women and the economics of family migration

8/4/2019 Women and the Economics of Family Migration

http://slidepdf.com/reader/full/women-and-the-economics-of-family-migration 1/10

Women and the Economics of Family MigrationAuthor(s): Steven H. SandellSource: The Review of Economics and Statistics, Vol. 59, No. 4 (Nov., 1977), pp. 406-414Published by: The MIT PressStable URL: http://www.jstor.org/stable/1928705

Accessed: 06/09/2010 11:14

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at

http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

http://www.jstor.org/action/showPublisher?publisherCode=mitpress.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of

Economics and Statistics.

http://www.jstor.org

http://www.jstor.org/action/showPublisher?publisherCode=mitpress

http://www.jstor.org/stable/1928705?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp



http://www.jstor.org/page/info/about/policies/terms.jsp

http://www.jstor.org/stable/1928705?origin=JSTOR-pdf




WOMEN AND THE ECONOMICS OF FAMILY

MIGRATION

Steven H. Sandell*

Woman is a greater migrant than man. This

may surprise those who associate women withdomestic life, but the figures of the census clearly

prove it.

OTWITHSTANDING this early state-ment by Ravenstein (1885, p. 196), the

separate study of geographic mobility amongwomen has been virtually ignored by studentsof migration.' The reason is obvious: womenare assumed to migratebecause their husbandsdo.2

While it is undoubtedly true that mostmigration involves family units (the migrationof husband and wife occurring jointly), thepossibility that the wife's welfare is consideredin the family's decision to migrate should notbe ruled out. It is at least desirable to test thehypothesis that the wife's employment isconsidered in the migration decision and toexamine the effect of that decision on women'searnings.

In this paper an economic model isdeveloped to explain the family's decision to

migrate and the effect of migration on thelabor market earningsof men and women. It is

based on the tenet that family utility, definedoperationally as the husband's and wife's labormarket earningsand leisure, is maximized. Themodel suggests that the wife's labor marketinvolvement is a significant consideration in a(husband-wife)family's decision to migrate.

The data from the National LongitudinalSurveys (NLS) are well suited for empiricaltesting of this model.3The surveys provide theopportunity to examine the change in labormarket earnings of families and individualsover a five-year period. Availability of data on

migratory status as well as on other personalcharacteristics of women and their familiespermitsthe direct testingof the model.

In section I a family utility maximizationmodel is used to derive implications withregard to the probability of migration by thefamily and the effect of migrationon individualand family earnings. These implications aretested in section II using multiple regressionanalysis and the NLS data for white womenwho were 35 to 49 years of age in 1972. Theimplications of the empirical estimates for the

economic welfare of women and for interpret-ing the observed earnings distribution arediscussed in section III.

I. A Theoryof Family Migration

The ModelIn the two-location, work-leisure choice

model developed in this section, nonpecuniarybenefits from working or living in eitherlocation are ignored. The family attempts tomaximize its utility (equation (1)), whichdepends on total family income, the wife'sleisure, and the husband's leisure.

The first three constraints (equations (2a)and (2b) and the first budget constraint,equation (3)) are similar to those established inthe conventional labor supply literature.4The

Received for publication October 14, 1975. Revisionaccepted for publicationMay 2, 1977.

* I am indebted to Scott Sutton, Dan Gressel, and RexJohnson for their very competent research assistance. Ihave benefited from comments on earlierdrafts from H.Parnes, A. Kohen, G. Nestel, H. Marvel, D. Brito, D.Parsons, A. Adams, A. Schwartz, and an anonymousreferee.Financial support for this study was provided bythe Center for Human Resource Research under acontract with the Manpower Administration, U.S. De-partmentof Labor. Any errors that remain are my ownresponsibility.

'Lansing and Mueller (1967), Gallaway (1969), and deBeauvior (1970) subscribe to the hypothesis of tiedmovement of women. Although Miller (1966), Masnik(1968), and Long (1974) have analyzed migrationrates ofmen according to their marital status and the employmentstatus of their wives, only relatively simple tabularanalyses and no explicit modeling is reported. Becker(1974, p. 1077), however, writingabout social interaction,illustratesa more general argumentabout decision makingof the head of household with the following statement:"For example, he would not move to another city if hisspouse'sor children's ncome would be decreasedby morethan his own income would be increased."

2For nonmarriedwomen who move, application of malemigrationmodels is presumablyappropriate.

3A description of these data can be found in Shea, Spitz,and Zeller (1970).

4Neither nonlabor income, the distinction betweenleisure and nonmarket work, age, nor education isconsideredin the theoreticalmodel in order to concentrateon the effect of the wife's employment on the migrationdecision.

[406]



WOMEN AND THE ECONOMICSOF FAMILY MIGRATION 407

present model differs from the standard laborsupply model in that the family is allowed tomigrate, therebychanging its budget constraint.The family could choose a budget constraintwith the set of wage rates available to it at the

new place of residence (equation (4)). If thefamily does migrate, moving costs are sub-tracted from total family income. Hence, thefamily chooses a budget constraint as well as apoint on it in orderto maximize its utility.

U= U(Lw,Lh, Yf) (1)

Dw+ Lw= TW (2a)

Dh+Lh=Th (2b)

Y Yw Yh WwDW+WhDDh (3)

Yf= YW Yh-M=

WWW+

WhDh-M

(4)where

U= family utility

Lw= the wife's leisure (including nonmarketwork)

Lh= the husband's leisure

Yf= total family (labor market)earnings

DW the wife's labor supplyDh= the husband's labor supply

Tw= the wife's total available time (a con-stant)

Th= the husband's total available time (aconstant)Yw the wife's (labor market)earningsYh= the husband's (labor market)earnings

Ww= the wife's wage rate

Wh= the husband's wage rate

M = moving costs

Y, Yw, Yh,D', D,, Ww, Wh, are the re-spective variables after migration hastaken place.

The choice of residence depends not only onthe

wagerates obtainable

bythe husband and

wife but also on their tastes for market work. Ahigh potential wage for the wife in a newlocation would not provide an incentive for thefamily to migrate if the wife would not chooseto work at that wage. Hence, for families wherethe wife would not work at any conceivablewage, the decision to migrate becomes afunction of the husband's labor market oppor-tunities only. If the wife is willing to work atcertainwage rates (the husband'swage is also adeterminant of the number of hours the wife

offers to the labor market), her labor market

opportunities become a consideration in thefamily's location choice. The greater utilityachieved in the new location for the migrantfamily can be associated with a change in itslabor supply.

Within a family context, the reductionin theearnings of a spouse is a cost of migration.Since this reduction is potentially quite largefor the husband, it often does not pay for thewife to search for a job in a distant area untilher husband has obtained some satisfactoryemployment there. Given the low marketwageopportunities for many married women, theirhusband's employment precludes their initia-tion of job searchoutside of the area of currentresidence.5Likewise, potential reduction in thewife's earningsis considered by the husband to

be a cost of a geographicaljob change on hispartand will constrainboth his searchbehaviorand actual family migration. Hence, we wouldexpect to observe, ceteris paribus, less geo-graphic movement among families wherehusband and wife are working and expect toremain in the labor market than among otherfamilies.6

As a consequence of migration, the familyfaces a new set of temporary and permanentmarketprices upon which it bases its behavior.

Since there arecosts to

job switching,and job

search often requires flexible hours, newlymigrant women might refuse low paying jobsthat would be immediatelyavailable in order tosearch the new labor market extensively. Inaddition, the increased value the family placeson the wife's time in setting up the newhousehold might initially keep her out of thelabor force. Therefore, we would expect toobserve higher unemployment rates and lowerlabor force participation among new migrantsthan among other marriedwomen.

The model can be extended to considerexplicitly the welfare of children and otherfamily members. Family migration imposes acost on children. Their schooling might beinterruptedand their friendshipsterminated.Inorder to minimize these costs, inter-city

5Following Stigler (1962), the optimal amount of jobsearch for husbands exceeds that for wives because theexpected labor force participation of married women isless than that of married men.

61t is possible that the wife will reduce her hours ofwork in the new location due to the increased earningsof

her husband or her own reducedwage offer.



408 THE REVIEW OF ECONOMICS AND STATISTICS

migration is likely to be timed to occur duringthe summer months when school is not insession. Families with school aged children areless likely to move than otherwise similarfamilies.

Family Income and the Migration Decision

In this section a model of the migrationdecision is presented based on the assumptionthat the family's objective is to maximize totalfamily income. Let the present value of thefamily's earnings streambe equal to the sum ofthe present value of the husband's labor marketearnings plus the present value of the wife'slabor market earnings.

t = R tt=R,

E Yfit(I + i) E-t(I + i)-'t=I t=1

t = Rh

+ E Yht(1l ) (5a)t= 1

or

F= W+H (5b)

where

Yf,= family's earnings in year t(without migration)

Y,( Yh,)= the wife's (husband's)earningsin year t

i= rate of discountR =year of retirement; Rw,

(Rh) iS the year of retire-ment for the wife (hus-band)

F, W,H = the present value of fam-ily, the wife's and thehusband's lifetime earn-ings (without migration)

M = the present value of the

moving cost

YJt,Yw,,Yh,= earningsafter migrationY', W', and H' = the present value of

earningsafter migration.

If a family acts rationally and decides tomove, it must expect the present value of thereturns to migration to exceed the cost ofmigration. Excluding nonpecuniary costs andreturns,this condition can be stated as

F'-M > F. (6)

If moving costs are positive and the familymoves, (6) implies

H'+ W'>H+ W (7a)

if both husband and wife are willing to work,

H'> H (7b)

if only the husband is in the labor market, or

W'> W (7c)

for the household with only the wife in thelabor market.7

That is, the expected earnings stream aftermigration must be greater than the expectedearnings without migration. For the householdwith two persons willing to work it is notpossible to say anything about the income

stream of either partner separately withoutadditional information.Maximization of familyearnings implies that the sum of the twopersons' income streams must increase. Thiscan happen if both increaseor if the increaseinthe income stream of one partner is greaterthan the reduction of the income stream of theother partner (plus the cost of moving). Themotivation for a family's migration could bedue solely to improvement of the husband'searnings if the negative effect on the wife'searnings is offset by the husband's improve-

ment.The model immediately yields a testable

hypothesis: migrant families expect their totallabor marketearningsstream after migrationtobe greater than their expected earnings wouldhave been without migration. If expectationsare met (in the aggregate) and earnings in asingle year can serve as a proxy for theearnings stream, the hypothesis can be testedusing the NLS data. When relevant personaland labor marketcharacteristicsare controlled,it is hypothesized that the increase in labormarketearningsof migrantfamilies (between ayear before and after migration) should be

7If the variance of expected family earnings rises withthe number of family workers and if people are riskaverters,then the labor force participationof the wife willreduce the family's propensity to migrate. The wife mightreceive a "rent" at her present job because of intense jobsearchif the husband considersit unlikely to migrate (say,for physicians in private practice) or if her labor forceparticipation is the result of her extraordinary jobopportunity.These factors, pointed out by an anonymousreferee, work in the direction of reducing the probabilityof migrationof families with workingwives.




greater than the increase for nonmigrantfamilies.8 For married women the relevantearnings figure is the sum of their own plustheir husbands' labor market earnings, whilefor single women only their personal earnings

are relevant.

II. EmpiricalResults

In this section, hypotheses developed fromthe model of family migrationare subjected toempirical tests. These involve two aspects ofmigration: the determinants of migration andthe effect of the geographic movement onfamily and individualearnings.

TheLikelihoodof Migration

The dependent variable used in the regres-sion analyses is the logit constructed from adummy variable with the value "1" if thefamily is migratoryand the value "O"otherwise(Theil, 1971, pp. 632-633). Solving the esti-mated logit equation we can determine theprobability of migration for a family with theobserved characteristics.A family is consideredto have migrated if it reports its county orSMSA of residence to be different in at leastone survey year (1968, 1969, 1971, 1972) than itwas in 1967.9

The probability of a family's moving de-pends on labor-market-relatedpersonal char-acteristics of each labor force participant. Ifmigrationis viewed as an investment, it is clearthat the incentive to move should decrease withage since the length of time for the person toreap benefits from moving decreases and thepsychic costs of moving probably increase withage. Inasmuch as the geographic scope of thelabor market is likely to be largerfor the morehighly educated, migration is expected to be

positively related to education.'0 In addition,

the presence of school aged children isexpected to inhibit family migration.

For our purposes, however, the abovevariables may be considered control variables;our chief interest lies in examining the effect ofthe wife's labor force commitment on themigration decision. Because it has been shownthat a family is probably less likely to improveits economic position by migration if twopersons rather than one are working, thepropensity of the family to move is expected tobe inversely related to the labor forcecommitment of the wife. Thus, coefficients of

the dummy variable for the 1967 employmentstatus and tenure with the 1967 employer arecrucial.

The regressionresults are presented in table1. The regression coefficients exhibit theexpected signs. The significant (at the 1% evel,one-tail test) negative signs of the regressioncoefficients for labor force commitment

TABLE 1.-LOGIT EQUATIONS OF THE LIKELIHOOD OF

FAMILY MIGRATION 1967-1972

Variable (1) (2) (3) (4)

Constant - 1.99 - 1.80 - 1.97 - 1.83(-23.98)a (-2.93)a (-26.60)- (-3.01)a

Employed -0.356 -0.266wife, 1967 (-2.5 1)a (- 1.84)b(dummy)

Husband's -0.044 - 0.040age, 1967 (..3.53)a (-3.26)a

Husband's 0.134 0.132education, (5.90)a (5.85)a1967

Children -0.188 -0.235aged 6-18 (-1.04) (-1.30)c(dummy)

Wife's job -0.148 -0.135tenure, (-3.20) (- 2.69)a1967

Wife's 0.004 0.004tenure (1.55)c (1.47)csquared

Pseudo R2d .006 .057 .017 .066

Likelihoodratio test 6.50 66.67 19.47 77.09

Universe: 2,322 white married women, spouse present.Note: t-statistics are in parentheses. The author will send

summary statistics on request.aSignificant at a < .01.b Significant at a < .05.CSignificant at a < .10.dEqual [1-exp{Z(L,-L,)/T})/[l-exp(Z(L -Lm)/T)]. L.

is the maximum of the log likelihood function using a constant; L,is the maximum using all variables; Lm is the maximum possible.

8Factors other than migration (e.g., level of educationand age) affect the change in a person's earnings.Theoretical explanations of the effects of these variablescan be found in Becker(1964).

9Approximately 11%(248) of the families of white,marriedwomen (same spouse present all surveyyears) areconsidered to be migrants under this definition. Between1968and 1971 (the only period in which data on distancemoved are available), 68%of the migrants moved morethan 100 miles, and 81%moved more than 50 miles. In1971 78% of the 1967-1971 migrants were living in thesame Census division as they did in 1967.

'?Bowles (1970) and Schwartz (1968) explain the

positive correlation between migration and education byhypothesizing that the more highly educated have greateraccess to job market informationin distant regions.



410 THE REVIEWOF ECONOMICSAND STATISTICS

variables when used separatelyin the equationsconfirm our hypothesis. That is, the families ofwomen who work are less likely to move thanare families of other marriedwomen, and thelikelihood of migration decreases the longer

they have worked for their 1967 employer.When the only independent variable in theregression equation is employment status, itsregression coefficient can be interpretedas thegross effect of working on the naturallogarithm of the odds in favor of familymigration. The respective net effects of em-ployment status and tenure on family migra-tion are the coefficients of these variables inthose equations where the husband's age andeducation are also included as independentvariables. The observed positive differential

between the gross and net effect of the wife'slabor force participation on migration is anindication of the correlation of some of theother independent variables with both thedependent variable (migration) and the em-ployment status of the wife. In particular,greater husbands' educational attainment isassociated with lower wives' labor forceparticipationand a higher probabilityof familymobility.

We evaluate the logit of equation (4) for a

(white) family with the husband's meaneducation (11.8 years) and husband's mean age(40.4 years) and find that the likelihood offamily migration between 1967 and 1972 was13.2% if there were no children in the

household and the wife did not work in 1967.The likelihood was only 4.4% if there wereschool aged children present and the wife's1967 job tenure was 10 years. Not only doesfamily migration vary inversely with the wife's

employmentstatus,but this inverserelationshipis strongerthe longer she has worked at her job(peaking at 17 years).

The Effect of Migration on Earnings of

Husband- Wife Families

The coefficient of the dummy variablerepresenting migration status in a regressionequation where the dependent variable ischange in wife's, husband's or family's (hus-band plus wife) labor market earnings repre-

sents the effect of migration on earnings. Bycontrolling for personal characteristics i.e., ageand education) and base year income it ispossible to isolate the net effect of migrationonearnings." Table 2 shows the regressionresultswhen change in the husband's earnings andchange in the wife's earningsbetween 1966 and1971 are the dependent variables. Table 3shows the effect of migration on familyearnings.

The control variables in the regressionequation are worthy of some discussion. The

negative coefficient for the husband's age

TABLE 2.-REGRESSIONS OF CHANGE IN HUSBAND'S AND WIFE'S EARNINGS 1966-1971, BY YEAR, FREQUENCY,

OR REASON FOR MIGRATION

Independent 1967-1971Migrants Multiple Migrants Intrafirm Transfers

Variables ffusband Wife Husband Wife Husband Wife Husband Wife

Constant 2891 - 968 2966 - 959 2965 -1029 2922 - 1022

Education 154 156 )a 153 ) 154 ) 156 )a 155 )a 156 gy 154 aA (4.18)a (5.52)a (4.16)a (5.45)a (4.28)a (5.48)a (4.28)a (5.46)a

Age, 1967 -8.7 Oa10.3 -60.1 )a

10.4 -60.1)a

11.9 -58.4)a

11.4(-3 30)a (0.68) (337)a (0.68) (-3.38)a (0.78) (-3.28)a (0.75)

Earnings, 1966 0.087 - 0.099 0.087 - 0.099 0.086 -0.100 0.082 - 0.098(2.64)a ( 2.73)a (2.64)a (2.76)a (2.62)a (- 2.77)a (2.48)a (-2.71)a

Migration dummy, 832 -372 3229 - 1289 1869 - 1201967-1971 (2.32)a ( 1.58)c (2.99)a (1 I81)b (2.65)a (-0.26)

Migration dummy, 980 246

1967-1969 _2 24)b (0.86)

Migration dummy, -29 9471969-1971 (-0.05) (- 2.67)a

K2 .053 .028 .052 .031 .056 .028 .055 .026S.E.E. 3398 2246 3400 2242 3393 2245 3396 2248F-ratio 17.7 9.4 14.1 8.7 18.7 9.6 18.1 8.8

Universe: 1,186 white married women, spouse present.

&,b,cSee table I for significance levels.

"IThe estimates of the effect of migrationon earningsare similarwhen base year earningsare not included as anindependentvariable.




<s~~ ~~~' _xXIO> ONO 4nw

.dC4 0 I

Cd -rqt-{,llo Nb

o~~~~~~~~~~oW

Z C

-~~~~~C4o

_ d n 00q o N

z

Cd, o _ONot ?8 O

z 3> C N_oo

Z~~~~~~~W M

cd - r- en C, 6)0 o o 4 0O , 0 'It -o0 mO

Z 00 O>OfN

CV>s mo,..gO

Cd C o > 8g ;0 B

(experience)and the positive coefficient for thevariable reflecting the number of years ofeducation are consistent with the theory ofhuman capital. Since the dependent variable isthe change in earnings, we are actually

examining an experience/earnings profile.Theory suggests that investment in on-the-jobtraining is positively associated with educationand negatively associatedwith age; therefore,itis expected that younger individuals and moreeducated individuals will exhibit, ceteris pari-bus, faster earnings growth than their olderand/or less educated counterparts.Thus, ourfinding, which employs the longitudinal panel,is consistent with the cross-sectional results ofother researchers.'2

Tables 2 and 3 show the net effect ofmigration between 1966 and 1971 on theseparate and combined labor market earningsof the husband and wife. The earnings of1967-1971 migrant husbands increased morethan those of nonmigranthusbands, and familyearningsof migrantsincreasedmore than thoseof nonmigrant families for our measures ofmigration. However, the earnings of non-migrant wives went up faster than those ofmigrant wives.13 Although migration seems tolead to an improvementin the earnings of the

family unit, which implies that the move iseconomically rational, the earnings of the wifedo not increaseas a resultof the move.

Separating 1967-1969 migrants from 1969-1971 migrants and regressing change inearnings on both dummy variables (as well asthe control variables) sheds additional light onthe effect of migration on earnings. Migrantwives living in a new geographicarea less thantwo years experienced $950 less growth inannual earnings than nonmobile wives, whilethe difference between the 1966 to 1971

12As we have seen in the likelihood of migrationequations, the probability of migration is positivelyassociated with education and negatively associated withage. Hence, the omission of age and education from thechange in earnings equation would lead to an overstate-ment of the returns to migration.

13Although the difference between the earnings ofmigrant and nonmigrant wives barely reaches statisticalsignificance, it is clear that the change in the earningsposition of mobile wives is significantlydifferentthan thalof mobile husbands.The difference between the change inthe earnings of mobile spouses is about three times thestandard error of the husband's mobility coefficient andfive times the standard error of the wife's mobilit)coefficient.




earnings growth wives who moved before the1969 survey was not statistically different fromthat of nonmigrant wives. Husbands whomoved between 1967 and 1969 experiencedearnings growth of $980 per year more thannonmigrant men, while the earnings growth of1969-1971 migrant husbands was not statisti-cally different from that of nonmovers. It isapparent that while migrant wives only recouptheir relative earnings position after two years,their husbands reap significant improvementsin earnings in the same time period.

To provide some insight into the source ofthe earningsloss to migrant wives, we regressedthe change in the number of weeks worked onthe migration dummy variables and thenumber of weeks workedin 1966 (table 4). Thestatistically significant negative coefficients forthe migration dummies in these equationsindicate that the slower growth in the earningsof migrant wives as compared to nonmobilewomen is due to reduced market work.Multiplying the wives' average 1966 weekly

earnings ($167) by the decline in weeks workedfollowing migration (5.7), we can explain theapparent decline in wives' earnings shown intable 2.

An examination of the change in weeksworked for 1967-1969 compared to 1969-1971migrants shows that the difference in weeksworked between migrants and nonmigrantsnarrows with the passing of time. This impliesthat the initial reduced work effort representsacost of migration for the wife rather than achange in taste for work by migrants. It seems

to be optimal from the point of view of the

family for the migrant wife to forgo marketwork in order to set up the new household aswell as search for a desirable job. After twoyears in their new residence the labor supply ofmigrant wives is not significantly differentfromthat of nonmigrant wives.

The regression results (see tables 2 and 3)show that for families that moved more thanonce between 1967 and 1971 (multiplemigrants) and for those families that movedbecause the husband received an intrafirmtransfer between 1968 and 1971, labor market

earnings grew substantially faster than theearnings of other migrant families. The reasonfor the above average gain can be traced to theimprovement in the earnings of the husbandssince the wives in these groups fared slightlyworse than the wives of all other migrants.

Marital Status and the Effect of Migration on

Women's Earnings and Labor Supply

A clear implication of the model is that forsingle women (all one-person families) migra-

tion will occur only if the move is expected tolead to an increase in utility. Since thiscondition does not necessarily hold for marriedwomen (or any individual members of multi-person households), we would expect toobserve, on average, a greater increase in thepersonal welfare due to migration for singlewomen compared to married women. Whileown earnings may not be a good proxy forwelfare of all married women, change inearnings may be regardedas a first approxima-tion change in welfare of those women who

desire to work full time. Hence, changes in

TABLE 4.-LEAST SQUARES REGRESSIONS OF CHANGE IN WIFE'S WEEKS WORKED 1966-1971, BY YEAR,

FREQUENCY, OR REASON FOR MIGRATION

Intrafirm TransfersIndependent Variable 1967-1971 Migrants Multiple Migrants (1968-1971)

Constant 15.11 14.87 14.66 14.71(19.o9)a (18.82)a (19.01)a (19.00).

Weeks worked, 1966 -0.47 -0.47 -0.47 -0.47(_ 17.94)a (_17.93)a (17.83)a (17.84)a

Migration dummy, 1967-1971 - 5.66 -7.26 -5.35( - 2.64)a (-1.1l) ( -1.26)

Migration dummy, 1967-1969 2.02(0.77)

Migration dummy, 1969-1971 - 10.24(- 3.15)a

R2 .21 .22 .21 .21S.E.E. 20.53 20.51 28.58 28.58F-ratio 162.7 109.7 159.1 159.3

Note: See table 2 for universe.aSee table 1 for significance level.




earnings and weeks worked of migratorywomen who worked more than 1,400 hours in1966 have been analyzed.

To examine the differential effect of migra-tion on the labor market earnings and labor

supply of married versus single women,regression analyses were performed using asample containing both married (spouse pres-ent) and never married women. Although wefound that single migrants fared much betterthan married migrants in terms of changes inearnings partly due to their greater number ofweeks worked after migration, there were only10 single women in the sample who migratedbetween 1967 and 1971. As a consequence, theempirical support for the model was notstatisticallysignificantand is not reportedhere.

111. Conclusions

The empirical results are consistent with thetheory. On the one hand, the labor marketorientation of the wife seems to be taken intoconsideration in the decision of a family tomigrate. On the other hand, the migration ofthe family increases the earnings of thehusband but does not increase the labor marketearnings of the wife. In contrast, the earningsof never married women increased after

moving. Since family earnings have beenshown to increase as a result of migration, thedecision to migrate is rational from theviewpoint of the family.

It seems that the contribution of the wife tofamily income is considered, but the positiveeffect of migrationon husband'searningsoftenoutweighs the (initial) negative effect ofmigration on the wife's weeks worked and,consequently, her earnings. This is not to saythat migration is involuntary for them in theusual sense, but to emphasize that what isbeneficial to the welfare of the family (and thewife as a family member and consumer offamily income) is nevertheless consistent withlower labor market earnings of the wife. Theinterruption of women's careers is often aneffect of migrationand the maximizationof theutility of the family unit. If the participationofwomen in the labor force continues to increase,this may have a limiting effect on thegeographicmobility of the male labor force. Tothe extent that female employment becomes

less casual and women develop greater at-

tachment to their jobs (i.e., there is morefirm-specifictraining and concomitant earningspremiums),this effect could be intensified.

This study documents the effects of migra-tion on the earnings of married women. We

have uncovered no evidence that the labormarket earnings of the husband are a moreimportantconsiderationthan those of the wife.Our data only tell us that, given the jobs heldby men and those held by women, the earningsimprovement for men resulting from geo-graphic movement is large enough to offsettheir wives' losses in market earnings.Furthermore,the wives' losses seem to be onlytemporary, a consequence of reduced marketwork in the period immediately following themove to allow for job search and household

establishment.Finally, it seems that we have shown an

additional reason for differences in the earn-ings of men and women. Family decisionmaking often restricts the wife's choice of joband reduces her continuity of employment. Anemployer's awareness of the possibility of herleaving her current residence and thereforehercurrent job, in spite of pay premiums whichwould make this job the best available to her,will be likely to lower his investment in her

human capital. Even if, on average, the tenureof males in particularfirms is no greater thanthat of females, the lack of influence ofdifferential salary payments on the behavior ofsome marriedwomen employees might ration-ally lead employers to treat male and femaleemployees differently.On the other hand, if thewoman's geographicalmobility is restricted bythe permanence of her husband's job, theemployer is able to discriminate and pay herlower wages than she could be receiving at analternativejob (in a different geographic area).

REFERENCES

Becker,Gary, HumanCapital:A Theoretical nd EmpiricalEnquiry with Special Referenceto Education(NewYork: Columbia University Press for NationalBureauof Economic Research, 1964)., "A Theory of Social Interaction," Journal ofPolitical Economy82 (Nov./Dec. 1974), 1063-1093.

Bowles, Samuel, "Migration as Investment: EmpiricalTests of the Human Investment Approach toGeographical Mobility," this REVIEW 52 (Nov.1970),356-362.

de Beauvior,Simone, TheSecondSex (New York: Bantam

Books, 1970),400-455.




Gallaway, Lowell E., GeographicLabor Mobility in theUnited States 1957 to 1960, Social SecurityAdministration, Departmentof Health, Education,and Welfare (Washington,D.C.: U.S. GovernmentPrinting Office, 1969).

Lansing, John B., and Eva Mueller, The GeographicMobility of Labor(Ann Arbor: Institute for Social

Research, 1967).Long, LarryH., "Women's Labor Force Participationandthe Residential Mobility of Families," Social Forces52 (Mar. 1974), 342-349.

Masnick, George, "EmploymentStatus and Retrospectiveand Prospective Migration in the United States,"Demography (1968), 79-85.

Miller, Ann R., "MigrationDifferentials in Labor ForceParticipation: United States, 1960," Demography3(1966), 58-67.

Ravenstein, E., "The Laws of Migration,"Journal of theRoyal Statistical Society 48 (June 1885), 167-235.

Schwartz, Aba, "Migrationand Life Span Earnings n theU.S.," Ph.D. dissertation, University of Chicago,1968.

Shea, John R., Ruth S. Spitz, and Frederick A. Zeller,Dual Careers, vol. I, Manpower Research Mono-

graph no. 21 (Washington, D.C.: U.S. GovernmentPrintingOffice, 1970).Stigler, George J., "Informationin the Labor Market,"

Journal of Political Economy 70 (Oct. 1962),supplement,94-105.

Theil, Henri, Principlesof EconometricsNew York: JohnWiley and Sons, 1971).

women and the economics of family migration

Documents