labor supply A major concern with examining labor supply comes from conflating
the supply decisions of a worker with the labor demands from the employer In most
industries workers have little control over the number of hours they work each weekmdash
service industry workers are put on a schedule office workers typically work either full-
or part-time (40 or 20 hours per week) and contract workers work until they complete
their task In these industries labor supply determinants can be interpreted as affecting
the workerrsquos willingness to agree to the contract In agriculture most non-managerial
jobs give workers considerable flexibility in their labor provision because most work
without contracts and without consequence for missing work days
Workers in the NAWS are first asked where they were born and foreign born workers
are later asked for their specific legal status Based on these legal statuses workers are
categorized as citizens green card holders undocumented or other work authorized I
2For this analysis I use the NAWS restricted access data so I can include controls and instruments for
potentially endogenous variables at the state rather than region level
5
2 DATA
exclude workers who are categorized as lsquoother authorizedrsquo The number of respondents
in this category is low (750) making it challenging to examine their behavior separately
from other groups of foreign born workers Rather than grouping them together with
foreign born workers I remove them from the sample I additionally divide citizens into
native and foreign born and group together foreign born citizens and green card holders
as foreign born legals3 The final legal status groups that I use are native citizens
foreign born legals and undocumented
A limitation to this study is that I do not examine the demographics or labor
provision of H2A Visa Holders (this is the work visa for temporary agricultural work)
These workers are not captured in the NAWS (the lsquoother authorizedrsquo category does not
include H2A workers) or any other nationally representative survey but are becoming
increasingly important for US agriculture As such my findings are only relevant for
currently employed workers who are not visiting on the agricultural visa
In my analysis I use information on worker demographic characteristics that in-
cludes age gender marital status parental status nativity and legal status I also use
information on employment that includes the payment type crop task wage employer-
offered benefits and weekly hours worked for the current farm job as well as annual
weeks worked in agriculture I exclude workers with missing values for these variables
from my analysis I summarize these variables in Tables 1 and 2
Table 1 shows the mean and standard deviation of the metrics of labor supply job
characteristics and employer provisions The first column shows information for the
entire sample and the remaining three columns divide the sample based on worker legal
status The two labor supply variables I use are hours worked per week and weeks
worked per year Hours per week are the number of hours spent working in the week
prior to the interview for the current farm employer This does not contain farmwork
for other employers or non-farmwork Weeks per year are the number of weeks spent
working in farmwork during the previous year Respondents are told to count all weeks
where they worked at least one day This includes weeks worked for fall farm employers
but does not include for non-farm jobs I remove workers from the sample who worked
zero weeks in the prior year (these are generally recent arrivals) The average worker
in the sample works 37 weeks per year and 44 hours per week This varies across legal
statuses with natives working the least (42 hours per week and 35 weeks per year) and
foreign born legals working the most (45 hours a week and 39 weeks per year)
3Foreign born legals are comprised of 88 green card holders and 12 foreign born citizens
6
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
1 INTRODUCTION
the first comprehensive study of how the changing demographic profile of agricultural
workers will affect industry labor supply I find that native born citizens work fewer
hours per week and fewer weeks per year than foreign born legal or undocumented
workers Mid-aged workers (25 - 44 year olds) work more hours per week than younger
or older workers (16 - 24 and ge 45 respectively) but older workers work more weeks
each year Parents work significantly more weeks per year than non-parents but work a
similar number of hours Males work significantly more hours and weeks than females
In light of ongoing trends in these demographic characteristics my results imply that
the way in which the agricultural workforce is aging will cause the hours and weeks of
labor provided by employed farmworkers to increase while the changes in gender and
family composition will cause labor supply to decrease
Second this paper provides the first empirical evidence on multidimensional differ-
ences in labor supply across legal statuses and key demographic characteristics Existing
evidence has focused on separately identifying effects of legal status gender age mar-
ital status and parental status on labor supply This is the first study to examine
how these interact and provides important insights into heterogeneity in labor supply
across worker legal statuses Among my sample of US agricultural workers I find
surprisingly little heterogeneity across legal statuses and within demographic groups
This suggests that while on average natives foreign born legals and undocumented
workers have different labor market behavior they behave similarly within age gender
marital status and parental status groups Two notable exceptions to this are single
women with children who are legal immigrants work 65 more hours per week than na-
tives and single men with children who are both legal and undocumented immigrants
work 55 more weeks per year than natives
Finally this paper provides empirical evidence on how specific employer actions can
influence the labor supply among those employed in agriculture I examine whether
employers can increase hours or weeks of work by offering higher wages health benefits
or pay bonuses I show evidence from naive OLS regressions that these employer policies
are significantly and positively correlated with labor supply outcomes However causal
evidence from an IV approach indicates that little of this correlation is causal Among
these policies I find that bonuses are the only ones that causally affect labor-hours and
labor weeks I find that offering a bonus causes the average worker to increase weekly
hours of labor by ten percent and to increase annual weeks working in agriculture by
65 weeks I also find evidence that bonuses paid to workers for staying through the
end of the season increases weeks of work (by 35 weeks per year) but not hours
The paper proceeds as follows In the next section I introduce the data used for this
study and provide summary statistics on worker demographics employment employer
4
2 DATA
behavior and variation in state minimum wages In section 3 I introduce the data and
show recent trends in worker demographics and employment that are relevant to this
study In section 4 I introducte my empirical methodology and show estimates of labor
supply across demographic characteristics In section 5 I discuss employer options for
increasing intensive margin labor supply Section 6 summarizes the findings discusses
policy implications and suggests directions for future research
2 Data
The data for this paper come from the National Agricultural Workers Survey (NAWS)
The NAWS is an employment-based repeated cross-sectional survey administered an-
nually by the US Department of Labor The survey began in 1989 and data are
currently available through the 2016 survey round The survey is randomized and pro-
vides a nationally representative sample of workers employed in US crop production
The survey includes questions on household and worker demographics income legal
status wages hours worked weeks worked and various characteristics of a workerrsquos
current job Because NAWS provides detailed information on worker demographics and
employment it is the best available data for this analysis2
In my analysis I drop the first four years of the survey as several relevant variables
were not collected until the 1993 survey round I remove workers from the sample
who fall under the lsquosupervisorrsquo category and those who are salaried I keep low-skilled
and semi-skilled workers who are paid hourly piece rate or a combination of hourly
and piece rate I focus on these workers to minimize demand-side drivers of short-run
labor supply A major concern with examining labor supply comes from conflating
the supply decisions of a worker with the labor demands from the employer In most
industries workers have little control over the number of hours they work each weekmdash
service industry workers are put on a schedule office workers typically work either full-
or part-time (40 or 20 hours per week) and contract workers work until they complete
their task In these industries labor supply determinants can be interpreted as affecting
the workerrsquos willingness to agree to the contract In agriculture most non-managerial
jobs give workers considerable flexibility in their labor provision because most work
without contracts and without consequence for missing work days
Workers in the NAWS are first asked where they were born and foreign born workers
are later asked for their specific legal status Based on these legal statuses workers are
categorized as citizens green card holders undocumented or other work authorized I
2For this analysis I use the NAWS restricted access data so I can include controls and instruments for
potentially endogenous variables at the state rather than region level
5
2 DATA
exclude workers who are categorized as lsquoother authorizedrsquo The number of respondents
in this category is low (750) making it challenging to examine their behavior separately
from other groups of foreign born workers Rather than grouping them together with
foreign born workers I remove them from the sample I additionally divide citizens into
native and foreign born and group together foreign born citizens and green card holders
as foreign born legals3 The final legal status groups that I use are native citizens
foreign born legals and undocumented
A limitation to this study is that I do not examine the demographics or labor
provision of H2A Visa Holders (this is the work visa for temporary agricultural work)
These workers are not captured in the NAWS (the lsquoother authorizedrsquo category does not
include H2A workers) or any other nationally representative survey but are becoming
increasingly important for US agriculture As such my findings are only relevant for
currently employed workers who are not visiting on the agricultural visa
In my analysis I use information on worker demographic characteristics that in-
cludes age gender marital status parental status nativity and legal status I also use
information on employment that includes the payment type crop task wage employer-
offered benefits and weekly hours worked for the current farm job as well as annual
weeks worked in agriculture I exclude workers with missing values for these variables
from my analysis I summarize these variables in Tables 1 and 2
Table 1 shows the mean and standard deviation of the metrics of labor supply job
characteristics and employer provisions The first column shows information for the
entire sample and the remaining three columns divide the sample based on worker legal
status The two labor supply variables I use are hours worked per week and weeks
worked per year Hours per week are the number of hours spent working in the week
prior to the interview for the current farm employer This does not contain farmwork
for other employers or non-farmwork Weeks per year are the number of weeks spent
working in farmwork during the previous year Respondents are told to count all weeks
where they worked at least one day This includes weeks worked for fall farm employers
but does not include for non-farm jobs I remove workers from the sample who worked
zero weeks in the prior year (these are generally recent arrivals) The average worker
in the sample works 37 weeks per year and 44 hours per week This varies across legal
statuses with natives working the least (42 hours per week and 35 weeks per year) and
foreign born legals working the most (45 hours a week and 39 weeks per year)
3Foreign born legals are comprised of 88 green card holders and 12 foreign born citizens
6
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
2 DATA
behavior and variation in state minimum wages In section 3 I introduce the data and
show recent trends in worker demographics and employment that are relevant to this
study In section 4 I introducte my empirical methodology and show estimates of labor
supply across demographic characteristics In section 5 I discuss employer options for
increasing intensive margin labor supply Section 6 summarizes the findings discusses
policy implications and suggests directions for future research
2 Data
The data for this paper come from the National Agricultural Workers Survey (NAWS)
The NAWS is an employment-based repeated cross-sectional survey administered an-
nually by the US Department of Labor The survey began in 1989 and data are
currently available through the 2016 survey round The survey is randomized and pro-
vides a nationally representative sample of workers employed in US crop production
The survey includes questions on household and worker demographics income legal
status wages hours worked weeks worked and various characteristics of a workerrsquos
current job Because NAWS provides detailed information on worker demographics and
employment it is the best available data for this analysis2
In my analysis I drop the first four years of the survey as several relevant variables
were not collected until the 1993 survey round I remove workers from the sample
who fall under the lsquosupervisorrsquo category and those who are salaried I keep low-skilled
and semi-skilled workers who are paid hourly piece rate or a combination of hourly
and piece rate I focus on these workers to minimize demand-side drivers of short-run
labor supply A major concern with examining labor supply comes from conflating
the supply decisions of a worker with the labor demands from the employer In most
industries workers have little control over the number of hours they work each weekmdash
service industry workers are put on a schedule office workers typically work either full-
or part-time (40 or 20 hours per week) and contract workers work until they complete
their task In these industries labor supply determinants can be interpreted as affecting
the workerrsquos willingness to agree to the contract In agriculture most non-managerial
jobs give workers considerable flexibility in their labor provision because most work
without contracts and without consequence for missing work days
Workers in the NAWS are first asked where they were born and foreign born workers
are later asked for their specific legal status Based on these legal statuses workers are
categorized as citizens green card holders undocumented or other work authorized I
2For this analysis I use the NAWS restricted access data so I can include controls and instruments for
potentially endogenous variables at the state rather than region level
5
2 DATA
exclude workers who are categorized as lsquoother authorizedrsquo The number of respondents
in this category is low (750) making it challenging to examine their behavior separately
from other groups of foreign born workers Rather than grouping them together with
foreign born workers I remove them from the sample I additionally divide citizens into
native and foreign born and group together foreign born citizens and green card holders
as foreign born legals3 The final legal status groups that I use are native citizens
foreign born legals and undocumented
A limitation to this study is that I do not examine the demographics or labor
provision of H2A Visa Holders (this is the work visa for temporary agricultural work)
These workers are not captured in the NAWS (the lsquoother authorizedrsquo category does not
include H2A workers) or any other nationally representative survey but are becoming
increasingly important for US agriculture As such my findings are only relevant for
currently employed workers who are not visiting on the agricultural visa
In my analysis I use information on worker demographic characteristics that in-
cludes age gender marital status parental status nativity and legal status I also use
information on employment that includes the payment type crop task wage employer-
offered benefits and weekly hours worked for the current farm job as well as annual
weeks worked in agriculture I exclude workers with missing values for these variables
from my analysis I summarize these variables in Tables 1 and 2
Table 1 shows the mean and standard deviation of the metrics of labor supply job
characteristics and employer provisions The first column shows information for the
entire sample and the remaining three columns divide the sample based on worker legal
status The two labor supply variables I use are hours worked per week and weeks
worked per year Hours per week are the number of hours spent working in the week
prior to the interview for the current farm employer This does not contain farmwork
for other employers or non-farmwork Weeks per year are the number of weeks spent
working in farmwork during the previous year Respondents are told to count all weeks
where they worked at least one day This includes weeks worked for fall farm employers
but does not include for non-farm jobs I remove workers from the sample who worked
zero weeks in the prior year (these are generally recent arrivals) The average worker
in the sample works 37 weeks per year and 44 hours per week This varies across legal
statuses with natives working the least (42 hours per week and 35 weeks per year) and
foreign born legals working the most (45 hours a week and 39 weeks per year)
3Foreign born legals are comprised of 88 green card holders and 12 foreign born citizens
6
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
2 DATA
exclude workers who are categorized as lsquoother authorizedrsquo The number of respondents
in this category is low (750) making it challenging to examine their behavior separately
from other groups of foreign born workers Rather than grouping them together with
foreign born workers I remove them from the sample I additionally divide citizens into
native and foreign born and group together foreign born citizens and green card holders
as foreign born legals3 The final legal status groups that I use are native citizens
foreign born legals and undocumented
A limitation to this study is that I do not examine the demographics or labor
provision of H2A Visa Holders (this is the work visa for temporary agricultural work)
These workers are not captured in the NAWS (the lsquoother authorizedrsquo category does not
include H2A workers) or any other nationally representative survey but are becoming
increasingly important for US agriculture As such my findings are only relevant for
currently employed workers who are not visiting on the agricultural visa
In my analysis I use information on worker demographic characteristics that in-
cludes age gender marital status parental status nativity and legal status I also use
information on employment that includes the payment type crop task wage employer-
offered benefits and weekly hours worked for the current farm job as well as annual
weeks worked in agriculture I exclude workers with missing values for these variables
from my analysis I summarize these variables in Tables 1 and 2
Table 1 shows the mean and standard deviation of the metrics of labor supply job
characteristics and employer provisions The first column shows information for the
entire sample and the remaining three columns divide the sample based on worker legal
status The two labor supply variables I use are hours worked per week and weeks
worked per year Hours per week are the number of hours spent working in the week
prior to the interview for the current farm employer This does not contain farmwork
for other employers or non-farmwork Weeks per year are the number of weeks spent
working in farmwork during the previous year Respondents are told to count all weeks
where they worked at least one day This includes weeks worked for fall farm employers
but does not include for non-farm jobs I remove workers from the sample who worked
zero weeks in the prior year (these are generally recent arrivals) The average worker
in the sample works 37 weeks per year and 44 hours per week This varies across legal
statuses with natives working the least (42 hours per week and 35 weeks per year) and
foreign born legals working the most (45 hours a week and 39 weeks per year)
3Foreign born legals are comprised of 88 green card holders and 12 foreign born citizens
6
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
2 DATA
Table 1 Sample Summary Statistics Labor Supply and Job Attributes
All workers Natives Foreign born legal UndocumentedLabor Supply
hours per week 436767 424323 452110 433080(1326) (1460) (1286) (1283)
weeks per year 368574 345758 385333 367902(1514) (1665) (1275) (1562)
Payment Typehourly 08461 09466 08422 08089
(036) (022) (036) (039)piece rate 01333 00442 01318 01689
(034) (021) (034) (037)combined hourlypiece rate 00206 00092 00260 00222
(014) (010) (016) (015)Task
pre-harvest 02200 02226 02017 02291(041) (042) (040) (042)
harvest 02671 01649 02448 03206(044) (037) (043) (047)
post-harvest 01166 01549 01093 01051(032) (036) (031) (031)
semi-skilled 02619 02515 03335 02267(044) (043) (047) (042)
other 01344 02061 01107 01186(034) (040) (031) (032)
Crop Categoryfield crops 01468 02759 01176 01125
(035) (045) (032) (032)fruits amp nuts 03583 01423 04473 03953
(048) (035) (050) (049)horticulture 01926 02813 01443 01833
(039) (045) (035) (039)vegetables 02452 02138 02418 02591
(043) (041) (043) (044)miscellaneous 00571 00867 00491 00499
(023) (028) (022) (022)Employer Provisions
wage ($hour) 77947 81674 80100 75270(273) (298) (282) (254)
bonus 03096 04408 03637 02232(046) (050) (048) (042)
season bonus 00978 01293 01236 00692(030) (034) (033) (025)
health coverage (on the job) 08397 08774 08907 07924(037) (033) (031) (041)
health coverage (off the job) 01274 01977 01652 00740(033) (040) (037) (026)
Standard deviation in parentheses
7
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
2 DATA
Workers in the sample are paid either hourly piece rate or a combination of the
two 85 of all workers are paid hourly and 13 are paid by the piece Almost all
natives are paid hourly (95) while lower proportions of foreign born legal and undoc-
umented workers are paid hourly (84 and 81 percent respectively) A larger proportion
of undocumented workers are paid piece rate (17) compared with natives and foreign
born legals The NAWS records the specific job of each worker interviewed and then
categorizes workers into six broader categories After removing workers categorized as
supervisors from the sample the five remaining task codes are pre-harvest harvest
post-harvest semi-skilled and lsquootherrsquo Most workers report being employed in harvest
or semi-skilled tasks and fewer report doing post-harvest tasks The biggest difference
in tasks between worker legal statuses is for harvesting A larger proportion of undocu-
mented workers are employed in harvesting tasks (32) while a very small proportion
of natives do these tasks (16) The NAWS also collects information on the crop the
farmworker is currently working with and codes these into the five categories listed in
Table 1 Not surprisingly given its high labor requirements fruit and nut production
employs the largest proportion of workers (36) followed by vegetable production and
horticulture This is not the case for native workers who are primarily employed in
horticulture and field crops (28) Relatively few natives work in fruit and nut pro-
duction (14) compared with foreign born legals (45) and undocumented workers
(40)
The last statistics in Table 1 summarize the variables on provisions from the current
farm employer Wage is the self-reported (before tax) hourly wage rate if the worker
is paid by the hour and the imputed hourly equivalent for workers paid by the piece
or combined hourly and piece rate4 The average worker in the sample receives 780
$hour from their current farm employer This is lower among undocumented workers
(750) and higher among foreign born legals and natives (800 and 815)
Bonus is an indicator variable for whether the worker receives any money besides
wages from their current employer Season bonus is an indicator variable for whether
this bonus is an end of season bonus (ie the money is paid to them if they work for
the employer through the end of the season) About 30 of workers receive some form
of bonus and 10 of workers receive an end of season bonus Both types of bonus are
more common for natives and foreign born legals than for undocumented workers 44
of natives receive any bonus and 13 receive an end of season bonus 36 of foreign
born legals receive any bonus and 12 receive an end of season bonus but only 22 of
undocumented workers receive a bonus and 7 receive an end of season bonus
4The hourly equivalent is estimated using the reported piece rate wage average daily productivity and
average hours worked
8
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
2 DATA
Finally the variables health coverage (on the job) and health coverage (off the job) are
indicators for whether the employer provides heath care coverage for injuriesillnesses
that occur on and off the job respectively The majority of workers (84) receive
health coverage for on-the-job issues while only 13 receive coverage for off-the-job
issues Undocumented workers are least frequently covered by both forms of health care
79 of undocumented workers report coverage for on-the-job illnesses and 7 report
coverage for off-the-job Comparatively 88 of natives report on-the-job coverage and
20 report off-the-job coverage
Table 2 shows summary statistics of the worker characteristics I include in this
study About 20 of workers are natives 29 are foreign born legals and 51 are
undocumented There is considerable variation in the education of workers in the three
legal statuses The majority of natives have completed high school (56 have 12 or
more years of education) whereas the majority of foreign born legals and undocumented
workers have not completed elementary school (72 and 66 have less than 8 years of
education respectively) On average the workforce is close to evenly divided between
the four age categories but this varies across the legal statuses Foreign born legal
workers are older than both natives and undocumented workers 72 of foreign born
legal workers are 35 and older and only 7 are under 25 Comparatively 70 of
undocumented workers are less than 35 years old and only 11 are 45 and older
The family composition variables show that most workers are either single and
childless (34) or married with children (45) Large proportions of natives and un-
documented workers are single and without children (48 and 40 respectively) while
comparatively few foreign born legal workers fall into this category (15) Foreign born
legals have the highest proportion of workers married with children (61) followed by
undocumented workers (44) and natives (26) Across all legal statuses single work-
ers with children are the least common family structure The majority of all workers
are men (80) Natives have the lowest proportion of men (73) and undocumented
workers have the highest (83)
9
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
2 DATA
Table 2 Sample Summary Statistics Worker Characteristics
All workers Natives Foreign born legal UndocumentedLegal Status
Native citizen 02031(040)
Foreign born legal 02868(045)
Undocumented 05100(050)
Educationle 8 years 05691 01015 07226 06592
(050) (030) (045) (047)8 - 11 years 02510 03410 01826 02563
(043) (047) (039) (044)ge 12 years 01799 05576 00948 00844
(038) (050) (029) (028)Age
16 - 24 02443 02813 00659 03321(043) (045) (025) (047)
25 - 34 02889 02072 02104 03658(045) (041) (041) (048)
35 - 44 02253 01968 03002 01938(042) (040) (046) (040)
ge 45 02416 03147 04235 01083(043) (046) (049) (031)
Family Compositionsingle no kids 03429 04758 01547 03968
(047) (050) (036) (049)single kids 00669 00871 00569 00639
(025) (028) (023) (024)married no kids 01376 01794 01827 00950
(034) (038) (039) (029)married kids 04526 02577 06058 04442
(050) (044) (049) (050)Male 08024 07288 08027 08322
(040) (044) (040) (037)
Observations 54126 10870 15350 27295
Standard deviation in parentheses
10
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
3 Trends in Worker Characteristics
Descriptive evidence from several sources of data show that the farm workforce is aging
becoming more settled and is increasingly female5 Here I show trends in these worker
demographics as represented in the NAWS Consistent with existing work I show that
the farm workforce is aging I find that this is driven by decreases in the proportion of
workers under 25 and increases in the the proportion of workers above 45 I show that
the proportion of male workers is declining and that this is driven by decreases in the
proportion of undocumented males I show that an increasing proportion of workers
are married with children I find that this is driven by a decrease in the proportion of
single childless men and an increase in the proportion of married women with children
Finally I show that despite widespread claims that increased immigration enforcement
will reduce the number of undocumented workers in the agricultural labor force the pro-
portions of native born citizens foreign born legal workers and undocumented workers
has remained stable over the sample period
Figure 1 shows trends in the age of crop workers surveyed in the NAWS Panel (a)
shows that the average age of workers has been increasing steadily over the sample
period Over the sample period the average age of crop workers has increased by
roughly 25 increasing from 31 years old in 19931994 to 38 years old in 20152016
This is equivalent to an increase of roughly one percent per year Panel (b) shows that
this increase in the average age of workers is driven by a decrease in the proportion
of younger workers (under 25) and an increase in the proportion of older workers (45
and older) The proportion of workers in the middle age groups has remained relatively
stable Over the sample period the proportion of younger workers has decreased by
nearly 20 percentage points falling from 365 in 19931994 to 176 in 20152016
The proportion of older workers has increased by roughly the same amount rising from
147 in 19931994 to 335 in 20152016 This is equivalent to approximately a one
percent increase in workers 45 and older and a one percent decrease in workers under
25 each year
5For example see the summary of American Community Survey Data made available by the Eco-
nomic Research Service (available at httpswwwersusdagovtopicsfarm-economyfarm-labor)
the summary of NAWS data by Martin (2017) or the summary of the California Farm Bureaursquos 2017
Agricultural Labor Availability Survey (available at httpwwwcfbfuswp-contentuploads201710
CFBF-Ag-Labor-Availability-Report-2017pdf)
11
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
Figure 1 Trends in Age
(a) Average Ages
(b) Age Groups
Average worker ages are computed using self-reported ages and survey weightsprovided in the NAWS Workers under 16 years old are dropped from the sample(this removes 343 workers) Averages are estimated within each NAWS cycle (ieevery two fiscal years) because of the NAWS sampling technique
12
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
Figure 2 Trends in Age by Legal Status
(a) Average Age (b) Native born citizens
(c) Foreign born legal (d) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign born TheNAWS categorizes workers into four work authorization categories citizen green card holderother work authorized (including temporary work visas) and undocumented Here workers areseparated into native born citizens foreign born legal workers (including green card holders andcitizens) and undocumented workers Due to the small number of observations rsquoother workauthorizedrsquo workers are removed from the sample
Figure 2 shows trends in average age and age compositions across legal status groups
Panel (a) shows that the increasing average age of the workforce is driven by foreign born
workers The average age of native citizen workers was increasing from 1995 through
2006 but has remained relatively stable since Until very recently undocumented
workers have had the lowest average age followed by native born citizen workers and
with foreign born legal workers having the highest However in the last survey round
the average age of undocumented workers surpassed that of native born citizens
Panels (b) (c) and (d) show the age composition for native born citizens foreign
born legal workers and undocumented workers respectively Panel (b) shows that the
trends in the average age of native citizens were driven by the share of native born
citizens aged 45 and older The proportion of these workers decreased until 2006 and
has been increasing since The age composition of native born citizens looks remarkably
13
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
similar in 2016 and 1994 with a small increase in the proportion of older workers and
a small decrease in the proportion of younger workers
Comparatively the age composition of foreign born legal workers and undocumented
workers has changed substantially over the period The majority of foreign born legal
workers are aged 45 and older which might reflect some self-selection into legal status
ie foreign workers who have been in the country longer are more likely to have obtained
citizenship or a green card Over the sample period the proportion of foreign born legal
workers aged 45 and older has risen by nearly 40 percentage points while the proportion
of middle-aged workers (aged 25-44) has fallen by nearly the same amount While the
average age of undocumented workers has also been rising over the sample period the
compositional changes driving these averages is quite different At the start of the
sample workers under 25 make up roughly half of the undocumented population while
at the end of the sample they account for only thirteen percent This decrease in the
proportion of younger workers is accompanied by an increase in workers aged 35 and
older In all Figure 2 suggests that there are important differences in the trends in
worker ages across legal statuses If these trends persist they are likely to influence
aggregate labor market behavior and outcomes
Figure 3 shows trends in the household composition of farmworkers Panel (a) shows
trends in the marital and parental composition of the workforce Workers are categorizes
as married (a parent) based on whether they report having a spouse (child under 18)
regardless of whether the spouse (child) lives with them The most striking trend in
the marital-parental composition of the workforce is the decline in farmworkers who are
single without children This proportion has decreased by almost 15 percentage points
over the sample period falling from 44 percent in 199394 to 30 percent in 201516
This decrease is accompanied by a ten percent increase in the proportion of single
workers with children a three percent increase in the proportion of married workers
with children and a one percent increase in the proportion of married workers without
children
14
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
Figure 3 Trends in Family Composition
(a) Separated by Marital and Parental Status
(b) Separated by Gender and MaritalParental Status
Marital status and parental status are defined based on whether the farmworkerhas a spouse or child under 18 respectively An alternative approach is to examinetrends based on whether the spousechild live in the same household as the workerbut results (and trends) are similar under this family composition structure Theproportions are estimated for each NAWS cycle using the survey-provided sampleweights 15
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
Figure 4 Trends in Family Composition by Legal Status
(a) Native born citizens (b) Foreign born legal
(c) Undocumented
Legal status is assigned based on work authorization and whether the worker is foreign bornMarital status and parental status are defined based on whether the farmworker has a spouse orchild under 18 respectively
Existing literature on the labor supply effects of marital and parental status suggest
heterogeneity across gender More specifically the literature suggests that the labor
supply of married women is lower than that of single women or men and that chil-
dren lead to a reduction in labor supply for women but not men (Angrist amp Evans
1998 Blau amp Kahn 2007 Eissa amp Hoynes 2004 McClelland amp Mok 2012 Meyer
amp Rosenbaum 2001) Because of this I further divide family composition based on
worker gender Panel (b) of Figure 3 shows trends across each of these dimensions
ie gender marital and parental status Panel (b) shows that the decrease in single
workers without children is driven by males The proportion of single males without
children decreases by almost 15 percentage points over the period falling from 37 per-
cent in 199394 to 23 percent in 201516 while the proportion of single females without
children remains close to seven percent over the sample period The increase in single
16
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
workers with children is driven equally by females and males (both increasing by five
percentage points over the period Trends for males and females diverge when exam-
ining married workers with children The proportion of married women with children
increases by six percentage points over the sample while the proportion of males in
this category falls by three percentage points
Figure 4 shows trends across family composition and legal status Similarly to trends
in worker age and gender the trends in family composition varies significantly across
the three legal statuses Panel (a) shows that family composition has been stable across
the sample period for native born citizens This is not the case for foreign born legal
workers or undocumented workers who are depicted in Panel (b) and (c) respectively
Among foreign born legal workers the proportion of married men with children has
decreased by 15 percentage points over the sample period and the proportion of single
men without children has decreased by 9 percentage points while all other groups have
grown proportionally The largest compositional changes have occurred among undoc-
umented workers Panel (c) shows that the proportion of single men without children
has decreased by 25 percentage points over the sample period while the proportions of
married women with children and married men with children have increased by 13 and
4 percent respectively
Figure 5 shows trends in worker genders Unlike the trends for worker ages and
parental status the changes in gender composition of the workforce have emerged re-
cently The percentage of male farmworkers remained near 80 percent until 2006 and
has been steadily declining since In the most recent survey round approximately 67
percent of the workforce is male a 13 percent decrease over the last ten years Panel
(b) shows these trends separately by worker legal statuses Interestingly the stable
proportion of males in the farm workforce until 2006 is simultaneously driven by a
decreasing proportion of undocumented males and an increasing proportion of native
males Following 2006 the proportion of native males has stabilized while the propor-
tion of undocumented males has continued to fall
Finally Figure 6 shows trends in the legal status of crop workers Despite evidence
that increased immigration enforcement decreases immigrant presence (eg Kostandini
Mykerezi amp Escalante 2014) the legal status composition of employed farmworkers
has changed remarkably little over the sample period
17
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
3 TRENDS IN WORKER CHARACTERISTICS
Figure 5 Trends in Gender
(a) All Workers
(b) By Legal Status
This shows the proportion of farmworkers who are male The proportion is esti-mated for each NAWS cycle using the survey-provided sample weights
18
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
Figure 6 Trends in Legal Statuses
Citizens include native born citizens and foreign born (naturalized) citizens Thisdoes not include workers categorized as having rsquoother work authorizationrsquo due tothe limited coverage of agricultural visa holders in the NAWS
4 Predictors of Labor Supply
In this section I estimate heterogeneity in labor supply across the demographic char-
acteristics summarized in the previous section The main estimating equation is an
OLS regression with state year and state-by-year fixed effects This equation can be
written
yi = α+ β1legal statusi + β2age groupi + β3education groupi
+ β4(marriedi times parenti times genderi) +Eprimeiγ + λs + λt + λst + ui
(1)
Where yi is the measure of labor supply either logged weekly hours worked or weeks
worked each year legal statusi is a categorical variable indicating whether the worker
is native foreign born legal or undocumented age groupi is a categorical variable for
worker age divided into four groups (lt25 25-34 35-44 and ge45) education groupi is a
categorical variable for education level (lt9 years 9-12 years and gt12 years) marriedi
is a binary indicator variable for a workerrsquos marital status parenti is an indicator
19
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
variable for whether the farmworker has children6 genderi is a binary indicator variable
equal to one if the worker is male and zero if they are female and Ei is a vector of
control variables for employment characteristics Employment characteristics include
categorical variables for the payment scheme task and crop of the workerrsquos current
farm job λs λt and λst are state year and state-by-year fixed effects These fixed
effects ensure that the results are not driven by time trends or spacial heterogeneity
and instead are estimated from differences in worker behavior within each state and
year
Table 3 shows coefficient estimates for Equation 1 and Table 4 shows the associated
average adjusted predictions (predictive margins) In both tables the first column
shows results from the regression with logged hours of work as the outcome variable
Table 3 shows that both foreign born legal and undocumented workers on average
provide more hours of labor per week than native workers Foreign born legals work
75 more hours and undocumented work 43 more Workers aged 25 to 44 provide
roughly 3 more hours of labor each week than those under 25 and those 45 and
older Married women with children work significantly fewer hours than single childless
women Men regardless of family composition work significantly more hours than
women On average I find that men work 14 more hours each week than women I find
that married men work significantly more hours per week than their single counterparts
and that married men with children work more than those without children (though
this difference is not significant) This behavior is consistent with the broader labor
supply literature which finds that the effect of an additional child for married couples
is negative for women (ie women decrease labor supply) and insignificant or positive
for men (ie men increase labor supply) (Angrist amp Evans 1998 Lundberg amp Rose
2002)
The first column of Table 4 translates these coefficients into the predicted hours of
work for workers of a given demographic holding all else constant These predictive
margins show that after netting out effects from all other predictive variables the
average foreign born legal worker works 41 hours per week undocumented workers
work 395 and natives work 38 Workers aged 25 to 44 work 40 hours a week while
those under 25 and older than 44 work 39 hours Married women with children work
355 hours a week while single women without children work 37 Men generally work
395 to 415 hours per week much higher than females who work 35 to 37 hours Men
who are married with children work the most at 415 hours each week while women
6I examine the effect of the farmworker being a parent regardless of whether they live with their child
The results are similar if I instead include an indicator variable for the farmworker being a parent and living
with their child Importantly the trends in parental status and parentalliving with the child are also similar
20
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
who are married with children work close to the fewest at 355 hours each week
The second columns of Tables 3 and 4 show results from the regression with number
of annual weeks working in a farm job as the outcome variable As shown in the second
column of Table 5 foreign born legal workers and undocumented workers work signif-
icantly more weeks per year than natives All workers 25 and older work significantly
more weeks than those under 25 and workers 45 and older work the most Married
women with children work significantly fewer weeks per year than single women with
children Men regardless of family composition work significantly more weeks than
women Men with children regardless of marital status work with most weeks per
year (but the difference between males is not statistically significant) Generally the
sign and relative magnitude of these coefficients mirror those in the first column but
there is one notable difference While the relationship between age and hours of work
is quadratic (ie increasing then decreasing) the relationship between age and weeks
of work is increasing (at a decreasing rate) Workers aged 25 to 44 work the most hours
per week but those 45 and older work the most weeks per year
21
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
Table 3 Labor Supply Differences Across Demographics
Outcome variablelog(hours) weeks worked
Foreign born legal 00745lowastlowastlowast 2421lowastlowastlowast(00186) (0750)
Undocumented 00429lowastlowast 1419lowast(00168) (0808)
Age 25 - 34 00346lowastlowastlowast 5708lowastlowastlowast(00100) (0318)
Age 35 - 44 00279lowast 7140lowastlowastlowast(00139) (0330)
Age ge 45 00000358 7366lowastlowastlowast(00135) (0585)
Single times child times female -00347 1017(00232) (0615)
Married times no child times female -00525 0418(00378) (1125)
Married times child times female -00375lowast -1120lowast(00191) (0645)
Single times no child times male 00700lowastlowastlowast 3661lowastlowastlowast(00202) (0724)
Single times child times male 00877lowastlowastlowast 4973lowastlowastlowast(00315) (1195)
Married times no child times male 0118lowastlowastlowast 4423lowastlowastlowast(00222) (0637)
Married times child times male 0121lowastlowastlowast 4990lowastlowastlowast(00212) (1047)
Constant 3694lowastlowastlowast 2643lowastlowastlowast(0216) (2829)
Observations 53406 54338R2 0250 0252Notes Robust standard errors in parentheses clustered at state The out-come variable log(hours) is the log of the number of hours worked in theweek prior to the interview for the workerrsquos current farm employer Theoutcome variable weeks worked is the total number of weeks in which theworker worked at least one day in farm work in the last year The base cate-gorical variables are as follows salaried workers for payment type less thanhigh school for education native born citizens for legal status and ages 16 -24 for age All regressions include state year state-by-year task and cropfixed effects Differences in the number of observations come from missingvalues (nonresponse) for the outcome variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast
p lt 001
22
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
Table 4 Labor Supply Differences Across Demographics Predictive Margins
Average adjusted predictions983142hours 983142 weeks
Legal StatusNative 3786 2947
Foreign born legal 4077lowastlowastlowast 3189lowastlowastlowast
Undocumented 3953lowastlowast 3088lowast
AgeAge 16 - 24 3882 2604
Age 25 - 34 4021lowastlowastlowast 3175lowastlowastlowast
Age 35 - 44 3992lowast 3318lowastlowastlowast
Age ge 45 3882 3340lowastlowastlowast
Family Composition amp GenderSingle X no child X female 3678 2750
Single X child X female 3555 2852
Married X no child X female 3492 2792
Married X child X female 3545lowast 2638lowast
Single X no child X male 3945lowastlowastlowast 3117lowastlowastlowast
Single X child X male 4017lowastlowastlowast 3248lowastlowastlowast
Married X no child X male 4143lowastlowastlowast 3193lowastlowastlowast
Married X child X male 4151lowastlowastlowast 3250lowastlowastlowast
Notes Predicted labor supply based on estimates from the fixed effect re-gressions presented in Table 3 Values for predicted hours are computedusing e
983142log(hours) Significance levels come from baseline regression and indi-cate that values are significantly different from the base (first) category ofeach variable lowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
The second column of Table 4 translates these coefficients into the predicted number
of annual working weeks These predictive margins show that the average foreign born
legal worker works 32 weeks per year undocumented workers work 31 and natives
work 29 Workers under 25 years old work 26 weeks per year those 25-34 work 32 and
those 35 and older work 33 Single women and married women without children work
roughly 28 weeks per year while married women with children work 26 Single men
without children work the fewest weeks per year (31) followed by married men without
children (32) and then men with children regardless of marital status (325) The
largest differences in annual weeks of labor supplied within these demographic groups
is between the youngest and oldest workers and workers (a seven week difference) and
between men and women who are married with children (a six week difference)
Results from Equation 1 shed light on differences in labor supply within these dis-
23
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
tinct demographic groups Because my analysis in the previous section suggests that
the trends in these characteristics are heterogeneous across legal status groups I am
also interested in examining labor supply differentials within demographic and legal
status groups The equation that estimates these effects can be written
yi = α+ β1age groupi + β2education groupi + β3(marriedi times parenti times genderi)
+ micro1(legal statusi times age groupi)
+ micro2(legal statusi timesmarriedi times parenti times genderi)
+Eprimeiγ + λs + λt + λst + ui
(2)
Where the new parameters legal statusitimesage groupi and legal statusitimesmarrieditimesparenti times genderi are the legal status and demographic characteristic interaction The
coefficients on these parameters capture differential labor supply across legal status
and within the demographic group For example a significant value of micro1 for foreign
born workers aged 45 and older indicates a statistically significant difference from native
born workers aged 45 and older
Table 5 shows coefficient estimates for Equation 2 and Table 6 shows the associated
predictive margins Again the first columns shows results from the regression with
logged hours of work as the outcome variable and the second columns show results with
annual weeks worked These results show little significant variation across legal statuses
The first column of Table 5 shows that among workers aged 45 and older both foreign
born legal and undocumented work more hours than natives Among single women with
children foreign born legals work significantly more hours than native citizens Finally
among married males with children undocumented immigrants work significantly fewer
hours than native citizens The first column of Table 6 shows that the average foreign
born legal and undocumented workers aged 45 and older work 40 and 39 hours per
week respectively This is roughly two hours more per week than native workers of the
same age Among single women with children legal immigrants work 65 hours more
per week than native citizens And among married males with children undocumented
immigrants work one hour less per week than natives
The second column of Table 5 shows no significant heterogeneity across legal statuses
within age groups but some significant differences within family composition groups
Compared with native citizens in the same groups legal immigrants who are female
married and do not have children work fewer weeks while legal immigrants who are
male single with children and undocumented immigrants who are male and single
regardless of parental status work more weeks
24
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
Table 5 Labor Supply Differences Across Demographics and Legal Status
Outcome variablelog(hours) weeks worked
Foreign born legal times Age 25 - 34 00389 1901(00400) (1645)
Foreign born legal times Age 35 - 44 00340 0550(00321) (1750)
Foreign born legal times Age ge 45 00756lowastlowast -1610(00372) (1692)
Undocumented times Age 25 - 34 00300 -0118(00315) (1266)
Undocumented times Age 35 - 44 000929 -0500(00272) (1301)
Undocumented times Age ge 45 00727lowastlowast -2007(00303) (1246)
Foreign born legal times Single times kids times female 0117lowast 0208(00646) (2031)
Foreign born legal times Married times no kids times female 00779 -3265lowast(00622) (1645)
Foreign born legal times Married times kids times female 00521 -2267(00617) (1576)
Foreign born legal times Single times no kids times male -00217 1220(00363) (1262)
Foreign born legal times Single times kids times male -00532 4463lowastlowast(00353) (2074)
Foreign born legal times Married times no kids times male 00206 1309(00415) (1765)
Foreign born legal times Married times kids times male -00413 1424(00343) (1676)
Undocumented times Single times kids times female 00510 2800(00651) (2143)
Undocumented times Married times no kids times female 00472 -1235(00585) (1570)
Undocumented times Married times kids times female 00424 -0465(00633) (1500)
Undocumented times Single times no kids times male -00201 2276lowast(00224) (1143)
Undocumented times Single times kids times male -00383 5288lowastlowast(00382) (2024)
Undocumented times Married times no kids times male -00271 -1371(00313) (1474)
Undocumented times Married times kids times male -00748lowastlowastlowast -1237(00272) (1400)
State-year FE yes yesTask amp crop FE yes yesObservations 53406 54338R2 0253 0260
Notes Robust standard errors in parentheses clustered at state Outcome variables andbase categorical variables same as prior regression (see Table 3 description) All regressionsinclude state year state-by-year task and crop fixed effects Differences in the number ofobservations come from missing values (nonresponse) for the outcome variable lowast p lt 010 lowastlowast
p lt 005 lowastlowastlowast p lt 001
25
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
4 PREDICTORS OF LABOR SUPPLY
Table 6 Legal Status Interactions Margins
Predicted values983142hours 983142 weeks
Status times AgeNative times Age 16 - 24 3659 2300Native times Age 25 - 34 4023 3114Native times Age 35 - 44 4032 3287Native times Age ge 45 3764 3414Foreign born legal times Age 16 - 24 4013 2681Foreign born legal times Age 25 - 34 4142 3381Foreign born legal times Age 35 - 44 4131 3419Foreign born legal times Age ge 45 4020lowastlowastlowast 3331Undocumented times Age 16 - 24 3954 2694Undocumented times Age 25 - 34 4024 3136Undocumented times Age 35 - 44 3951 3270Undocumented times Age ge 45 3929lowastlowastlowast 3247
Status times Family Composition amp GenderNative times single times no child times female 3539 2693Native times single times child times female 3243 2633Native times married times no child times female 3240 2833Native times married times child times female 3276 2684Native times single times no child times male 3833 2900Native times single times child times male 3965 2857Native times married times no child times male 3999 3152Native times married times child times male 4211 3234Foreign born legal times single times no child times female 3770 2809Foreign born legal times single times child times female 3884lowast 2769Foreign born legal times married times no child times female 3731 2622lowast
Foreign born legal times married times child times female 3677 2572Foreign born legal times single times no child times male 3995 3137Foreign born legal times single times child times male 4005 3419lowastlowastlowast
Foreign born legal times married times no child times male 4348 3398Foreign born legal times married times child times male 4304 3492Undocumented times single times no child times female 3743 2737Undocumented times single times child times female 3609 2957Undocumented times married times no child times female 3593 2753Undocumented times married times child times female 3615 2681Undocumented times single times no child times male 3973 3171lowast
Undocumented times single times child times male 4036 3430lowastlowastlowast
Undocumented times married times no child times male 4116 3058Undocumented times married times child times male 4133lowastlowastlowast 3154
Notes Predicted labor supply based on estimates from the fixed effect regressions pre-sented in Table 5 Values for predicted hours are computed using e
983142log(hours) Signif-icance levels come from baseline regression and indicate that values are significantlydifferent from native citizen workers within the same age or family composition categorylowast p lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
26
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
5 EMPLOYER STRATEGIES
The second column of Table 6 shows that married childless women who are legal
immigrants work 26 weeks per year while natives work 28 Single men with children
who are both legal and undocumented immigrants work 34 weeks per year while natives
work 285 Finally single men without children who are undocumented work 32 weeks
per year while natives work 29
Overall my results show that intensive margin labor supply varies significantly over
age family composition and legal status However I find little significant difference in
the labor supply of workers of different legal statuses within age and family composition
groups This suggests that future changes in the availability of labor-hours and labor-
weeks will be driven by average industry shifts in terms of age gender marital status
parental status and legal status with a few exceptions One of the most prominent
trends in the age composition of the workforce is the increase in the proportion of
foreign born legal workers ages 45 and above My results from estimating Equation 1
show that workers in this age group work more weeks per year and similar hours per
week compared with the youngest workers The results from estimating Equation 2
show no significant differences in the weeks worked between workers within age groups
and between legal status groups but do show significant increases in hours of work
Specifically I find that workers in the fastest growing age-legal status group (foreign
born legal workers ge 45) work significantly more hours per week In all my results
suggest that the way in which the workforce is aging will result in a labor force that is
willing to work more hours per week and weeks per year in agriculture
One of the most prominent trends in the gender-marital-parental composition of the
workforce is the decline in the proportion of undocumented single men without children
The results in Table 5 show that these workers provide among the highest weeks of
labor each year This suggests that the changing gender and family composition of the
workforce will result in a labor force that is willing to work fewer weeks per year in
agriculture Rising prevalence of married legal immigrant women without children will
contribute to this falling labor supply
5 Employer Strategies
The effects of the changing agricultural workforce on future labor supply is of interest to
industry stakeholders researchers and policy makers However a question of particular
interest to industry stakeholders is what can employers do to increase labor-hours and
labor-weeks Here I examine the viability of several alternative employer offerings for
increasing labor supply Naive regressions that do not account for the endogeneity
of these employer policies would suggest that offering health coverage bonuses and
27
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
5 EMPLOYER STRATEGIES
higher wages are all viable options However after accounting for the inherent bias
between these employer policies and labor supply (eg employers might pay higher
wages to more motivated workers) I find that bonuses are the only employer policy
that significantly affect labor supply Offering a bonus causes workers to work 10
more hours within a week and 65 more weeks within a year
I use an instrumental variable approach to estimate the causal effect of these em-
ployer policies on labor-hours and labor-weeks The regression equation can be written
in two stages
employer policyi = α+ θ1Pminusi +Dprimeiβ +Eprime
iγ + λs + λt + 983171i
yi = α+ θ1 983142employer policyi +Dprimeiβ +Eprime
iγ + λs + λt + ui(3)
Where the employer policyi is one of an indicator variable for the worker receiving
employer health coverage for off-the-job illnessesinjuries an indicator variable for the
worker receiving a monetary bonus an indicator variable for the worker receiving a
bonus at the end of the season or the logged hourly wage rate Di and Ei are vectors
of controls for the same worker demographic characteristics and employment charac-
teristics in Equation (1) yi is either logged hours of work or number of weeks worked
I include state and year fixed effects
Pminusi is the instrumental variable and represents the average value of the employer
policy for all other workers (ie not including the focal worker i) within the same
state-year group This instrument corrects for the endogeneity of these employer poli-
cies in the labor supply regression Here the concern with an OLS regression comes
from omitted variable bias and potentially reverse causality Some employers likely
offer higher wages bonuses and health coverage to a select group of workers These
workers may put in more hours than their peers (reverse causality) or they may have
some unobserved characteristic (eg motivation) that makes employers like them more
and causes them to work more hours (omitted variable bias) By instrumenting the
employer policy faced by the focal worker with the average of the policy for all other
workers within the state-year the estimated effect of the employer policy is unrelated
to these worker characteristics Instead the effect is measured through increases in
the likelihood of the focal worker receiving the employer policy based on other workers
receiving it Intuitively a worker is more likely to have an employer cover health care
costs if that employer or nearby employers offer their workers health coverage
Table 7 shows the results from this specification as well as the OLS regressions I
use separate first and second stage regressions to analyze each of the employer policies
In addition to the IV approach outlined in Equation (3) Table 7 also shows results
from an alternative instrument (logged state minimum wages) for logged wages The
28
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
5 EMPLOYER STRATEGIES
coefficients on the instrumental variables in the first stage regressions are all positive
and significant (ranging from 06 to 08) and F-statistics are sufficiently large
Table 7 Effect of Employer Policies
log(hours) weeks workedOLS IV OLS IV
Health coverage 00680lowastlowastlowast -00518 4698lowastlowastlowast 2408(000967) (00843) (0362) (2749)
Bonus 00950lowastlowastlowast 01026lowast 7360lowastlowastlowast 6465lowastlowastlowast(000765) (00563) (0252) (0704)
Season bonus 00828lowastlowastlowast 00592 3626lowastlowastlowast 3362lowast(00111) (00612) (0333) (0851)
log(wage) 01549lowastlowastlowast -0095 11004lowastlowastlowast 2960(00168) (02474) (05797) (77053)
State minimum wage IV 00561 -1578(00797) (3343)
State-year FE yes no yes noAll controls yes yes yes yesIV no yes no yesNotes Robust standard errors in parentheses clustered at state Reported coefficientsare from separate regressions containing the same set of fixed effects and control vari-ables but changing the predictor variable of interest IV regressions use the proportionof other workers within the same year and state as the focal worker who receives theemployer policy Because instruments vary at the state-year level IV regressions do notinclude state-by-year fixed effects and instead include separate year and state fixed ef-fects For log(wage) I additionally use state minimum wages as the instrument (shownin the bottom row of the table) The state minimum wage IV includes state fixedeffects and a time trend lowastp lt 010 lowastlowast p lt 005 lowastlowastlowast p lt 001
Results from the OLS regressions shown in the first and third columns of Table 7
suggest that all of these policies are positively correlated with labor supply Workers
who receive off-the-job health coverage work 7 more hours per week and 5 more weeks
per year than those who do not Workers who receive a (season end) bonus work 95
(8) more hours and 7 (35) more weeks than those who do not Finally the naive
regressions suggest that a 10 higher hourly wage rate is associated with working
15 more hours per week and one more week per year The IV results however
suggest that the correlations in the OLS regressions are primarily driven by unobserved
characteristics correlated with labor supply and the likelihood of receiving the employer
policy The second and third columns of Table 7 show results from the second stage
of the IV regressions These results show that employer policies have little significant
effect on either labor-hours or labor-weeks The only employer policies causally linked
with higher labor supply are bonuses and end of season bonuses In general receiving
29
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
6 DISCUSSION AND CONCLUSION
a bonus cause workers to increase weekly hours of work by 10 (note that this is only
significant at the 10 percent level) and to increase annual weeks worked by 65 weeks
End of season bonuses have no significant effect on weekly hours of work but increase
annual weeks of work by 35 Unlike the results from the IV regressions for health
coverage and logged wages these coefficients are similar to the OLS coefficients but
the standard errors are larger
6 Discussion and Conclusion
This paper is the first comprehensive study of the labor supply of US agricultural
workers I estimate differences in short-run labor supply for workers across several
demographic characteristics to show the likely effect of current workforce trends on
future labor supply If the workforce continues to age and in particular if the proportion
of workers aged 45 and older continues to rise I find that this will result in a labor force
that is willing to work more hours per week and weeks per year in agriculture However
if current trends in the gender and family composition of the workforce continue it will
result in a labor force that is willing to work less
This paper is one of few that examines heterogeneity in labor supply across worker
legal statuses I find that among employed US crop workers foreign born legals work
the most hours per week and weeks per year followed by undocumented workers and
then US natives To the best of my knowledge it is the only paper that examines
heterogeneity within worker legal statuses with respect to demographic characteristics
While I find significant variation in labor supply across legal status and demographic
groups separately I find little evidence of heterogeneity within the demographic groups
Notable exceptions are foreign born workers aged 45 and older work 40 hours per week
while natives work 38 single women with children who are legal immigrants work 39
hours per week while natives work 325 single men with children who are both legal
and undocumented immigrants work 34 weeks per year while natives work 285 and
single men without children who are undocumented work 32 weeks per year while
natives work 29
Finally this paper sheds light on the viability of four producer alternatives for in-
creasing labor-hours and labor-weeks To the best of my knowledge this is the first
causal evidence on the effects of these policies on the labor supply of employed agricul-
tural workers I find that bonuses have a positive and significant effect on both hours
and weeks of labor while offering higher wages or off-the-farm health coverage have
no significant effect My results suggest that on average offering workers a monetary
bonus increases labor-hours by 10 and labor-weeks by 65
30
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
REFERENCES REFERENCES
References
Angrist J D amp Evans W N (1998) Children and Their Parentsrsquo Labor Supply Ev-
idence from Exogenous Variation in Family Size The American Economic Review
88(3) 450ndash477
Blau F D amp Kahn L M (2007) Changes in the Labour Supply Behvaiour of Married
Women 1980-2000 Journal of Labour Economics 25(3) 393ndash438
Blundell R amp Macurdy T (1999) Chapter 27 Labor supply A review of alternative
approaches In O C Ashenfelter amp D Card (Eds) Handbook of Labor Economics
volume 3 chapter 27 (pp 1559ndash1695) Elsevier 1 edition
Borjas G J (2017) The Labor Supply of Undocumented Immigrants Labour Eco-
nomics 46 14ndash15
Boucher S R Smith A Taylor J E Fletcher P L amp Yuacutenez-Naude A (2012)
Immigration and the US farm labour supply Migration Letters 9(1) 87ndash99
Eissa N amp Hoynes H W (2004) Taxes and the labor market participation of married
couples The earned income tax credit Journal of Public Economics 88(9-10)
1931ndash1958
Emerson R D amp Roka F (2002) Income Distribution and Farm Labour Markets In
J L Findeis A M Vandeman J M Larson amp J L Runyan (Eds) The Dynamics
of Hired Farm Labour Constraints and Community Responses chapter 11 (pp 137ndash
150) New York NY CABI Publishing
Fallick B amp Pingle J (2010) The Effect of Population Aging on the Aggregate Labor
Market In K G Abraham M Harper amp J Pingle (Eds) Labor in the New
Economy (pp 31ndash80) National Bureau of Economic Research
Fallick B amp Pingle J F (2006) A Cohort-Based Model of Labor Force Participation
Technical report Federal Reserve Board Finance and Economics Discussion Series
Washington DC
Hernandez T Gabbard S amp Carroll D (2016) Findings from the National Agricul-
tural Employment Profile of Workers Survey United States Farmworkers (NAWS)
2013-2014 Technical Report Research Report No 12 US Department of Labor
Washington DC
Kaushal N (2006) Amnesty Programs and the Labor Market Outcomes of Undocu-
mented Workers The Jounral of Human Resources 41(3) 631ndash647
Kossoudji S A amp Cobb-Clark D A (2002) Coming Out of the Shadows Learning
About Legal Status and Wages from the Legalized Population Journal of Labour
Economics 20(3)
31
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32
REFERENCES REFERENCES
Kostandini G Mykerezi E amp Escalante C (2014) The impact of immigration en-
forcement on the US farming sector American Journal of Agricultural Economics
96(1) 172ndash192
Lofstrom M Hill L amp Hayes J (2013) Wage and mobility effects of legalization
Evidence from the new immigrant survey Journal of Regional Science 53(1) 171ndash
197
Lundberg S amp Rose E (2002) The Effects of Sons and Daughters on Menrsquos Labor
Supply and Wages The Review of Economics and Statistics 84(May) 251ndash268
Maestas N Mullen K amp Powell D (2016) The Effect of Population Aging on
Economic Growth the Labor Force and Productivity
Martin P amp Calvin L (2010) Immigration reform What does it mean for agriculture
and rural America Applied Economic Perspectives and Policy 32(2) 232ndash253
McClelland R amp Mok S (2012) A Review of Recent Research on Labor Supply
Elasticities
Meyer B D amp Rosenbaum D T (2001) Welfare the earned income tax credit
and the labor supply of single mothers Quarterly Journal of Economics 116(3)
1063ndash1114
Pena A A (2010) Legalization and Immigrants in US Agriculture The BE Journal
of Economic Analysis amp Policy 10(1)
Rivera-Batiz F L (1999) Undocumented Workers in the Labor Market An Analysis
of the Earnings of Legal and Illegal Mexican Immigrants in the United States
Journal of Population Economics 12(1) 91ndash116
Sheiner L (2014) The Determinants of the Macroeconomic Implications of Aging
American Economic Review Papers amp Proceedings 104(5) 218ndash223
Taylor J E (2010) Agricultural Labor and Migration Policy The Annual Review of
Resource Economics 2 369ndash93
Taylor J E amp Thilmany D (1993) Worker Turnover Farm Labor Contractors
and IRCArsquos Impact on the California Farm Labor Market American Journal of
Agricultural Economics 75(2) 350ndash60
Zahniser S Hertz T Dixon P Rimmer M amp Go W W W W W E R S U
S D A (2012) The Potential Impact of Changes in Immigration Policy on US
Agriculture and the Market for Hired Farm Labor A Simulation Anay Technical
report Economic Research Service Report Number 135 Washington DC
32