Institut für Arbeitsmarkt- und Berufsforschung
In the Right Place at the Wrong Time
The Role of Firms and Luck inYoung Workers’ Careers
Till von Wachter, Columbia University and IZA
Stefan Bender, Institute for Employment Research (IAB)
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
• First Job may be very important in determining the long termcareer development.
• It is difficult to estimate the long-run impact of early labor market experience, because
- not only worker differ and firms choose among them,- firms differ as well (sorting into firm at initial hiring).
• Topics of the talk:
- long-term effects of an early job loss,- role of firms in the process of early job mobility,- how data on firms can be used for understand job and wage dynamics of young workers.
Introduction
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Long-Term Effects of Early Displacements?
Displaced Worker Studies:
• Data: NLSY (Kletzer & Fairlie 2001, Gustafson 1998)• Approach: compare wage changes of movers and stayers• Results:
After 5 years, significant loss in earnings of 10-15%
Issue - Potential Biases:
• Negative Selection: Displaced workers least able• Initial Assignment: Workers sort by firms’ turnover rates• Positive Selection: Voluntary mobility
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Apprentice Training in Germany
POSITIVESELECTIONNEGATIVESELECTION
POSITIVESELECTION
NEGATIVESELECTION
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
A Statistical Model of Young Workers’ Wages
itiVtiItiit VIaw 00
wage loss loss
ability involuntarymover dummy
voluntarymoverdummy
000 iii VID mobilitystatus
only Di0
is observed
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
itiVtiItiit VIaw 00
individualability
itiItVtiItififiit VDaaaw 00)()(
average abilityat training firm
because only Di0 is observed
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Wage Determination and Theories of Job Mobility
Adverse Selection (Gibbons/Katz 1991; Greenwald 1986):
• Displacement may be perceived as a negative signal about a worker by other employers.
• Implies that movers are likely to be the least able workers within a firm.
itiItVtiItififiit VDaaaw 00)()(
Negative Selection:
least able workers are moving overestimate effect of moving 0,cov 0 iifi Daa
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Wage Determination and Theories of Job Mobility
Sorting:– If less able workers are hired by firms with higher
turnover rates, then movers and stayers are not comparable across different types of firms.
– Firms with lower turnover rates offer better career opportunities.
itiItVtiItififiit VDaaaw 00)()(
Sorting across firms:
workers with low ability are moving overestimate effect of moving 0,cov 0 iif Da
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
• If one observes a fully informative pre-displacement wage, the bias from perfect initial sorting or negative selection can be eliminated by analyzing wage changes (for example Jacobson, Lalonde, Sullivan 1993).
Might not be a good strategy for young workers:
• Young workers receive training wages that are below their ability levels.
• If there is asymmetric information between employers about workers´ ability pre-displacement wages do not reflect productivity (Gibbons and Katz 1993).
• Firms and workers themselves only gradually learn about their abilities and preferences. More able workers should leave less attractive firms once their ability becomes known.
Wage Determination and Theories of Job Mobility
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Wage Determination and Theories of Job Mobility
Voluntary Job Mobility (Topel and Ward 1992):
• Is an important feature of early careers in the U.S.• Early mobility plays an important role in other countries as
well (for Germany: Euwals and Winkelmann 2001).
itiItVtiItififiit VDaaaw 00)()(
Positive selection:
voluntary mobility underestimate effect of job loss
0,cov 00 ii DV
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
itiItVtiItififiit VDaaaw 00)()(
Model captures potential biases of OLS regression of wit on Di0:
Negative Selection:
least able workers are moving overestimate effect of moving
0,cov 0 iifi Daa
Sorting across firms:
workers with low ability are moving overestimate effect of moving 0,cov 0 iif Da
Positive selection:
voluntary mobility underestimate effect of job loss 0,cov 00 ii DV
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Need Alternative Estimates of Wage Differences
• Two tired strategy:
- It uses firm fixed effects to control for systematic differences of workers between firms. Thereby, non-displaced workers at the same training firm function as a comparision group.
- It uses firms´ retention rates of other young graduates finishing apprenticeship in the same year as a displaced worker as an instrument for the probability of a displacement.
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Need Alternative Estimates of Wage Differences
(1) Firm Fixed Effects – Change Comparison Group
OLS: all movers compared with all stayers- we have discussed OLS already
OLSFE: compare movers to stayers at same training firm- including training firm fixed effects- takes care of initial sorting
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Need Alternative Estimates of Wage Differences
(2) IV: Fraction of Movers Among OtherGraduates at Same Firm as Instrument
IV - Level of the fraction as an instrument; puts more weight on workers moving from high turnover firms
IVFE - Deviations of firms‘ retention rates of young workers from the firm specific average: workers moving due to temporarily high turnover
IVFE2 - Workers moving from firms with high exit rates into non-employment
1
fc
ifcifc n
mz
Institut für Arbeitsmarkt- und Berufsforschung
German Data and Institutions
the employment statistics register
base is the integrated notification procedure for the health, pension and
unemployment insurance ( freelancers, self employed persons and civil servants are not included)
• notifications have to be submitted to the social security agencies at least
once a year
• notifications include for example information about
employees‘ entries and exits gender, qualification and current occupation (3-digit) gross daily precise wages (including all bonus payments) an unique plant identifier and the industry code since 1999: regional information
each person has an unique identifier (social insurance number)
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
German Data and Institutions
Main Apprentice Sample: Graduates from 1992-1994
• 5 years of potential labor market experience
• Basic restrictions: min duration of training, participation in dual system, no prior work experience, West Germany only
• Substantial Restriction on firm and program size:
at least 50 employees and 5 annual graduates
• Minimal restrictions on real log daily wages, LF participation
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Wage Effects of Moving Out of Training Firm - Full Sample
Wage Differenceby Year of
RawDifferences
OLS withControls
OLS onlyFirm Fixed
Effects
IV without IV with
Firm FixedEffects
Firm FixedEffects
Experience (1) (2) (3) (4) (5)
1st Year -0.094 -0.065 -0.065 -0.121 -0.108
(0.0038) (0.0029) (0.0025) (0.0077) (0.026)
3rd Year -0.093 -0.071 -0.064 -0.108 -0.035
(0.0011) (0.0011) (0.0028) (0.0071) (0.0203)
5th Year -0.092 -0.072 -0.063 -0.098 0.009
(0.0011) (0.0011) (0.0030) (0.0075) (0.0221)
T-Statistics
H0: 1=3 -0.7 4.1 -0.8 -3.4 -3.6
H0: 3=5 -0.1 0.5 -0.8 -3.4 -3.5
H0: 3=5 -1.3 3.7 -1.2 -5.0 -4.7Notes: The dependent variable is the log real daily wage. Standard errors clustered at the establishment level are in parentheses. All Regression have worker, firm, and training variables as well as experience*cohort and occupation dummies. Each regression has 991,004 observations and 13,009 establishments.
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
OLS and IV Estimates of Effect of Moving From Training Firm, With and Without Firm Fixed Effects - Main Sample
Lo
w W
age D
ifference
Experience
1 3 5
-0,13
-0,11
-0,09
-0,07
-0,05
-0,03
-0,01
0,01
RAW OLS OLSFE IV IVFE
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Implications of Estimates for Models of Job Mobility
1. Initial Assignment Matters- firm fixed effects reduce wage differences: OLSFE > RAW- IV in levels increase wage differences: IV < RAW
2. Job Search Matters- ‘random’ move bigger initial effect: IVFE1 <<RAW- ‘random’ move no long-term effect: IVFE5 ~ 0
3. Negative Selection -> Initial Assignment Not Perfect- firm fixed effects alone not enough: OLSFE < 0
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Open Questions: Training Quality, LM Indicators & The Role of Firms
• Indicators of labor market attachment• Amount of job, industry and occupation changes (in comparison to
previous firm/ industry/ occupation)• Time spent in employment, unemployment and out of labor force
(i.e. self-employment, maternity, military) in days per potential experience year
• Role of firm characteristics for mobility (for example size, average wage)
with Katja Görlitz (RWI, Essen)
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
BACKUP
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Basic Characteristics at End of Training – Part 1
All Graduates 1992-94
Workers With Valid Wage Observations
By Mobility Status at Graduation
Stayers Movers
Average Age at End of Training 20.9 20.9 21.0
Fraction Male 0.87 0.88 0.86
Fraction Training Duration >2Years 0.83 0.84 0.81
Cohort 1992 0.37 0.39 0.34
Cohort 1993 0.32 0.32 0.33
Cohort 1994 0.30 0.28 0.34
Observations 295653 177855 117798
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Basic Characteristics at End of Training – Part 2
All Graduates
1992-94
Workers With Valid Wage ObservationsBy Mobility Status at
Graduation
Stayers Movers
Fraction Moving At End of Training 0.40 0 1
Average Fraction MoversAmong other Apprentices
0.40(0.30)
0.28(0.23)
0.59(0.30)
Average Training Firm AnnualEmployment Growth
-0.03(0.30)
-0.02(0.30)
-0.05(0.29)
Fraction High School Degree 0.17 0.18 0.15
Training Firm Size 500+ 0.54 0.56 0.50
Average Log Training Wage 3.91(0.30)
3.96(0.30)
3.85(0.29)
Average Log Real Daily Wage (Time Variying)
4.88(0.26)
4.92(0.23)
4.80(0.29)
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Conclusions: Effects of Job Loss and Early Mobility
Role of Initial Assignment and Frictions• Initial sorting matters: differences in firms’ turnover rates• Job search matters: voluntary mobility is high• Negative selection: adverse selection?
No Persistent Effects• Controlling for potential biases show initially large but
temporary effects of early mobility
Have to address stepwise sorting and selection process when studying job and wage mobility of young workers
LoWER 2006
Institut für Arbeitsmarkt- und Berufsforschung
Open Questions: Training Quality & The Role of Firms
Are retention rates correlated with differences in training quality?
Distinguish initial assignment from differences in training Permanent effects from choice of training firm
What do large firms do: Are there internal labor markets? Use data for multiple firms to evaluate internal markets Cohort effects within firms: Selection at entry?
Institut für Arbeitsmarkt- und Berufsforschung
- individual data on employment are added by unemployment notifications
- the federal unemployment register contains for example information about
start and end of each benefit notification
reason for submitting the notification
type and amount of the benefit
age, gender, nationality, childs (yes/no), married (yes/no)
both kind of data can be integrated; single notifications are
appended to the (un-)employment histories of persons
each person has an unique identifier (social insurance number)
German Data and Institutions
Josh Rouse - "Subtitulo