Transcript שקופית 1 - LUISS Guido Carli

INCENTIVES TO INVEST IN STUDYING THE
NATIVE LANGUAGE OF THE HOST COUNTRY
Erez Siniver
Department of Economics
College of Management, Israel
2
ABSTRACT
Cross-sectional analyses show that immigrant earnings tend to rise faster than
those of natives. One reason for this phenomenon is that immigrants' wages
rise as they acquire the host country's native language. Immigrants can
improve their knowledge of the native language simply by interacting with
native speakers or by taking formal language courses. The present study
inquires whether immigrants with the highest expected benefits from studying
Hebrew will tend more to invest in learning the language by taking the basic
Hebrew course.
INTRODUCTION
3
The economic literature indicates a positive relationship between immigrants‘
knowledge of the native language of the host country and their
earnings.Chiswick and Repetto (2001) and Chiswick (1998), using the 1972
and 1983 census of Israel, respectively, found that Hebrew speaking skills
and Hebrew literacy increase with the level of schooling and duration in Israel
and that earnings increase with the acquisition of both writing and speaking
Hebrew skills. Other studies [e.g, Beenstock (1996), Berman, Lang and
Siniver, (2003), Beenstock, Chiswick, Repetto (2001), Beenstock, Chiswick,
Paltiel, (2005)] also found that earnings of immigrants to Israel increase with
being more proficient in Hebrew. Studies conducted by Carliner (1981) and
Lazear (1995) found that immigrants are most likely to learn English when
they live in communities having small proportions of individuals from their
home country. Immigrants living in communities with large proportions of
compatriots will tend to learn English more slowly. This finding is explained by
the fact that immigrants who live in ethnic enclaves obtain lower returns for
knowing the native language than do immigrants who live in communities
having small proportions of compatriots.
METHODS
4
The lifetime earnings of Russian immigrants who have taken the course in
Hebrew and of Russian immigrants who have not taken the course were
compared. In each period t there is a probability of employment that
depends, among other factors, on Hebrew skills, and a wage that depends
on Hebrew skills and multiplicatively on experience (t-s), where s is either 0
or 12 months depending on whether the immigrant did not take the course or
did take it, respectively. Adding discounting, I have that the immigrant will
take the course if the sum from 12 months to retirement with Hebrew set at
its highest level is greater than the sum from 0 to retirement with Hebrew
adjusting with time in the labor market.
In order to arrive at lifetime earnings, I estimate the following parameters:
a. How quickly immigrants learn Hebrew without taking the course in Hebrew.
b. The employment probability as a function of Hebrew knowledge, labor market
experience and other factors.
c. Earnings as a function of Hebrew knowledge, potential experience and other
factors.
d. Demographic characteristics of Russian immigrants who have taken the
course in Hebrew.
e. The present value of earnings with and without taking the course for each
immigrant.
DATA
5
The data were obtained from the Survey of Recent Immigrants (SRI)[1]. The
data were based on a sample of nearly 1,200 households, migrants from
the former Soviet Union. These households contain 2715 individuals aged
16-65. The information I derived from the survey is:
(1) Personal details such as: Gender, coded as 1 for male and 0 for
female; marital status coded as 1 for married and 0 for single; age,
years of education and year of immigration to Israel.
(2) Details about employment and current earnings. Respondents were
asked whether they were employed and if they were employed what
were their current earnings.
(3) Details regarding the immigrants' ability to speak and write Hebrew.
Respondents were asked to classify their ability to speak Hebrew as
"fluently”, "with difficultly" or "cannot speak Hebrew at all", which were
coded also as 1, 2 and 3, respectively. Respondents were asked to
classify their ability to write Hebrew as "fluently," "with difficultly" or
"cannot write Hebrew at all" coded also as 1, 2 and 3, respectively.
[1] Israel Central Bureau of Statistics, 1993. Monthly Bulletin of Statistics,April 1994. Jerusalem: ICBS. (Hebrew).
Table 1 – Descriptive Statistics
6
Russian Immigrants Aged 16-65
All Samples
Employed
Unemployed
Age
38.83
(0.25)
38.98
(0.33)
38.69
(0.38)
Years of education
13.66
(0.05)
13.81
(0.07)
13.51
(0.08)
Experience
19.18
(0.25)
19.17
(0.33)
19.18
(0.37)
Year since Migration
25.28
(0.07)
25.55
(0.06)
25.00
(0.13)
Currently married
0.82
(0.01)
0.85
(0.01)
0.79
(0.01)
Male
0.53
(0.01)
0.52
(0.01)
0.54
(0.01)
Speak Hebrew
1.44
(0.01)
1.40
(0.02)
1.48
(0.02)
Write Hebrew
1.79
(0.02)
1.79
(0.02)
1.79
(0.02)
-
1949.36
(33.74)
-
Employed
50.06%
-
-
Unemployed
49.94%
-
-
Studied Course
15.80%
13.39%
17.92%
Did not studied Course
84.35%
Wage
No. of cases
2715
86.61%
1359
82.08%
1356
Table 1a – Descriptive Statistics
7
Russian Immigrants Aged 16-65 with 13+ years of education
All
Samples
Employed
Unemployed
Age
41.30
(0.27)
40.62
(0.35)
42.04
(0.41)
Years of education
15.35
(0.04)
15.34
(0.05)
15.36
(0.06)
Experience
19.95
(0.26)
19.27
(0.34)
20.68
(0.40)
Year since Migration
25.22
(0.09)
25.53
(0.08)
24.88
(0.17)
Currently married
0.91
(0.01)
0.92
(0.01)
0.90
(0.01)
Male
0.54
(0.01)
0.51
(0.02)
0.57
(0.02)
Speak Hebrew
1.40
(0.01)
1.36
(0.02)
1.44
(0.02)
Write Hebrew
1.73
(0.02)
1.72
(0.02)
1.74
(0.03)
-
1977.61
(42.10)
-
Employed
52.31%
-
-
Unemployed
47.69%
-
-
Studied Course
19.20%
16.10%
22.53%
Did not studied Course
80.82%
83.90%
77.47%
1758
919
839
Wage
No. of cases
8
THE PROBABILITY OF ACHIEVING ROFICIENCY IN HEBREW
WITHOUT A FORMAL COURSE
To estimate the probabilities of achieving moderate or fluent proficiency in
Hebrew within any given period of time for immigrants of different
characteristics, two ordered probit estimations were run. The dependent
variables for the first and second order probit stimation are the ability to
speak and to write Hebrew, respectively. Immigrants were asked to classify
their ability to speak/write Hebrew as "fluently," "with difficultly" or "cannot
speak/write Hebrew at all", which were coded as 1, 2 and 3, respectively.
The independent variables were: marital status (a dichotomous variable,
where 1 = married and 0 = single), gender (a dichotomous variable,
where 1 = male and 0 = female), education (in number of schooling years),
duration in Israel (in months of residence), and age.
9
Table 2 – The Probabilities of Achieving Fluent Proficiency in Hebrew
Without Taking a Formal Course.
Dependent
variable –
ability to
speak
Hebrew
Dependent
variable –
ability to
write
Hebrew
(1)
(2)
Gender
-0.056*
(0.022)
0.109*
(0.041)
Marital status
0.389*
(0.192)
1.033*
(0.147)
Education
-0.262*
(0.017)
-0.309*
(0.017)
Age
0.113*
(0.005)
0.096*
(0.004)
Duration in Israel
(month)
-0.206*
(0.032)
-0.176*
(0.034)
Duration in
Israel^2
0.002*
(0.0008)
0.003*
(0.0008)
Cutoff1
-2.0139
(0.463)
-2.812
(0.471)
Cutoff2
0.742
(0.453)
-0.707
(0.464)
2715
-1668.546
2715
-2258.244
# of observation
Log likelihood
10
THE PROBABILITY OF EMPLOYMENT
To calculate the probability that an immigrant will be employed, the following
probit regression was run, using employment as the dependent variable
(a dichotomous variable, where 1 = employed and 0 = unemployed).
The independent variables entered into the equation were gender, marital
status, education, experience, experience^2, residence in Israel, ability to
speak Hebrew and ability to write Hebrew.
Table 3: Probabilities of Employment
Dependent variable –
Employment
Gender
-0.120
(0.078)
Marital status
0.379*
(0.140)
Education
0.002
(0.016)
Experience
0.023*
(0.010)
Experience^2
-0.0006
(0.0003)
Duration in Israel
(month)
0.003*
(0.0009)
Ability to Write2
0.065
(0.107)
Ability to Write3
0.085
(0.147)
Ability to Speak2
-0.179*
(0.065)
Ability to Speak3
-0.515*
(0.206)
Constant
-1.209
(0.365)
Number of
Observations
Log Likelihood
2714
-1851.465
11
12
THE EARNINGS ESTIMATION
There is a vast international evidence that speaking the language of the host
country fluently has a positive effect on the immigrants' earnings. Indeed,
Table 3 shows that immigrants who improve their ability to speak Hebrew
also improve their probability of finding a job. This implies that the OLS
estimates might be biased. In our case, it might be that those with low
potential wage chose not to participate in the workforce, which creates an
upward bias in the OLS equation for wage.
To estimate the earnings equation controlling for self-selection I use (1) The
inverse Mill's ratios in a standard two-stage Heckman model; (2) The
Maximum Likelihood estimation, Newton-Raphson maximization.
13
Table 4 – Earnings – Equation Estimates.
2-step Heckman
(1)
Maximum Likelihood estimation
Newton-Raphson maximisation
(2)
Dependent variable – Employment
Probit selection equation
Gender
-0.074
(0.049)
-0.077
(0.049)
Marital status
0.235*
(0.087)
0.236*
(0.087)
Education
0.0009
(0.010)
0.0009
(0.010)
Experience
0.018*
(0.009)
0.014*
(0.009)
Experience^2
-0.0004*
(0.0002)
-0.0004*
(0.0002)
Duration in Israel
(month)
0.002*
(0.0006)
0.002*
(0.0006)
Ability to Write2
0.102
(0.067)
0.081
(0.067)
Ability to Write3
0.140
(0.092)
0.103
(0.092)
Ability to Speak2
-0.113*
(0.042)
-0.113*
(0.032)
Ability to Speak3
-0.324*
(0.128)
-0.308*
(0.128)
Table 4 – Earnings – Equation Estimates.
2-step Heckman
(1)
14
Maximum Likelihood estimation
Newton-Raphson maximisation
(2)
Dependent variable – Ln Wage
Outcome equation
Gender
0.009
(0.237)
0.011
(0.032)
Marital status
0.072
(0.772)
0.068
(0.059)
Education
0.007
(0.008)
0.007
(0.006)
Experience
0.009*
(0.003)
0.009*
(0.004)
Experience^2
-0.0002
(0.001)
-0.0002
(0.0001)
Duration in Israel
(month)
0.013
(0.065)
0.013*
(0.006)
Ability to Write2
-0.017
(0.317)
-0.015
(0.043)
Ability to Write3
-0.044
(0.442)
-0.038
(0.058)
Ability to Speak2
-0.020*
(0.003)
-0.018*
(0.004)
Ability to Speak3
-0.149*
(0.050)
-0.147*
(0.057)
InvMillsRatio
0.405
(5.032)
Sigma
0.619
0.608*
(0.036)
rho
0.655
0.623*
(0.106)
Number of
Observations
Log Likelihood
2715
2715
-2912.574
15
DEMOGRAPHIC CHARACTERISTICS OF IMMIGRANTS
WHO INVEST IN THE FORMAL COURSE
In this section, I discuss the relationship between the demographic
characteristics (gender, years of education, age, marital status) of the
Russian immigrants and the incentives for them to invest in studying
Hebrew.
The dependent variable is study of the native language (a dichotomous
variable, where 1 = taking the course and 0 = not taking the course).
The independent variables are: gender (1 = male, 0 = female), marital status
(1 = married, 0 = single), education (years of schooling), age, and age^2.
16
Table 5 – Demographic Characteristics of Immigrants Who Invest in Formal Course.
Dependent variable –
Study in Ulpan
Dependent variable –
Study in Ulpan
Probability of taking
a formal course
in Hebrew
(1)
Probability of taking
a formal course
in Hebrew
(2)
Gender
0.0136*
(0.050)
0.130*
(0.060)
Marital status
-0.533*
(0.201)
-0.575*
(0.206)
Education
0.141*
(0.022)
0.845*
(0.141)
Age
0.018*
(0.0033)
0.022*
(0.0033)
Age^2
-0.00005*
(0.00004)
-0.00004*
(0.00004)
Benefit
-
-
-4.095*
(0.557)
-2.782*
(0.565)
2715
-1138.640
2715
-1141.268
Constant
Number of
Observations
Log Likelihood
17
BENEFITS GAINED WHEN TAKING
THE COURSE IN HEBREW
I estimate the PV of lifetime earnings for Russian immigrants who have taken a course in
Hebrew and for those who have not taken the course. If the difference in earnings is
higher than the foregone earnings, the immigrants are better off if they take the
course.
The PV of lifetime earnings for immigrants who have not taken the course in Hebrew is
calculated as:
PV1 
65*12
3
3
[Pt (e) Pt (S  i,W  j) * Et (S  i,W  j)]/(1  r )t
t 0
i 1 j 1
Where S is the ability to speak Hebrew, W is the ability to write Hebrew and Pt(e) is the
probability that the immigrant is employed. After each specified number of months,
Pt(S=i, W=j) is the probability that a Russian immigrant will have achieved S(i) and
W(j); Et(S=i, W=j) is the earnings given that the immigrants' ability to speak Hebrew is
level i, and the immigrants' ability to write Hebrew is level j.
The PV of lifetime earnings for immigrants who have taken the course in Hebrew (which
extends 12 months) is calculated as:
PV 2 
65*12
[ P (e) * E (S  1,W  1)]/(1  r )
t 12
t
t
t
The data show that immigrants who had taken the course in Hebrew could speak and
write Hebrew fluently (i.e, S=1, W=1).
18
BENEFITS GAINED WHEN TAKING
THE COURSE IN HEBREW
In order to test whether the decision to take the course in Hebrew is driven by
the benefit that each immigrant gains when taking the course, I added to the
probit regression the independent variable benefit (a dichotomous variable,
where 1 = immigrants whose PV2/PV1 is in the top 15.8 percent of all the
immigrants and 0 = otherwise). If the decision to take the course is driven by
the benefit each immigrant gains when taking the course in Hebrew, I
expect to find that only the coefficient for the independent variable benefit is
significant and the coefficients for the other independent variables are not.
19
Table 6 – Benefit Gained When Taking the Course in Hebrew
Benefit Gained When Taking the
Course in Hebrew
Test for Benefit as a
Single Motivation
Dependent variable –
Benefit
Dependent variable –
Study in Ulpan
r = 1%
(3)
r = 2%
(4)
r = 4%
(5)
r = 6%
(6)
r = 10%
(7)
r = 4%
(8)
3.526*
(1.040)
3.528*
(1.071)
3.543*
(1.013)
3.728*
(1.071)
3.643*
(1.013)
0.130
(0.100)
Marital status
-29.833*
(5.321)
-30.819*
(5.545)
-29.169*
(5.161)
-31.819*
(5.545)
-30.179*
(5.261)
-0.444
(0.428)
Education
-29.951*
(5.211)
-31.352*
(5.516)
-28.961*
(5.001)
-32.452*
(5.616)
-29.971*
(5.011)
0.135*
(0.022)
Age
12.749*
(2.252)
13.403*
(2.399)
12.282*
(2.152)
13.8403
*
(2.499)
13.280*
(2.252)
0.037
(0.038)
Age^2
-0.069*
(0.013)
-0.074*
(0.014)
-0.067*
(0.012)
-0.084*
(0.014)
-0.077*
(0.015)
-0.00015
(0.0004)
Benefit
-
-
-
-
-
0.267*
(0.068)
118.536
*
(20.487)
123.415
*
(21.497)
115.129*
(19.765)
120.415
*
(21.550)
116.129*
(20.765)
-4.293*
(0.595)
2715
2715
2715
2715
2715
2715
-1138.142
Gender
Constant
Number of
Observations
Log Likelihood
20
BENEFITS GAINED WHEN TAKING
THE COURSE IN HEBREW
Uneducated immigrants benefit more from taking the course than educated
immigrants; however, taking the course is more common among educated
immigrants than among uneducated immigrants. It might be that the greater
tendency of educated immigrants to take the course may reflect lower
psychic cost of education for this group or it might be that the coefficients for
the independent variables ability to speak Hebrew and ability to write
Hebrew are probably biased upward for less educated immigrants (Berman,
E., Lang, K., Siniver, E. 2003). To deal with this problem, I have redone the
entire analysis only for immigrants with 13+ years of education, the results
of which are shown in Table 7.
21
Table 7 – Immigrants with 13+ Years of Education.
Demographic Characteristics of
Immigrants Who Invest in
Formal Course
Benefit Gained When Taking
the
Course in Hebrew
Test for Benefit as a
Single Motivation
Dependent variable – Study in
Ulpan
Dependent variable –
Benefit
Dependent variable –
Study in Ulpan
(1)
r = 4%
(2)
r = 4%
(3)
Gender
0.093*
(0.011)
0.354*
(0.100)
0.101
(0.124)
Marital status
-0.451*
(0.248)
-38.234*
(1.429)
-0.417
(0.348)
Education
0.063*
(0.032)
33.759*
(10.639)
0.019
(0.048)
Age
0.033*
(0.004)
14.269*
(2.872)
0.065
(0.045)
Age^2
-0.0009*
(0.0002)
-0.770*
(0.282)
-0.0004
(0.0005)
Benefit
-
-
0.609*
(0.214)
Constant
-3.269*
(0.931)
-2.698*
(0.902)
-3.263*
(0.932)
Number of
Observations
Log Likelihood
1758
-846.653
1758
1758
-1195.336
22
SUMMARY AND ONCLUSIONS
1.
2.
3.
4.
5.
The length of time it takes immigrants who do not take a formal course to
attain moderate or fluent ability to speak Hebrew is longer for male,
uneducated, older and married immigrants than for female, educated,
younger and single immigrants.
Immigrants who improve their ability to speak Hebrew also improve their
probability to be employed as well as their earnings.
Male, uneducated, older and single immigrant workers benefit more from
taking the formal course in Hebrew than do female, educated, younger
and married immigrants.
Taking the course in Hebrew is more common among male, educated,
older and single immigrant workers than among female, uneducated,
younger and married immigrant workers.
When including only immigrants with 13+ years of education, the results
show that the decision to take the course is driven only by the benefit
each immigrant gains when taking the course in Hebrew. The conclusion
is that the earnings maximization model does a good job of predicting
who will take the Hebrew course, especially when including immigrants
with 13+ years of education.