Migration and Self-Selection: Lessons from the Israeli Kibbutz

Ran Abramitzky
June 9, 2006

Abstract
I employ a longitudinal dataset of individuals linked across the 1983 and 1995 Israeli censuses of
population to test Borjas hypothesis that migrants self-selection depends on the dierence in earnings
inequality between the origin and the destination. The nature of the environment and dataset are
especially suited to test the selection hypothesis. The Kibbutz is a voluntary cooperative based on
equality in earnings distribution, while Israeli cities oer much higher earnings inequality. The findings
provide the first empirical support, at the individual-level, to Borjas hypothesis. Individuals who left
the Kibbutz are positively selected in their education and skills compared to stayers, and entrants into
the Kibbutz are adversely selected in their pre-entry earnings compared to non-entrants. Nevertheless,
the findings reveal that the type of migrants selection with respect to the population of origin may
dier from that with respect to the population of destination. Whereas less educated Kibbutz-to-city
migrants outperform natives with similar education level, more educated Kibbutz-to-city migrants do
not.

Stanford University. I am grateful to Joel Mokyr, Fabio Braggion, Adeline Delavande, Zvi Eckstein, Joe Ferrie, Avner

Greif, Reuben Gronau, Timothy Guinnane, Nisan Langberg, Charles Manski, Jacob Metzer, Chiaki Moriguchi, Rob Porter,
William Rogerson, Kathy Spier, Luis Vasconcelos and seminar participants at Harvard University, Stanford University, North-
western University, Yale SOM, Harvard Business School, University of Maryland, University of California at San Diego,
University of California at Irvine, University of California at Chicago and the University of British Columbia for helpful
discussions and comments. I thank Financial support from Northwestern Dissertation Dean Fellowship and the Economic
History Association Dissertation Award is gratefully acknowledged.
1 Introduction

Are migrants positively or negatively selected from their source country? How do migrants perform in

terms of their earnings compared to the native population? These questions have generated a great deal of

attention and dispute over the last few decades (prominent contributions include Chiswick (1978), Borjas

(1987), and Chiquiar and Hanson (2005)). In this paper, I test a well-known and much-disputed hypothesis,

according to which a positive self-selection of migrants is expected when the origin has a more equal earnings

distribution than the destination, while a negative one is expected when the origin has a more unequal

earnings distribution (Borjas (1987, 1991, 1994)).

According to Borjas, positive (negative) selection occurs when individuals with high (low) skill levels

leave the country of origin and outperform (underperform) the natives in the country of destination in terms

of their earnings. The basic idea is that a country with a low level of earnings inequality insures low-

skilled individuals and taxes high-skilled individuals. However, the empirical relevance of this conjecture

has been widely criticized (Chiswick 1999, Chiswick 2000, Liebig and Sousa-Poza 2003, Chiquiar and

Hanson 2005). First, migrants are only observed at the destination. This may bias the analysis both

because the destination may have policies favoring migrants with certain skills and because the migrants

self-selection should be compared to that of non-migrants.1 Second, Borjas measure of income inequality

may be poorly related to relative skill dierentials (Chiswick 1999, p. 184). Third, almost no study has

found evidence for the negative selection of migrants, regardless of whether they come from origins with

high or low income inequality.

I analyze a longitudinal dataset of individuals exiting from and entering to the Israeli Kibbutz to test

Borjas hypothesis. The Kibbutz is a good environment to test the selection hypothesis since it is a voluntary

cooperative aimed at equality in earnings distribution, and as such, it oers lower skill premia than do

Israeli cities. Thus, Kibbutz-to-city migrants are moving from an origin with low earnings inequality to a

destination with high earnings inequality, and city-to Kibbutz migrants are moving from high inequality

origin to low inequality destination.

1 See Chiswick (1978), Carliner (1980), Borjas (1987), Chiquiar and Hanson (2005), and the surveys by Borjas (1994),

Chiswick (1999), Burda, Hrdle, Mller, and Werwatz (1998) and Liebig and Sousa-Poza (2003) overcome these limitations
by employing data on the intention to migrate.

1
 The longitudinal data of individuals linked across the 1983 and 1995 Israeli population censuses allow me

to test Borjas hypothesis directly while avoiding the above-mentioned critiques. First, unlike international

migrants, there are no policy restrictions on Kibbutz-to-city migration, allowing a pure test of self-selection

in exit.2 In entry, the Kibbutz does impose restrictions on new applicants, but the census data contain

information on pre-entry wealth and earnings that are unobserved by the Kibbutz, allowing a test of adverse

selection. Second, I observe migrants in both the origin and the destination and can, thus, compare them

both to non-migrants and to the population of the destination. That is, I test whether Kibbutz-to-city

migrants are positively selected in their education and skills with respect to stayers, and whether, once in

the city, they earn higher wages than native workers with similar characteristics.

This analysis provides the first empirical support, at the individual level, to Borjas hypothesis. I

find that individuals who left are positively selected from the Kibbutz population in terms of both their

observable characteristics (education, occupation) and unobservable characteristics that are associated

with higher earnings outside the Kibbutz (such as ability, motivation). Negative selection of entrants to

the Kibbutz is suggested by their lower earnings, ceteris paribus, than non-entrants and by their lower

pre-entry wealth.

Yet, the analysis reveals that a positive selection with respect to the population of origin does not nec-

essary imply a positive selection with respect to the population of destination. More specifically, Kibbutz-

to-city migrants are not generally positively selected compared to native workers of the city. Whereas

less educated Kibbutz-to-city migrants outperform natives with similar education level, the performance of

more educated Kibbutz-to-city migrants is similar to that of educated workers in the Israeli labor market.

Besides the migration literature, this paper contributes to the literature on adverse selection. Following

the approach developed in Finkelstein and McGarry (2006) and Finkelstein and Poterba (2006), I test for

the presence of adverse selection by employing observable characteristics that are not used by the Kibbutz

in the application process. Although the importance of adverse selection is widely accepted, relatively

little empirical work tests for its presence. The diculty in conducting a direct test is that the researcher

usually has access to the data available to the less-informed party. To test for the presence of asymmetric
2 Gabriel and Schmitz (1995) also study internal migration and, thus, do not face policy restrictions. Their study, however,

only observes individuals at one point in time.

2
information, studies had to rely on evidence on prices (e.g., Greenwald and Glasspiegel (1983), Genesove

(1993)) or the correlation between unobservables in the contract choice and ex-post behavior (Chiappori

and Salani 2000a, Chiappori and Salani 2000b, Cardon and Hendel 2001, Finkelstein and Poterba 2004).

I have access to information regarding applicants productivity that the Kibbutz did not have. Specifically,

the census data contain information on pre-entry wealth and earnings that were unobserved or overlooked

by Kibbutz ocials. Hence, I possess information available only to the more informed party, which enables

a direct test of adverse selection. I find that, after controlling for information that is observable by the

Kibbutz, entrants are adversely selected.

Last, this paper contributes to the literature on the Kibbutz. The widely accepted view that the

success of the Kibbutz depends on its ability to keep productive members and to attract new ones has

weak empirical basis.3 Yet, many scholars and observers have attributed this ability to solidarity and

ideology rather than to economic forces. My findings reveal the role of economic forces in the decision by

individuals to live in the Kibbutz. The recent shift of the Kibbutz away from equality can be viewed as a

rational response to the brain drain and adverse selection documented here.4

Section 2 provides a brief description of the environment. Section 3 describes the theoretical framework

relating migrants selection to earnings inequality. The data source used in the analysis is described

in section 4 and section 5 tests the hypothesis that Kibbut-to-city migrants are positively self-selected

compared to non-migrants. Section 6 tests whether once in the city, Kibbutz-to-city migrants outperform

the Israeli labor force in terms of their earnings. Section 7 tests the hypothesis that entrants to the Kibbutz

are negatively selected from the Israeli labor force in terms of their pre-entry earnings, and section 8 tests

whether asymmetric information leads to adverse selection in entering the Kibbutz. Section 9 concludes.
3 The mainly sociological research is based on a small number of individuals and Kibbutzim. This research is mainly based
on talent tests that are given to movers and stayers (Natan, Shnable, and Peskin (1982) and Leviatan and Orchan (1982)),
or on former teachers evaluations of the talent level of movers and stayers (Helman 1982). Leviatan (1993) examines exit
rates of members who held major positions in their Kibbutz. No convincing evidence of a brain drain was found.
4 See Abramitzky (2005) for a theoretical and empirical analysis of the Kibbutz as an economic institution that provides

members with optimal insurance contracts.

3
2 Brief Background of the Environment

The Kibbutz is a unique society based on collective ownership of property and equality in consumption as

reflected in the Marxist idea from each according to his ability to each according to his needs. Before the

1950s, a main source of population growth had been net migration to the Kibbutz through youth movements

from Israel and abroad. Since then, however, the main source of population growth has been internal (i.e.,

Kibbutz-born individuals who have stayed in their Kibbutz).5 This implies that a Kibbutz ability to retain

its members has been critical to its stability and population growth. Each Kibbutz member can leave the

Kibbutz and earn a premium for his ability and eort outside the Kibbutz, but movers cannot take with

them their share of the Kibbutz (house, plants, public goods), which makes moving costly. Individuals

from outside the Kibbutz could apply to membership, which requires the approval of the Kibbutz (more

later).

The first Kibbutz (Degania) was established in 1910 primarily as a communal farm but today most

Kibbutzim have factories and an industrial base alongside agriculture. The Kibbutz in 1995 consisted

of 120,000 members in 268 Kibbutzim, located all over Israel, and it accounts for 2.6% of the Jewish

population in Israel. It had always been perceived that the standard of living of Kibbutzim was at least

as high as the rest of the Jewish population of Israel (Barkai 1977, Klinov and Palgi 2006).

The data used in this paper provide information on the same individuals at two points in time, in

1983 and in 1995, a period in which the Kibbutz was still based on full earning equality. This is a

period known in Israels historiography as the Kibbutz crisis period, since the Kibbutz was hit severely

by the 1985 government stabilization program as well as by the sharp decline in the world prices of

agriculture. As a result, many Kibbutz members in this period left their Kibbutz and membership declined

substantially. Table 1 presents the development over time of the total number of Kibbutzim and their

population. The Kibbutz population grew continuously throughout the century until the mid-1980s, when

it shrank substantially.6 As will be shown later, in the period studied in this paper (19831995), about
5 Barkai (1977) for the period before 1970 and my data set for the period after.
6 The recent decline in membership has been motivated by a number of factors including the decline in members ideology,
the financial crisis that forced Kibbutzim to reduce the facilities oered to members and the changes in the economic and
political environment facing Kibbutzim.

4
25% of members between the age of 15-54 left their Kibbutz.

3 Theoretical Framework

Borjas (1987) and Chiquiar and Hanson (2005) show that migrants selection depends on the dierences

in earnings inequality between the origin and the destination. Migrants from origins with low earnings

inequality are expected to be positively selected (i.e. they have higher-than-average skill) and migrants

from origins with high returns to skill and earning inequality are negatively selected (lower-than-average

skill). To illustrate, assume that individuals from the Kibbutz (indexed by 0) choose whether to migrate

to an Israeli city (indexed by 1). The Kibbutz is based on full earning equality:

ln(w0 ) = 0 (1)

where w0 is the wage in the Kibbutz and 0 is the base wage in the Kibbutz. If the entire population of

the Kibbutz were to migrate to the city, they would earn

ln(w1 ) = 1 + 1 s (2)

where w1 is the wage for Kibbutz migrants in the city, 1 is the base wage in the city, s is the level of skill

and 1 is the return to skill. Similarly to Chiquiar and Hanson (2005), I focus mainly on migrants selection

over observable skills, that is education and occupation, but I also account for selection in unobservable

skills, that is migrants unobservable (to the researcher) characteristics that are associated with a higher

(or lower) level of earnings.

Kibbutz members migrate to the city if

I = ln(w1 ) ln(w0 c) 1 + 1 s 0 c > 0 (3)

where c is the time equivalent measure of the cost of migrating from the Kibbutz to the city. The cost

of migrating from the Kibbutz include both direct costs and indirect costs due to the fact that migrants

cannot take with them their physical assets (e.g., their house and their share of the public goods and

plants). Equation 3 defines a cuto skill above which Kibbutz members would migrate to the city. That

5
is, we expect Kibbutz-to-city migrants to be positively selected from the Kibbutz population. This is

illustrated in Figure 1, which shows that individuals with a skill level below s remain in the Kibbutz, and

individuals with a skill level above s to the city.

The city-to-Kibbutz migration can be described in a similar way. Since the Kibbutz oers lower returns

to skill and higher earning equality, we expect city-to-Kibbutz migrants to be negatively selected from

the city population. The main dierence is that the Kibbutz is well-aware of the tendency of low-skilled

individuals to attempt to enter and does not admit applicants with (observed) skill level below s . However,

individuals may have private information on their skills (that is unobserved by the Kibbutz) and enter if

their privately observed skills is low (adverse selection). The census data used here contain information on

pre-entry wealth and earnings that were unobserved or overlooked by Kibbutz ocials. Hence, I possess

information available only to the more informed party, which enables a test of adverse selection in entering

the Kibbutz.

4 The Data

The data set used here is a random representative sample of individuals linked between the 1983 and the

1995 Israeli Censuses of Population (the link was done by the Israeli Central Bureau of Statistics). They

contain all Israeli citizens who answered the extensive questionnaire in both years, which was given at

each census to 20% of the households in a way that adequately represents the entire population. Thus,

the matched sample account for representative 4% of the Israeli population (including representative 4%

of Kibbutz members). The data identify individuals who live in a cooperative rural settlement, in which

production, marketing and consumption are organized in a cooperative manner, which uniquely defines

Kibbutz members. I employ three main sub-samples: (i) 1983 Kibbutz members observed also in 1995,

(ii) outside of Kibbutz residents observed in 1995 (including individuals who migrated from the Kibbutz

between 1983 and 1995) and (iii) outside of Kibbutz residents observed in 1983 (including individuals

who would later migrate to the Kibbutz between 1983 and 1995). The first sample allows me to compare

Kibbutz-to-city migrants with Kibbutz members who stayed in the Kibbutz. The second sample allows me

to analyze how well those Kibbutz-to-city migrants fare in the labor market relative to natives, and the

6
third sample allows me to compare city-to-Kibbutz migrants with individuals who stayed in the city.

To make these comparisons meaningful, I concentrate on the Jewish population of Israel and focus

on individuals who were in a relevant age to make a migration decision. Thus, in the first two samples,

the focus is on Jewish individuals who were between the age of 1554 in 1983 (and thus between the age

of 2766 in 1995). A total of 397 out of 1655 individuals in the sample left the Kibbutz between 1983

and 1995, more than 20%. The availability of census information on the relevant variables restricts the

number of movers in the second sample to 378. The second sample consists of 24,948 Jewish individuals

who lived outside the Kibbutz in 1995, out of whom 24,570 lived outside the Kibbutz in both 1983 and

1995, and 378 emigrated from the Kibbutz between 19831995. The latter are the same Kibbutz-to-city

migrants of the first sample, but here they are compared to the Jewish population outside the Kibbutz

rather than to the Kibbutz population. All 397 are analyzed in the first sample, but results are essentially

the same when only the 378 migrants, analyzed in the second sample, are considered. In the third sample,

only Jewish individuals who were between the age of 1949 in 1983 (3166 in 1995) are considered (as

opposed to individuals between the age of 1554 in the first two samples), since individuals before their

army service face dierent rules of entry than the one described in this paper and individuals over 50 years

old are not considered by the Kibbutz. The third sample also restricts attention to Israeli-born individuals

(as new-immigrants to Israel may be allocated temporarily to the Kibbutz) for whom the census provides

information on education, occupation and a relevant measure of wealth. There are 13,981 individuals who

satisfy these criteria, out of whom 125 entered a Kibbutz in this period. Since the census has more missing

observations on some wealth variables (wages) than on others (apartment size, television ownership), the

number of observations is dierent for each measure of wealth and is reported in Table 10 for each wealth

specification. The results are the same when I restrict attention to the smaller sample that contains

information on all three measures of wealth.

5 Positive Selection of Movers vs. Stayers

Over 30 years ago, Bettelheim (1969) argued that the Kibbutz breeds mediocrity and that the best and

worst members would tend to leave. Nevertheless, the sociological literature on the Kibbutz has not

7
found convincing evidence that this is the case.7 Moreover, there is no evidence that the less ideological

members were the first to leave.8

The census data at hand allow a test of the relative education and skill of those members who choose to

leave the Kibbutz. High-Skill occupations are defined as academic and managerial occupations. Low-

Skill occupations are unskilled industrial professions and service workers. A third omitted control group

contains all other occupations such as clerks, sales persons and workers in agriculture.

Summary statistics of the variables used in the regression are presented in Table 4, for both movers

and stayers. Individuals who left the Kibbutz are on average more educated than stayers (54% of migrants

have at least a high school diploma compared to 48% of stayers), and they are less likely to have a low-

skilled occupation (13% compared to 23%). Individuals with bigger families tend to stay, in part because

a Kibbutz member does not fully internalize the cost of having a big family as this cost is spread over the

entire Kibbutz population. Foreign-born individuals seem to stay in the Kibbutz, but this may simply be

a result of the fact that many movers are young and many of the young were born in Israel. A regression

analysis will disentangle these eects.

A Probit regression that assigns 1 to movers and 0 to stayers is presented in the second column of

Table 5. The regression provides further evidence for a brain drain from the Kibbutz. The more educated

members were more likely to leave, such that having at least a high school education increases the exit rate

by a third (from 20% in average to 30%). Members with high-skill occupations are nine percentage points

more likely to leave the Kibbutz and low-skill members are nine percentage points less likely to leave.

One potential problem with tests of the eect of skills on migration is that they ignore the eect of

returns to skills in host economies on investment in human capital in the home economy. Thus, the fact
7 Natan, Shnable, and Peskin (1982) followed an entire high school class for seven years. Before graduating, the minors
were asked to fill in a CPI questionnaire (personality questionnaire) combined with a talent test. Seven years later, the
researchers found no dierences between movers and stayers in either the talent test or in their education and their job
positions. Similarly, Helman (1982) found no such evidence in movers and stayers talent level as evaluated by their former
high-school teachers. Leviatan and Orchan (1982) based their research on members under 35 years old from 10 Kibbutzim
and found no dierence in a self-assigned talent level between movers and stayers. Leviatan (1993) considers the increase in
the percentage of movers who are older than 30 as evidence for the exodus of good members, as these people are generally
more experienced and more educated due to their age. He also found a 4% exit rate of members who held major positions in
their Kibbutz over a period of five years and viewed this as evidence that the best members tend to leave.
8 Leviatan (1996) evaluates the relative role materialism and ideology play in members migration decisions. He found

no evidence that members who are less ideological are more inclined to leave, or that members are more inclined to leave
Kibbutzim that are more materialistic and less ideological. Note that the materialistic level of a Kibbutz is proxied by the
level of a members satisfaction from the material level in their Kibbutz. Ideological Kibbutz is proxied by the level of a
members satisfaction from the level of Kibbutz values provided by the Kibbutz.

8
that migration from low to high earnings inequality economies is skill-biased could be attributed either to

selection or to selective investment in human capital. The application studied in this paper allow me to

alleviate this problem. The period of study is one of increased migration (25% of Kibbutz members in the

sample left between 1983 and 1995) following an exogenous crisis that hit the Kibbutz (for more details

on the crisis, see Abramitzky (2005)). Under these circumstances, the skill-bias of movers (compared to

stayers) more likely reflects the eect of skill on migration (the selection hypothesis) rather than dierent

investment in human capital of migrants compared to non-migrants. As a robustness check, I conduct

the same analysis on subsamples of individuals who were above 30, 35, and 40 years old in 1983. Older

individuals had already invested in their human capital years before the period studied here (probably

with the intention to stay, otherwise they would leave much earlier) and their decisions less likely represent

selective investment in human capital. The last fact is even more true in the Kibbutz, where one cannot

save privately and cannot take her share of the Kibbutz (e.g. her house, her share in the Kibbutz plants,

public goods) with her upon exiting, making the move more costly at older age. The results of the Probit

regression (1 if left the Kibbutz) of the subsamples are reported in Table 6 (the first column considers

all individuals over 15 years old in 1983. The second column considers only individuals who are at least

30 years old in 1983. The third column considers individuals over 35 and the fourth column considers

individuals over 40) . The regression results suggest that (under all specifications), more educated and

high-skill workers are more likely to exit and low-skill workers are less likely to exit. That is, the eect

of education and skill on the probability to migrate is economically large and statistically significant even

when older individuals who probably only migrated as a result of the crisis are considered.

To further examine the eect of income gain on the migration decision and to account for migrants self-

selection over unobservable characteristics, I employ a switching regression model (a Type 5 Tobit Model),

which is based on Heckmans (1979) two step selection model. This model allows for two regimes (in my

case stay in the Kibbutz or move to city), a criterion function (a migration rule that determines

whether or not an individual will leave the Kibbutz) and two regression equations describing income

determination in the Kibbutz and in the city. This framework has been used to study migration (see, for

example, Robinson and Tomes (1982) and Ferrie (1999) returns to schooling (Willis and Rosen 1979), and

9
the impact of union membership on wages (Lee 1978). One dierence between the Kibbutz-case and the

above examples stems from a defining feature of the Kibbutz, namely that the income earned by a member

is not a function of his personal characteristics.

Assume that an individual chooses whether or not to migrate (Di ) based on her income gain from mov-
95 95
ing from the Kibbutz to the city (ln(IiC ) ln(IiK )) as well as other individual and location characteristics

(Zi ) The individual migrates if an unobservable criterion function is positive:

Di = 1 Zi + 2 [ln(IiC
95 95
) ln(IiK )] + ui (4)

where E(ui ) = 0 and var(ui ) is normalized to 1 without loss of generality. For every Kibbutz-to-city
95
migrant, we observe the 1995 income level ln(IiC ) in the city. We do not observe the counterfactual city

income that a stayer would have got had he chosen to move to the city instead of staying in her Kibbutz.

However, the city income can be written as a function of personal characteristics Xi reflecting expected

productivity, which are observable for both movers and stayers. I assume that
95

ln(IiK ) = K
(5)
95 0
ln(IiC ) = C Xi + i

where E(i ) = 0 and var(i ) = c2 . The census does not provide information about the earnings of each

Kibbutz member. Yet, since the Kibbutz equated in the distribution of earnings, income is exogenous to
95
individual member and does not depend on his personal characteristics, i.e., ln(IiK ) = K. This assumption

can possibly bias the estimation of the structural equation 4. First, K may be a function the characteristics

of the average Kibbutz member and thus indirectly depend on members characteristics Xi . Note that even

if K is a function of individuals characteristics such that an individuals income in the city is positively

correlated with her income in the Kibbutz, then the estimated 2 underestimates the eect of income

dierences on migration and provides a lower bound for the true parameter. In this sense, my finding

that members leave the Kibbutz to seek higher income is underestimated. Second, dierent Kibbutzim

provide members with dierent levels of income. Unfortunately, the dataset does not allow me to identify

the specific Kibbutz in which individuals live. The assumption that income is constant across Kibbutzim

may over or underestimate the eect of income dierences on migration.9

9 On the one hand, members from less wealthy Kibbutzim are more likely to leave, resulting in an overestimation of 2 under

10
 The individuals decision rule Di is unobservable, but his choice between staying in the Kibbutz (D = 0)

and moving to the city (D = 1) is observable. More specifically,

0 0
D = 1 if Di = 1 Zi 2 K + 2 C Xi + (2 i + ui ) = Wi + i 0
(6)
0 0
D = 0 if Di = 1 Zi 2 K + 2 C Xi + (2 i + ui ) = Wi + i < 0.
where Wi contains all the elements of X and Z. To capture the dependence between individuals migration

decisions and their subsequent income, I assume that (i , i ) are i.i.d and distributed Bivariate Normal

with corr (i , i ) = c and covariance 12 .10 It follows that:11

95 0 ( 0 Wi )
E(ln(IiC )) = C XiC + 12 (7)
( 0 Wi ) i
( 0 Wi )
where is the normal pdf, ( 0 Wi ) represents the normal cdf and ( 0 Wi ) is the Inverse Mill Ratio. The

12 indicates the direction of selection over characteristics that are unobservable to the researcher. A

positive (negative) 12 suggests that a migrants unobservable characteristics are associated with a higher

(lower) level of income outside the Kibbutz.

The vector X includes 1983 individual characteristics thought to influence the income of Kibbutz

members who left to cities: age and age squared, occupation (high-skill or low-skill), education level (high

school diploma or higher) and gender. The elements of Z include age and age squared, family size, whether

the Kibbutz of origin is located in the far north or south of Israel, education level, foreign birth and gender.

The parameters 1 , 2 , and C are identified under the restrictions imposed by the model. To identify 1

and 2 , there must be at least one variable in X not in Z and vice versa. Both conditions are satisfied: Z

includes variables not included in X (family size, whether the Kibbutz of origin is located in the far north

or south of Israel and foreign birth), and X includes variables not in Z (the occupation variables), which

are assumed to aect the migration decision only through their eect on income.12

The structural parameters of the model were estimated via a three-step procedure.13 First, the reduced
0
b. Then, using
form equation Wi +i is estimated by Probit maximum likelihood to find b, we obtain the
my assumption. On the other hand, members from less wealthy Kibbutzim are expected to leave even when oered a lower
income in the city, resulting in underestimation of 2 under my assumption.
1 0 Although the distributional assumptions used in the analysis are standard, they are strong and might have substantial

eects on the estimation (Heckman 1990, Manski 1990, Newey, Powell, and Walker 1990).
1 1 See Greene (1997, Theorem 20.4, p. 975)
1 2 The exclusion restrictions were also tested using an F test and t tests, which rejected the hypothesis that they are omitted

variables. That is, there is no evidence of Zi terms not in Xi mattering apart from  their eect through the Mill ratio, and no
evidence of Xi terms not in Zi mattering apart from their eect through ln IiC 95 .
1 3 See Maddala (1983, pp. 223228) and Amemiya (1985, pp. 399402).

11
 (e
Wi )
Inverse Mill Ratio for each observation: Wi ) .
(e Second, equation 7 is estimated by OLS. That is, I run an
( 0 Wi )
95
OLS regression of ln(IiC ) on XiC and ( 0 Wi ) and get c
0 \ 95 c0
C and ln(IiC ) = C XiC . Estimates of the standard

errors of the coecients are corrected using Whites method to account for both the heteroscedasticity

introduced by the Inverse Mill ratio and the fact that is computed rather than known. Note that by
95
assumption ln(IiK ) = K is known for each individual, which means that the counterfactual for a mover
\
is known and does not have to be estimated. Third, equipped with ln(I 95 95
iC ) and ln(IiK ), I estimate the

structural decision equation 4 to find the determinants of migration.14

Table 5 presents the estimation results. The first column displays the variable means. The second

column presents the reduced-form Probit equation and the third column presents the structural Probit

equation (equation 4). The fourth column presents the OLS regression of the determinants of earnings
Prob(move)
in the city (equation 5). The marginal eects X are evaluated at the mean of X for continuous

variables and are Prob(D = 1/X = 1) Prob(D = 1/X = 0) for discrete variables (and calculated at the

means of the other Xs).

The results provide further evidence for brain drain from the Kibbutz in the period 19831995. The

most productive members were more likely to leave the Kibbutz during this period.

As already discussed, the reduced-form Probit regression (column 2) suggests that Kibbutz-to-city

migrants were more educated and more skilled compared to stayers. The structural Probit equation (column

3) reveals the importance of the expected income in the migration decision. The large positive coecient
\
on ln(I 95
iC ) indicates that individuals anticipating higher income outside the Kibbutz were much more likely

to leave. An anticipated increase of 2500 NIS (about $600) increases the probability to leave by thirty

percentage points, making a member more than twice more likely to leave.

Furthermore, the OLS regression (that was corrected for selection) presented in column 4 suggests

that Kibbutz-to-city migrants were also positively selected from the Kibbutz population in terms of their

characteristics unobservable by the researcher. The positive and large coecient on the Mill ratio shown

in Table 5 (fourth column) suggests that members who left the Kibbutz were positively selected in the

sense that their unobservable (by the researcher) personal characteristics were associated with a higher
1 4 Standard errors were corrected by bootstrapping through data resampling with 1000 repetitions to account for the fact
 95 
that ln IiC is estimated rather than observed.

12
income level outside the Kibbutz. That is, migrants did better outside the Kibbutz than predicted by their

observable characteristics.15

The results corroborate the observation made by a secretary of Kibbutz Givat Brenner in 2001 who

summarize the process that has taken place in his Kibbutz during the last two decades:16

I dont want to shout it out loud, but there is a negative selection process at work here. We

have lost some of the best and brightest of our young adults....We do not have enough members

in the twenty-five to forty age group, and frankly, those who have stayed are not the best.

The aggregate data collected by Kibbutz ocials are also suggestive that a brain drain from the Kibbutz

has taken place.17 Table 2 provides evidence that the percentage of those who leave the Kibbutz without

a diploma has decreased dramatically. Whereas on the eve of the crisis, almost 90% of those who left did

not have a diploma, in 1999 this percentage decreased to about 60%. The number of educated members

leaving the Kibbutz has increased continuously over the last fifteen years. Table 3 presents the field of

study of those who left the Kibbutz in the last fifteen years. It suggests that since the 1980s, increasingly

educated members leave the Kibbutz and that as time goes by, more economists, engineers and teachers

leave and fewer members who studied humanities and education.

Other factors aect the decision to stay in the Kibbutz (column 3). Young members are more likely

to leave. As discussed, members with bigger families prefer to stay in the Kibbutz and let other members

subsidize them, rather than leave and finance their big family outside the Kibbutz. An additional family

member decreases the exit probability by two percentage points (from 20% on average, to 18%). Overseas

immigrants are four percentage points more likely to leave the Kibbutz than Israeli-born individuals,

although this parameter is not measured precisely. Members from Kibbutzim in the far north and south

of Israel were five percentage points more likely to leave their Kibbutz. It is also interesting to note

dierences by gender in the migration decision. Summary statistics and the reduced form equation do not

reveal dierences in behavior between men and women and even suggest that more men than women have
1 5 By construction of the model, those who left the Kibbutz would have earned the same (as those who stayed) had they

decided to stay instead of leaving the Kibbutz.

1 6 Gavron (2000, p. 68)
1 7 Nevertheless, they also reflect a general trend in Israel towards acquiring a higher level of education (specifically in fields

such as economics and engineering), which is the reason for the need of longitudinal micro-level data that allow a comparison
between movers and stayers.

13
left the Kibbutz in the late 1980s and early 1990s. On the other hand, the structural equation suggests

that men tend to earn more than women outside the Kibbutz, which is why we see more men leaving.

When controlling for this income eect, it turns out that men are 20% less likely to leave the Kibbutz than

are women. This may be a result of more women marrying out of the Kibbutz than men, but this issue

requires further investigation.

Wages of former Kibbutz members outside depends on a number of factors (column 4). As expected,

more educated individuals fare better and so do individuals with academic and managerial occupations.

Unskilled workers in industry and service workers who left the Kibbutz could expect to earn less outside

the Kibbutz than do other industry and service workers. Income is larger for older individuals, but the

rate of increase decreases with age. Men earn higher incomes than women outside the Kibbutz.

6 Relative Performance of Kibbutz-to-city Migrants Outside the

Kibbutz

The previous section provided evidence of positive selection of Kibbutz-to-city migrants compared to mem-

bers who stayed in the Kibbutz. This section tests whether Kibbutz-to-city migrants are positively selected

compared to the Israeli labor force outside the Kibbutz. That is, I evaluate how former Kibbutz members,

who left the Kibbutz between 1983 and 1995, perform outside the Kibbutz in 1995 relative to individuals

already in the city. Unlike in the previous section, whether a member left her Kibbutz can be treated as

exogenous and thus there is no need to use sample selection. I estimate the following regression model:18

ln (Ii ) = + 0 Xi + 1 EXIT i + 2 (EXIT Educated )i + i (8)

where Ii is the monthly income earned outside the Kibbutz in 1995, Xi is a vector of ex-ante (i.e. 1983)

characteristics that are believed to aect ex-post (i.e. 1995) monthly income and includes age, age squared,

gender, a dummy for Israeli born, education (a dummy for high school education or higher) and occupation.

EXIT i is a dummy variable indicating a former Kibbutz member who left the Kibbutz between 1983 and

1995, and (EXIT Educated )i indicates an educated worker who left the Kibbutz.

The results are presented in Table 7. The first column presents the mean characteristics of the whole
18 A similar regression model was analyzed in the seminal work of Chiswick (1978) and Carliner (1980).

14
sample. The second column displays the mean characteristics of the 378 Kibbutz-to-city migrants and

the third column presents the mean characteristics of individuals who lived outside the Kibbutz in both

periods. The last column shows the determinants of wages outside the Kibbutz in 1995.

The first three columns of Table 7 suggest that the average Kibbutz-to-city migrant in 1983 (before

migration) is younger (26 years old) than the average Israeli Jew (30 years old) and more educated (52%

have at least a high school diploma compared to 45% in the Jewish population outside the Kibbutz), but

less likely to have a high-skill occupation (8% as compared with 12%), and more likely to work in a low-skill

occupation (13% as compared with 8%). Table 8 shows that, on average (and without controlling for indi-

viduals characteristics), Kibbutz-to-city migrants earn similar wages to the Jewish average. Nevertheless,

whereas less-educated Kibbutz-to-city migrants earn higher wages than less-educated natives (5627NIS

compared with 5121NIS), more-educated migrants earn lower wages than more-educated natives (6984NIS

compared with 7656NIS).

The regression results (the fourth column of Table 7) reveal that, ceteris paribus, Kibbutz-to-city

migrants earn higher wages than the general Jewish population (i.e. 1 is positive and significant), but that

this dierence is entirely due to less-educated Kibbutz-to-city migrants. That is, whereas more-educated

migrants earn similar wages to the general Jewish population, less-educated migrants earn a substantial

wage premium (i.e. 2 is negative and significant and similar in magnitude to 1 ). These results make

sense since less-educated workers who were willing to leave the Kibbutz and forgo sharing their income with

more-educated workers, are expected to be more talented than average less-educated workers. Following

the same logic, more-educated workers who leave the Kibbutz are not expected to earn more than average

more-skilled workers. These results also suggest that the type of selection with respect to the population of

the destination might dier across individuals with dierent observable characteristics (such as educational

level).

7 Negative Selection in Entering a Kibbutz

In this section, I test the type of selection of city-to-Kibbutz migrants compared to non-entrants. The

positive selection hypothesis has it that migrants are more able and more highly motivated than natives

15
(Chiswick 1978, p. 900). If motivation and ability are positively correlated with both propensity to migrate

and labor market outcome, one expects to observe that potential migrants earn, ceteris paribus, more than

individuals who do not migrate. This observation forms the bases of the work by Gabriel and Schmitz

(1995), who study internal migration in the US and found that migrants were positively selected.

On the other hand, according to Borjas hypothesis, migrants tend to come from the lower tail of

the home countrys income distribution (Borjas 1987, p. 534) if the origin has a more unequal income

distribution than the destination. If negative selection is present and entrants to the Kibbutz are less able

than non-entrants, than one expects potential entrants to the Kibbutz to earn, ceteris paribus, less than

individuals who do not enter. Nevertheless, supporting this conjecture has so far not been successful.19

I evaluate whether in 1983 (prior to entry) prospective migrants to the Kibbutz earned, ceteris paribus,

a lower wage than Jewish Israelis who did not migrate to the Kibbutz. Specifically, I conduct the following

regression analysis (similar type of regression as in Mellow (1981), Mincer (1978), and Gabriel and Schmitz

(1995)

ln (W83 ) = + 0 X83 + ENTER +

where W83 is the individuals wage earned outside the Kibbutz in 1983, X is a vector of characteristics that

are believed to aect wages and includes age, age squared, gender, a dummy for Israeli born, education

and occupation. ENTER is a dummy indicating whether the individual entered a Kibbutz between 1983

and 1995. A positive will indicate a favorable selection of city-to-Kibbutz migration and a negative will

indicate an unfavorable selection of city-to-Kibbutz migration.

The results presented in Table 9 provide strong evidence of negative selection. The coecient on the

entry variable is negative, big and significant, suggesting that prospective (ex-ante) migrants have lower

wages than nonmigrants. That is, those who subsequently entered a Kibbutz earned, ceteris paribus, a

lower wage than non-entrants.

1 9 Borjas (1987) found, using US data, some evidence for negative selection of immigrants to the US from countries with a

higher level of income inequality. His findings and methodology was criticized by Chiswick (1999) and others.

16
8 Adverse Selection in Entering a Kibbutz

Even if Kibbutz ocials are well-aware of this tendency, adverse selection might still transpire as individuals

who seek to enter may have private information regarding their productivity. The census data contain

information on individuals pre-entry wealth and hours worked, which are not observable (apartment size,

ownership of some durable goods, hours worked) or dicult to verify (wage) by Kibbutz ocials. Having

relevant information about the informed party allows me to test for adverse selection.

A central body, the Kibbutzs Department of Absorption and Demographic Growth, is in charge of the

screening process of individual applicants to the Kibbutz and imposes strict rules on entry.20 Applicants

must be between twenty and fifty years old, having served in the army, and have at least twelve years

of education.21 They must be married with no more than three children and both the husband and wife

must have an occupation that will allow them to find a job within the Kibbutz or its area immediately.

Moreover, applicants have to go through a lengthy interview, to fill out forms about their and their

childrens physical and mental health, to give a statement about their education level, occupation and

army experience and to write a Curriculum Vita, which is tested by a graphologist. They also have to

answer a long questionnaire meant to check whether the individual is personally suited to live in a Kibbutz.

Only about one third of the Kibbutzim absorbs new members from outside the Kibbutz. These Kib-

butzim are typically average Kibbutzim (neither the rich Kibbutzim nor the poor ones) and they are

more likely to be located in Israels far south and far north.22 Some Kibbutzim require an entry fee, but

this can often be recovered in part upon leaving. Applicants can list their preferences to join particular

Kibbutzim, but the decision takes into account both applicants preferences and Kibbutzim needs, so one

can be assigned to any Kibbutz.

I do not have census data on applicants to the Kibbutz, but only on individuals who actually entered.
2 0 This section draws on the application forms and interviews with all members of the Kibbutzs Department of Absorption
and Demographic Growth. I am grateful to them for sharing the data and ideas with me.
2 1 In the early 1980s, the leaders of the Kibbutz wanted to restore the central place of the Kibbutz in Israeli society and

decided to engage in a massive absorption. Since then, however, the Kibbutz has put more eort into the screening process
and decided to concentrate on quality rather than quantity.
2 2 Poor Kibbutzim often cannot absorb outsiders, as they lack the resources to build new apartments for children who

decide to staystill the main source of population growth. Poor Kibbutzim also face unemployment, which prevent them
from absorbing new individuals from the outside. Rich Kibbutzim often also do not accept outsiders, as they face high
demand for apartments from their own children. Kibbutzim in the center of Israel and those close to the sea are crowded
with members and often do not have room for new housing.

17
This makes it harder to document adverse selection since it only includes applicants who were accepted by

the Kibbutz and exclude those who were rejected (the application process is described below). Hence, my

results provide a lower bound for the actual adverse selection.

I conduct a Probit analysis of entry to the Kibbutz in 19831995, in which I evaluate whether entrants

tend to be less productive in terms of both characteristics that are observable to the Kibbutz (such as

education and occupation) and ones that are unobservable-to-Kibbutz (such as wealth). Evidence for the

latter is interpreted as adverse selection. More specifically, I test the hypothesis that entrants to the

Kibbutz are adversely selected in terms of their pre-entry wealth, which is a proxy for their privately held

information regarding their productivity and ability.23 I consider a few measures of wealth in my regression.

The first (column 3 in Table 10) is an individuals wage prior to entry. The second measure (column 4)

is an individuals number of rooms if they own an apartment (the measure equals zero if the individual

does not own an apartment). A third measure (column 5) uses both wages and the number of rooms (if

they own an apartment) as regressors. To account for the possible endogeneity of apartment ownership,

I instrument the number of rooms in the individuals own apartment with television ownership and wage

(column 6). The characteristics that are public information (i.e. observable to both the individual and the

Kibbutz) are age, gender, family size, education and occupation prior to entry.

Xi is defined as the vector of publicly observed characteristics that aect individuals decision of whether

to join a Kibbutz. As discussed above, these include age, gender, family size, education and occupation

prior to entry. Yi is the privately observed characteristics (various measures of wealth) that aect this

decision. Let Di denote the (unobservable) reduced form outcome generated by both the individuals

decision and the Kibbutzs decision whether to accept the individual:

Di = 10 Xi + 20 Yi + i (9)
2 3 Hours worked may be a more problematic measure of ability and is thus excluded from the analysis that follows.

Nevertheless, including it only strengthens my claim: I find that individuals who worked more hours prior to entry are less
likely to enter the Kibbutz (and all the other results presented below remain the same). Results are available upon request.

18
 Denote Di the observable decision:

Di = 1 if Di > 0 (10)

Di = 0 otherwise

These two equations imply that:

prob(Di = 1) = prob(i > 10 Xi 20 Yi ) = 1 F (10 Xi 20 Yi ) (11)

where F is the cdf of i . Assuming i N (0, 1), the Likelihood function is:24

L = Di =0 (10 Xi 20 Yi )Di =1 [1 (10 Xi 20 Yi )] (12)

Then, maximizing the Likelihood function we find an estimator for 0 = (10 , 20 ). We need to find:

(10 Xi +20 Yi )
1. Xik = (10 Xi +20 Yi )1k . That is, how does the probability of entering a Kibbutz change as

individual characteristics that are observable by the Kibbutz change? In particular, do less educated

individuals enter? Do individuals with less-skilled occupations enter? If so, this would indicate

negative selection that is either due to self-selection or due to policy restrictions imposed by the

Kibbutz

(10 Xi +20 Yi )
2. Yik = (10 Xi + 20 Yi )2k . That is, how does the probability of entering a Kibbutz change

as individual characteristics that are not observable by the Kibbutz change? In particular, given X,

do individuals with lower wealth enter a Kibbutz? If so, this would indicate adverse selection.

Table 10 presents the mean characteristics of both individuals who entered the Kibbutz between 1983

and 1995 (column 1) and those who were outside the Kibbutz in both periods (column 2) as well as the

regression results (columns 36).

The average individual who enters a Kibbutz is younger than the general population (28 vs. 31 years

old) and has a smaller family (3.18 vs. 3.81 members). An entrant is more educated than the general

Jewish population (75% possess at least a high-school diploma vs. 53% in the general population). He
2 4 The proportion of entrants to the Kibbutz is small, implying that the assumption of a normal distribution might be

strong. To check for robustness, I tried two other specifications: a Logit regression (with a logistic distribution) and a linear
probability model. Results are qualitatively the same in all specifications.

19
is equally as likely to work in a high-skill occupation as the average worker, and is also equally likely to

work in an low-skilled occupation. However, his wage is substantially lower than that of the general Jewish

population (20735 Shekels vs. 24921 Shekels) and his apartment is smaller (2.07 rooms vs. 2.81).

The Probit regression (Table 10 columns 36) provides no evidence of negative selection over observable-

to-Kibbutz characteristics, but strong evidence of adverse selection. That is, whereas younger and more

educated individuals are more likely to enter, entrants to the Kibbutz are, ceteris paribus, much less wealthy

than non-entrants (in all measures of wealth).

9 Conclusion

Employing and analyzing a unique longitudinal data set of individuals linked across the Israeli 1983 and

1995 censuses, I show that Kibbutz-to-city migrants are positively selected compared to those who stayed

in their Kibbutz and that entrants to the Kibbutz are adversely selected from the Jewish population of

Israel.

The results provide supportive evidence, at the individual level, to Borjas hypothesis that the type

of selection (positive or negative) depends on the relative inequality in income distribution of the origin

and destination. Nevertheless, the results also reveal the limitations of his hypothesis. More specifically, I

find that positive selection of migrants with respect to non-migrants does not necessarily mean a positive

selection with respect to the population of the destination across education levels. Whereas less-educated

Kibbutz-to-city migrants outperform individuals with similar characteristics, more-educated Kibbutz-to-

city migrants do not outperform individuals with similar characteristics.

The analysis of entry to the Kibbutz provides a direct test for adverse selection. I show that, controlling

for publicly observable information, entrants to the Kibbutz are adversely selected in terms of their privately

observed pre-entry wealth.

Recently (after the period analyzed in this paper), many Kibbutzim have introduced substantial reforms

and shifted away from full equality. The new Kibbutz presents dierentiated wages and related material

incentives to reward ability.25 In light of the papers findings, the recent institutional change of the Kibbutz
2 5 See Abramitzky (2005) for a detailed discussion of the reforms.

20
can be viewed as a rational response to the brain drain and adverse selection it faces.26 The reforms allow

high ability individuals to receive a premium for their ability and thus reduce the incentives of skilled

individuals to leave the Kibbutz and of unskilled individuals to enter a Kibbutz.

References

Abramitzky, R. (2005): The Limits of Equality: An Economic Analysis of the Israeli Kibbutz, Stanford

University mimeo.

Amemiya, T. (1985): Advanced Econometrics. Harvard University Press, Cambridge, MA.

Barkai, H. (1977): Growth Patterns of the Kibbutz Economy. North-Holland Pub. Co., Amsterdam; New

York.

Bettelheim, B. (1969): The Childrren of the Dream. The McMillan Company, New York.

Borjas, G. J. (1987): Self-Selection and the Earnings of Immigrants, The American Economic Review,

77(4), 531553.

(1991): Immigration and Self-Selection, in Immigration, Trade and the Labor Market, ed. by

J. Abowd, and R. Freeman. The University of Chicago Press, Chicago.

(1994): The Economics of Immigration, Journal of Economic Literature, 32(4), 16671717.

Burda, M., W. Hrdle, M. Mller, and A. Werwatz (1998): Semiparametric Analysis of German

East-West Migration Intentions: Facts and Theory, Journal of Applied Econometrics, 13(5), 525541.

Cardon, J., and I. Hendel (2001): Asymmetric Information in Health Insurance Markets: Evidence

from the National Medical Expenditure Survey, RAND Journal of Economics, 32(3), 408427.

Carliner, G. (1980): Wages, Earnings, and the Hours of First, Second and Third Generation American

Males, Economic Inquiry, 18(1), 87102.

2 6 See Abramitzky (2005) for a theoretical and empirical study of the Kibbutz as a self-enforcing institution that balances

insurance and incentives.

21
Chiappori, P.-A., and B. Salani (2000a): Testing Contract Theory: A Survey of Some Recent Work,

in Advances in Economics and Econometrics: Theory and Applications: Eighth World Congress, ed. by

M. Dewartripont, L. P. Hansen, and S. Turnovsky, vol. 1. Cambridge University Press, Cambridge.

(2000b): Testing for Asymmetric Information in Insurance Markets, Journal of Political Econ-

omy, 108(1), 5678.

Chiquiar, D., and G. H. Hanson (2005): International Migration, Self-Selection, and the Distribution

of Wages: Evidence from Mexico and the United States, Journal of Political Economy, 113(2), 239281.

Chiswick, B. (1978): The Eect of Americanization on the Earnings of Foreign-Born Men, Journal of

Political Economy, 86(5), 897921.

(1999): Are Immigrant Favorably Self-Selected?, The American Economic Review, 89(2), 181

185.

(2000): Are Immigrants Favorably Self-Selected? An Economic Analysis, in Migration Theory:

Talking Across Disciplines, ed. by C. D. Brettell, and J. F. Hollifield, pp. 6176. Routledge, New York.

Ferrie, J. P. (1999): Yankeys Now: Immigrants in the Antebellum United States, 18401860. Oxford

University Press, New York.

Finkelstein, A., and K. McGarry (2006): Private Information and its Eect on Market Equilibrium:

New Evidence from Long-Term Care Insurance, Forthcoming in the American Economic Review.

Finkelstein, A., and J. M. Poterba (2004): Adverse Selection in Insurance Markets: Policyholder

Evidence from the U.K. Annuity Market, Journal of Political Economy, 112(1), 183208, Part 1.

(2006): Testing for Adverse Selection with Unused Observable, NBER Working Papers 12112,

National Bureau of Economic Research, Cambridge, MA.

Gabriel, P. E., and S. Schmitz (1995): Favorable Self-Selection and the Internal Migration of Young

White Males in the United States, Journal of Human Resources, 30(3), 460471.

22
Gavron, D. (2000): The Kibbutz: Awaking from Utopia. Rowman & Littlefield Publishers, Lanham, MD.

Genesove, D. (1993): Adverse Selection in the Wholesale Used Car Market, Journal of Political Econ-

omy, 101(4), 644665.

Greene, W. H. (1997): Econometric Analysis. Prentice-Hall, Upper Saddle River, NJ, third edn.

Greenwald, B. C., and R. R. Glasspiegel (1983): Adverse Selection in the Market for Slaves: New

Orleans, 18301860, Quarterly Journal of Economics, 98(3), 479499.

Heckman, J. (1979): Sample Selection Bias as a Specification Error, Econometrica, 47, 153161.

(1990): Varieties of Selection Bias, American Economic Review, 80(2), 313318.

Helman, A. (1982): The Economic Motive in Leaving the Kibbutz, Discussion Paper 10, Rupin Insti-

tute.

Klinov, R., and M. Palgi (2006): The Standard of Living in Kibbutzim A Comparison to Urban

Households, .

Lee, L.-F. (1978): Unionism and Wage Rates: a Simultaneous Equations Model with Qualitative and

Limited Dependent Variables, International Economic Review, 19(2), 415433.

Leviatan, U. (1993): When Do the Best Ones Leave?, Kibbutz Reseach Centre mimeo [In Hebrew].

(1996): Negative Selection Among Youngsters in the Kibbutz, Discussion paper, The Institute

for Study and Research of the Kibbutz, University of Haifa, In Hebrew.

Leviatan, U., and E. Orchan (1982): Kibbutz Born Ex-Members and Their Adjustment to New Life

Outside the Kibbutz, Interchange, 31, 1628.

Liebig, T., and A. Sousa-Poza (2003): How Does Income Inequality Influence International Migra-

tion?, ERSA conference papers ersa03p472, European Regional Science Association.

Maddala, G. S. (1983): Limited-Dependent and Qualitative Variables in Econometrics. Cambridge Uni-

versity Press, Cambridge.

23
Manski, C. F. (1990): Nonparametric Bounds on Treatment Eects, American Economic Review,

80(2), 319323.

Mellow, W. (1981): Unionism and Wages: A Longitudinal Analysis, Review of Economics and Statis-

tics, 63(1), 4352.

Mincer, J. (1978): Family Migration Decisions, Journal of Political Economy, 86(5), 749773.

Natan, M., A. Shnable, and H. Peskin (1982): Together and Alone: Kibbutz-born Individuals Who

Graduated High School in 1969Followup After 10 Years, HaKibbutz, 8, 105127.

Newey, W. K., J. L. Powell, and J. R. Walker (1990): Semiparametric Estimation of Selection

Models: Some Empirical Results, American Economic Review, 80(2), 324328.

Robinson, C., and N. Tomes (1982): Self-Selection and Interprovincial Migration in Canada, Canadian

Journal of Economics, 15(3), 474502.

Willis, R. J., and S. Rosen (1979): Education and Self-Selection, Journal of Political Economy, 87(5),

S7S36, Part 2.

24
 TABLE 1
Kibbutz Population, 1910-1999.
Year Number of Kibbutzim Population
1910 1
1920 12 805
1930 29 3877
1940 82 26554
1950 214 66708
1960 229 77955
1970 229 85100
1980 255 111200
1987 268 127000
1990 270 125100
1999 267 117400
Source: Pavin, The Kibbutz Movement: Facts and Figures (2001).
 TABLE 2
Education of Movers (% of movers)
year no diploma technicians practical engineers teachers other diploma BA MA Ph.D
1985 87.5 0.1 0.5 0.5 5.4 0.4 0.2
1986 84.8 0.2 0.3 0.8 0.3 5.6 0.8 0.3
1987 77.9 0.2 0.6 1 11.9 6.2 0.7 0.2
1988 73.3 0.2 1 2.2 14.3 8.2 0.8 0.1
1989 70.5 0.2 1.2 2.6 15.5 8.9 0.9 0.3
1990 70.8 0.2 1.1 2.4 15.3 8.6 0.8 0.7
1991 73 0.3 1.8 2.5 13.2 8.1 0.8 0.3
1992 70.2 0.3 1.5 2.7 14.1 8.9 1 0.8
1993 65 0.9 1.6 2.5 12.7 13.6 1.9 0.4
1994 59.6 0.3 2.8 3.6 14.7 15.2 1.6 0.7
1995 63.5 0.6 2.5 3.1 14.6 13.2 1.4 0.5
1996 62.4 0.4 2.8 3 12.8 16.5 1.2 0.4
1997 61.8 0.3 2.9 2.5 13.9 15.9 1.6 0.6
1998 60.4 0.3 2.6 2.8 15.2 16.4 1.2 0.7
1999 61.4 0.5 2.4 2.9 13.5 17.2 1.1 0.5
100 70 0.3 1.6 2.3 12.2 10.2 1 0.4
Total 17338 79 394 564 3016 2673 257 107
Source: Pavin, The Kibbutz Movement: Facts and Figures (2001).
 TABLE 3
Field of Study of Educated movers (% of educated movers)
year nature sciences engineering exact sciences social sciences economics and management medicine art humanities education
1985 5.9 10.9 4.2 14.3 8.4 10.9 7.6 16.8 21
1986 8.8 12.2 3.4 10.2 10.2 12.9 4.1 15.6 23.1
1987 9.5 14 4.1 4.5 8.1 12.7 7.2 12.7 20.8
1988 8.6 12.4 4.9 9 9 8.6 8.2 12.7 22.1
1989 8.9 15.4 3.8 9.6 9.6 10.6 7.2 9.6 21.5
1990 7.9 16.2 4.1 12.8 12.8 11.4 5.5 10 18.6
1991 8.5 18.1 2.6 11.1 11.1 8.1 7 12.6 19.6
1992 10.1 15.2 4.1 12.2 12.2 12.2 8.8 8.8 18.9
1993 9.9 21.3 3.5 10.8 10.8 9.6 10.5 7 15.2
1994 9.4 19 3.6 10.2 10.2 10.4 11.5 7.8 17.4
1995 9.2 20.4 4 11.5 11.5 10.6 6.6 7.8 18.4
1996 10.1 18.5 2 12.4 12.4 8.7 10.1 7.6 16.3
1997 6.8 20.6 3.1 13.2 13.2 9 9.3 8.7 14.1
1998 7.4 17.5 3 11.8 11.8 11.5 11.8 7.4 15.9
1999 7.5 21.7 3.1 11.1 11.1 9.2 9.5 7.5 13.6
Source: Pavin, The Kibbutz Movement: Facts and Figures (2001).
 TABLE 4
Summary Statistics of Movers vs. Stayers, 1983-1995.
Stayed in Kibbutz Left the Kibbutz
Variable Mean Mean
Age 34.94*** 27.23***
Man 0.5 0.54
Family Size 3.57*** 2.88***
At Least High School Diploma 0.48** 0.54**
High-skilled 0.08 0.09
Low-skilled 0.23*** 0.13***
Born in Israel 0.68*** 0.78***
Israel's North region 0.53 0.54
Israel's South region 0.19* 0.23*

Observations 1258 397

t-test for difference in means significant at ***1% **5% *10%
Note: all variables are measured in 1983.
 TABLE 5
Mover/Stayer Analysis of Kibbutz Members, 1983-1995.
1 2 3

Pr ob(Move) Pr ob(Move) ln (I ic )
95

Variable Mean X X X
Specification Reduced Structural OLS
form form
Personal characteristics:
Age 33.09 -0.025*** -0.051*** 0.082***
Age squared 0.001 0.0007*** -0.002***
Male 0.51 0.009 -0.200*** 0.681***
Family size 3.41 -0.018*** 0.018***
At Least High School Diploma 0.49 0.100*** -0.03 0.413***
High-skilled 0.08 0.090** 0.357***
Low-skilled 0.21 -0.093*** -0.249***
Born in Israel 0.71 -0.039 -0.039
ln(I) 7.35 0.312***

Kibbutz characteristics:
Israel's North region 0.53 0.044* 0.044*
Israel's South region 0.2 0.053* 0.055*

Selection parameter:
Mills ratio 0.778***

Predicted Probablity 0.203 0.203

Observations 1655 1655 397
R 0.326
Adjusted R 0.314
F 26.92***
LR 255.49*** 254.30***
Coefficient significant at ***1% **5% *10%.
Notes: the dependent variable in columns 1 and 2 is equal to 1 if person moved from the Kibbutz.
the dependent variable in column 3 is the natural log of earnings in 1995.
All other variables are measured in 1983.
"High-skilled" are individuals working in either academic or managerial occupations.
"Low-skilled" are individuals working in either unskilled occupation in industry or as service
workers. A third control group contains all other occupations.
"Israel's North region" includes the following districts: Sefad, Kinneret, Yizrael, Akko and Golan.
"Israel's South region" includes Ashkelon and Beer-Sheva districts.
 TABLE 6
Probit analysis of exit from the Kibbutz between 1983-1995, by age

Variable Mean reg 1 reg 2 reg 3 reg 4

Specification age>15 age>30 age>35 age>40

Personal characteristics:
Age 33.09 -0.025*** 0.028 -0.048 -0.163*
Age squared 0.001 -0.0004 0.0004 0.002
Male 0.51 0.009 0.024 0.007 0.157
Family size 3.41 -0.018*** -0.003 -0.003 -0.012
At Least High School Diploma 0.49 0.100*** 0.072*** 0.081*** 0.068***
High-skilled 0.08 0.090** 0.076** 0.087** 0.124***
Low-skilled 0.21 -0.093*** -0.069** -0.057** -0.023
Born in Israel 0.71 -0.039 -0.058*** -0.044** -0.018

Kibbutz characteristics:
Israel's North region 0.53 0.044* 0.029 0.014 -0.009
Israel's South region 0.2 0.053* 0.017 -0.022 -0.004

Predicted Probablity 0.203 0.126 0.095 0.058

Observations 1655 1029 734 430

LR 255.49*** 62.62*** 52.38*** 35.77***

Coefficient significant at ***1% **5% *10%.
Notes: Dependent variable is equal to 1 if the person left the Kibbutz.
All explanatory variables are measured in 1983. The coefficients reported are marginal probabilities.
"High-skilled" are individuals working in either academic or managerial occupations.
"Low-skilled" are individuals working in either unskilled occupation in industry or as service
workers. "Israel's North region" includes the following districts: Sefad, Kinneret, Yizrael,
Akko and Golan. "Israel's South region" includes Ashkelon and Beer-Sheva districts.
 TABLE 7
Analysis of Kibbutz-to-city migrants' 1995 Wage Premium

ln( Wage )
Variable Mean (1) Mean Mover (2) Mean City (3) X (4)
Age 30 26 30 0.566***
Age squared -0.015***
Male 0.558 0.532 0.558 0.566***
Born in Israel 0.61 0.775*** 0.607*** 0.095***
At Least High school diploma 0.457 0.524*** 0.456*** 0.313***
High-skilled 0.121 0.082** 0.122** 0.407***
Low-skilled 0.079 0.127*** 0.079*** -0.251***
EXIT 0.016 0.117**
EXIT*At Least High School Diploma 0.008 -0.110*

Constant 6.905***

Monthly Income (NIS) 6277.7 6338.1 6276.8

Observations 378 24,570 24,948
R 0.2584
Adjusted R 0.2581
F 965.42***
Column (2) and (3): t-test for difference in means significant at *** 1% ** 5% * 10%
Column (4): coefficient significant at *** 1% ** 5% *10%
Notes: "High-skilled" are individuals working in either academic or managerial
occupations. "Low-skilled" are individuals working in industry or construction, or
as service and sales workers. "EXIT" is an individual who lived in a Kibbutz
in 1983 and outside the Kibbutz in 1995. "EXIT*At Least High-School Diploma" is an individual who left
the Kibbutz with at least a high school diploma.
Notice that the dependent variable (wage) is measured in 1995 and all explanatory variables are measured in 1983.
 TABLE 8
Wage (NIS) of Kibbutz-to-city Migrants vs. the Rest of the Jewish Population, 1995.
All education levels More educated Less educated
Kibbutz-to-city migrants 6338 6984 5121*
Rest of Jewish Population 6277 7656 5627*
t-test of difference in means significant at ***1% **5% *10%
Note that a "more educated" individual is one with at least a high-school diploma
and a "less educated" individual is one with less than a high-school diploma
 TABLE 9
Wage Level Equation, 1983 (OLS).
ln (W83 )
Variable: X
Age 0.503
Age Squared -0.025
Male 0.338
BA degree or more 0.131
High-skilled 0.089
Low-skilled -0.113
ENTER -0.112

Observations: 11010 (out of which 67 entered a Kibbutz subsequently)

R 0.2564
Adjusted R 0.256
F 542.02
Coefficient significant at *** 1% ** 5% * 10%
High-skilled are individuals working in either academic or managerial occupations.
Low-skilled are individuals working in either unskilled occupation in industry or as
service workers or in craft. "ENTER" indicates an individuals who joined a Kibbutz
between 1983 and 1995.
 TABLE 10
Probit Analysis of Entry to the Kibbutz, 1983-1995.
Pr ob(enter )
X
measure of wealth used:
(1) (2) (3) (4) (5) (6)
Variable Mean entrants Mean non-entrants wage apartment size (3) and (4) IV for (4)
Observables:
Age 28*** 31*** -0.0011** -0.0027*** -0.0008 -0.0011**
Male 0.57 0.51 0.0031** 0.0029** 0.0032** 0.0026*
Family size 3.18*** 3.85*** -0.0013*** -0.0007 -0.0007 -0.0009*
At Least High School Diploma 0.75*** 0.53*** 0.0052*** 0.0053*** 0.0058*** 0.0053***
High-skilled 0.08 0.15 -0.0014 -0.0044* -0.0026 -0.0028
Low-skilled 0.22 0.18 0.0011 -0.0014 0.0016 0.0016

Unobservables:
wage (Old Shekels) 20735** 24921** -0.0006** -0.0007**
apartment size (if own) 2.07*** 2.81*** -0.0008** -0.0007*
IV for apartment size 2.7*** 2.88*** -0.0047***

Observations:
Entrants 51 8479 67 125 51 51
non-Entrants 10943 13856 8530 8479

Predicted Probablilty 0.0043 0.0066 0.004 0.0042

LR 48.81*** 69.56*** 43.81*** 41.83***
Notes: columns (1), (2): t-test of difference in means significant at ***1% **5% *10%
Column (3): coefficient significant at *** 1% ** 5% * 10%
"High-skilled" are individuals working in either academic or managerial occupations. "Low-skilled" are individuals working in either unskilled occupation in industry
or as service workers or in craft. "apartment size" is the number of rooms in own apartment (equals 0 if the individual does not own apartment). The instrumental
variable (IV) is a predicted value of the regression of apartment size on television ownership and wage. Summary statistics shown for smallest sample.
Regression run on the smallest sample are very similar and the main result holds. All explanatory variables are 1983 characteristics
 FIGURE 1

wage 1-c+1s

1-c

s* s (skill)