i‘
E! in:
a

.
It 1 .AI.
\r.‘......:w;m“a€u, “an.
J fifinzigahfmq
7:... i. :23: a .2. .
kt: . a. . F .4“;ng
: .. s. .“31 o
. in x haw. . $1.“.
$1 I. , . 39.3.»...3: z . .
:x ..

7.. .7....v
3.11.9

. 2.:
-2: . .
5.2.5::

111v
. o. i

. . panning? .. .mwh
£5 {.95.} Inn...“- . Qt
33 90mm"

. . .

30¢th

.3. v:
:3... .c. s

3...... o .t:i.
4‘: . 1‘ $ 5.
3:...

aura-3:

2"

1731~

and.“

(97-:th -

 

«if!

"VI"

.0

ix 3.?
Ii 11 I’VE: \\ v)
{if . 9.. {n .
3. I: thetuf .xl}
V is
.. 3.117;}:
in? «122.1?

'2;

 

m“; 8. § ‘ i . t' MA
fl ~Pv. ‘Jlnkdfiu
«

.w

m fifingéé fiflfifiafifi.

 

Illlllllllllllllllll‘llllllHillllllll‘llllllllllllllllll

3 1293 010466

This is to certify that the

dissertation entitled

Three Essays on the Labor-Market Characteristics

of Immigrants
presented by

Alan Michael Barrett

has been accepted towards fulfillment
of the requirements for

Eh. D . degree in Economics

[/amfl/Mc //

a 101' professor

Date August. 26. 1995

MSU is an Affirmative Action/Equal Opportunity Institution 0« 12771

A;
4 _ 4- ._____——.'

_v

.~—_.——‘-
- 4 w,——.———

 

 

 

 

LIBRARY
Michigan State
University

 

 

 

PLACE II RETURN BOXtonmwoflibchodthmm your“.
TO AVOID FINES Mun on or Moro dd. duo.

DATE DUE DATE DUE DATE DUE

      

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU icAn Afflunltivo Action/E Opportmity lnotituion
W mm:

THREE ESSAYS ON THE LABOR-MARKET CHARACTERISTICS OF
IMMIGRANTS

BY
Alan Michael Barrett

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

Doctor of Philosophy

Department of Economics

1994

ABSTRACT

THREE ESSAYS ON THE LABOR-MARKET CHARACTERISTICS OF
IMMIGRANTS

BY
Alan Michael Barrett

This dissertation is a theoretical and empirical analysis
of the ,labor-market characteristics of different immigrant
groups.

The first two groups considered are skill-based and
family-based immigrants. By estimating the earnings difference
between these two groups from a range of countries, I develop
an indicator of the nature of the immigrant flows from those
countries. A small difference between the two immigrant groups
from a country indicates a family-based group that is quite
skilled, so the immigrant selection process of that country
leads to the immigration of the high skilled. The opposite
applies if there is a large difference in skill levels.

I expand on the analysis by examining whether the
variation in my indicator of labor-market quality can be
explained by country-specific characteristics. In so doing, I
test the model developed by Borjas in his M article of
September 1987.

The next two groups I compare are family-based immigrants

and those who won a greencard in one of the lotteries of

recent years. These groups differ in that one group joins
family members while the other does not. My analysis includes
how these groups differ in their locational choices.

The final two groups of immigrants that I compare are
those who arrived in the years 1979 to 1980 and 1989 to 1990.
This comparison helps to determine whether the decline in
immigrant quality of recent decades continued into the 1980's.

The data I use is from the Immigration and Naturalization
Service. I find that the relative skill levels of family-based
and skill-based immigrants differ across countries. Part of
this variation can be explained by differences in income
inequality, so the Borj as model receives support. The lottery
immigrants are more skilled than the family-based immigrants
and are more geographically concentrated. Finally, the data
show that the decline in immigrant quality did not persist in

the 1980 ' s.

ACKNOWLEDGEMENTS

I would like to begin by thanking my advisor, Professor
Dan Hamermesh. I consider myself very fortunate to have
written my dissertation under his guidance. His help and
encouragement were simply invaluable. I have learned much from
him and am eternally in his debt.

I would also like to thank the other members of my
committee, Professors Jeff Biddle and Harry Holzer. Both
provided valuable comments on earlier drafts of the
dissertation. They also helped me with the most intimidating
part of the process, the mainframe.

My colleagues Leslie Schenk and Steve Hughes have played
an important role in my graduate career and I would like to
thank them for the support they have provided. Whenever I hear
the words "a pitcher of Labatts please" I will think of you.
Sanjay Padaki also deserves thanks, for his friendship and for
drawing the figures in this dissertation.

I reserve special thanks for Rosella Gardecki. She has
given me love and support and so much more. Thanks Zelli

Finally, I want to thank my parents, Jim and Dympna
Barrett. Without their love and support over the years, I
would never have arrived at this point. Whatever I may have

achieved, they deserve more than half the credit.

iv

TABLE OF CONTENTS

List of Tables page vii
List of Figures ix
Chapter 1: Introduction

1. Statistics on Immigration 1

2. The Issues Involved in Immigrant Labor-Market

Quality 4
3. Literature Review 6
4. Dissertation Outline 10

Chapter 2: The Effect of Class of Admission on Immigrant
Selection across Countries

1. Introduction 15
2. The Model 17
3. Outline of the Empirical Approach 30
4. Empirical Results, Part 1 33
5. Empirical Results, Part 2 36
6. Summary and Conclusion 46

Chapter 3: Evidence on Immigrant Selection from the
Greencard Lottery Winners

1. Introduction 49
2. The Model 51
3. Estimating the Lottery/Family Wage Differential 60

4. The Locational Decisions of the Lottery Immigrants 68

5. Summary and Conclusion 79

Table of Contents (cont'd)

Chapter 4: Using the 1988-1990 Immigrant Cohort to Learn
More about Immigrant Selection

1. Introduction 82

p 2. Skill-Based and Family-Based Immigrants Revisited 83

3. The 1979-80 and 1989-90 Immigrant Cohorts Compared 93

4. Summary 105
Chapter 5: Summary and Conclusions

1. A Review of What Has Been Done in the Dissertation 107

2. A Review of the Results 109
3. Policy Conclusions 112
4. Ideas for Further Research 115

Appendix A: Sources of Country Specific Data Used in the
Regressions Presented in Table 4 117

Appendix B: A Further Explanation of the Relationship
between Mean Income in the Country of Origin

and Selection 118
Appendix C: Rising Inequality and Rising Immigrant

Quality 119
List of References 120

vi

Table 1:

Table 2:

Table 3:

Table 4:

Table 5:

Table 6:

Table 7:

Table 8:

Table 9:

Table 10:

Table 11:

Table 12:

Table 13:

LIST OF TABLES

The Flow of Immigrants into the United States
1901-1990 3

Regression which Estimates the Coefficients of
the Variables Shown Using Data from all 41
Countries 35

Estimating the Wage Premium for the Skill-Based
Immigrants 37

Using Country-Specific Characteristics to Explain
the Variation in the Coefficient of SKILL 43

Regressions Estimating the Lottery/Family Wage
Differences for Lottery Winning Countries 62

Regressions Estimating Lottery/Family Wage
Differences with Lottery Groups Broken into
Different Lotteries 66

The Geographic Distribution of Irish Male
Immigrants by Census Region and Class of
Admission 70

Tobit Analysis of the Location Decisions of
Irish Lottery Immigrants 73

The Regressions of Table 8 with the Lottery
Group Broken into Three Waves 77

Estimating the Wage Premium for the Skill-Based
Immigrants from the 1988-90 Cohort 87

Using Country-Specific Characteristics to
Explain the Variation in SKILL for 1988-90 90

Regression Using Changes in Inequality and
Changes in Per-Capita GNP to Explain Changes in
the SKILL Coefficient 90

Regressions Using All Male Immigrants Aged
22-64 Admitted in the Years 1979-80 and 1989-90 96

vii

List of Tables (cont'd)
Table 14: Regressions for Seven World Regions Using Male

Immigrants Aged 22-64 Admitted in the Years
1979-80 and 1989-90 101

viii

Figure 1:
Figure 2:

Figure 3:

LIST OF FIGURES

Positive Selection
Negative Selection

Refugee Sorting

Figure 4(a): Lottery Immigrants with Different Means and

Smaller Foreign Country Income Variance

Figure 4(b): Lottery Immigrants with Different Means and

Figure 5:

Figure 6:

Bigger Foreign Country Income Variance

Positive Selection and Lottery Immigrant

Negative Selection and Lottery Immigrants

ix

26

27

28

55

56

57

58

Chapter 1
Introduction
Section 1: Statistics on Immigration
Although the United States is a nation of immigrants
there has been a long debate in this country on the
desirability of admitting more immigrants. As a reflection of
this, U.S. immigration policy has, since the 1880's, attempted
"to restrict the numbers, race and national origin composition

of the immigrant flow". 1

The first of two arguments which underlie concerns over
additional immigration is that those admitted will compete in
the labor'market with U.S. nationals and drive down employment
and wages. Underlying this argument is the assumption that
immigrants and natives are substitutes in production. The
second argument against allowing additional immigration is
that the characteristics of immigrants are such that they will
not make a positive contribution to U.S. life, but rather will
create social tension and put pressure on public budgets
through their use of publicly provided programs and
facilities. The assumption implicit here is that immigrants do
not possess the skills required for success in the U.S. labor
market or for assimilation into U.S. life.

It is the second argument on immigrant characteristics
which motivates this dissertation. The specific characteristic

that I consider is labor-market quality. Central to the

 

' Borjas, 1990 p4

2
discussion is whether, and how, immigrant labor-market quality
differs across sending countries. I also consider labor-market
quality across classes of admission and over time.

Before proceeding with a discussion of the economic
issues in the area of immigrant labor-market characteristics,
I will put the general issue of immigration in context by
considering the number of immigrants coming to the U.S. and
the trend in immigration during the century. Table 1 presents
the number of legal immigrants coming to the U.S. over the
period 1901 to 1990, along with the annual rate per 1000 of
the U.S. population.

While the numbers admitted dropped in the middle part of
this century, the trend is clearly toward an increase in
immigration. When the numbers currently being admitted are
considered in the context of low growth in the native
population, it is clear that immigrants will make up an
increasing proportion of the population and of the labor
force.

Along with this growth in the numbers coming to the U.S.
there has been a shift in the last fifty years in the source
of immigrants. In the period 1941 to 1950, 17.7 percent of
legal immigrants came from Latin America and the Caribbean,
while 76.6 percent came from Europe and Canada. In the period
1981 to 1990, these:numbers‘were 47.1 percent and 12.5 percent

respectively.2 If it is the case that immigrants from

 

2 Rolph, p24

3

Table 1: The Flow of Immigrants into the United States 1901-
1990

 

 

Period Immigrants IRate' Percentage of
Admitted Population
(000,000) Foreign-bornb

1901-1910 8.8 10.4 -

1911-1920 5.7 5.7 13.2

1921-1930 4.1 3.5 11.6

1931-1940 0.5 0.4 8.8

1941-1950 1.0 0.7 6.9

1951-1960 2.5 1.5 5.4

1961-1970 3.3 1.7 4.7

1971-1980 4.5 2.1 6.7

1981-1990 7.3 3.0 4. 8.6 _‘

 

 

 

Source: Elizabeth S. Rolph,

the 12§Q'§ ggd Is sues {go on the 1229's, Rand 1992

' Annual rate per 1,000 U.S. population

" Includes both ‘undocumented and. documented foreign-born
residents

4
different regions have different characteristics then this
shift will have a large effect on the overall nature of the
immigrant flow.

It should also be pointed out that the vast majority of
immigrants locate in a small number of states. Of all
immigrants legally admitted between the years 1982 and 1990,
74 percent intended to locate in one of six states:
California, Florida, Illinois, New Jersey, New York or Texas.
This means that whatever impact immigrants will have, it will
be concentrated in these areas.

Section 2: The Issues Involved in Immigrant Labor-Market
Quality

The main issue I address in this dissertation is
immigrant labor-market quality. The primary question here is
whether the skill levels of immigrants are greater or less
than those of people born in the United States. The importance
of this question stems from the effect labor-market quality
has on the impact of immigrants on the U.S. If immigrants are
highly skilled and highly educated the potential exists for
the impact of immigration to be positive. The influx of such
a group would mean the provision of a productive body of
people who will contribute to taxes and.not be,a drain.through
their use of public assistance programs. Such people are also
more likely to assimilate into American life, lessening the
potential social strife that can accompany the co-existence of
different cultural groups in a society. Furthermore, given the

shift in the demands of the U.S. labor market toward more

5
highly skilled workers, it is possible that high-skilled
immigrants will have a growing positive effect.

Alternatively, if immigrants are a low-skilled group and
have low education levels the potential exists for immigration
to have a negative impact. Such immigrants will not contribute
much in terms of productivity or taxation and may have a
greater propensity to take advantage of public assistance
programs. Likewise, the potential for greater geographic
concentration, which tends to hinder assimilation, seems
greater among low-skilled immigrants is greater, making the
above mentioned social strife more possible.

Although studies such as Borjas (1987a) and Lalonde and
Topel (1991) find that immigrants do not depress the wages or
employment of native workers, more recent work has produced a
different finding. Borjas, Freeman and Katz (1992) and Topel
(1994) find that immigration has reduced the wages of low-
skilled native workers, presumably because of labor-market
competition between low-skilled immigrants and low-skilled
native workers. This creates an additional concern over the
immigration of the low-skilled. No such labor-market effect
has been found for the high-skilled.so the same concern is not
present for the immigration of the high-skilled.

Rather than immigrants in general being low-skilled or
high- skilled, the potential exists for the skill levels of
immigrants to differ across countries. The economic issue is
whether it is possible to identify country characteristics

which determine the nature of the immigrant flow from a

6
country. The policy issue is then whether it is possible to
target admission policies toward countries that send the high-
skilled immigrants.

Another issue that arises in this context is the
administrative method of granting permanent legal residency to
people. In order to ensure that immigrants have desirable
labor- market qualities some countries have had immigration
policies that deliberately favored the admission of the highly
skilled. Immigration law in the United States currently does
not in general grant permanent residency with reference to
labor-market quality but focuses instead on the humanitarian
goal of family re-unification, leaving a'broad avenue of entry
for low-skilled immigrants.

Section 3: Literature Review

Previous research on the economics of immigration has
looked at both the effect of immigration on the U.S. economy
and at the characteristics of immigrants. As the latter topic
is of relevance here the focus of this review is on papers
which deal with immigrants' characteristics.

One of the more important papers in this area is Chiswick
(1978), which presents a view of immigrants containing the
following elements. When immigrants first arrive in the U.S.
they have less knowledge of the U.S. labor market and less
U.S. specific education and training than native-born workers.
This leads to a wage disadvantage for foreign-born workers
relative to comparable native-born workers. Over time the

foreign-born workers learn about the U.S. labor market and

7

acquire U.S. specific training and education. This produces a
reduction in the difference between the wages of the foreign-
born and the native-born. Were the two groups similar in terms
of innate ability and motivation, wage differentials, adjusted
for the usual characteristics, should disappear over time.
Chiswick, however, hypothesized further that immigrants may be
more motivated and have more innate ability than the average
native-born citizen. If this is true, the wages of immigrants
should not just rise to equal those of the native-born but
should rise above them.

Using data from the 1970 Census, Chiswick shows that
immigrant wages were below those of the native-born at the
time of entry but rose with years in the U.S. and eventually
overtook the wages of the native-born. Such a finding put
immigrants in a favorable light.

This view of immigrants is also seen in a paper by
Carliner (1980). In explaining his observations on the
relative earnings of immigrants he makes reference to the
higher motivation of immigrants and to their greater
preference for money over family ties, leisure and easy work.

A series of papers by Borjas (1985, 1987b, 1991) produced
arguments and results that cast doubt on the Chiswick/Carliner
findings and on their general view that immigrants had
favorable labor-market characteristics. In the 1985 paper
Borjas brings up the possibility that the use of a single
cross-section to draw implications on the earnings growth of

immigrants could suffer from a serious flaw. Although Chiswick

8

concludes that the observed earnings growth of immigrants is
the result of labor-market assimilation, Borjas suggests that
the observation could be a result of a decline in the labor-
market quality of successive immigrant cohorts. The empirical
work in his 1985 paper shows that the use of a single cross-
section did indeed overestimate the true earnings growth of
immigrants and that within cohort earnings growth was much
lower than previously estimated.

This finding implies that there was a decline in the
labor- market quality of immigrants across cohorts. This in
turn casts doubt on the view that immigrants are always
selected from the upper end of a country's ability
distribution. In his 1987b and 1991 papers Borjas explores
further the issue of immigrant selection and shows
theoretically how country-specific characteristics can lead to
the immigration of the low skilled from.some countries and the
high skilled from other countries. Whereas Chiswick and
Carliner see selection as being always "positive" in the sense
that the best choose to immigrate, Borjas shows how selection
could. be "positive" or "negative". Furthermore, he
demonstrates that relative income inequality between the home
and the host countries has a crucial effect on the nature of
the immigrant flow from a particular country.

Borjas' empirical work shows that the quality of
immigrants does differ across countries of origin and that
certain country-specific characteristics can explain the

differences. One of the important implications of this is that

9
a shift in the national- origin mix of immigrants can alter
the quality of the immigrant pool. Borjas maintains that the
shift in immigration into the U.S. from a European to a Latin
American origin has produced the decline in immigrant quality
that has been observed in recent decades.

The issue of the interrelation between immigrant earnings
and country of origin is also considered, by' Jasso and
Rosenzweig (1986). As in Borjas (1987b), Jasso and Rosenzweig
show that differences in the earnings of immigrants across
countries can be explained not just by dummy variables for
different countries but by country-specific characteristics.
They show that the effects of country dummy variables on the
earnings of immigrants practically disappear if such factors
as economic conditions, travel costs and the level of
information available about the U.S. are controlled for.

Lalonde and Topel (1991) look at the decline in the skill
levels of immigrants between ‘the 1970's and 1980's and
conclude, like Borjas, that the shift in the origin of
immigration was almost entirely responsible for the decline.
They go on to show, however, that these immigrants assimilate
quickly and eventually earn as much as comparable natives.
When combined with their observation that immigrants do not
have negative effects on the labor-market outcomes of natives,
they conclude that the concerns generated by observations of
declining immigrant skill levels are overstated.

Characteristics of immigrants other than earnings have

also been studied. Borjas and Trejo (1991 and 1993) examine

10

the propensity of immigrants to use the welfare system and how
this propensity differs across countries of origin. As with
the work on earnings, their results show that country
characteristics can explain much of the variation in welfare
use. In particular, recent increases in welfare use by
immigrants can be explained by the changing national-origin
mix of immigrants.

Bartel and Koch (1991) have studied the propensity of
immigrants to migrate within the U.S. This is important
because it determines the extent to which any effects of
immigrants are dispersed or localized. Their main finding is
that there is little evidence that immigrants become more
dispersed over time. The only national groups to show some
dispersion were Indians, Japanese and Koreans. Younger and
more educated immigrants were also found to be more likely to
move.

Section 4: Dissertation Outline

The purpose of this dissertation is to add to the
insights already produced in the area of immigrant labor-
market characteristics. As with other work, I explore
variation in the labor-market quality of immigrants across
countries. However, an important and useful departure is
contained in the work to be presented here. In both the
theorethical and empirical analyses I use, in an unprecedented
way, the fact that immigrants are admitted under different
legal classes of admission. Most work in this area treats all

immigrants as a single group, regardless of the legal basis on

11
which they were admitted. By separating legal immigrants into
those admitted on the basis of their skills, family contacts,
refugee status and lottery winning, important insights will be
gained into the nature of the immigrant flows from different
countries.

The structure of the dissertation is a follows. In
Chapter 2 I develop a model which shows how the labor-market
quality of skill-based immigrants relative to the family-based
immigrants differs across countries. This quality difference
between the two groups of immigrants reflects the general
nature of the immigrant flow from a country, which is of
immediate interest. A small difference between the skill-based
and the family-based immigrants from a particular country
means that the family-based must be reasonably high skilled,
so such a country, in general, sends high quality immigrants.
A big difference between the skill- based and family-based
means that the family-based must be low skilled and in this
case the country sends low skilled immigrants, in general. I
estimate these differences for a range of countries using data
on immigrants from the years 1978 to 1980. These estimates
have immediate interest since they allow us to identify which
countries send which types of immigrants. The analysis does go
a step further, however, as I to explain variation in the wage
premium of the skill-based immigrants relative to the family-
based across countries, using country-specific characteristics
consistent with the model. In this way it becomes possible to

predict which types of countries will send which types of

12
immigrants.

The purpose of Chapter 2 is similar to that of Borjas
(1987b, 1991) and.JassoIand.Rosenzweig (1986) but.the approach
is very different, exploiting the existence of the different
classes of admission. The results show how a legislative
shift toward skill requirements for more immigrants would have
a different effect on the nature of the immigrants flows
across countries. For some countries, skill requirements would
have little effect since those who choose to immigrate are
quite skilled anyway. For other countries, such requirements
would increase the quality of immigrants by excluding the low
skilled who wish to immigrate.

In Chapter 3 I turn my attention to those immigrants who
gained legal admission to the United States through the
lottery programs of the late 1980's and 1990. Again, an effort
is made to use the relative earnings of two groups of
immigrants to gain some insight into the nature of the
immigrant flow from a particular country. I develop a model to
show how the migration decision of the lottery immigrants
differs from that of the family-based immigrants. Home-country
economic conditions are then incorporated into the model and
jpredictions about the relative skill levels of the lottery and
family-based immigrants are derived. I estimate the skill
differences between the two groups for some countries and
consider the results in the context of the predictions of the
model. In this chapter I also consider the location decisions

of a group of lottery immigrants to see how such decisions

13
differ from those of the family immigrants. Location of
immigrants is an important issue when the pressures faced by
areas of high immigrant concentrations are considered. Hence,
additional insights into the location decision are useful.'

The lottery immigrants are a potentially valuable
resource in the examination of the immigration decision.
Immigrants who are admitted because of family ties or because
of their skills do not necessarily give researchers a random
sample of those who want to immigrate. The desire of these
individuals to immigrate had to be supported by the legal
ability to do so, hence the possible non-random nature of the
immigrant pool. The lottery immigrants are, however, a random
sample of those who wish to immigrate and so studying them.can
produce a truer picture of those who find it optimal to
immigrate.

In Chapter 4, I compare the family-based and skill-based
immigrants again, this time using data on immigrants admitted
in the years 1988 to 1990. In Chapter 2, support is found for
the ideas of the model, most notably that the level of
inequality in the sending country has an important effect on
the quality of the immigrants coming from that country. The
empirical analysis is done again to see if this support is
maintained using data from the later period. The idea that
inequality in the home country would influence the nature of
the immigrant flow was first proposed by Borjas (1987b). The
model used in Chapter 2 builds on the Borjas' ideas by

incorporating class of admission, so the empirical work of

14
Chapters 2 and 4 can be seen as further testing those ideas.
The final part of Chapter 4 contains an examination of the
relative labor-market quality of immigrants admitted in the
years 1979 to 1980 and 1989 to 1990. Borjas has suggested that
rising income inequality in the United States may lead to an
increase in the quality of immigrants into the U.S. Given the
rise in inequality over the 1980's and data on immigrants
admitted at the beginning and end of the period, I am able to
test this hypothesis. As the skill level of immigrants
declined from the 1960's to 1980, it is important to see if
this trend continued through the 1980's. It is also possible
to see if any change in immigrant skill levels is the result
of a change in the country of origin mix of immigrants, as was

the case between the 1960's and 1980.

Chapter 2
The Effect of Class of Admission on Immigrant Selection
across Countries
Section 1: Introduction

A common finding in the literature on the characteristics
of immigrants is that immigrants are not merely a random
sample from their countries of origin. They are the products
of a selection process which leads the immigrant group from a
country to differ from the population of that country in its
labor-market characteristics. Another common finding is that
certain economic characteristics of each country of origin
influence the nature of the group who choose to migrate.
Hence, immigrants from different countries will differ in
their labor-market characteristics.

This reported difference in the labor-market quality of
immigrants across countries has important implications fer
immigration policy. Current U.S. policy pays very little
attention .to the labor-market characteristics of immigrants as
a ‘criterion in determining eligibility for permanent
residency. As such, the actual labor-market quality of the
immigrant pool depends to a great extent on the outcomes of
the immigrant selection process across countries and the
numbers coming from various countries. An understanding of the
selection process and how national characteristics influence
the outcome of the process can help to identify which sorts of
immigrants will come from which sorts of countries.

Borjas (1987b) and Jasso and Rosenzweig (1986) have

15

16
addressed these issues, examining the labor-market
characteristics of immigrants from various countries and the
relationship between the labor-market quality of immigrants
from a country and the characteristics of that country.

In this chapter, I add to the examination. of the
difference in the labor-market quality of immigrants across
countries. I do so in a way that exploits a useful element in
the structure of U.S. immigration policy. A small proportion
of permanent residency visas are granted each year to
immigrants who are classified as highly skilled. Although the
recipients of these visas are the products of the immigration
selection process of their various countries, they are
constrained by U.S. law to be in the upper region of the U.S.
earnings distribution. As such, this group will be similar
across countries. A much larger proportion of permanent
residency visas are awarded on the basis of family ties to
U.S. citizens or permanent residents. As labor market criteria
are not used for this group in allowing immigration, the
position of this group in the U.S. earnings distribution is
unconstrained by U.S. law. The position of the group in the
earnings distribution depends on the nature of the immigrant
selection process and will differ across countries as the
parameters underlying the process differ.

The existence of these two groups, skill- and family-
based immigrants, allows me to examine the immigrant selection
process across countries. A small difference in labor-market

quality between the skill- and family-based immigrants from a

17

country indicates that the selection process of that country
tends to lead to the migration of high quality people.
Alternatively put, the group that is unconstrained by law in
its labor-market characteristics is similar to the group that
is constrained to be in the upper end of the earnings
distribution. A large difference between the two indicates
that the selection process tends to result in the migration of
the low-skilled. Here the unconstrained group is quite
different from the group who are admitted because of their
high skill levels.

The approach I take in this chapter will be as follows.
I first outline a model showing how the divergence between the
skill-based and family-based groups in labor-market
characteristics may differ across countries. In testing the
model, I show that there is a divergence across countries in
the difference between the labor-market quality of the skill-
based and family-based immigrants. I then attempt to explain
the pattern across countries using country-specific variables
consistent with the model presented. Finally, I present a
summary and some conclusions.
Section 2: The Model

In the years which I examine in the empirical part of

this chapter, 1978, 1979 and 1980, about 500,000 immigrants
gained permanent residency annually. The 500,000 were admitted
under a number of "classes of admission". Of these, some
290,000 were subject to numerical restrictions. The group

which was not subject to numerical restrictions was made up of

18

refugees.and spouses, parents and minor children of adultLU.S.
citizens. The numerically restricted group was broken up into
seven preference categories and included unmarried adult
children of U.S. citizens, spouses and unmarried children of
permanent resident. aliens, etc. .Another’ group 'which ‘was
included in the numerically restricted category was the
professional and highly skilled group. Ten percent of the
290,000 'visas *were reserved for’ such individuals, their
spouses and their children. It is to this group that I am
referring when I write about the skill-based immigrants. The
family-based immigrants are those admitted because of family
ties. Refugees are not dealt with at any length in this
chapter.

I now present a model which deals with the difference
between the labor-market quality of skill-based and family-
based immigrants. Consider first the skill-based group. For
now, I assume that the mean earnings of this group are equal
across countries. Given that individuals in this group are
admitted because of their high skill levels, the mean earnings
of this group will be greater than the mean earnings of the
U.S. population. Denoting the mean earnings of the immigrants
by E(y;) and the mean earnings of the U.S. population by u,,
let

30'.) = u. + 6 (2.1)
where 6 is some positive number.

The focus of this chapter is on the difference between

the labor-market quality of the skill-based and the family

19

based immigrants across countries. The extent of the
difference between the two reflects the extent to which the
immigrant selection process of a country tends to lead to the
immigration.of high- or low-skilled.people when the outcome of
that process is unconstrained by U.S. law. Denoting the mean
earnings of the family-based group by E(y}), the focus is on
the variation in the following quantity across countries:

Ely.) - Em)- (2.2)

As E(y;) is assumed to be constant across countries, the
variation in (2) will come about through variation in E(y}).
Even if this assumption of a fixed E(ys) is relaxed, it
appears that the variation in E(yf) would be greater than that
in E(y;) due to the fact that E(y}) can vary over the entire
range of the earnings distribution. E(y;) can only vary over
the smaller region at the upper end of the earnings
distribution. Given this, the variation in (2) across
countries will be driven by variation in E(y,) , so that a
model explaining variation in E(y}) is required.

The model I use is that of Borjas (1987b). Due to
differences in the data sets used in that paper and this
chapter, the empirical interpretations of the models differ
slightly. I will point out the relevant differences at the
appropriate place. The Borjas model specifies wages to be
distributed in the country of origin (country 0) as follows:

In w0 = “o + 60 (2.3)
where ‘o is normally distributed N(0,aoz) . Borjas takes "0 to

be earnings due to observed socioeconomic characteristics and

20

co to be earnings due to unobserved characteristics. As the
data set used in this chapter only has information on age, sex
and marital status by way of "observed socioeconomic
characteristics", no is taken to be average earnings due to
those characteristics. The distribution of earnings about no
is then determined by the distribution of so, which reflects
the distribution of other characteristics in the population
such as education, health, innate ability, etc.

Wages in the country of destination (which Borjas labels
the U.S. or country 1), were all individuals in country 0 to
migrate, are distributed as follows:

In W1==ih + 61 (2.4)
where 61 is normally distributed N(0,of). The definitions of
n1 and 51 are similar to those of no and so. The correlation
coefficient between co and e, is p. Borjas also takes n1 to be
the average earnings of U.S. natives by assuming that
differences in skills across countries have been standardized.
This assumption has strong implications for the
interpretations of the selection outcomes which Borjas goes on
to derive. While I keep the assumption in this use of the
Borjas model, I will point out, where appropriate, its
implications.

Given the income distributions in the two countries, the
migration decision is determined by the sign of the index
function

I = ln(w1/(wo + C)) (2.5)

=(“1'uo'x)-(€1'€o)

21
where C represents costs of migration and 1t gives a "time
equivalent" measure of these costs. The emigration rate which
arises is given by

P = Pr[v > -(n1- no -1r)] (2.6)

= l - ((2)
where v = 61 - co: 2 = -(n1 - no -1r)/ov: and 4 is the standard
normal distribution function.

This model of the immigration decision can next be used
to predict selection outcomes of the immigration process. In
order to obtain these predictions the following conditional
means are required:

E(onI>0) = no + (aoo1/av) (p - ao/01)L (2.7)

E(w1|I>0) = Em) = u, + (com/0,) (01/00 - p)L (2-8)
where L ¢(z)/P and (P is the density of the standard normal.

It can readily be seen from equations (2.7) and (2.8)
that the position of immigrantsin their home country earnings
distribution is determined by whether (Jo/o1 ’< p. Similarly,
their position in the host country earnings distribution is
determined by whether oo/oo ’< p, again, assuming n1 is the mean
earnings of both immigrants and the U.S. population. Defining
Qo to be the earnings differential between the average person
and the average emigrant in country 0, Q1 to be the earnings
differential between the average immigrant and the average
native in country 1, and k = o1/ao, three selection outcomes
can be identified.

The first selection outcome is "positive selection",

whereby Qo > 0 and Q1 > 0. This is where immigrants come from

22
the upper end of the home country distribution and are in the
upper end of the host country distribution. The necessary and
sufficient conditions for this to arise are derived from
equations (2.7) and (2.8):

p > min [l/k, k] and k > 1. (2.9)
The intuition underlying this result is that as long as the
ranking of skills is sufficiently similar across countries
(i.e. p sufficiently large), those in the upper end of the
home country distribution will migrate to a country with a
greater income spread so as to avoid being "taxed" because of
their position in the home country distribution. Those in the
lower end of the home country distribution will not migrate to
the country with the greater earnings spread because in their
home country they are being "insured" against lower earnings.
An example of this sort of country is a Western European
social democracy with high tax rates and generous social
benefits.

The importance of the n1 assumption can be seen here.
Rather than the average skill level being the same in the home
and host countries, suppose the average skill level in the
home country is higher. A random draw from this population
would produce an immigrant pool that is more skilled on
average than the U.S. population. It would be a mistake,
however, to see this immigrant group as being the product of
the selection process described above.

The second selection outcome which can be identified is

"negative selection", whereby Qo«< 0 and Q1‘< 0: those in the

23
lower part of the home country distribution migrate and are
located in the lower part of the host country distribution.
The necessary and sufficient conditions for this toiarise are:

p > min [1/k, k] and k < l. (2.10)
In this case the low earners can insure themselves against
lower earnings by migrating to the country with the narrower
earnings spread. The high earners stay where earnings are more
spread and avoid the "tax" which would apply were they to
migrate. A country with very low marginal tax rates and little
provision for those in the lower end of the earnings
distribution is an example of a negative selection country.

A violation of the n1 assumption, again, has implications
for the interpretation. of ‘the source of this selection
outcome. If a country's population is, on average, less
skilled than the U.S. population, its immigrants may be less
skilled than the U.S. population. This may be a random
outcome, however, and not the result of a selection process
driven by differences in inequality.

A third selection outcome is labeled "refugee sorting"
and arises when Qo < 0 and Q1 > 0. In this case, although the
immigrants come from the lower part of the home country
distribution, they are in the upper part of the host country
distribution. The necessary and sufficient condition for this
to arise is

p < min [l/k, k]. (2.11)
The intuition here is that skills valued little in one country

may be valued highly in another, thus attracting the migration

24
of owners of these skills. An example is a communist country
which rewards bureaucrats well and entrepreneurs poorly. The
remarks made about the notassumption apply here also.

With an expression for E(yf) and different potential
values outlined, I return to the quantity E(yo) - E(y}). This
quantity is equal to:

in. + 61 - in, + (com/o.) (we, - m (2.12)
= 6 - (00°1/avH01/ao - p) (2.13)
In the case of positive selection, 01 > co and 01/00 > p.
Therefore, the second quantity in expression (2.13) is
positive, so the total quantity is less than 6. In the case of
negative selection, this same second quantity is negative, so
the total quantity is greater than 6. Finally, in the case of
refugee sorting, the total quantity is less than 6. The clear
conclusion from this is that the difference between the skill-
based and family-based immigrants from.positive selection and
refugee sorting countries will be less than the difference for
the same groups of immigrants from negative selection
countries. In addition, it is clear from the model that.a.high
degree of earnings inequality leads a country to exhibit
negative selection and hence a large difference between its
skill-based and family-based immigrants. Similarly, low
earnings inequality in other countries leads to positive
selection and a smaller difference between the skill-based and
the family-based.
In order to see this, consider Figures 1-3. Figure 1

shows two earnings distributions, one for a foreign country

25

and one for the U.S. In the U.S. earnings are less equally
distributed and, following the Borjas model, this foreign
country will exhibit positive selection, assuming a
sufficiently high correlation between the return to various
skills in the two countries. The positive selection comes
about because individuals between points A and B in the
foreign earnings distribution wish to migrate, and arrive
between points A and D in the U.S. distribution.

Notice that for the foreign country in Figure 1, above
average quality immigrants will come to the'U.S. regardless of
whether labor-market characteristics are a criterion for
immigration. That being the case, family-based immigrants from
a positive selection country will have above average earnings
in the U.S., compared to the native population of the U.S. In
the diagram the mean earnings of this group is represented by
E(yo). This assumes that these immigrants are responding to
economic incentives when migrating and are not migrating for
strictly family reasons.

The skill-based immigrants will be located between points
C and D in the U.S. earnings distribution, and their mean
earnings are represented by E(yo) . The wage premium for the
skill-based relative to the family-based immigrants for this
positive selection country is then the distance indicated in
the figure.

Figure 2 is the negative selection case, with earnings
more widely spread in the foreign country than in the U.S.

Those who immigrate from the foreign country, in the absence

Cm u :38.— mnsna
m. u mac—in:
Minn n 58: ”sac—ca arm: can»:

m3; u 33: 58:8 35:7. acme.—

Cm

Ems—.6 —. $2.353 mango:

 

- -
act .56

Mona—fin

98

27

mace—Sm

m3

III-qud

segue—em 95ch Z .N 253%

ace
—

38.. E53 2:85 :8... u 15m
32:. :2... 2:85 :8... u 3cm
cums—o..— u h

_ 8.5m 33:: u m3

28

mus—Ecum—

QE 1
—

 

m3

O

U

IllllllllllllllldlllllIIIIIIIIIIIIIIIII I

was

956m cow—:3— .m 6.53..—

 

<

m

9.8 See .58.: :8... u 1 its
8...... a... 2:85 :8... u at...
53.8% n h

8.5m tog—.3 u m:

29
of labor-market immigration criteria, are those in the lower
end of the earnings distributions. They are located between
points A and B in the home country distribution and arrive
between points E and B in the U.S. distribution. Their mean
earnings are as shown and are below mean earnings in the U.S.
population.

From the theory developed so far it is likely that the
numbers of skill-based immigrants from negative selection
countries will be small, given that they can earn more in
their'home country. However, assuming that some will choose to
immigrate, their mean earnings should be the same or at least
very similar to the mean earnings of the skill-based
immigrants of positive selection countries. Hence, they too
will be located between points C and D in the U.S.
distribution. This being the case, the wage premium for the
skill-based relative to the family-based immigrants for
negative selection countries should be greater than the
corresponding premium for the skill-based immigrants from
positive selection countries.

The skill-based versus family-based wage premium in the
case of refugee sorting is shown in Figure 3. Those from the
lower end of their home country earnings distribution migrate
and arrive in the upper end of the host country distribution.
This is represented in the diagram as people between points A
and B in their own country's earnings distribution moving to
the area between points E and G in the U.S. distribution. In

this situation the family-based immigrants have mean earnings

30
above the host country's mean. The skill-based premium is thus
similar to the positive selection case and smaller than the
negative selection case.

In summary, for positive selection and refugee sorting
countries the influence of class of admission on the labor-
market quality of immigrants will be less than the influence
for negative selection countries. Regardless of the
immigration criterion faced, immigrants from positive
selection and refugee sorting countries will arrive in the
upper end of the U.S. distribution. For negative selection
countries, however, immigrants admitted without reference to
labor-market quality, i.e. , family-based immigrants, will come
from the lower end of both their home and the U.S. earnings
distributions. Skill-based immigrants from.negative selection
countries will be in the upper end of the U.S. earnings
distribution.

Section 3: Outline of the Empirical Approach

I do the empirical analysis in two stages. The first
stage is to determine whether or not across country
differences between the family-based and skill-based
immigrants exist. I do this by estimating wage equations for
a range of countries and including in the regression a dummy
variable equal to one if the individual is a skill-based
immigrant and equal to zero if they are a family-based
immigrant. The coefficient of this dummy variable is then
interpreted as measuring the wage premium of the skill-based

relative to the family-based immigrants. If the model is

31

correct this coefficient should vary across countries in a
predictable way. The second stage of the testing process will
be to determine whether or not any observed variation in the
measure of the wage premium of the skill-based is consistent
with.the theory'presented.earlier; I do this by regressing the
measure of the skill-based premium on some country-specific
variables to see if these variables do indeed explain the
observed pattern in this premium.

In the first stage I use data from the Immigration and
Naturalization Service. Information on all aliens legally
admitted to the United States for permanent residence in the
fiscal years 1978 to 1980 is available on tape and contains
variables such as occupation, class of admission and
nationality. Other information such as age, marital status and
sex is also contained, as is information on whether the alien
is a new arrival in the U.S. or is adjusting to permanent
residence from a non-immigrant status.

For aliens admitted on the basis of job skills, the
occupation reported is that which they will perform in the
United States. The occupation reported for all other aliens is
that which they were doing in their most recent country of
residence. Those admitted under the skill class of admission
would probably practice their particular skill in their own
country, so no major inconsistency should arise due to this
way of reporting occupation. Use of these data does, however,
require the assumption that occupational attainment is highly

correlated across countries of origin and the U.S.

32

I selected males aged 22 to 64 from 41 countries for
analysis. Using only males adds credence to treating family-
based immigrants as if they are responding to economic as
opposed to purely family considerations. The same
consideration motivated the omission of the very old. The very
young are omitted because the occupations they report will
less accurately reflect their true potential labor-market
quality. The 41 countries used are those for which country
specific data are available.

In order to perform analysis, the occupational variable
had to be converted into a wage variable. I did this by
imputing earnings for each 3-digit occupation code, using
earnings data from the 1980 Census. I matched the median
earnings of full-time full-year males in each occupation with
the occupation for each individual in the sample.

The equation I estimated for each country is

LWAGE = bo + b1 AGE + be AGESQ + 133 MARRIED + bo D78 4»

b5 D79 + b6 D80 + b, ADJUSTER + bo YEARS IN U.S.

+ b9 SKILL + b", REFUGEE + e
where EWAGE is the logarithmtof'the earnings of the occupation
reported by the immigrant: AGESQ is age squared: MARRIED is a
dummy variable equal to 1 if the individual is married and 0
otherwise: D78, D79 and D80 are dummy variables for each year
of admission (the omitted year is 1977): ADJUSTER is a dummy
variable equal to 1 if the immigrant was already in the U.S.
as a non-resident alien before adjusting to permanent

residence and equal to 0 for new arrivals: YEARS IN U.S. is

33

the number of years spent in the U.S. before gaining permanent
residence (it equals 0 for new arrivals): SKILL is a dummy
variable equal to 1 if the immigrant gets permanent residence
on a skill-based class of admission and 0 otherwise: REFUGEE
is a dummy variable equal to 1 if the immigrant gets permanent
residence on a refugee. based class of admission and 0
otherwise.

The coefficient of greatest interest is that of SKILL.
This represents the wage premium of the skill-based relative
to the family-based immigrants for each country. The larger
the coefficient for a country, the larger is the difference
between the skill-based and the family-based immigrants from
that country. The theoretical discussion suggests that this
should increase with the variance of the home country earnings
distribution.

I use AGE, AGESQ and MARRIED as standard socio-economic
controls. The year dummy variables, D78, D79 and 080, are
added to capture any possible changes in immigrant cohort
quality over these years. The variables ADJUSTER and YEARS IN
U.S. are intended to capture any possible difference between
the position of new arrivals and those who gain permanent
residency after having been in the‘U.S. for a number of years.
Finally, I use the variable REFUGEE to control for the
experience of refugees who presumably have immigrated for non-
economic reasons.

Section 4: Empirical Results, Part 1

Table 2 presents results from.a pooled regression of all

34

41 countries with 40 dummy variables included to represent
countries. This equation gives an overview of the value of
these coefficient estimates. The table shows that the signs on
most coefficients are as expected. The SKILL coefficient shows
that on average across countries those admitted on a skilled
class of admission are in occupations which pay 32 percent
more than the occupations performed by those admitted on a
family class of admission. Of passing interest is the negative
coefficient on REFUGEE, the refugee:dummy variable. That it is
significantly different from the family group indicates that
the migration motive for this group is truly different from
that of other migrants and that they are not merely economic
refugees in disguise.

The negative coefficient on the variable YEARS IN U.S.
for those adjusting to permanent residency may appear at odds
with other results in this area ‘which ShOW’ a positive
relationship between years in the U.S. and. immigrant earnings.
This apparent conflict is explained by the use of individual
wage rates in other studies but occupational wage rates in
this study. Individuals may earn more as they spend longer in
the U.S. but the occupation they report may place them in the
lower end of the earnings distribution. This problem may be
particularly relevant for immigrants whose work opportunities
were limited by their visa status prior to gaining permanent
residency. They may have had to work illegally or, in the case
of students, restrict their work time to a small number of

hours. In both cases, occupational attainment will be low.

35

Table 2: Regression which Estimates the Coefficients of the
Variables Shown Using Data from all 41 Countries

 

 

 

Dependent Variable: log wage
Parameter Parameter Standard Error
Estimate
INTERCEPT 9.364 0.0103
AGE 0.014 0.0005
AGESQ -0.0001 0.000006
MARRIED 0.019 0.0017
D78 -0.013 0.0023
D79 -0.018 0.0024
D80 -0.019 0.0025
ADJUSTING T0 0.012 0.0022
PERMANENT
RESISENCY
YEARS IN U.S. -0.001 0.0004
SKILL 0.323 0.0036
REFUGEE -0.106 0.0048
Adjusted R2 = .1858 N=249,102

 

 

Note: also included in the regression but not reported here
were 40 country dummy variables.

36

Table 3 contains a limited presentation of the results
from the 41 country regressions. It is apparent from the
values of the SKILL coefficient that considerable variation
does exist across countries. This finding supports the idea
that the family-based immigrants from some countries are not
too unlike the skill-based immigrants from the same country,
while for other countries there is a considerable difference.
The low European values such as the Netherlands (.1849) and
Sweden (.1662) and the high South and Central American values
such. as Mexico (.6827) and. Haiti (.9090) also give an
indication that something systematic may be giving rise to the
different outcomes across countries.
Section 5: Empirical Results, Part 2

The regressions presented in Table 4 have as the
dependent variable the coefficient of SKIII.estimated earlier
for each country and reported in Table 3. The first two
independent variables reported in Table 4 are intended to
reflect economic conditions in the countries of origin.
INEQUALITY is the ratio of the income of the top 10 percent of
households to the income of the bottom 20 percent of
households. It captures thelextent.of income inequality in the
countries of origin. Clearly, given the crucial role of income
inequality in the theoretical model, the sign and significance
of the estimated coefficients of this variable are of
paramount interest.

The second variable which I use to capture the economic

conditions in the country of origin is LOG GNP. This is the

37

Estimating the Wage Premium for the Skill-Based

Table 3 :
Immigrants

i T

Dependent Variable: Log wage

 

 

 

 

Country Skill Standard # Skill- N
Coefficient Error Based
‘Austria .3451 0.0658 28 293
Czechoslovakia .2410 0.0889 21 621
Denmark .2793 0.0592 32 329
France .3013 0.0395 94 1433
Germany .2178 0.0262 163 1813
Greece .4440 0.0481 60 5048
Hungary .1667 0.0963 11 529
.Ireland .4627 0.0469 65 964
Italy .4842 0.0375 81 4454
Netherlands .1849 0.0305 112 1004
Norway .1973 0.0723 24 336
Poland .3669 0.0453 47 3022
Portugal .4643 0.0648 26 6905
Romania .3287 0.0848 18 1272
Spain .5479 0.0616 32 1266
Sweden .1662 0.0388 71 458
Switzerland .2281 0.0476 58 590
U.K. .2876 0.0085 1396 16970
U.S.S.R. .2363 0.0863 13 805
Yugoslavia .3473 0.0699 25 1495
China .3540 0.0153 948 13984
Egypt .1408 0.0271 217 2232
India .2322 0.0092 2470 15318
Israel .2588 0.0218 210 3323
Japan .3638 0.0286 201 2596
Korea .1273 0.0169 346 13737
Philippines .3981 0.0178 544 17977
Argentina .6354 0.0395 86 2458

 

 

38

Table 3 (cont'd):

 

i

Dependent Variable: Log wage

 

Country Skill Standard # Skill- N
Coefficient Error Based

Canada .4639 0.0105 1370 12809
Colombia .4265 0.0436 42 6241
Cuba -0.0400 0.1489 3 12861
Dominican Rep. .7028 0.0884 10 13499
Ecuador .6036 0.0864 10 3817
Guatemala .7765 0.1136 6 2205
Haiti .9090 0.0806 14 5246
Jamaica .4367 0.0566 26 11712
Mexico .6827 0.0334 62 48691
Panama .6148 0.1438 5 1397
Trinidad and .6447 0.0722 12 3511

 

 

Tobago

39
logarithm of 1980 per-capita GNP'in dollars and is intended to
represent the mean income in the countries of origin.

The next three variables are constructed to capture the
political conditions in the countries of origin. POLITICALLY
COMPETITIVE is a dummy variable equal to 1 if the country had
a competitive party system during the entire period 1950 to
1973, and 0 otherwise. RECENT LOSS OF FREEDOM is a dummy
variable equal to 1 if the country had a competitive party
system at the beginning of the 1950 to 1973 period but did not
have one at the end, and 0 otherwise. NUMBER.OF.ASSASSINATIONS
is equal to the number of politically motivated murders or
attempted murders of high government officials or'politicians
in the period 1950 to 1973.

I use these political variables to capture empirically
the theoretical conditions which give rise to refugee sorting.
Countries without competitive party systems and with political
unrest are more likely to be countries which reward skills
differently than does the U.S. Returning to the example used
earlier, one can think of a communist state in which a
bureaucrat is well compensated while an entrepreneur is not.

The next variable used in the regressions presented in
Table 4 is ENGLISH. It is measured as the fraction of the
1975-1980 immigrant cohort from each country that speaks
English well or very well. The ENGLISH variable is included
because of the findings in other studies that it plays an
important role in determining labor-market outcomes for

immigrant groups [McManus et al. (1983) and Borjas (1987b)].

40

The next variable is DISTANCE, measured as the number of
air-miles between the country's capital and the nearest U.S.
gateway of New York, Miami or Los Angeles. I include this
variable to reflect costs of migration and because it.has been
found to have an important effect on migration decisions and
outcomes (Schwartz 1973). This variable is meant to capture
both the pecuniary and the psychic costs of immigration. The
pecuniary costs are picked up through the more expensive
travel from more distant locations. The psychic costs arise
from the greater difficulty of returning home for visits, the
lower likelihood of visits from friends and family and the
greater likelihood of cultural differences. The DISTANCE
variable can also be viewed as a proxy for information
available to potential immigrants, further distances being
associated with less information.‘

The final variable I include is HIGHER EDUCATION and is
equal to the percentage of the population aged 20 to 24 that
is enrolled in higher education. This variable is included to
deal with the issues associated with the violation of the
assumption of standardized skills across countries. As I
pointed out when discussing the three selection outcomes, it

could be that a small difference between the family-based and

skill-based immigrants is simply the result of the family

 

‘ Most of the data used in this section of the paper were
provided.by George Borjas. A.description.of his sources can be
found in Appendix 1. The values of HIGHER EDUCATION were taken

from the W of 1982-

41

immigrants being drawn randomly from a highly skilled
population as opposed to the systematic selection process
described in the model. This variable will measure the skill
level of the group most likely to migrate and so will control
for the the effect I describe, to a limited extent. This
variable does not, however, control for the average level of
skills in the population of the home country. As such, I can
not claim to have fully accounted for differences in skill
levels across countries and the results should be considered
in this light.

The regressions presented in Table 4 are weighted least
squares regressions, because the dependent variable is a
coefficient estimate from an earlier set of regressions. The
variance of the dependent variable will differ across
observations because these are coefficient estimates. In order
to correct for this I weighted the observations by the inverse
of the standard errors of the SKILL coefficient estimates from
the earnings equations.2

As I noted earlier, the sign and significance of the
variable INEQUALITY are of great importance if the results are
to be interpreted as giving support to the Borjas model from
this new perspective that I have developed. In all four models
INEQUALITY is significant, and. the positive sign is as
predicted. According to the model, negative selection is a

result.of high income inequality, which should in turn lead to

 

2 I would like to acknowledge the advice provided to me
on this point by Jeff Wooldridge.

42

a big difference between the family-based and skill-based
immigrants from such a country. Similarly, low income
inequality is associated with positive selection and, by
extension, is likely to lead to a smaller difference between
the skill-based and family-based immigrants. Hence, the
relationship between INEQUALITY and the measure of the skill-
based premium should be positive, again, assuming n1 is the
mean earnings of both immigrants and the U.S. population.

The coefficient on the variable LOG GNP is positive in
each of the three models and is significant in two of them.
This means that a higher mean income in a country tended to
lead to a greater difference between the family-based and
skill-based immigrants. Such a result may strike some as being
counter-intuitive if the assumption held is that higher mean
incomes are associated with positive selection and hence a
smaller difference between family-based and skill-based
immigrants. The theoretical model does, however, allow for
such a result. A higher mean income reduces the average
quality of immigrants from a negative selection country by
removing from the group the most skilled. As such, the family-
based mean income is dragged further away from the mean. (A
further explanation of this point can be found in Appendix 2).

As can be seen from the adjusted R2 of Model 1, the two
economic variables alone explain almost 30 percent of the
variation in the wage premium measure. I take this to be
reasonably impressive and quite satisfying, especially when

combined with the results for the INEQUALITY variable.

43

Table 4: Using Country-Specific Characteristics to Explain the
Variation in the Coefficient of SKILL

i

 

Dependent Variable: Coefficient of SKILL

 

Variable Model 1 Model 2 Model 3

VINTERCEPT -0.l7l7 -0.2474 0.3375

(0.1844) (0.2277) (0.2268)

INEQUALITY 0.0233 0.0222 0.0129

(0.0056) (0.0057) (0.0051)

LOG GNP 0.0439 0.0572 0.0272

(0.0199) (0.0247) (0.0222)

POLITICALLY - -0.0649 -0.0089

COMPETITIVE (0.0494) (0.0576)

RECENT LOSS OF - -0.0036 0.0545

FREEDOM (0.0712) (0.0661)

NUMBER OF - 0.0031 0.0112

ASSASSINATIONS (.0059) (.0051)

ENGLISH - - 0.2196

(0.1153)

DISTANCE - - -0.0392

(0.0105)

HIGHER - - -

EDUCATION
Adjusted R2

.2792

.2619

.5167

Model4
0.1595
(0.2911)

0.0146
(0.0057)

0.0456
(0.0322)

-0.0373
(0.0706)

0.0212
(0.0857)

0.0152
(0.0063)

-O.1621
(0.1358)

-0.0368
(0.0121)

-0.0015
(0.0035)

.5081

from Table 3 n

1

(Standard errors are in parentheses: N for models 1-3 = 41, N
for model 4 = 37.)

44

I add the political variables in Model 2. The only one
which is significant in any of the models is NUMBER OF
ASSASSINATIONS. A possible interpretation of the positive
coefficient is as follows: a country may experience political
unrest if income inequality is large in that country: but
income inequality is precisely the factor which leads to
negative selection and a large wage premium for skill-based
immigrants relative to family-based. In this way, the positive
coefficient on NUMBER OF ASSASSINATIONS makes sense. I would
suggest that the negative coefficient on POLITICALLY
COMPETITIVE is a result of the correlation of lower income
inequality and politically competitive systems in Western
Europe.

The addition of ENGLISH and DISTANCE in Model 3 increases
the explanatory power quite dramatically, bringing the
adjusted R2 to .52. That ENGLISH is significant, albeit at the
90 percent level, is consistent with work referred to earlier.
Its negative sign makes sense when one considers the different
impact English language skills have across skill groups. A
country whose immigrants have, on average, low English-
language skills is more likely to have a large group of low-
skilled immigrants among the family-based immigrant group.
This is because the lack of good English skills will be less
of a penalty for low-skilled than for high-skilled workers.
Such a country will therefore exhibit a big difference between
the skill-based and family-based immigrants. This explains the
negative relationship between English language skills of the

45
immigrant population and the wage premium for the skill-based
immigrants.

That the coefficient on the DISTANCE variable is
significant and negatively signed is also consistent with
other work (Schwartz 1973). The negative sign indicates that
the further away a country is, the smaller will be the wage
premium for the skill-based relative to the family-based
immigrants. In other'words, the further away is the country in
question, the higher is the quality of the family-based
immigrants. DISTANCE is being used to get a handle on costs,
so I will consider first the pecuniary cost argument. The
further away the country, the higher is the cost of transport
to the U.S. As such, the greater will be the resource
constraint for potential immigrants from further away places.
Only those who can overcome this constraint can immigrate from
far-away places, i.e., the high earners or the high- skilled.
From the psychic cost perspective, it can be argued that the
greater the distance, the greater are the psychic costs. If it
is the case, though, that psychic costs are lower for high-
skilled people, these will be less of a deterrent for them.
The family-based immigrant group from far away places will
then have a higher concentration of high-skilled people, and
this will reduce the premium.

The final point to be made on the DISTANCE variable
concerns the information argument. The further away a country
is, the lower is the level of information about the U.S.

However, deficient information is likely to be less of a

pr:
00‘.

til

we

46
problem for high-skilled people, given education, professional
contacts, etc. This, once again, leads to a situation in which
the family-based immigrant groups from far-away countries lose
a number of lower-skilled immigrants, thus raising the average
quality of the family-based group and lowering the skill-based
wage premium.

The addition of HIGHER EDUCATION in Model 4 does not
alter the significance of the INEQUALITY coefficient, so the
idea that the SKILL coefficient is merely a product of the
distribution of skills in a country is not supported. The only
effect of the HIGHER EDUCATION variable is to reduce the
significance of the ENGLISH coefficient. This is hardly
surprising, since a high correlation should exist between the
level of education in a country and the English language
ability of its people. This variable does not, however,
measure average skills in the populations of the home
countries, so it is still possible that the distribution of
skills plays a role in determining the SKILL coefficient.
Section 6: Summary and conclusion

The first finding in this chapter is the variation across
countries in the difference in labor-market quality between
the skill-based and family-based immigrants. This supports the
idea that the underlying national framework, in which
individuals make migration decisions, influences the nature of
the immigrant group from that country. The second finding, and
the more interesting one, is that income inequality in the

sending country has an influence on the make-up of the

47
immigrant group, a finding that gives support to the Borjas
model. Additional evidence has also been found which indicates
the importance of distance and English language skills in
determining the quality of immigrants.

Some tentative policy lessons can be drawn. It may be
thought that an immigration policy that does not use labor-
market characteristics as immigration criteria will lead to
the immigration of the low-skilled. The analysis here
indicates that for some countries such as Sweden, with its
SKILL coefficient of .1662, and the Netherlands, with its
SKILL coefficient of .1849, such a view is untrue. For these
countries, the absence of labor-market immigration criteria
does not lead to an inflow of unskilled people. Rather, the
immigrants from these and.other similar countries will tend to
be of higher quality regardless of immigration criteria.
Restricting immigration from such countries does not keep out
the low-skilled, against whom so much anti-immigration
rhetoric is directed. Instead, it keeps out the relatively
high-skilled.

For other countries, the absence of labor market criteria
in immigration policy does lead to the influx of the low-
skilled. The Mexican SKILL coefficient of .6827 shows how far
below the skill- based immigrants Mexican family-based
immigrants are. A policy which favors the immigration of those
from negative selection countries will lead to a pool of
immigrants with less desirable labor-market characteristics.

If the labor market objective of immigration policy gains

48
precedence over the family re-unification objective, then the
need for labor market criteria for negative selection
countries is clear. For positive selection countries, however,
such criteria appear to be less necessary in an effort to
restrict immigration to the high-skilled. Whatever the
objective of immigration policy one point seems clear: the
effect of immigration policy across countries will differ
because of the different nature of the potential immigrant

groups from each country.

Chapter 3
Evidence on Immigrant Selection from the Greencard Lottery
winners

Section 1: Introduction

Since 1987 the U.S. State Department has run a series of
lotteries in which the prizes are permanent residency in the
United States. Citizens from various countries could enter
these lotteries by simply sending a piece of paper with their
name, address and one or two other details to a post office
box. From the pile of entries, the winners are randomly drawn
and informed that they are eligible for permanent residency.

The lotteries came about because of the shift in the
national origin mix of immigrants to the United States in the
last 20 to 25 years. The decline in the numbers coming from
the former "Old World" countries means that these countries do
not "have access to immigration channels primarily because
sponsorship links have been depleted".1 The purpose of the
lotteries is to provide a quick and simple way "to promote
diversification in (the) legal immigration system". This is
done by restricting the lotteries to citizens of countries
"that have sent the United States relatively few immigrants
over the past several years".

The first lottery was authorized by the Immigration
Reform and Control Act of 1986. This act provided for the

issuance of 5,000 greencards through lottery in each of 1987

 

1 Quotes in this paragraph are taken from House of
Representatives, Report 100-1038

49

50

and 1988. This was later extended by the Immigration
Amendments of 1988 to 15,000 in each of 1989 and 1990. The
Immigration Amendments of 1988 also set up a second program to
distribute 10,000 visas in each of 1990 and 1991. A third
lottery program was established under the Immigration Act of
1990 which provided for the issuance of 40,000 visas in each
of 1992, 1993 and 1994.

This chapter exploits this natural experiment in
immigrant selection. Prior to the lotteries, immigrants coming
to the United States were not a random sample of people in a
country who wanted to migrate. In order to migrate their
desire to migrate had to be supported by the legal ability to
do so. In practical terms only those with close family ties in
the U.S., those with highly sought skills or refugees could
immigrate. The resulting pool of immigrants did not
necessarily give a complete and unbiased picture of those who
wanted to migrate, although economists have studied them as if
they did.

The lottery immigrants are, however, a random draw from
the pool of people in a country who want to migrate. In this
chapter I develop a model to show how the relative skill
levels of lottery-winning and family-based immigrants.provide
us an insight into the nature of the pool of people in a
particular country who want to immigrate into the United
States. I then show how the relative shapes of the income
distributions in the U.S. and a sending country will influence

the relative skill levels of lottery and family immigrants. I

51

then examine the actual relative skill levels of lottery and
other immigrants from a range of countries to determine if
there is support for the model, using data from the
Immigration and Naturalization Service. Finally, I consider
the’locational choices of different immigrant groups to see if
there is evidence supporting one of the model's assumptions.
Section 2: The Model

In this section I develop a model which will help to
predict how lottery immigrants may differ in labor-market
quality from immigrants who gain admission into the United
States through family contacts. It will show how the relative
skill levels of the two groups can provide insight into the
nature of the potential immigrant pool from a country.

When choosing whether or not to migrate to a particular
country people compare their utility levels in the two
countries. Having adjusted for the costs of migration, they
choose the location in which utility is maximized. In this
model, utility comes from two sources: wages and the presence
of ., family. Denoting the utility from wages as w and the
utility from the presence of family as f, the decision on

whether or not to migrate is based on the comparison:

>
<

in + f1-c: v“ + fo.

The subscript 1 means a value in the U.S. , and 0 means a value
in the home country. The person will want to migrate if the
argument on the left is greater, i.e., if utility in the U.S.
less the cost of migration is greater than utility at home.

Consider this comparison for a person with close family

52

in the U.S. who can gain entry into the U.S. through
sponsorship by these family members. Assume that the value of
f for this person is positive in both.the U.S. and in the home
country. The subscript r is used.to denote a family immigrant,
i.e., an immigrant who is sponsored by relatives. This person
will choose to migrate if

wr1 + fr1 - c > wr0 + fm°
For the marginal family immigrant this will be an equality:

wr1 + fr1 - c = “w + fw‘
If we assume that the utility from family is the same in both
the home country and the U.S. for family immigrants, i.e. fr1

= fw' then for the marginal family immigrant:

W -C=Wro.

n
This says that the marginal family immigrant will want to
migrate as long as no pecuniary loss is suffered. The loss in
utility through leaving family in the home country is offset
by the gain in utility from.the presence of family in the‘U.S.
The slightest monetary gain will lead to migration. (While
this assumption on family utility may appear extreme, all the
results hold once there is greater utility from the presence
of family for family immigrants relative to lottery
immigrants.)

Now consider an immigrant without these sort of family
contacts who can only immigrate through the lottery. For such
an individual assume that the utility from family is 0 in the

U.S. and positive at home. If the subscript l is used to

denote lottery immigrant, the following holds for the marginal

53
lottery immigrant:
"n ' c 3 Wm + fw

or wl1 - c > wm.

This says that for the marginal lottery immigrant wages in the
U.S. less costs of migration will have to be greater than
‘wages at home to compensate for the loss in utility from being
away from family.

The next issue is how this difference between the two
groups on the margins of immigration will affect their
relative labor market quality. Consider first a situation in
which all in a country have a monetary incentive to migrate.
This could arise when the mean income in a country is so far
below that of the U.S. that all could improve their wages,
adjusting for the cost of migrating, by moving to the U.S. In
such a situation, all those with family ties will choose to
immigrate. They improve their utility by improving wages,
without losing utility from the presence of family. Hence, the
average labor-market quality of the family group will be the
same as the average labor-market quality in the population of
the sending country.

The situation for those without family ties, the lottery
group, is different. As stated earlier, only those who will
receive a pecuniary gain sufficient to compensate for the loss
of family will migrate. To see how the dispersion of wages in
the home country 'will influence who immigrates through
lottery, consider Figures 4(a) and 4(b). These figures show

the distribution of earnings in the home country and in the

54

U.S., were all to migrate to the U.S. In 4(a) income is more
equally distributed in the home country relative to the U.S.
In this case the greater financial gain from migration accrues
to those at the upper end of the earnings distribution. This
can be seen in the greater distance between the potential
earnings in the two locations for upper-income individuals
relative to lower income people. This greater earnings gain
produces a greater probability that upper-income people will
gain sufficiently from migration to compensate for the loss of
family and will thus be more prepared to migrate through
lottery. In this situation the lottery immigrants will be more
skilled than the family immigrants.

In Figure 4(b) earnings are more equally distributed in
the U.S., so the distance between the home earnings and the
U.S. earnings is greater for those at the lower end of the
distribution relative to those at the upper end. As such, 4(b)
shows a situation in which the greater financial gain from
migration accrues to individuals in the lower end of the
earnings distribution. The lottery group will be less skilled
than the family immigrants, since the lottery group is
selected from those with the greatest financial incentive to
migrate and the family group is selected from across the full
range of the earnings distribution.

Now consider the situation in which the means between the
two countries are sufficiently similar that not all have a

monetary incentive to migrate. As discussed in Chapter 2, two

55

353...; 2:3:— Eecscu swung..—
so=§=m use 9:32 Both:— :33 macs—ESE— .Eozou— .6... 8.5»:—

 

95 as”.
. n
u u
u I
a n
u n
n a

n
n
u
u
-
u
-

 

.895 5.85 .885 588 u ac”.—

 

 

.htezc. 2:3... :3... .I. 15H m

cuts..— u m

«8.3m 18...: .I.. m:

3:55..» 2:8:— Eacscu 55.3%

.5me use 2:52 .55.:5 .5? 3:3me5— Date.— .e. 6.5»:—

 

56

 

 

95 1.3
1 i
u n
.
a n
.
.
.
.
.
u
.
.
moilIv " a
u .522. 2.5:.— 0562: :5:— u AW:—
" >536. 2:8... :5:— u it”.—
.

 

  

5.9:...— u m

nee-3m 682.: u m:

57

Seesaw—E:— m5=od ecu sauce—em «>525 .m 55w?—

8556 2:5. ace

1. .

 

 

m:

 

.522. >=E£ 5:52: 55:. u AVE
528. 2:85 :8: u 95
c398..— u h

5.5m 55:: u m:

3:53—55 #5:?— EB gage—om ozaawoz .e ouawmh

3553‘

15”.. :3

 

 

T

58

 

 

 

coma: >=E£ 2:85 53:. n AWE
mD E3...— 9585 :3:— u AWE
5.9.8..— " r.—

935m 33:: n m:

59

selection possibilities exist, positive and negative.2
Consider first the positive selection case which is drawn in
Figure 5. Note that wages are more unequally distributed in
the U.S. Immigrants come from the upper end of the earnings
distribution. In the case of the lottery immigrants, only
those who get a sufficient monetary benefit will choose to
immigrate, so the area from which this group is selected is
even more skilled than would be the case if only the monetary
incentive to immigrate was being considered.3

Among the family immigrants all those who make the
slightest pecuniary gain from migration will migrate. As the
family group comes from a wider range of the distribution than
do the lottery immigrants, the lottery immigrants from such a
country should be more skilled than the family immigrants.

Now consider the case of negative selection, as shown in
Figure 6. The monetary gainers here are located at the lower
end.of'the:distributionm The lottery immigrants will come from
this end of the distribution.but the cutoff will be at a lower
level than would be the case if only monetary considerations
were involved. Family immigrants also come from the lower end,
but the cutoff will be at a higher level than the cutoff for
the lottery immigrants. Again, even the slightest pecuniary

gain will bring about the migration of the family immigrants

 

2 The third selection outcome, refugee sorting, is
omitted to keep the analysis more compact.

3 The intuition here is based on the model of George
Borjas, A.E.R. 1987

60
since no compensation for loss of family is required. In this
situation the model suggests that the family group will be
more skilled than the lottery group.

To summarize the ‘model so far, lottery and family
immigrants differ in their migration decisions because of the
utility derived from the presence of family. Lottery
immigrants leave family and must be compensated in order to
make migration attractive. Family immigrants substitute the
family members in the home country for the family members in
the U.S. so no such compensation is required.

The relative labor market quality of lottery and family
immigrants depends on which individuals in a country achieve
the greatest pecuniary gain from migration and hence the
compensation sufficient for them to leave their families. The
lower is the level of inequality in the home country relative
to the host, the more likely it is that the lottery immigrants
come from the upper end of the income distribution and are
more skilled than the family immigrants.
section 3: Estimating the Lottery/Family Wage Differential

In this section I estimate the labor-market quality
differences between the family-based and lottery-winning
immigrants from nine countries. The data are from the
Immigration and Naturalization Service and include all
immigrants legally admitted to the United States between 1988
and 1990, except for those admitted under the amnesty program
of IRCA. For each individual the data include occupation

performed in the country of last residence, nationality, age,

61

sex, marital status and class of admission, i.e., family,
lottery, skill etc. I assigned a wage to each immigrant by
matching their occupations with the occupational median weekly
earnings of full-time wage and salary workers from EEELQXEQDS
ang_fig;ning§ (January 1990). Although the I.N.S. data include
all those legally admitted, I include only males aged 22 to 64
in these regressions. The wage data I use are thus male
earnings. I selected the nine countries on the basis of their
large representation among the lottery winners. Together they
account for over 80 percent of the lottery winners in this
age/sex category.

In order to estimate the skill difference between the
family and lottery immigrants I estimate wage equations which
include a dummy variable equal to one if the individual is a
lottery immigrant and zero otherwise. A dummy for those who
were admitted on the basis of their skills is also included.
As family-based immigrants are the omitted class, the
coefficients on the lottery and skill dummy variables can be
interpreted as the wages of these groups relative to those of
the family immigrants.

The results of the regressions are presented in Table 5
along"with. an inequality"measure for’ each country. The
inequality measure is the ratio of the income of the highest
10 percent to that of the lowest 20 percent and is taken from
the flgzlg_ngyglgpmgn;_3gpgrt of 1993. As described in Chapter
2, the size of the coefficient on the skill dummy variable can

be taken as an indicator of whether each country exhibits

62

lllllllxllsliniulllilllvl I.)

 

 

 

 

 

musmmorl
>m.~ n H~.n mm.~ mn.~ n~.v auuamsumcH
mucmumHaEH
mme m¢n ooh av» emm omen mumuuoq
«Han mnam mnma mmom bump mmmwa z.
~man.° ammo.o mmaa.o mmHH.o muoo.o o¢¢H.o ~mu.no<
.oao.oc Aamo.ov AmHo.ov Amao.oc Aoao.oc .noo.oc .
mHH.o ~mo.ou mHH.o m~o.o voa.o oma.o amassed
Aoao.oc Aamo.oc Aauo.ov Ammo.ov .Hmc.ov .moc.oc
mmn.o ann.o mmn.o mo¢.o ome.o hon.o quxm
Anao.oc Anao.ov Aamo.oc Amao.ov Aooo.oc Apoo.oc
aha.o cmo.on voo.o p~o.ou ano.o oHo.o omummaz
Aboooo.oc Amoooo.ov Amoooo.ov “moooo.ov Anoooo.cv .moooo.oc nameoom
«coo.ou ~ooo.ou ~ooo.o nooo.ou ~ooo.on nooo.on mom
Amoo.oc Aqoo.oc Ahoo.oc .moo.oc Anoc.oc Amoo.oc
omo.o -o.o nac.on buo.o «Ho.o nmo.o woe
.Hoa.oc Anho.oc AHnH.oV Apaa.ov Ameo.o. Ammno.oc
me.m ~m.m mne.m omm.m ~mn.m me>.m smuommeza
,
canon coumwxom mammcoocH smoomamcom ocmaom momcmu manmfiuo>r
. mflfihnvflfiou
mmocmuouuuo coma anasom\»uouuon on» ocwumaaumu mcoammmuvmm an ounce

guess“: muwuuoa you

63

 

.mummMMImmmammdMNwmIUAumm on» no cowuaoo nodaumm co Bonn coxou ma mucmau man» u

 

.- ; !.-!--!- w! ! .lwmawlmms J
.>.m .m .m¢.n aufiamsvmnu
mucmuwweaH
o- mom Hmama aumuuoq _
eomw whom women 2
mmo.o eeH.o camo.c ~m .noc
Ammo.ov .mmo.oc Aooo.oc
Heo.ou ona.o evo.o amassed
Apmo.ov Ammo.oc .wo.oc
mv~.o men.o ¢H¢.o AAme
Abao.oc Amao.oc Amoo.oc
omH.ou moo.o Ho.o omammez
Anoooo.ov «moooo.ov Anoooo.ov
«coo.ou ncoo.ou noooo.o.. ommaoom mom
Amoo.oc .moo.o. Amoo.o.
mno.o nmo.o mnoo.o mo<
“NHH.oc .umo.oc A~¢o.oc
mmm.m «an.m no.m emuomuazu
woman ucfiusouud manoauo>

ocoaouH

«Ac.acoo. u wanna

64

positive or negative selection. A large value indicates a big
difference between the family-based and skill-based
immigrants, so the bulk of the country's immigrants come from
the lower end of the distribution, i.e., negative selectiOn.
A small difference between the family-based and the skill-
based immigrants means that the bulk of a country's immigrants
come from the upper end of the distribution, i.e., positive
selection. In order to put the inequality measure and SKILL
coefficient into context, note that the value of the
inequality measure for the‘U.S. is 5.319 and the average value
of the skill coefficient from 38 countries (as reported in
Chapter 4) is .3522.

Of the nine countries that are examined six have LOTTERY
coefficients that are positive and significant. Five of these
six have inequality measures that indicate lower levels of
inequality than the United States so these five alone support
the ideas in the model. The sixth country however, Argentina,
has a greater level of inequality, which runs counter to the
predicted effect.

The only other country that has a level of inequality
greater than that of the United States is Egypt. While its
LOTTERY coefficient is insignificant at conventional levels,
it is negative. This gives some support to the model, but the
size of the SKILL coefficient erodes this support. I predicted
that a country with a high inequality level will have a
negative LOTTERY coefficient but also a large SKILL

coefficient. Since Egypt has the lowest SKILL coefficient, its

65
values of the SKILL and LOTTERY coefficients are contradictory
by the logic of the model. The small SKILL coefficient
indicates that the family immigrants are relatively highly
skilled, so the lottery immigrants must be relatively highly
skilled also.

The negative IOTTERY coefficient of Pakistan is also
contrary to the predictions of the model. Pakistan has a lower
level of inequality than the U.S. so the model predicts that
its lottery immigrants should.be:more skilled.than the family-
based immigrants. This result, along with those of Argentina
and Egypt, leaves the model with little empirical support.

From a purely descriptive point of view it is interesting
that, in general, the lottery winners seem to be more skilled
than the family immigrants. In order to investigate this
difference further'it.is possible to examine whether or not it
is present across lotteries or whether one particular wave is
driving the results.

For all countries the data include information on which
immigrants were admitted under the first and second lotteries.
For Ireland, the data also include information on those
admitted under the first wave of the third program. Table 6
presents results from regressions in which the individuals
from the lottery groups of Table 5 are broken down into the
particular lotteries under which they were admitted. Once
again, the reference category is the family group..Only six of
the countries are reported because two of the remaining three

were not included in more than one lottery and Egypt had only

66

 

 

 

 

 

Amoo.ov
«\z ~ooo.o «\z «\z «\z ¢\z n humane;
Ammo.oc Ano.o. .mmo.oc Ammo.o. .Hno.oc .omo.o.
Hmo.o nno.o «wa.o an.o Hoa.o mea.o N mumuuoq
Ammo.oc “moo.oc .mao.ov .Hmo.oc .oao.ov Anoo.ov
noa.c moc.o nafl.o naa.o «oH.o o~a.o H aumuuoq
acwucmvud osmHmuH cacao mwmocoocn condom momcmu

 

moduouuoq acououmwa ovcw
coxoum mosouo humvuoq sud3 moocououuwo mums >H«30m\muouvoa msuuoaqumm unawmmoummm no ounce

67
one immigrant in one of the lotteries.

In four of the six countries the skill levels of the
lottery immigrants relative to the family immigrants are
greater for both lottery groups. This supports the idea that
the greater skill levels of the lottery groups are the result
of some systematic selection process as opposed to a random
occurrence. That the skill level of the Irish and Argentinian
lottery winners converges to that of the family immigrants
dilutes this finding. For the Irish, part of the explanation
for this convergence may lie in the fact that in order to
qualify for a greencard under the third lottery, the winners
had to have an offer of one year's employment in the‘U.S. This
provision may have favored illegals already in the U.S. If the
occupations reported by these illegals are the jobs they are
doing illegally in the U.S., this explains this part of the
convergence. No ready explanation exists for the convergence
seen under lottery 2. However, as only 76 individuals come
under this heading out of an Irish lottery total of 12,151,
results for this wave are of less interest than the other two
waves. In the case of Argentina, the lack of significance for
the second lottery may also be due to the low number of people
admitted under that lottery, only 30.

The observation that the lottery immigrants are, in
general, more skilled than the family immigrants suggests that
the selection of lottery immigrants may be influenced by a
process other than the one described in the model. One

possible alternative explanation of the relative skill levels

68

views the presence of family less in a utility sense and more
as a labor market resource. If families offer support to
immigrants either through providing employment or through
directing immigrants toward employment opportunities, a
difference will be generated between those who immigrate with
family present and those without family, i.e., the lottery
immigrants. If this sort of support from families is
particularly important for those at the lower end of the
income distribution, the lottery group will have a smaller
proportion of low-skilled people. One can think in terms of an
unskilled worker who will migrate only if he will have help in
finding a job, and a highly skilled worker who will migrate
independent of any help.
Section 4: The Locational Decisions of Lottery Immigrants

In the model I assume that the utility from the presence
of family members would be higher in the U.S. for family
immigrants relative to lottery immigrants. In order to test
this assumption, I use information on the location decisions
of the various immigrant groups. The implications of the
analysis go beyond testing an assumption of the model.
Location decisions of immigrants have implications for the
social make-up of regions and for public budgetary situations.
Location decisions of immigrants also reveal information on
the extent of assimilation into U.S. life.‘

As with all economic goods, whenever a particular good is

 

" See Ann Bartel and Marianne Koch in Abowd and Freeman (eds.)

69

not available individuals will seek a substitute. For lottery
immigrants the good that is not available is the presence of
family members in the U.S. and the substitute which is
proposed is the presence of people of the same national
origin. The result of this should be a greater concentration
of lottery immigrants in areas where they find large numbers
of their compatriots. Family immigrants, however, are less in
need of the substitute good. They will be more likely to move
away from the concentration of compatriots, especially if they
are marrying Americans or coming to family members who have
been in the United States for a while and have moved away from
the hub. Even if the primary purpose of the family in
migration is to provide the labor-market support described
above, the utility dimension of family can still exist, so the
reasoning here is still valid.

In order to test this, I focus on Irish immigrants. The
main reason for this is that the Irish are concentrated in one
area of the country and this concentration facilitates the
econometric analysis presented later. The data have, for each
individual, the zip code of intended residence. Table 7 shows
how the Irish are distributed across the U.S. as a group and
under different classes of admission. The numbers show that
67.9 percent of all Irish immigrants intend to reside in the
New England and Middle Atlantic regions. Of the family
immigrants 62.4 percent intend to reside in the two regions
mentioned above. The corresponding figure for the lottery

immigrants is 69.4 percent, so the predicted greater

7O

 

 

 

 

 

 

 

! a I !!! !!! |!! | ! ! J
cod HmH~H oca «mm cod «mam Hence
~.¢H own” m.m~ on o.mH can onuaomm
m.o pm >.~ n m.H mm canvases
Hmuucwu
H.H and m.e ma m.~ om suzom uwoz_
Hmuucoo
m.o and m.a m m.a en suuoz ummz
Hmuucoo_
e.n com o.HH mm o.m «ma :uuoz ummu
Hmuucoo_
m.o «a m.H e m.c ma bosom unam_
caucuauc
o.m mum ~.HH an 6.5 «ma runom.
caucnau<
H.m¢ «awn m.m~ on «.66 com «Hood:
ocoamcm
n.H~ qmmm o.HH mm o.mH Han awz
» a a a a * :ofimmm
shunned Haexm Nassau
COHmdenu‘

mo mmoao can cosmom monsoo >3 mucnumwaSH oamz.smfiuH mo sagasnauumwo Dwnmcuooou one up wanna

71
concentration of the lottery immigrants is evident.

I performed two chi-squared tests to see if the different
proportions of family and lottery immigrants in the different
regions are statistically significant. One test used the nine
regions and produced x2 = 68.925. As {395(8) = 15.5, the null
hypothesis that the proportions were the same was easily
rejected. The second test considered just two areas. The
Middle Atlantic and.New England were treated as one region and
the other seven regions as another region. In this test x2"
9.593, again allowing the easy rejection of the null
hypothesis that the proportion of family and lottery
immigrants in each region is the same, since x235(1) - 3.84.

In order to ensure that the different locational pattern
between lottery and family immigrants is not solely an Irish
phenomenon, I also considered the locational pattern of Polish
immigrants. Polish immigrants are concentrated in two areas,
the East North Central and the Middle Atlantic. Of the lottery
immigrants 87.9 percent are in these two regions. The
corresponding figure for the family-based immigrants is 80.6
percent. I performed. a chi-squared test to see if the
proportions of lottery and family immigrants were
significantly different. In this test x2 = 95.627. As x2.”(8)
- 15.5, the null hypothesis of equal proportions was once
again rejected.

To examine further the hypothesis that the lottery
immigrants will be more heavily concentrated in the area where

compatriots are found I performed the following test. As the

72

Boston/New York area is the center of the Irish community in
the U.S., the zip code for each immigrant is matched with a
measure of distance from New York City. This distance measure
is set equal to zero for Massachusetts, Connecticut, New
JerSey, New York and Rhode Island, so that no distance is
imputed to those who remain in what amounts to the Irish hub.
For each individual there now exists a measure of the extent
to which they moved to areas away from the Irish hub of the
Boston/New York area and, as such, a measure of the need for
the substitute good forche presence of family. The usefulness
of the single area concentration of the Irish for the present
task should now be apparent. If it is true that the family
immigrants are less in need of the substitute of the presence
of compatriots than the lottery immigrants, we should observe
the family group moving out of the Irish hub to a greater
extent than the lottery group.

Table 8 presents results from Tobit type regressions in
which the dependent variable is distance in miles from New
York City to the largest city in the State where the immigrant
intends to reside. Tobit analysis is used because over half
the values of the dependent variable are zero. Results from
similar estimates are presented in Table 9, the difference
between the two being the pooling of all lottery immigrants in
Table 8 and the breaking up of this group into three waves in
Table 9. The distances are taken from the Bang;uguglly_gggg
Atlas of 1993. Only the continental United States is included,

which causes the loss of a tiny proportion of the sample. The

 

 

Table 8 :

Lottery Immigrants

73

Tobit Analysis of the Location Decisions of Irish

 

 

N 3 14,442

 

 

Variable MODEL 1 MODEL 2 MODEL 3
INTERCEPT -1725.139 ~1700.613 -2335.170
(358.763) (363.021) (444.338)
AGE 62.153 56.515 79.649
(20.885) (21.066) (23.173)
AGE SQUARED -0.677 -0.596 -0.708
(0.283) (0.286) (0.291)
SKILL 967.673 976.313 2298.677
(155.471) (156.331) (710.769)
LOTTERY -315.412 -263.243 359.047
(64.959) (72.576) (256.807)
MARRIED -l3.654 -35.462 -9.681
(52.998) (53.898) (54.834)
YEARS in U.S. - 66.786 70.492
(28.752) (28.777)
LOTTERY*AGE - - -19.459
(7.720)
SKILL*AGE - - -40.388

(20.910)

 

74

set of independent variables is similar to those used in the
regressions presented in Table 5. One variable that is added
here is YEARS IN U.S. For immigrants who are already in the
U.S. on non-immigrant visas, the data include information on
the year of non-immigrant admission. For these individuals
YEARS IN U.S. is equal to the year in which they got their
greencard minus the year they arrived on a non-immigrant visa.
For those newly arrived in the U.S. this variable equals zero.
YEARS IN U.S. may also be equal to O for many illegals, since
they may not want to reveal that they were residing in the
U.S. This variable is included to capture possible
assimilation effects which would lead immigrants to move
further from the hub.

In Table 8, Model 1, the negative coefficient on the
LOTTERY variable and its statistical significance are as
predicted by the model, i.e. , the lottery immigrants move out
of the hub to a lesser extent than the family immigrants. In
Model 2, YEARS IN U.S. is added and the negative and
significant coefficient on lottery remains. The sign and
significance of the coefficient of YEARS IN U.S. is as
expected. 4

In Table 8, Model 3 two interaction terms are added,
producing an interesting twist in the results. The coefficient
of LOTTERY loses its significance, but the LOTTERY*AGE
interaction term is negative and significant. This indicates
that the observed greater concentration of the lottery

immigrants in the Northeast relative to the family immigrants

75
works through older lottery immigrants settling more in this
area. This result also indicates that the need for the
substitute for family is stronger for the older lottery
immigrants. If youth is more of a time for freedom and
independence, this could explain the result. As they age,
familiarity may be desired. If one thinks in terms of families
providing support in the labor market, the greater
concentration of the lottery immigrants could be a result of
them looking to compatriots to provide the network that family
immigrants get from family'members already in the‘U.S. If such
networks are more important for older workers, then the
greater concentration of older lottery immigrants makes sense.

Table 8 also shows that the skill-based immigrants are
less geographically concentrated than the family-based
immigrants. The reason for this may be that sponsorship by an
employer is required by this group when permanent residency is
being granted. The location of a skill-based immigrant is
therefore determined by the location of the employer. The
negative coefficient of SKILL*AGE indicates that older skill-
based immigrants will only move if the prospective employer is
in the Irish hub.

Earlier in this chapter we saw that the observed higher
skill levels of the Irish lottery immigrants relative to the
family immigrants was the result of the higher skill levels of
the first wave of lottery immigrants. The third wave did not
share this feature, possibly because of the presence of

illegals among this group. In order to see if the first and

76
third waves also differ in their location choices I
reestimated the distance equations, breaking the lottery group
into three waves. Since the first wave of Irish lottery
winners was more typical of the lottery winners in general, it
may be that the experience of this group is more generally
applicable. The results are reported in Table 9.

The pattern of results in the models in Table 8 is
repeated.to a great extent in Table 9. In Model 1 the negative
coefficient of LOTTERY is repeated in the coefficients of
LOTTERY 1, 2 and 3. The lack of statistical significance for
LOTTERY 2 may be a function of the relatively small number of
immigrants in this category, a mere 76. The statistical
significance for LOTTERY 1 and 3 remains in Model 2 where
YEARS IN U.S. is added. Finally, the twist produced by the
addition of the interaction terms is repeated.

While the negative coefficients on the lottery variables
are consistent with the utility arguments in the model, an
alternative interpretation could be given. Those living in the
Boston/New York area, both illegals and non-resident aliens,
might have had better information on the lottery programs
because of the big Irish community and so a greater proportion
of applicants may have come from.here. Thus the concentration
of the lottery winners could result from information as
opposed to preferences.

One implication of this alternative interpretation is
that the lottery winners who were already in the U.S. should

show a greater concentration than lottery winners who are new

Table 9: The Regressions of

Broken into Three Waves

77

Table 8 with the Lottery Group

 

Variable

 

 

MODEL 1 MODEL 2 MODEL 3
INTERCEPT -1710.386 -1682.745 -2387.375
(355.507) (359.442) (447.174)
AGE 59.383 54.252 79.686
(20.811) (20.992) (23.190)
AGE SQUARED -0.651 -0.577 -O.710
(0.283) (0.285) (0.291)
SKILL 978.384 989.135 2335.872
(155.009) (155.889) (710.652)
LOTTERY 1 -299.122 -233.273 391.310
(65.416) (73.220) (266.837)
LOTTERY 2 -249.230 -182.208 -141.527
(315.317) (318.643) (1156.702)
LOTTERY 3 -304.791 -273.400 457.402
(70.637) (78.417) (297.921)
MARRIED 18.125 3.357 25.591
(43.628) (44.318) (45.049)
YEARS in U.S. - 64.653 68.673
(28.300) (28.753)
LOTTERY 1*AGE - - -l9.564
(8.139)
LOTTERY 2*AGE - - -0.197
. (36.539)
LOTTERY 3*AGE - - -23.079
(9.192)
! SKILL*AGE - - -41.235
(20.915)

N I 14,442

 

78
arrivals. This implication can be tested by estimating a
distance regression using only the lottery winners and
including a dummy variable indicating those who were already
in the U.S. and are adjusting to permanent residency. Such a
regression shows no statistical difference between adjusters
and new arrivals.

Since illegals may appear to be new arrivals if they deny
having been in the U.S., this result on the adjuster/new
arrival dummy variable cannot be considered conclusive. It
still could be that the illegals in Boston and New York have
been in the U.S. for a while and have better information on
the lottery through the Irish community.

There are two empirical reasons to doubt that the
illegals are driving the results on the location of lottery
winners. If the first and the third waves differ in the
proportions of illegals present, the location choices of the
two should differ if illegals are driving the result. No such
difference is evident. The coefficients of the age/lottery
interactions are also informative. If illegal immigrants are
younger than other immigrants, as seems to be the case, and
the concentration of illegals in the Boston/New York area is
driving the lottery result, the age/lottery interaction
coefficient should be positive: younger lottery immigrants
should show more concentration. The coefficients on the
interactions Show that the younger lottery immigrants are
actually more dispersed. It is older lottery immigrants who

jproduce the result that lottery immigrants move out of the

79
Northeast to a lesser extent. This makes it less likely that
illegals are driving the result.

While there are empirical reasons to doubt that legal and
illegal aliens already in the U.S. cause the lottery distance
results on their own, it is difficult to ascertain their joint
effect. Fundamentally, the problem is one of inaccurate data
arising from possible inaccuracies in recording whether
illegals who won greencards were already present in the U.S.
Section 5: Summary and Conclusion

In this chapter I have used the existence of an unusual
category of immigrants to gain some insight into the nature of
the migration decision and selection outcome. I developed a
model that shows how the decision to migrate differs at the
margin for family and lottery immigrants. I also showed that
if the lottery immigrants from a country are more skilled than
the family immigrants, it must be because those who can make
the greatest monetary gain through migration are located in
the upper end of the skill distribution. Likewise, if the
lottery immigrants are less skilled than the family
immigrants, those with the greatest gain from migration are
from the lower end of the distribution. The potential gain for
different groups is influenced by the relative levels of
income inequality between two countries.

The empirical results showed that the lottery groups from
a range of countries differ in their skill levels from the
family groups. Furthermore, this difference persists over

different lottery programs, so it is likely to be the product

80

of a systematic selection process. The idea that the relative
skill levels of family and lottery immigrants is a result of
the interaction between private utility maximization and the
level of inequality in home countries receives only mixed
support. The role played by family in providing labor-market
information to immigrants might be a more powerful determinant
of relative skill levels between the two groups.

For Irish immigrants the location decision of lottery
immigrants differs from that of the family immigrants. This
indicates that the lottery immigrants substitute the company
of compatriots for the presence of family.

The stated goal of the lottery programs was not related
to labor market outcomes. From the evidence here, however,
admitting immigrants in this manner affects the labor-market
quality of the immigrant pool. For the countries considered
the general result is that admission through lottery produces
more skilled immigrants. Without family in the U.S.,
immigration is more difficult for lottery immigrants, so that
only' those ‘who 'will do particularly ‘well migrate. This
supports the view of immigrants contained in the work of
Chiswick, i.e., that they are a group of well motivated and
capable individuals who take the risks inherent in migration
in order’ to better' themselves. Such a selection result
justifies admission by lottery beyond the stated goal of
diversification in the national make-up of the immigrant pool.

The results on location found in this chapter put the

lotteries in a less favorable light. The concentration of

81
immigrants puts pressure on certain areas in terms of the
costs associated with the presence of a large group of
immigrants. Such concentration also reduces the chances of
assimilation. If the lotteries produce greater concentration,

then this may offset the benefits the lotteries produce in the

form of higher quality immigrants.

Chapter 4
Using the 1988-1990 Immigrant Cohort to Learn More about
Immigrant Selection
Section 1: Introduction

In Chapter 2 of this dissertation I examined the labor-
market characteristics of immigrants from the years 1978 to
1980 in an effort to identify the quality of immigrants coming
from various countries. Furthermore, I attempted to identify
country characteristics which could help predict the relative
skill level of immigrants coming from a particular country.

In this chapter I repeat the analysis undertaken for the
1978-80 cohort using the 1988-90 cohort. Once again, I take
the skill difference between the skill-based and family-based
immigrants to be an indicator of the nature of the outcome of
each country's immigrant selection process. I then use
country-specific characteristics use in an effort to explain
differences in the skill indicator across countries.

In the model developed in Chapter 2 the variable of
greatest interest.in explaining the'variation in the skills of
immigrants across countries was income inequality. The
theoretical prediction concerning the effect of income
inequality on ‘the skill indicator’ was supported. by' the
empirical results. The primary purpose therefore in repeating
the analysis with the later cohort is to see if the
theoretical prediction continues to be supported. The
availability of values of the skill indicator and income

inequality for a number of countries for both periods also

82

83
allows an extension of the testing.

The availability of data on immigrants from the late
1970's.and late 1980's, combinednwith.rising income inequality
in the U.S. over this period, makes it possible to consider
the effect of growing income inequality in the destination
country on the make-up of the immigrant group. This rise in
inequality and some proposed explanations have been summarized
by Levy and Murnane (1992) . They write that the year 1979
”marked the beginning of a sharp acceleration in the growth in
earnings inequality." They also point out that the effect of
the increasing inequality in the period 1979 to 1987 was to
increase the proportion of men earning more than $40,000 (in
1988 dollars) while at the same time increasing the proportion
of men earning less than $20,000 (in 1988 dollars). I address
the possible effect of increased inequality in the destination
country on immigrant quality in this chapter.

Section 2: Skill-Based and Family-Based Immigrants Revisited

In this section, I review the model of Chapter 2. I then
present the results of the empirical analysis using the 1988-
90 immigrant cohort. The fundamental point of the model is
that relative income inequality between a country of origin
and a destination country will influence the nature of the
immigrant flow. If income in the home country is more equally
distributed relative to the destination country, those at the
upper end of the home country's income distribution have an
incentive to migrate. Conversely, if income is more unequally

distributed in the home country, those at the lower end have

84
an incentive to migrate.

Immigrants can gain permanent residency in the U.S. if
they are highly skilled or if they have family ties in the
U.S. Immigrants who are admitted because of their high skill
levels will be located at the upper end of the U.S. income
distribution, regardless of their country of origin. The
position of the family-based immigrants in the income
distribution depends, however, on the immigrant selection
process of each country. A country with a high level of income
inequality will, by the logic of the last paragraph, tend to
send lower-skilled immigrants. This produces a relatively
large difference in the skill levels of the skill-based and
family-based immigrants from such a country. A country with a
relatively high level of income equality will tend to send
high- skilled ‘workers, so that for such a country the
difference in skills between the skill-based and family-based
immigrants will be relatively small.

In order to test this model it is necessary to estimate
the difference between the skill levels of the family-based
and skill-based immigrants for’ a range of countries. I
estimate the wage premium of the skill-based relative to the
family-based. I then regress the wage premium against country
characteristics to see if the variation across countries is
explained.

As in Chapter 2, the data I use in estimating the skill
difference between the family-based and skill-based are from

the Immigration and Naturalization Service. For each

85
individual the data set contains information on occupation,
age, class of admission, sex, marital status and nationality.

In the data from the period 1978 to 1980 the reporting of
occupation was quite disaggregated and produced over 400
occupational categories. Unfortunately, the reporting in this
data set is much more aggregated and only 25 occupational
categories are produced. This lead to a concern that little
variation would be captured in the empirical analysis, but
positive results did emerge.

I transformed the occupational variable into a wage
measure by matching the occupation of each individual with the
median weekly earnings of full—time wage and salary workers in
those occupations. I took these data from W
Earnings of January 1990. I included only males aged 22 to 64
so the earnings data are male earnings. I selected the
countries because of the availability of data. Of the 39
countries, 25 were also considered in Chapter 2.

The equation I estimate for each country is the same as
that in Chapter 2, but it is shown here for convenience, with
explanations of the variables:

LWAGE .. be + b, AGE + 1::2 AGESQ + b3 MARRIED +‘ 8, D88 +

b5 D89 + b6 D90 + b, ADJUSTER + 8, YEARS IN U.S.
+ b, SKILL + b", REFUGEE + e
where LWAGE is the logarithm of the earnings associated with
the occupation reported. by ‘the immigrant: AGESQ is age
squared; MARRIED is a dummy variable equal to 1 if the

individual is married and 0 otherwise: D88, D89 and D90 are

86

dummy variables for each year of admission (1987 is the
omitted year); ADJUSTER is a dummy variable equal to 1 if the
individual was already in the U.S. as a non-resident alien
and is adjusting to permanent residence and 0 otherwise: YEARS
IN U.S. is years in the U.S. for adjusters to permanent
residence (it equals 0 for new arrivals): SKILL is a dummy
variable equal to 1 if the individual is a skill-based
immigrant and 0 otherwise: REFUGEE is equal to 1 if the
individual gains admission as a refugee and 0 otherwise.

The coefficient of SKILL can be interpreted as the wage
premium of the skill-based relative to the family-based. For
the reasons given above this is taken to be an indicator of
the nature of the flow from a country's immigrant-selection
process.

Table 10 presents the results from the regressions
performed for each country. As was the case with the
regressions using the 1978-80 cohort, there is clear variation
in the coefficient of SKILL. For Danish immigrants, the skill-
based are in occupations which pay 16 percent more than the
occupations of the family-based. For Sweden this figure is
19.6 percent and for Switzerland it is 18.2 percent. At the
other end of the spectrum, the skill-based immigrants from
Mexico are in occupations which pay 52 percent more than those
of the family-based immigrants. This figure is also 52 percent
for Costa Rica and 50.3 percent for Jamaica.

Table 11, Models 1-3, presents results from regressions

in which the dependent variable is the value of the SKILL

87

Table 10: Estimating the Wage Premium for the Skill-Based
Immigrants from the 1988-90 Cohort

Dependent Variable: Log wage

 

N !

 

 

Country’ SKILL Standard # Skill-
Coefficient Error Based
Belgium .2559 .0345 136 403
Denmark .1600 .0433 76 408
Finland .2762 .0588 44 152
France .2977 .0206 372 2217
Germany .2563 .0168 488 2459
Hungary .3421 .0443 38 871
Italy .4496 .0272 155 1986
Netherlands .2344 .0256 259 1030
Norway .2178 .0604 35 402
Poland .4662 .0211 194 7014
Portugal .5030 .0469 33 3237
Spain .3207 .0356 80 1045
Sweden .1956 .0277 190 812
Switzerland .1820 .0338 133 654
U.K. .2838 .0063 3365 18506
China .3911 .0114 1596 24735
India .3117 .0101 3434 19914
Indonesia .3646 .0298 167 601
ISrael .2931 .0150 563 3490
Japan .3255 .0162 650 2614
Malaysia .3246 .0247 341 957
Pakistan .3935 .0221 351 4827
Philippines .3988 .0178 439 24923
Thailand .4457 .0322 98 1909
Brazil .3585 .0257 181 1691
Canada .3122 .0080 1864 10402
Chile .4131 .0338 85 1294
Colombia .2976 .0216 117 6431

 

Table 10 (cont'd):

88

Dependent Variable: Log wage "

 

 

. Dominican
iRepublic

' Guatamala
!Honduras
‘Jamaica
[Mexico

! Panama

' Peru

EAustralia

fNew Zealand

iMorocco

.4794

.4810
.4291
.5031
.5196
.3778
.4680
.2964
.2984

.4813

.0585

.0507
.0737
.0327
.0113
.0784
.0252
.0230
.0402

.0840

25

22
10
69

18

Country' SKILL Standard # Skill- N
Coefficient Error Based
Costa Rica .5203 .0705 15 896

14822

4189
2767
12924
70118
1075
4972
1643
718
708

 

89

coefficient for each country from Table 10. As the dependent
variable is a regression estimate, the standard error will
differ across observations. In order to correct for this
weighted least square regressions are estimated, the weights
being the inverse of the standards errors of the SKILL
coefficients. The dependent variable in Table 12 is the
difference in the SKILL coefficients for each country for the
periods 1978 to 1980 and 1988 to 1990.

The results in Models 1 and 2 of Table 11 support the
Borjas model, since the sign and significance of the
coefficient of INEQUALITY are as predicted. The coefficients
of PER CAPITA GNP in Models 2 and 3 are negative and
significant which is in contrast to the results for the period
1978 to 1980. For that period the coefficients on this
variable were either positive or insignificant. One possible
explanation for the differing results is connected with the
measurement of GNP. For the later period a purchasing power
parity measure is used, whereas in the earlier period the GNP
measure was exchange-rate based. To the extent that the
purchasing power parity measure may give a truer reflection of
the relative conditions faced by immigrants, the estimates for
the later period may be more believable. The result that
higher per-capita income is associated with a smaller
difference between the family- and skill-based immigrants and
hence positive selection is consistent with the results of
Borjas (1987b).

In Model 3 a variable controlling for distance is added.

90

Table 11: Using Country-Specific Characteristics to Explain
the Variation in the Coefficient of SKILL for 1988-90

 

 

 

 

! —

Dependent variable: Coefficient of SKILL from Table 10 n
Variable Model 1 Model 2 Model 3
INTERCEPT 0.307 0.711 0.889

(0.022) (0.121) (0.143)
INEQUALITY 0.007 0.005 0.003

(0.003) (0.0027) (0.003)
PER CAPITA - -0.044 -0.057 9 )
GNP (0.013) (0.014)
DISTANCE - - -0.012

(0.006)

 

{Adjusted R2

 

Table 12: Regression Using Changes in Inequality and Changes

in Per-Capita G.N.P. to Explain Changes in the SKILL
Coefficient

 

 

—
“Variable Model 1 Model 2
!INTERCEPT -0.048 -0.048
; (0.016) (0.015)
5 CHANGE in 0.001 0.001
! INEQUALITY (0.003) (0.003)

 

- 0.013
(0.010)

!Adjusted R? 0.002
.N - 25

 

91

This DISTANCE variable is measured as the distance in air-
miles from the capital of each country to the nearest of New
York, Los.Angeles or Miami. With.the addition of this variable
the significance of 'the INEQUALITY coefficient is lost,
although the sign is still as predicted. The sign of the
DISTANCE coefficient is the same as it was in the earlier
analysis. This result, that greater distance is associated
with higher-quality immigrants, is also consistent with other
work (Schwartz 1973). One possible explanation for the
negative sign is that the further away a country is, the
greater the expense in coming to the U.S. If the low-skilled
face an income constraint, their immigration is curtailed.
This leaves a pool of high-skilled immigrants and hence' a
smaller difference between the family-based and skill-based
immigrants. The DISTANCE coefficient can also be explained in
terms of the psychic costs of migration. The further away a
country is, the greater are the costs due to the greater
difficulty of return visits or greater cultural differences.
Assuming psychic costs to be higher for the low-skilled, the
disincentive attached to distance will be greater for them.
Again, the resulting pool of immigrants will be relatively
high-skilled.

Table 12, Model 1, presents the result.of a regression in
which the change in the inequality levels over the period is
regressed on the change in the SKILl.coefficient over the same
period. In Model 2 the change in GNP, taken from the 39:14
ngyglgpmgnt_3gpgzt of 1982 and 1990, is added. The values of

92
SKILL for the period 1978 to 1980 were re-estimated using
weekly earnings data instead of the yearly earnings data that
had previously been used. This makes the measures of SKILL
consistent across the periods.

From the comparative statics in Borjas (1987b) we know
that, in general, countries which experienced rising
inequality over the period should also show a greater
difference between the family- and skill-based immigrants.
While the signs of the coefficients of CHANGE IN INEQUALITY
are as predicted by this general result the estimates are
insignificant.

There are a number of possible explanations for these
results. The explanation most damaging to the model is that
fixed effects are causing the cross-section estimates of SKILL
and not inequality. However, with only 25 observations in the
differenced regressions, it could also be that the lack of
significance is a result of the small sample. It is also
possible that the small number of occupations in the 78-80
data set made insignificant results likely. Another possible
explanation can be found in the comparative statics of the
model. Borjas has shown how, under certain circumstances, a
rise in inequality in the home country can lead to a rise in

1

immigrant quality: .A fuller explanation of this comparative

static result can be found in Appendix 3.

 

‘ Borjas, m 1987, page 538

93
Section 3: The 1979-80 and 1989-90 Immigrant Cohorts Compared

The focus of the work so far has been on the effect of
income inequality in sending Countries on the characteristics
of the immigrants coming from those countries. In this section
I reverse the focus. As income inequality rose in the United
States over the 1980's, it is possible to see if a change in
the skill levels of immigrants coming to the U.S. accompanied
it. One reason to expect a change is that rising inequality
increases the incentive to migrate to the U.S. for those at
the upper end of the distribution. Likewise it reduces the
incentive for those at the lower end. This change in relative
incentives can be seen if the changing earnings experience of
different education groups is considered. Levy and Murnane
(1992) note that for males aged.25 to 34 the earnings of’high-
school graduates fell by 12 percent over the period 1979 to
1987, while the earnings of college graduates grew by 8
percent. As Borjas (1991) points out, an increase in the
variance of income in the United States will produce a
composition effect whereby immigrant quality will rise.

As immigrants respond to changes in relative conditions
in the home and host countries, the effect of the rise in
inequality in the U.S. could be negated by rising inequality
elsewhere. Katz and Loveman (1990) show that inequality rose
in the United Kingdom in the 1980's so this it is possible
that this negating influence is present. If I find no quality
difference between the immigrants from the two cohorts, it

could be because of a rise in inequality elsewhere or because

94
of other region-specific influences on immigrant quality.
Whatever the underlying causes may be, it is of interest to
see what the trend is in immigrant quality.

A comparison of the immigrants from the late 1970's and
the late 1980's has been done by Funkhouser and Trejo using
CPS data. According to their findings "these data suggest that
male immigrants who entered the United States during the late
1980's are more skilled than those who arrived earlier in the
decade". They caution, however, that because of small sample
sizes their work cannot be seen as conclusive. Borjas (1994)
has also compared these immigrant cohorts, using data which
includes the 1990 Census. He finds the opposite to Funkhouser
and Trejo. According to his results immigrant quality declined
over the 1980's, although at a slower pace than during the
1970's.

In the comparison in this chapter I use yet another data
set. The data I use are part of those used earlier, i.e., the
I.N.S. data on legal immigrants admitted during the years 1978
to 1980 and 1988 to 1990. I omit the years 1978 and 1988 in
order to make the data set more manageable. This still leaves
around 400,000 males aged 22 to 64 for analysis.

One group that is not included in the data are those who
were granted permanent residency under the legalization
provisions of the Immigration Control and Reform Act of 1986.
This was an amnesty program for illegals who had resided in
the U.S. for a number of years. As this group is very large,

880,372 aliens in 1990, and is made up of former illegals,

95
inclusion would have produced a biased picture of the relative
skill levels of the two cohorts. As such, I consider their
exclusion an advantage for the present purpose.

As before, the occupation of each individual has to be
matched with a wage. To do this I use data from the Bureau of
Labor Statistics. For the 1989-90 group median I use weekly
earnings in 1990 of full time wage and salary workers. For the
earlier group I use the corresponding earnings from 1983, the
earliest year for which comparable data were available. After
adjusting for inflation, the earnings data show that median
earnings in the U.S. population were about 1 percent lower in
1990 than in 1983. Any changes in the earnings of immigrants
can be assessed in this context.

Table 13 shows the mean wages of the two cohorts. Also
shown are the results from four regressions in which all males
aged 22 to 64 who were admitted in the years 1979 to 1980 and
1989 to 1990 are included and the dependent variable is the
logarithm of wages. In each regression a dummy variable is
included which is equal to one if the individual was admitted
in the years 1989 to 1990 and zero otherwise. The coefficient
of this dummy variable measures the extent to which the two
cohorts differ in their labor market quality.

The mean wages are essentially the same for the two
cohorts. Since wages in the U.S. population fell by 1 percent
over the period, this amounts to a very modest rise in
immigrant quality. More important, these means imply that the

decline in.immigrant quality of the 1970's did not continue in

!! !!!! !!!! !! !! !! !!!! J

666.666 u z. 6666 a 66-66 «6666 m m6-66 mwceemwa;mums

96

 

 

m
.6.
666.6 666.6 666.6 666.6 60065666
2666.66 66-66
666.6- - - - 4 6666666
1666.61 66-66.
666.6 - - - 6 qus6_
mmHz=66_

1 mm» 66» oz oz oeonomm

_ 1666.61 .666.66 .666.6c

1 666.6- 666.6- 666.6- - 6666662

_ 1666.66 1666.66 6666.6.

1 666.6 666.6 666.6 - quxm

I .666.66 1666.62 1666.61 1666.65 emomoo

_ 666.6 666.6 666.6 666.6 66-66

H 2666.61 .666.6c 1666.66 1666.61 ,

1 666.6 H66.6 666.6 66666.6- omHmmez,

,

_ 1666666.6V 1666666.6c 1666666.6. “666666.61 omm<oom
6666.6- 6666.6- 6666.6- 6666.6- -664
.6666.6c 16666.65 26666.66 16666.61 .

1 666.6 666.6 666.6 666.6 666

1 1666.65 .666.6c .666.6c .666.6c

, 666.6 666.6 666.6 666.6 samommsza

_ 6 6666: n 6066: N 6066: a 6660: manaeun>

66-666H

use oonohma once» on» as nevus—Gd 66.66 pond mucoumwasH was: .33 0:66: mcofimmmumom and 0.369

97
the 1980's. This is in contrast to the findings in Borjas
(1994).

In the first regression reported in Table 13, Model 1,
age and marital status are controlled for. The modest rise in
immigrant quality is now seen in the 89-90 cohort dummy
variable and is larger than in the case of the simple means.

One possible source of a change in the skill level of
immigrants is an increase in the number admitted on the basis
of their skills. In the data set there is an increase in the
number of these admittees, from 4 percent in 1979 to 1980 to
6.2 percent in 1989 to 1990. Similarly, evidence presented
earlier in this dissertation shows that refugees are
relatively less skilled than other immigrants, so a change in
their proportion could alter earnings. Again, their proportion
did.change over the period, from 13.8 percent in 1979-80 to 12
percent in 1989-90. This increase in the proportion of the
relatively high-skilled and decrease in the proportion of the
relatively low-skilled.may explain the overall skill increase
seen in Model 1. As such, in Model 2 I add dummy variables to
control for class of admission, but the positive and
significant coefficient on the 1989—90 cohort remains.

Another possible explanation for the change in skill
levels is a change in the national origin mix of immigrants.
Of particular note is a fall in the proportion of immigrants
coming from Latin America. In the 1979-80 cohort 38.5 percent
are from Latin America whereas in the 1989-90 cohort 35.4

percent are from this region. Much work on the economics of

98
immigration finds immigrants from Latin America to be less
skilled than those from elsewhere, so a decline in their
proportion might explain the increase in skill levels.

In Model 3 I add dummy variables to control for regions,
with Africa being the omitted region. The coefficient on the
1989-90 cohort dummy stays the same. This indicates that the
change in immigrant quality occurred for reasons other than
changes in the class of admission mix or the national origin
mix.

In model 4 I add interaction terms between the skill and
refugee classes of admission and the 1989-90 cohort. Their
inclusion controls for the possibility that the positive
cohort effect is a result of the later skill-based immigrants
being more skilled than the earlier skill-based immigrants or
the later’ refugees. being ‘more skilled. than the. earlier
refugees. The coefficient on the 1989-90 cohort dummy variable
remains positive and significant after the addition of the
interaction dummy variables. This seems to support the idea
that there was an increase in the general skill levels of
immigrants over the period.

Care must be taken not to read too much into the various
coefficients because of the use of occupational wages and the
increase in inequality' over’ the jperiod. The SKILL*89-90
coefficient merely implies that the occupations held by skill-
based immigrants paid more in the late 1980's. This could be
because the skill-based immigrants became more skilled or

because the returns to having particular skills rose. The

99

reason for including dummy variables for skill and refugee
immigrants and the interactions was to isolate the trend in
non-skill, non-refugee immigrants, i.e., the family
immigrants. Assuming that family immigrant occupational
attainment is either similar to or lower than the U.S.
average, the persistence of a positive 89-90 cohort
coefficient, while controlling for the other categories,
indicates an improvement in the skill level, of family
immigrants over the period.

One additional potential concern in the interpretation of
the coefficients is that any change in skill levels may have
been a result of changes in I.N.S. skill requirements as
opposed to changes in private behavior brought about by
changing economic conditions. In her review of 1980's
immigration legislation Elizabeth Rolph refers to the U.S.
Congress altering the numbers of immigrants admitted under
certain categories but she makes no mention of changing skill
requirements. The most recent piece of legislation to address
the issue of skills was the Immigration Act of 1990. Since the
data set does not include immigrants admitted after its
passage, it seems safe to assume that I.N.S. requirements are
not driving the results.

The result of this comparison is different from the
Borjas result. One possible explanation for the difference is
that the Borjas data includes the immigrants admitted through
the IRCA amnesty program, while my data set does not. The

inclusion of this group of former illegals may explain why

100
Borjas observes a decline in skill levels, whereas I find a
rise.

To explore the changes in the skill levels of immigrants
over the period further, I broke the data set into seven
regions and I performed separate analyses for each. I ran
Models 1, 2 and 4 from Table 13 for each region and the
results are reported in Table 14.

The most striking aspect of Table 14 is that three of the
seven regions show evidence of rising immigrant quality, three
show evidence of falling immigrant quality and one, Eastern
Europe, shows mixed evidence. Without a uniform effect there
is no general support for the argument that the rise in
inequality brought about an increase in the skill levels of
immigrants. However, if specific regions are considered,
namely Western Europe and Canada, some support is found.

The Western Europe cohort effect is the largest and is in
the direction expected. As Western Europeans may have greater
information regarding the U.S. , the positive cohort effect may
be a response to the inequality stimulus noted. What is more,
as Western European income levels are close to those of the
U.S., it seems reasonable that the second.moment of the income
distribution may have more of an influence on migration
decisions. Similar arguments can be made with regard to Canada
whose cohort effect, though not as large as the Western
European effect, is nonetheless positive.

The Eastern Europe case is interesting, as the

coefficient on the 89-90 dummy variable changes sign as the

101

 

 

 

66666666-
6666 66-66 6666 u 66-66 6666 66-66 6666 I 66-66 6666
666.6 666.6 666.6 666.66 666.66 666.66 2
666.6 666.6 666.6 666.6 666.6 666.6 6.6 6366666
6666.66 1666.66 66-66
666.6- - - 666.6 - - 6 6666666
6666.6. 1666.66 66-66
666.6 - - 666.6 - - 6 66666
1666.66 6666.66 .666.66 1666.66
666.6 666.6- - 666.6- 666.6- - 6666666
1666.66 6666.66 1666.66 .666.66
666.6 666.6 - 666.6 666.6 - 66666
“666.66 1666.66 2666.66 6666.66 .666.66 “666.66
666.6- 666.6- 666.6- 666.6 666.6 666.6 66-66
1666.66 .666.66 1666.66 1666.66 .666.66 6666.66
666.6 666.6 666.6 666.6- 666.6- 666.6- 666666:
666666.66 666666.66 .66666.6v 666666.66 666666.66 666666.66
6666.6- 6666.6- 6666.6- 6666.6- 6666.6- 6666.6- 6.66666
1666.66 6666.66 1666.66 1666.66 .666.66 “666.66
666.6 666.6 666.6 666.6 666.6 666.6 666
6666.66 .666.6. “666.66 6666.66 1666.66 .666.6.
666.6 666.6 666.6 666.6 666.6 666.6 966666626
MHGOOOO wflwcflu minnow-66>

o6-6666 666 66166.66 6.660»
we» :6 oouu6664 66166 6664 6666666666 666: so mcowoom oduoz :0666 now nco6mmouoomnuvn OHOOB

102

66666660

 

 

6666 n 66-66 6666 u 66-66 6666 u 66-66 6666 u 66-66 666:
666.666 666.666 666.666 666.666 666.666 666.666 2
666.6 666.6 666.6 666.6 666.6 666.6 66 66966666
1666.66 1666.66 66-66
666.6- - - 666.6- - - 6 66666
1666.66 6666.66 66-66
666.6 - - 666.6 - - 4 66666
6666.6. 6666.66 .666.66 6666.66
666.6 666.6- - 666.6- 666.6- - 6666666
.666.66 6666.66 1666.6. 6666.66
666.6 666.6 - 666.6 666.6 - 66666
6666.66 6666.66 .666.66 6666.66 6666.66 .666.66
666.6- 666.6- 666.6- 666.6 666.6 666.6 66-66
1666.66 .666.66 .666.66 1666.6. 6666.66 .666.6.
666.6- 666.6- 666.6- 666.6 666.6 666.6 666666:
6666666.66 6666666.66 “666666.66 .666666.6. 6666666.66 .666666.6c
6666.6- 6666.6- 6666.6- 6666.6- 6666.6- 6666.6- 6.66664
66666.66 16666.66 16666.66 66666.6. 66666.66 .6666.66
666.6 666.6 666.6 666.6 666.6 666.6 666
.666.66 6666.66 1666.66 1666.66 6666.6. 1666.6.
666.6 666.6 666.6 666.6 666.6 666.6 666666626
6066664 c6666

 

6664

 

6626666>

“66.66606 66 06666

 

 

 

 

 

103

 

 

 

 

 

 

DIVE }
H! !l !“l!
66666666
6666 u 66-66 6666 u 66-66 6666 a 66-66 6666 a 66-66 :66:
666.66 666.66 666.66 666.66 666.66 666.66 2
666.6 666.6 666.6 666.6 666.6 666.6 66 66666666
.666.66 1666.66 66-66
666.6- - - 666.6- - - 6 6666666
6666.66 6666.66 66-66
666.6- - - 666.6 - - 6 66666
1666.66 6666.66 6666.66 .666.6.
666.6- 666.6- - 666.6 666.6- - 6666666
1666.66 1666.66 6666.66 .666.66
666.6 666.6 - 666.6 666.6 - 66666
1666.66 6666.66 6666.66 6666.66 .666.66 6666.66
666.6 666.6 666.6 666.6 666.6 666.6- 66-66
1666.66 6666.66 .666.66 6666.66 .666.66 6666.66
666.6 666.6 666.6- 666.6 666.6 666.6 666666:
666666.66 .66666.6v 666666.66 666666.66 166666.66 666666.66
6666.6- 6666.6- 6666.6- 6666.6- 6666.6- 6666.6- 6.66666
6666.66 6666.66 .666.66 6666.66 1666.66 .666.6.
666.6 666.6 666.6 666.6 666.6 666.6 666
1666.66 6666.66 6666.6. “666.6. .666.66 .666.66
666.6 666.6 666.6 666.6 666.6 666.6 966666626
moousm snoumoz onousm cumumwn 66O6666>

6.6.66606 66 66666

 

104

man» a omlmo

nNm6mH

NNH.O

“666.66
666.6

6666.6.
666.6
6666.66
666.6:
6666.66
666.6
6666.66
666.6-
6666.66
666.6:
.66666.66
6666.6:

.666.66
666.6

6666.66
666.6

nNm.mH

OHH.O

.666.66
666.6:

“666.66
666.6
6666.66
666.6:
“666.66
666.6:
666666.66
6666.6:

.666.66
666.6

6666.66
666.6

«afim I omlmh

nNm.mH

Ono.o

.666.66
666.6:

.666.66
666.6:
“66666.6.
6666.6:

.666.6.
666.6

6666.66
666.6

66:6:666
:60:

z

66
66666666

omnmw
¢ mmwbhmm

omlmm
« HAme
fimwbhmm
AAHKm
omlmm

ONHmmdz

D.OMM04

66¢

BANUKWBZH

 

606666

 

 

0636666>

”66.66666 66 66666

105

class of admission and interaction dummy variables are added.
One possible explanation for the sign change between the first
and second models is that the 1989-90 cohort contained more
refugees as movement out of this area became easier as
Communism weakened. Without controlling for this the 1989-90
cohort looks less skilled. Once class of admission is
controlled for the positive cohort effectemerges. However,
the positive effect is even stronger when the interaction
terms are added. The REFUGEE*89-90 coefficient shows that the
negative cohort effect in the simplest model is a result of
the refugees in the 1989-90 group being particularly low-
skilled. This indicates that it was not so much that more
refugees were escaping, but that a different group of refugees
was escaping. The positive coefficient on REFUGEE in the third
model, when combined with the negative coefficient on
REFUGEE*89-90, indicates that this particular flow took on a
different character over this period.

Without data on relative inequality between the U.S. and
the regions considered, the task of establishing a link
between rising U.S. inequality and immigrant quality is a
difficult one. An important result has, however, emerged from
the analysis. The continued decline in immigrant quality found
in Borjas (1994) is not supported by these data.

Section 4: Summary

In this chapter I have repeated the analysis of Chapter

2 using data for a later cohort. The outcome is continued

support from cross-section regressions for the Borj as model in

106

that higher levels of income inequality tend to be associated
with larger differences between the skill levels of skill-
based and family-based immigrants. The continued support
emerged even though the range of occupation categories
reported fell from over 400 in the earlier data set to just 25
in the later set. The fixed effects model did not give support
to the Borjas model but this may be because of the small
sample of countries.

The:analysis comparing the 1979-80 and 1989-90 cohorts is
less conclusive. Although a modest increase in the skill level
of immigrants over the period was detected in the sample of
immigrants as a whole, the experience across regions was not
uniform. The regions that were most likely to respond to
changes in inequality did show positive cohort effects. This
gives some support to the idea that immigrants do respond in
a predictable way to changes in income inequality in the

source and receiving countries.

Chapter 5
Summary and Conclusions

In this concluding chapter I first review what has been
done in this dissertation. I next determine what has been
learned, and then draw some conclusions regarding policy.
Finally, I suggest some ideas for research that arise from
this dissertation.
Section 1: A Review of‘What Has Been Done in the Dissertation

This dissertation has been a theoretical and empirical
analysis of the differences in the labor-market quality of
different immigrant groups, designed to determine the nature
of the immigrant flows from different countries and how these
flows are affected by various factors. The first two groups I
compared were skill-based and family-based immigrants (Chapter
2 and Chapter 4, section 1). The idea behind examining these
two groups was that a smaller difference in labor-market
quality between the two for a given country would mean that
the family-based immigrants from that country were relatively
skilled. To the extent that the family-based immigrants
reflect the general characteristics of those in a country who
find it optimal to migrate, this skill-based/family-based
difference measures the general nature of the immigrant flow
from a country. An alternative view of the skill-based versus
family-based immigrant difference is that it shows the extent
to which a policy shift towards more skill requirements in
immigration law would alter the nature of the immigrant flow

from various countries.

107

108

The exercise in these chapters goes beyond simply seeing
which countries send which types of immigrants. It is also
designed to discover which types of countries send which types
of immigrants. As the model that I used to predict the
relevant country characteristics is based on the work of
Borjas, the tests in Chapters 2 and 4 can also be viewed as a
test of the Borjas model.

The next two groups of immigrants I compared were family-
based immigrants and those who gained permanent residency
through winning a greencard in one of the lotteries (Chapter
3). The model showed how the migration decisions of the two
groups differed. According to the model, family-based
immigrants were migrating to join family members, so no
pecuniary compensation for the loss of family was required.
Lottery-winning immigrants, however, were leaving family and
thus required compensation in order for them to experience a
utility gain from immigration. This greater need for financial
compensation by the lottery immigrants meant that they would
be selected from the group in the population with most to gain
financially from immigration. If the lottery immigrants were
more skilled than the family immigrants, those at the upper
end of the income distribution must have the most to gain.
Likewise, if the lottery immigrants were less skilled, those
at the lower end must have the most to gain. Little support
was found for this model but an alternative explanation for
the results in this chapter was given. This alternative

explanation was based on the notion of family as a provider of

 

109
an employment network.

I extended the comparison between the lottery and family-
based immigrants to consider’the location decisions of the two
groups. The model suggests that the lottery immigrants‘will be
more concentrated close to large numbers of compatriots.
Compatriots may act as a substitute for the loss of family,
which is a feature of the lottery immigrant experience
relative to the family-based immigrant experience.

The final two groups I compared were those admitted in
the years 1979 to 1980 and 1989 to 1990 (Chapter 4, section
2). The reason for examining these was to see how a change in
income inequality in a host country, i.e., the U.S., would
affect the nature of the immigrant flow into that country. As
inequality in 'the host country rises, the incentive ‘to
immigrate for those who would be at the upper end of the U.S.
income distribution rises. At the same time, the incentive to
immigrate is reduced for those at the lower end. Overall, the
rise in inequality should lead to an increase in the quality
of the immigrant flow. Investigating the changes in the skill
levels of immigrants over the 1980's is also of interest in
itself. The observed decline in immigrant labor market quality
from the 1960's until the early 1980's has been a major theme
of the literature in this area. It is therefore of interest to
see if this trend has continued.

Section 2: A Review of the Results

Five results can be isolated.

1) The wage premium of skill-based immigrants relative to

 

110

family-based immigrants varies across countries. This was
found to be true using data from the years 1978 to 1980 and
1988 to 1990. For some countries the skill difference between
the family—based and skill-based immigrants is relatively
small, indicating that the family-based immigrants are
reasonably skilled. For such countries, the imposition of
skill requirements for'a broader group of immigrants would not
significantly alter the nature of the immigrant flow. For
other countries there is a big difference between skill-based
and family-based immigrants, so additional skill requirements
would substantially alter the average character of the
immigrant flow.

2) The level of income inequality in a country of origin
affects the labor-market quality of those who choose to
migrate. Again using data from the periods 1978 to 1980 and
1988 to 1990, the level of home country inequality helped
explain the variation across countries in the labor-market
quality of skill-based relative to family-based immigrants.
This finding confirms the Borjas model (5.5.3. 1987) but may
also be due to the fact that inequality of income in a country
reflects an inequality of skills, rather than of the returns
to skills. The results on the effect of country of origin GNP
were contradictory between the two periods examined. The more
believable result, however, is that higher levels of GNP are
associated with smaller differences between skill- and family-
based immigrants. The level of English language ability in a

population also influences the quality of immigrants coming

 

 

111
from a country. Distance from the United States influences
quality also.

3) In general, lottery immigrants are more skilled than
family-based immigrants. The model suggested that home-country
inequality would help predict the relative quality of lottery
and family-based immigrants but this was not so. A possible
explanation of the greater skill levels of the lottery group
relative to the family group is that the family acts as a
provider of employment or information about employment for
family-based immigrants. This allows for the immigration of
low-skilled family members. The lottery immigrants do not have
this facility available to them, so the low-skilled may be
eliminated from the pool of potential lottery immigrants.

4) Relative to family-based immigrants, lottery
immigrants are more geographically concentrated close to large
numbers of compatriots. This is true at least for the Irish
and Polish immigrant groups. Further examination of the Irish
lottery immigrants reveals that this greater concentration of
the lottery immigrants is driven by older lottery immigrants.
This must mean that the need for the presence of compatriots
is stronger for older immigrants. This in turn could be
because of the utility derived from compatriots or because of
employment services provided by compatriots.

5) The data show a modest increase in the skill levels of
immigrants over the 1980's. This is true even when changes in
the mix of regions and changes in the mix of classes of

admission are controlled for. The importance of this finding

112

is that it is contrary to the recent trend in immigrant
quality and contrary also to the findings in Borjas (1994).
The evidence is not clear that this observed increase in skill
is a result of the increase in inequality in the U.S., since
the experience is not uniform across regions. The region that
would be most expected to show a reaction to the change in
U.S. inequality, Western Europe, does, however, show the
predicted movement in quality. This is some support for the
prediction about the impact of inequality.

Section 3: Policy Conclusions

One of the arguments against allowing additional
immigration is that immigrants do not possess characteristics
that will lead them to make positive contributions to the U.S.
economy. Such a concern was not much in evidence in the
structure of immigration law until very recently. As described
in Chapter 2, only a small proportion of greencards were
issued on the basis of labor-market characteristics. During
the 1980's, when almost 750,000 immigrants were being admitted
each year, only 27,000 visas were set aside for the highly
skilled. This was increased to 90,000 in the Immigration Act
of 1990, so policy has partially caught up with public
concerns.

A problem with this Act is that while it increases
opportunities for the highly skilled to immigrate, it does
little to discourage the immigration of the low-skilled or no-
skilled. The evidence presented here shows that the countries

which are the sources of these low skilled immigrants can be

 

9..

113

identified. As such, it would be possible either to restrict
immigration from these countries or to impose skill
requirements for immigrants from them. These skill
requirements need.not.be as stringent as the requirements that
exist for those who are currently admitted on the basis of
their skills but the law could impose some minimum
requirement.

Proposals to treat different national groups differently
may be politically unacceptable currently, making the proposal
to restrict immigration from certain countries unworkable. The
skill requirement approach could instead be imposed on all
countries. The evidence presented in Chapters 2 and 4 shows
that for some countries additional skill requirements would
not alter the quality of the immigrant flow and would not act
as a great impediment to immigration, since such countries
send relatively highly skilled immigrants. The real impact of
skill requirements would be felt by those countries which send
low-skilled immigrants. This would be an example of a policy
applying to all but affecting the few at whom it is really
targeted.

The argument that restricting immigration from some
countries and not from others is discriminatory is not
entirely fair. Were the U.S. to restrict immigration from a
particular country relative to other countries, it would not
be saying that the people of that country were inferior. All
that the U.S. would be saying is that the people of a country

who find it optimal to immigrate are not the ones who would.be

114

making a positive contribution to the U.S. This argument says
nothing about the character of the people from a particular
country in general. It merely says that relative conditions
between the U.S. and this other country lead to the
immigration of a particular sample. Restricting immigration
from a country may make rational economic sense and should not
necessarily be seen as discriminatory.

If immigration from certain countries is not restricted
and no skill requirement are imposed, the U.S. will either
have to live with an influx of unskilled immigrants or'develop
an immigration policy that extends to helping those who are
here. The cost of restricted immigration or skill requirements
is the dilution of the humanitarian goals of immigration
policy. The cost of helping those here is the tax dollars
required to set up training and education programs, etc. The
cost of doing nothing will be difficult to estimate, but
ideally the costs and benefits of each approach should be
evaluated so that a sound policy can be pursued.

The goal of the lottery programs was to add diversity to
the immigrant pool and specifically to generate additional
immigration from countries that had been underrepresented in
recent immigration waves. The change in the national origin
mix of immigrants brought about the decline in the skill level
of immigrants. The lottery programs, by altering the national
origin mix, may help to raise the skill levels of immigrants.
As was discovered in Chapter 3, however, the lottery programs

also have an effect on immigrant quality by encouraging the

 

115
immigration of people who are more skilled than the family
immigrants from the same country. These two positive effects
of the lottery programs should be weighed against the negative
one of the greater concentration of the lottery immigrants in
a few locations.
Section 4: Ideas for Further Research

The results on the relative labor market quality of
immigrants admitted under different classes of admission have
emerged from a‘data set that is quite limited. The data do not
have wage rates for each immigrant, nor do they have
information on education or experience. While data sets taken
from the Census and the C.P.S. have more detailed information
on immigrants, these data sets do not indicate the class of
admission under which the immigrant was admitted. A data set
that included both class of admission and detailed individual
information is needed to allow a much deeper investigation of
the sort of issues explored here.

More detailed data on the lottery immigrants could be
particularly valuable. This group represents a random sample
of those in a country who want to immigrate. This is not
necessarily true of those admitted under other classes of
admission. Questions regarding positive or negative selection
can be answered better using the lottery immigrants.

Even with the existing data advances can be made in this
area. Since the occupational distribution of the lottery
immigrants is available, it should be possible to compare the

distributions of different national lottery groups with the

 

116
occupational distribution in each home country. This would act
as a direct test of whether those who wish to immigrate are
selected from the upper or lower part of the national income
distribution.

The provision of the Immigration Act of 1990 which
increased the number of immigrants granted admission based on
high skill levels was intended to increase the average skill
level of immigrants. The approach used in this dissertation
could be used to examine the effect of this provision.

This dissertation began with a reference to the argument
against immigration that views immigrants as lacking labor-
market skills. The evidence presented here shows that such an
argument is overly simplistic. Immigrant labor-market quality
varies along a range of dimensions. The findings of this
dissertation, along with the findings of other work in this
area, should be used to inform the debate, so policy based on

fact can emerge.

APPEN DI CES

117

Appendix A

Sources of Country Specific Data Used in the Regressions
Presented in Table 4

Politically competitive system, recent loss of freedom and
number of assassinations: Cross-National Time-Series Archive
(CNTSA) , a data set created by Arthur Banks and available-
through the Inter-University Consortium for Political and

Social Research, Ann Arbor, MI 48106.

 

Income inequality: World Bank. MW.

various issues and United Nations, o 'um

£3321§212§. 1977-

Log per capita GNP: U.S. Arms Control and Disarmament Agency,
- 9 , ',ae A"!!. _ ‘= i!! :eu‘ a! ‘ - ° - ’1 ,
U.S.G.P.0. 1984.

Distance: Airline offices contacted by George Borjas

English proficiency: 5/100 sample of the 1980 Census

118

Appendix D

A Further Explanation of the Relationship between Mean Income
in the Country of Origin and Selection

On page 535 of his ALL paper Borjas presents the
derivative of 0,, the income differential between the average
immigrant and the average native person in the United States,
with respect to uo, mean income in the home country. The
derivative is:

dQ1/duo = moo/of (k - p) 6L/6z.

If k - p is negative the country is a negative selection
country (see the condition derived in section 1 above) and
this derivative is negative. Similarly, if k - p is positive
the country is a positive selection country (again, see the
conditions in section 1) and the derivative is positive.

In order to understand the intuition behind this result,
consider a negative selection country. The rise in mean income
reduces the rate of immigration. Those that no longer find it
optimal to migrate are those which were the most skilled of
the immigrant pool from this country. Hence, the average
quality of the pool falls. In the case of a positive
selection, the increase in mean income also reduces the rate
of immigration but this time it is the least skilled of the
immigrant pool who no longer find it optimal to migrate. Hence

the rise in average quality.

 

119

Appendix C

Rising Inequality and Rising Immigrant Quality

The following is an explanation of the comparative static
result from that the Borjas model (1987b) which shows how a
rise in inequality in a home country can lead to a rise in
immigrant quality.

Relatively high levels of inequality will produce
negative selection and relatively low levels will produce
positive selection. Furthermore, the composition effect of a
rise in inequality will always be to reduce immigrant quality.

The scale effect of a rise in inequality can produce a
complication. If average income in the U.S. is less than the
sum of average income in the home country plus the cost of
migration, the scale effect of a rise in inequality will be to
increase immigrant quality. For a rise in inequality to
produce an overall rise in immigrant quality the relative mean
income condition would have to hold and the scale effect would

have to be bigger than the composition effect.

 

LIST OF REFERENCES

LIST OF REFERENCES

1. Bartel, Ann and Roch, Marianne, "Internal Migration of U.S.
Immigrants", in Abowd, John and Freeman, Richard (eds.)

WW 1991-

2. Borjas, George, "Assimilation, Changes in Cohort Quality,
and the Earnings of Immigrants",
vol. 3 no. 4 (1985).

3. Borjas, George (1987a), "Immigrants, Minorities, and Labor
Market Competition" , ' t
vol. 40 no. 3 (April 1987).

4. Borjas, George (1987b) , "Self-Selection and the 4Earnings of

Immigrants" W vol 77 no. 4(September
1987).

5. Borjas, George, °
Immigrant§_gn_the_ulfil_fisgngmx 199°-

6. Borjas, George, "Immigration and Self-Selection", in Abowd
and Freeman (eds.) 1991.

7. Borjas, George, "What Happened to Immigrant Skills in the
1980's?" Himeograph March 1994

8. Borjas, George, Freeman, Richard and Rats, Lawrence, ”On
the Labor Market Effects of Immigration and Trade" in Borjas,

George and Freeman, Richard (eds. ) Immigratign__and__§hg

. ' 01011 91-‘9L‘! = 0‘ 1 ‘0. _ -. “ #9!

fiQEIE__AI§2§ 1992

9. Sorj as, George and Trejo, Stephen, "Immigrant Participation

in the Welfare System". BMW
vol. 44 no. 2 (January 1991).

10. Borjas, George and Trejo, Stephen, "National Origin and

Immigrant Welfare Recipiency". MW
vol. 50 no. 3 (March 1993).

11. Bureau of Labor Statistics, WW
January 1990.

12. Carliner, Geoffrey, "Wages, Earnings and Hours of First,

Second and.Third.Generation.Americanbnales", Eggngmig_1nggizy
vol. 18 (January 1980).

120

121

13. Chiswick, Barry, "The Effect of Americanization on the
Earnings of Foreign-born Men", 0 o 't o
vol. 86 no. 51 (1978).

14. Funkhouser, Edward and Trejo, Stephen, "The Labor Market
Skills of Recent Male Immigrants: Evidence from the Current
Population Surveys", Mimeograph, October 1993.

15. House of Representatives, Begg;;_1gg;1ggg.

16. Jasso, Guillermina and Rosenzweig, Mark, "What's in a
Name? Country-of-origin Influences on the Earnings of

Immigrants in the United states". Beeear2h_in_flnman_£asital

so! u—- eon-I .H.’ 1 0! y-ue: a! a at! D‘ ‘ oouf: V01.

4 (1986).

17. Rats, Lawrence 1and. Loveman, Gary, "An International
Comparison of Changes in the Structure of Wages: France, the
United Kingdom and the United States, Mimeograph 1990

18. LaLonde, Robert and Topel, Robert, "Immigrants in the
American Labor Market: Quality, Assimilation and

Distributional Effects". AIEIA1_£aner§_and_2roseedins§ (May

1991).

19. McManus, Walter, Gould, William and Welsh, Finis,
”Earnings of Hispanic Men: The Role of English Language

Proficiency". J2nrnal_of_Labor;E22n2miQ§‘vol- 1 (April 1983)-

20. Roplh, Elisabeth,‘ -
129Q_s_and_Issne§_for_the_1222_s Rand 1992-

21. Schwartz, Abe, "Interpreting the Effect of Distance on
Migration". W vol- 81 no. 5
(September/October 1973).

22. Topel, Robert, "Regional Labor Markets and the
Determinants of Wage Inequality" , W
Broseedinga May 1994

21. World Bank, Wg:1§_pgyglgpmgnt_3gpgrt, various issues.

CHIGRN STRTE UNIV LI

1111112 91191111191 11 1991111911111911119119111