“LABOR MOBILITY OF IMMIGRANTS: TRAINING, EXPERIENCE, LANGUAGE AND

OPPORTUNITIES”By

Sarit Cohen

Bar-Ilan University

and

Zvi Eckstein

Tel-Aviv University,

University of Minnesota and CEPR

2

Introduction

The transition pattern of immigrants to a new labor market is characterized by high wage growth, fast decrease in unemployment as immigrants first find blue-collar jobs, followed by a gradual movement to white-collar occupations.

3

• Focus on - Acquisition of local human capital in: training, experience and local language.

• Data: quarterly labor mobility since arrival of high skilled male immigrants who moved from the former Soviet Union to Israel.

• Main macro facts.

4

Actual Proportions in White Collar, Blue Collar and Unemployment

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Quarter since Migration

%

Unemployment Blue Collar White Collar

5

Participation in White Collar andBlue Collar Training

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Quarter since Migration

%

Training in White CollarTraining in Blue Collar

6

Formulate a dynamic choice model for: • blue and white-collar occupations• training related to these occupations• Unemployment

Labor market opportunities are random and

are affected by characteristics, past choices

and language knowledge.

Participation in training is affected by: the

mean wage return, the job offer probabilities,

preferences and lost of potential wages.

7

Main Results• The estimated model fits well the main patterns of

the labor market mobility.• Return to training: white-collar 19%; blue-collar

13%, for 78% of population and zero for the rest.

• High return to local experience and language, but –conditional on local human capital - zero return to imported schooling.

• Main return to training is by the increase of 100% of white-collar offer probability.

8

Main Results (cont.)

• Individual welfare gain at arrival from training programs is 1-1.5%.

• Aggregate growth rate of wages from the availability of the government provided vocational training programs is .85 percent.

• Main reasons: return to experience is high and utility from participating in training is low (liquidity constraint).

9

Table 3: Multinomial-logit on Employment by Occupation and Unemployment

VariableWhite-Collar

Unemployed

constant-4.4424

)0.5034(

-0.4753 )0.4804 (

Hebrew 0.9612 )0.0761(

0.1342 )0.0701(

English0.6563 )0.0428 (

0.0205 )0.0052(

age at arrival0.0331 )0.0212 (

0.0332 )0.0190(

Schooling0.0031

)0.0212(

0.0332

)0.0190(

training in WC0.9421 )0.1153 (

0.8183 )0.1658 (

training in BC-0.2101 )0.1594 (

0.9586 )0.1815 (

experience-0.0046 )0.0100 (

- 0.6807 )0.0233 (

occupation in USSR

1.4837

)0.1417(

0.2156 )0.1137 (

Num. Of Obs .5536

Log likelihood-3558.40

10

Table 4: OLS Wage RegressionDependent VariableLn hourly wage

white-collar occupation

Ln hourly wage Blue-collar occupation

Cons 1.091

) 0.407(

2.122

)0.120 (

Hebrew0.129

) 0.061(

0.050

)0.027(

English 0.132

) 0.036(

-0.011

)0.022 (

Age at arrival 0.013

)0.005(

-0.003

)0.002 (

Years of schooling 0.021

) 0.022(

0.008

) 0.006(

Training in WC 0.116

) 0.079(

-0.009

)0.062 (

Ttraining in BC-0.045

) 0.129 (

0.056

) 0.055(

Experience in Israel 0.017

) 0.009(

0.024

) 0.003(

Num. of Obs.132442

R20.2300.153

11

A Dynamic Choice Model

Choice set:

•Work in a White-Collar job (WC)•Work in a Blue-Collar job (BC)•Training related to White-Collar jobs (WT)•Training related to Blue-Collar jobs (BT)•Unemployment (UE)

12

Utility by Choice:

Wage Functions:

jit

jit

jit zKw ln

iSjiAjFiFj

HitHj

jitcjitejj0

jit edLLCEXK

00)( itit ueUUE 11)( itit wUWC

22)( itit wUBC 33)( itit trUWT

44)( itit trUBT

13

Transition Probabilities are limited by job-offer probabilities and training-offer probabilities:

Individual state and characteristics: last period choice r, experience in Israel, occupation in the country of origin, knowledge of Hebrew and English and training.

)2,1j(,}Qexp{1

}Qexp{P

ijt

ijtrjit

: offunction linear ijtQ

14

The Model

1.

UE

2.

UE

BC

3.

UE

BC

WC

BT

WT

20.

UE

BC

WC

BT

WT

Quarter SinceMigration:

Choices:

…….

Study Hebrew

15

Solution MethodThe value function

}.1d,t,S/4,...,0jfor),1t,S(Vmax{EU)t,S(V ritit1it

ji

ritit

ri

)}).1d,t,S/1t,S(V(max{E)g(PU)t,S(V jitit1it

a

1it

~A

1a

a1it

jitit

ji

ait

gait

P(g1

ofy probabilit lconditiona )1

outcomes feasible indicatesat vector tha1 a

itg

16

• The model is solved using backward recursion with a finite linear approximated value at the 21’th quarter as function of Si21.

• We use Monte Carlo integration to numerically solve for the Value Functions and the probability of the choices jointly with the accepted wages.

• By simulations we show that the model can capture the main dynamic aspects of the labor market mobility as depicted by the figure.

17

Estimation Method

• The model is estimated using simulated maximum likelihood (SML) (McFadden(1989))

• Given data on choices and wage, the solution of the dynamic programming problem serves as input in the estimation procedure.

• All the parameters of the model enter to the likelihood through their effect on the choice probabilities and wages. Wages are assumed to be measured with error. M=2.

mim

jo

mit

j

mit

jo

mi

j

mi

I

i

M

m

jo

mi

j

mi xmtypeSwdwdwdLii

),/,,....,,,,Pr()( 0221 1

11

18

Results Order

• Fit of labor market states

• Fit of transitions and wages

• Estimated parameters

• Interpretation of types

• Policy Implications on training

19

Actual and Predicted Proportions in Unemployment, Blue-Collar and White-

Collar*

0

10

20

30

40

50

60

70

80

90

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Quarter since Migration

%

UE - Actual UE - ML BC - Actual BC- ML WC- Actual WC - ML

20

Actual and ML Proportions inWhite Collar Training

0

1

2

3

4

5

6

7

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Quarter since Migration

%

Training in White Collar - ActualTraining in White Collar - ML

21

Actual and ML Proportions inBlue Collar Training

0

1

2

3

4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Quarters since Migration

%

Training in Blue Collar - ActualTraining in Blue Collar - ML

22

Fit results

• The estimated model fits well the pattern but a formal 2 test rejects the fit of the model.

• The 5’th year (20%)reduction in BC and increase in WC is explained by : Cohort and prior events (~10%); BC to WC transitions as unemployment reach minimum (~10%).

23

Table 6: Actual and Simulated Accepted Wages by Tenure and Training

WC occupationBC occupation

ActualModelObservations

ActualModelObservations

By quarters in Israel

1-421.76614.215410.47510.96864

5-815.06215.5634610.96811.687139

9-1218.86417.3762911.86812.65873

13-1620.44918.7382512.49713.71797

17-2021.52120.0372815.23214.77569

By training

No training17.93216.8409611.98512.211402

After training19.98117.8463612.66013.66640

24

Table 7: Estimated Wage Function ParametersWage parametersBCWC

Cons. type11.8799**1.6276

Deviation of type2

from type 1

*0.19300.1443-

Hebrew*0.1100*0.0964

English*0.0418-*0.1386

Age at arrival0.00008-0.0050

Years of schooling0.00900.0126

Accumulated experience*0.0187*0.0205

Trained in WC type1*0.1908

Trained in WC type 20.0004

Trained in BC type10.1275

Trained in BC type 20.00008

Proportion of type 1*0.781

25

Wage Function Results

• Very large return to local human capital accumulation: Experience – 2% per quarter, Training- 13 to 19 % by Type; Hebrew – 15 to 19%.

• Conditional on local human capital – no return to imported human capital.

26

Table 8: Estimated Job Offer Parameters

WC Offer Probability

J=1

BC Offer Probability

J=2

b01j1-worked in WC at t-1 type 1*15.9966*2.4980-

b01j2-worked in WC at t-1 deviation from type 1

0.0053-*1.7338

b02j1-worked in BC at t-1 type 1*2.9737-*14.0431

b02j2-worked in BC at t-1 deviation from type 1

1.1589-0.0082

b03j1- didn't worked at t-1

type 1

*1.7604-*0.4116-

b03j2- didn't worked at t-1

deviation from type 1

0.6392*1.3162

b11j-work experience in Israel 1-40.2761-*0.2421

b12j-work experience in Israel >5*0.8935-*0.2707-

b2j-training in occupation j*0.94240.2196

b3j – Age of arrival*0.0286-*0.0071-

b4j - Hebrew*0.0938-*0.1744-

b5 - English*0.2095

b6 – WC=1 in soviet union*0.5554

b7 - first period dummy*0.4881-

27

Table 9: Training and Job offer Probabilities (weighted by types)

To/FromWCBCWT

Experience01-45+01-45+01-45+

WCAfter training0110.0840.1030.066000

No training1110.0690.0850.0540.0.370.0370.037

BCAfter training0.0680.0520.029111000

No training0.0280.0210.0121110.0370.0370.037

UEAfter training0.2540.2060.12400.3500.4030 295000

No training0.1180.0930.0520.3050.3550.2550.0370.0370.037

28

Offer Probabilities

• Large positive effect of training on WC offers and on BC offers

• Very Low WT opportunities P=0.037

• Very low offers for WC from BC and higher , but low from UE.

29

Interpretation of Types

• Type 2 have unobserved characteristics that fit well the Israeli labor market – easily receive offers and do not need training. (22%).

• Type 1 – need the training to adjust but the cost is high (utility ~ liquidity problem).

30

Policy analysis by Counterfactual Simulations

Structural estimation enables to simulate the effect of alternative policy interventions on the choice distribution, wages, unemployment and the discounted expected utility (PV).

Policy Choices: Case 1: No training is available. Case 2: Only training in blue-collar (BT) is available.Case 3: Only training in white-collar (WT) is available.Case 4: Double the probability to participate in WT.

31

Table 12: Predicted Policy Effects on Mean Accepted Wages and Unemployment (4’th and 5’th years)

Policy ChangeNo Training is AvailableDouble WT Offer Rate

ImmigrantAccepted wage) %( ((Change)Accepted wage) %( (Change)

WCBCUEWCBCUE

BC in USSR schooling=12-1.1-0.103.52.50

WC in USSR schooling=15-0.8-0.103.42.60

32

Table 13: The Predicted Annual Effect of Training Availability on Mean Accepted Wages: Percent

Change Relative to an Economy without Training**Percent change of simulated mean accepted wages on the sample, comparing the training at the estimated model to a no

training economy.

AllWhite-Collar

Blue-Collar

Year 10.070.1460.035

Year 20.601.1720.239

Year 30.961.5590.318

Year 41.221.8830.396

Year 51.402.0290.492

All Years0.851.6050.261

33

Aggregate Wage Growth (Social Rate of Return)

• Aggregate wage growth is increasing overtime due to the permanent affect on job offers to WC.

• The social rate of return is above 1% mainly due to type 1 accepting WC jobs and type 2 BC jobs. Better process of job sorting.

• Double WT opportunities has a high (above 3%) social rate of return.

34

Table 14: Predicted Policy Effect on the Hourly Present Value (PV)

ExperimentBC in USSR, schooling=12

WC in USSR, schooling=15

age at arrival 30

age at arrival 45

age at arrival 30

age at arrival 45

Upon Arrival*

3,371.873,117.303,458.923,203.37

No Training-)1.11 (3,334.58

-)1.47 (3,071.45

-) 0.95 (3,425.98

-)1.35 (3,160.24

No WT-)1.11 (3,334.85

-)1.47 (3,071.45

-)0.95 (3,425.98

-)1.35 (3,160.24

No BT) 0.00 (3,371.87

)0.00 (3,117.30

)0.00 (3,458.92

)0.00 (3,203.37

Double WT offer

)0.96 (3,404.10

)1.24 (3,155.98

)0.84 (3,487.97

)1.16 (3,240.43

35

Table 15: Partition of the Gain from Training by Sources

ExperimentBC in USSR, schooling=12WC in USSR, schooling=15

age at arrival 30

age at arrival 45

age at arrival 30

age at arrival 45

No training)3,334.58()3,071.45()3,425.98()3,160.24(

No return in all sources

)3,334.57(

0.00

)3,071.43 (

0.00

)3,425.97(

0.00

)3,160.23(

0.00

Return in utility only)3,335.17(

1.6

)3,072.23(

1.7

)3,426.49(

1.6

)3,160.94(

1.6

Return in utility and terminal

)3,361.53 (

72.3

)3,105.20(

73.6

)3,448.90(

69.6

)3,190.00 (

69.1

Return in utility, terminal, job offer

)3,371.20(

98.2

)3,116.63(

98.6

)3,458.10(

97.5

)3,202.49(

98.0

36

Conclusions

• The model provided a way to estimate the social and the individual rate of return from alternative training programs.

• Most of the gain from training is due to increasing WC job opportunities over long time.

• Large fraction of wage growth is due to occupational mobility, experience and language learning.

• The return to imported imported human capital is zero conditional on the locally accumulated human capital.

37

TableA1. Summary Statistics

ObservationsPercentMeanSD

Schooling41914.582.74

Age at arrival41938.059.15

White-collar USSR28467.78

Blue-collar USSR12730.31

Did not work in USSR

81.91

Married36386.63

English4191.760.94

Hebrew before migration

5011.9

Ulpan Attendance38692.3

Ulpan completion33279.2

Ulpan Length )months(

3874.61.34

Hebrew1 )first survey(

4192.710.82

Hebrew2 )second survey(

3162.980.83