NBER WORKING PAPER SERIES

IS THE NEW IMMIGRATION

REALLY SO BAD?

David Card

Working Paper 11547

http://www.nber.org/papers/w11547

NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue
Cambridge, MA 02138
August 2005

I am grateful to Christian Dustmann, Thomas Lemieux, Ethan Lewis, Stephen Machin, and two anonymous
referees for helpful suggestions, and to Florence Neymotin for outstanding research assistance. Partial
funding for this work was provided by the NICHD. The views expressed herein are those of the author(s)
and do not necessarily reflect the views of the National Bureau of Economic Research.

©2005 by David Card. All rights reserved. Short sections of text, not to exceed two paragraphs, may be
quoted without explicit permission provided that full credit, including © notice, is given to the source.
Is the New Immigration Really So Bad?
David Card
NBER Working Paper No. 11547
August 2005
JEL No. J61

ABSTRACT

This paper reviews the recent evidence on U.S. immigration, focusing on two key questions: (1)

Does immigration reduce the labor market opportunities of less-skilled natives? (2) Have

immigrants who arrived after the 1965 Immigration Reform Act successfully assimilated? Looking

across major cities, differential immigrant inflows are strongly correlated with the relative supply

of high school dropouts. Nevertheless, data from the 2000 Census shows that relative wages of

native dropouts are uncorrelated with the relative supply of less-educated workers, as they were in

earlier years. At the aggregate level, the wage gap between dropouts and high school graduates has

remained nearly constant since 1980, despite supply pressure from immigration and the rise of other

education-related wage gaps. Overall, evidence that immigrants have harmed the opportunities of

less educated natives is scant. On the question of assimilation, the success of the U.S.-born children

of immigrants is a key yardstick. By this metric, post-1965 immigrants are doing reasonably well:

second generation sons and daughters have higher education and wages than the children of natives.

Even children of the least- educated immigrant origin groups have closed most of the education gap

with the children of natives.

David Card
Department of Economics
549 Evans Hall, #3880
UC Berkeley
Berkeley, CA 94720-3880
and NBER
[email protected]
 Over the past two decades economists’ perceptions of U.S. immigrants have shifted. In the

1970s, immigrants were viewed in a mainly positive light. Chiswick (1978) found that immigrant

men earned as much as natives, despite having less education, and concluded that investments in on-

the-job training made up for the gap in formal schooling. Grossman (1982) examined the impact of

immigration on native wages and concluded that the effects were small. Subsequent research – most

notably by Borjas (1985, 1995, 1999, 2003) – has chipped away at both conclusions and gradually led

to a more negative picture of U.S. immigration. The shift in perceptions has closely tracked changes

in the national origin of U.S. immigrants, often attributed to the 1965 Immigration Reform Act, and

a widening gap between the language and culture of natives and immigrants (Borjas, 1999; Lazear,

1999). Concerns over immigration have also been heightened by the decline in low-skilled wages in

the U.S., and the belief that some of this may be due to immigrant competition (Borjas, Freeman,

and Katz, 1997).

This paper presents an overview and update of the U.S. immigration literature, focusing on

two central questions: (1) Do immigrants harm the labor market opportunities of less skilled natives?

(2) How do today’s immigrants perform in the U.S. labor market, and are they successfully

“assimilating”? These questions are at the heart of the debate about immigration in many other

countries – including most European nations – and insights from the recent U.S. literature may

prove useful in answering the questions elsewhere. My conclusion is that the “revisionist” view of

recent U.S. immigration is overly pessimistic. The evidence that immigrants harm native

opportunities is slight, while the fear that post-1965 immigrants will never assimilate is belied by the

rather surprising educational success of their children.

I. The Characteristics of Immigrants

Most of the immigrants in Chiswick’s (1978) landmark study had entered the U.S. under the
provisions of the Immigrant and Nationality Acts of 1924, which established national origin quotas

with a strong bias in favor of Northern Europeans.1 In the 1970 data analyzed by Chiswick, 63

percent of immigrants were born in Europe or Canada (Card, DiNardo and Estes, 2000, Table 6.3).

The vast majority of working age immigrants in the U.S. today arrived after the 1965 Immigration

Act, which relaxed the quota system and established preferences for people with family members

already in the country. The new law, coupled with declining supplies of potential immigrants from

traditional source countries such as Britain, Germany, and Italy, and increasing potential supplies

from Mexico, Central America, and Asia, have led to a shift in the ethnic composition of

immigrants.2 In 2000, only 13.6 percent of adult immigrants in the U.S. were born in Europe, while

32 percent were born in Mexico, 16 percent in Central America or the Carribean, and 26.6 percent in

Asia.

As emphasized by Borjas (1985, 1995), the skill characteristics of immigrants in the U.S. are

strongly related to their country of origin. For example, country of origin dummies explain 30

percent of the variation in average education levels among immigrants in the 2000 Census.3

Reflecting the high fraction of immigrants from countries like Mexico, the Philippines, Vietnam, and

El Salvador (all of which supply relatively low-education immigrants) immigrants as a whole have

1
The law was influenced by research of Carl Brigham (1923), who classified immigrants into four
racial categories: “Nordic”, “Alpine”, “Mediterranean” and “Asian”, and argued that members of the Alpine
and Mediterranean races had lower intelligence than Nordics. The influence of Brigham’s work is illustrated
by a headline announcing the new law in the Los Angeles Times (13 April 1924): “Nordic Victory is seen in
Drastic Reduction.”
2
The 2000 U.S. Census has information on exact arrival year, and I used this information to examine
changes in the fraction of immigrants from different countries before and after 1965. The fraction of
Mexican immigrants, for example, is 20.1 percent for 1963-65 arrivals and 20.1 for 1966-68 arrivals. A sharp
impact of the law is not discernable in these data, though there is a trend between 1950 and 1975.
3
Education levels of immigrants are correlated with education levels in the home country, but there
are many interesting exceptions. For example, immigrants from India have the highest average education
(average of 15.6 years of completed schooling). Immigrants from Russia are a very close second.

2
lower average schooling than natives. Table 1 compares the education distributions of natives and

two subgroups of immigrants - those who had been in the U.S. at least 5 years at the time of the

2000 Census, and those who had arrived more recently. The recent arrival group is not very

different from the earlier arrivals, reflecting the relative stability of immigrant inflow composition

since the late 1970s. Nevertheless, both groups have a much higher fraction of people with very low

schooling than natives. The excess concentration of immigrants in the “less than high school”

category is balanced by relative shortfalls in the number who completed high school but have no

further formal schooling (24 percent of immigrants versus 39 percent of natives) and the number

with 1-3 years of college (about 15 percent of immigrants versus 24 percent of natives). At the

upper end of the education distribution immigrants and natives are very similar, though immigrants

are slightly more likely than natives to hold an advanced degree.

The patterns in Table 1 point to two important conclusions. First, labor market competition

from immigrants is most intense for natives with the lowest levels of education. While immigrants

comprised only 13 percent of the working age population in 2000, they made up 28 percent of the

population with less than a high school diploma, and over half of all those with less than 8 years of

schooling. For this reason, most studies of immigrant competition have focused on the impacts on

very low skilled natives.4 Second, the positive effect of immigrants on the relative supply of the

people with the lowest levels of education is offset by negative effects on the relative supply of

people in the middle of the education distribution, with no effect on the relative supply of those

with a bachelor’s degree or higher. Arguably, then, immigrant inflows have exerted upward pressure

on the wage gap between high school graduates and dropouts, and downward pressure on the

4
If immigrants had the same education and other skill characteristics as natives, and if capital is
elastically supplied to the relevant labor market, then standard economic models would predict no impact on
native wages – see Altonji and Card (1991).

3
college-high school wage gap.

II. Immigrant Competition and the Labor Market Outcomes of Low Skilled Natives

a. Conceptual Issues

There are two main approaches in the literature to estimating the impact of immigration on

native workers. The first – pioneered by Grossman (1982) – relates differences in the relative

structure of wages in different local labor markets to differences in the relative supply of

immigrants.5 The advantage of this approach is that there are many local labor markets in the U.S.

with different fractions of immigrants, and samples from the Decennial Censuses can be used to

estimate relatively rich models of the local wage structure. The disadvantage is that cities are not

isolated economies: people, goods, and services all flow between cities, and depending on how

sensitive these flows are to differences in local wages or prices, comparisons across cities may reveal

a lot or little about the underlying parameters that theoretically determine the effects of immigration

on native opportunities. The second approach is a time series methodology, relating changes over

time in immigrant densities to economy-wide measures of relative labor market outcomes. The

advantage of this approach is that it can potentially reveal the impact of immigration even when the

local markets approach “fails” because of intercity factor mobility or trade (Borjas, Freeman, and

Katz, 1996).6 The disadvantage is the absence of a clear counterfactual. Inferences from the macro

time series approach rely on assumptions about the trends in factors like the degree of skill bias in

5
This approach is closely related to work on internal migration and local wage structures, including
Sjaastad (1962), Topel (1986), and Dahl (2002).
6
Of course the same arguments about intercity trade and factor mobility also apply across countries.
Models of international trade often imply that relative wages in a country are independent of the relative
supplies of different skill groups, at least in some range. See Kuhn and Wooten (1991).

4
recent technological change.

Early studies using the local labor markets approach (Grossman, 1982; Borjas, 1987; Altonji

and Card, 1991; Lalonde and Topel, 1991) treated “immigrants” as one type of labor and

distinguished between various subgroups of natives in the same city. While simple and intuitively

appealing, there are two obvious problems with this framework. One is immigrant heterogeneity: in

some cities immigrants are actually more highly skilled than natives, whereas in others the reverse is

true.7 Thus, it is important to classify the immigrant populations in different cities according to their

skill levels. A second problem is that conventional economic models imply that immigrants should

only affect relative wages to the extent that they distort the relative supplies of different skill groups.

If inflows of unskilled immigrants cause unskilled natives to move out, for example, there may be

little discernable effect of immigration on the local wage structure, even though relative demand

curves at the local level are downward sloping.

A potentially better way to model the impact of immigration is to assign immigrants and

natives to skill groups and to assume that within skill groups, immigrants and natives are perfect

substitutes (e.g., Jaeger, 1996; Card, 2001; ).8 Following this approach, the first step in evaluating the

impact of immigration is to assess the effect of immigrants on the relative supplies of different skill

groups in different cities. The second step is to then relate the relative wages for different skill

groups to the relative supplies in the local labor market. The maintained assumption – that

immigrants and natives are perfect substitutes within skill groups – can be tested by examining the

7
Card (2001, page 23) notes that immigrant men earned more than native men in one third of the
largest U.S. cities in 1990.
8
An alternative approach is to assume that workers with different characteristics sell “bundles” of
skills, where the number of latent skills is small. This approach has been suggested to study the structure of
wages by age and education (e.g., Welch, 1969) but becomes complex once allowance is made for non-linear
pricing of the bundles (Heckman and Scheinkman, 1987).

5
stability of immigrant-native wage differences across different labor markets.9

b. Effects of Immigration on Relative Supply of Low-Skilled Labor

Some indication of the impact of immigration on the relative size of the dropout labor force

in different cities is presented in Table 2. The table shows the fractions of immigrants in all cities

and in 15 selected cities in 1980 and 2000, along with the fractions of immigrants and natives with

less than 12 years of completed schooling, and the overall fractions of the working age population in

each city with less than 12 years of schooling. The data for all cities in the first row of the table

reveals three interesting facts.10 First, the fraction of immigrants in U.S. cities has roughly doubled

since 1980, from 9.5 percent to 18 percent. Second, in both 1980 and 2000, slightly more than one-

third of immigrants had less than a high school education. Third, the fraction of natives with less

than a high school education has fallen sharply, more than offsetting the inflow of less-educated

immigrants.11 Thus, despite the upward pressure on the relative supply of dropouts caused by

immigration, the overall fraction of dropouts in urban areas fell from 24.3% in 1980 to 17.7% in

2000.

There is a lot of cross-city variation in these patterns, however. The dramatic increase in the

fraction of low-education immigrants in Los Angeles, for example, led to a rise in the fraction of high

school dropouts in the local population. In Pittsburgh and Cleveland, on the other hand, immigrant

9
Under the perfect substitutes assumption, for example, the wage gap between immigrants with less
than 12 years of schooling and native high school dropouts should be constant (controlling for age, time in
the U.S., etc.) Even if the perfect substitutes assumption is true, the wage gap could vary across cities if
immigrants in different cities possess different unobserved skills.
10
The set of all cities includes 272 Standard Metropolitican Statistical Areas in 1980 and 325
Metropolitan Statistical Areas in 2000.
11
This is largely a cohort effect, reflecting the steady rise across cohorts in the fraction of high school
graduates until cohorts born in the 1950s. See Card and Lemieux (2000, 2001).

6
densities are low and have fallen over the past two decades, so the the trend in the overall fraction of

dropouts closely parallels the trend among natives. Most high-immigration cities, including New

York, Houston, San Francisco, and Miami, experienced relatively small declines in the fraction of

dropouts between 1980 and 2000, whereas most low-immigration cities, including Philadelphia,

Detroit, and Atlanta, experienced bigger reductions.

The question of whether inflows of unskilled immigration have systematically affected the

relative supply of dropout labor in different cities is addressed in Figure 1. As motivation for this

figure, note that the share of dropouts in the local working age population in city c, sd(c), is the sum

of the share of native dropouts sdN(c) and the share of immigrant dropouts, sdI(c). An interesting

descriptive regression relates the overall dropout share in a city to the share of immigrant dropouts:

sd(c) = " + $ sdI(c) + e(c),

where e(c) is a residual. If inflows of less educated immigrants are offset by outflows of native

dropouts (or if less educated immigrants tend to move to cities where there is a bigger positive trend

in the educational attainment of the native population), immigration will have little impact on the

overall dropout share and the coefficient $ will be close to 0. If mobility flows of native dropouts

(and trends in native educational attainment) are uncorrelated with the inflow rate of low skilled

immigrants, the coefficient $ will be close to 1.

As suggested by the data in Table 2, the scatter of points in Figure 1 is more consistent with

a value of $=1 than $=0. For reference, the graph shows a restricted regression line in which the

slope is set to 1. This benchmark provides a reasonable fit, although it tends to under-predict the

fraction of dropouts in cities with few immigrant dropouts. In fact, a univariate regression across

325 cities yields an estimate of $ equal to 0.79 (with a standard error of 0.03). When controls are

added for city size and the fraction of blacks in the city population, the estimate of $ rises to 1.01

7
(with a standard error of 0.03).12

Findings similar to those in Figure 1 are reported in Card (2001), using data for 175 cities

from the 1990 Census, and defining low skilled workers as those who are predicted to work in low

wage occupations based on their age, education, gender, race, ethnicity, and country of origin. In

that paper I looked specifically at mobility responses of natives to recent immigrant inflows, and

concluded that each new immigrant in the lowest skill group adds about 1 to the net supply of low

skilled workers in a city. Focusing on longer term mobility, Card and DiNardo (2000) use a three

skill group taxonomy to examine the effect of immigrant inflows on native migration rates between

1980 and 1990. Again, the conclusion is that native mobility has virtually no offsetting effect on the

relative supply shocks created by immigration. Indeed, once controls are introduced for city-specific

trends in native population growth, the data suggest that native mobility responses may slightly

reinforce the relative supply effects of immigration (Card and DiNardo, 2000, Table 2).

A concern with the interpretation of the data in Figure 1 is that unskilled immigrants may be

drawn to cities where the relative demand for dropout labor is increasing. One way to partially

address this concern is to use historical immigration patterns as instruments for current inflows

(Altonji and Card, 1991; Card, 2001). For example, Card and Dinardo (2000, Table 2) report models

in which low skilled immigration inflows between 1980 and 1990 are instrumented by the fraction of

Mexican immigrants in the city in 1970. These models give no indication that the OLS estimates are

biased by local demand shocks.

Although the results in Figure 1 suggest that immigration has a powerful effect on local

labor markets, some researchers have argued the opposite. Borjas Freeman and Katz (1997), for

12
The size controls are the log of the adult population and its square. The regressions are estimated
by weighted least squares using the size of the population as weights.

8
example, claim that native mobility effectively undoes any local impact of immigrant inflows.13

Importantly, however, Borjas Freeman and Katz (1997) focus only on total population, not on the

relative size of different skill groups. Looking at the California and Texas cities in Table 2 it is very

hard to argue that immigration has not had some impact on the fraction of less-educated people in

the local labor market. To the best of my knowledge, in fact, all studies that have looked at the

relative supply impacts of immigration find very large effects on local labor markets.

c. Impacts on Less Skilled Natives

Once it is established that low-skilled immigration increases the relative supply of unskilled

workers in local labor markets, the next step is to measure the effects on the relative labor market

outcomes of less skilled workers. A simple theoretical framework for this analysis consists of a local

production function and a set of per-capita labor supply functions for members of each skill group.14

Such a model implies that the relative wages and relative employment rates of workers in any two

skill groups depend on the relative fractions of the groups in the local population. For example,

comparing high school graduates to dropouts, a relative supply/demand model implies that

(1a) log (wd/wH) = a1 + b1 log (sd/sH) + u1

(1b) log (Ed/EH) = a2 + b2 log (sd/sH) + u2,

13
Likewise Frey (1995, 1996) reports a strong correlation between immigrant inflows and native
outflows. Wright, Ellis, and Reibel (1997) re-examine Frey’s specifications and show that his results disappear
once controls for city size are added to the model.
14
See e.g., Card (2001). Specifically, consider a production function for local output
y = [ 3j (ejNj) (F!1)/F] F/(F!1) and supply functions log (Nj/Pj) = , log wj + Nj, where Nj is the number of
people employed in skill group j, wj is the wage of group j, ej is a relative demand shock, Pj is the population
of skill group j, and Nj is a local supply shock. These equations imply a relative labor demand curve
log (Nd/NH) = !F log (wd/wH) + (F!1) log (ed/eH) ,
and a relative labor supply curve
log (Nd/NH) = log (Pd/PH) + , log (wd/wH) + Nd!NH .

9
where wd and wH are the mean wages of dropouts and high school graduates in a city, Ed and EH are

the mean employment-population rates of the two groups, and sd and sH are the shares of dropouts

and high school graduates in the local population. The coefficients b1 and b2 depend on the elasticity

of substitution between skill groups (F) and on the elasticity of the per-capita labor supply functions

(,):

b1=!1/(F+,) , b2=!,/(F+,) .

If local labor supplies are perfectly inelastic (,=0) then equation (1a) reduces to the familiar model

used in studies of education-based wage gaps (e.g., Katz and Murphy, 1992).

Apart from the potential problem posed by unobserved relative demand and supply shocks,

the key problem for interpreting estimates based on equations (1a ) and (1b) is that the slope of the

local relative demand curve may be uninformative about the degree of substitutability between skill

groups. In particular, a Hecksher-Olin style model of local economies suggests that relative wages

may be uncorrelated with relative labor supplies, even though at the national level relative wages are

negatively related to relative supplies. I return to this point below.

Some simple reduced form evidence on the impact of unskilled immigration on relative

wages and relative employment of low-skilled natives is presented in Figures 2 and 3. Figure 2

shows the gap in mean log wages between native male high school graduates and native male

dropouts in each of the 175 largest U.S. cities in 2000, plotted against the fraction of immigrant

dropouts in the local labor market.15 Figure 3 shows the log of the ratio of the employment-

population rates of high school graduates and dropouts, also plotted against the fraction of

15
The wage gaps were estimated as follows. First, separate models were estimated for mean log wages
of native male dropouts and high school graduates, including unrestricted city dummies (for 325 MSA’s) a
quartic function of age, dummies for black race and Hispanic ethnicity, and interactions of the black and
Hispanic dummies with age and ages squared. The wage gaps are estimated as the differences in the city
dummies from these two models (re-normalized to have the same mean as the raw data).

10
immigrant dropouts.16 The graph for relative wages shows little connection between native wages

and the fraction of immigrant dropouts, while the graph for relative employment suggests a slightly

positive correlation. Estimated regression models fit over all 325 cities confirm this impression: the

estimated regression slope for wages is statistically insignificant (slope=!0.06; standard error = 0.06)

while the estimated slope for relative employment is significantly positive (slope = 0.07, standard

error =0.02), suggesting a small negative impact of more unskilled immigrants on native dropout

employment.

Formal estimation results for the local demand/supply system are presented in Table 3,

which shows both ordinary least squares (OLS) estimates and instrumental variables (IV) estimates

using the fraction of immigrant dropouts as an instrument for the relative supply of high school

versus dropout labor. The IV results are quite similar to the OLS results, and nearly as precise,

reflecting the strong first stage.17 The estimates suggest there is no relationship between the relative

supply of high school dropouts and their relative wages, but point to a small negative impact of

relative supply on relative employment. These findings are quite similar to the results in Card (2001)

using data for 1990 and occupation-based skill groupings.18 As in most of the previous work

looking at local labor market impacts of immigration, there is a surprisingly weak relationship

between immigration and less-skilled native wages (see Friedberg and Hunt, 1995 and Borjas, 1994).

16
The local employment population rates of the two groups were estimated as the city dummies in
separate linear probability models for the event of working in the previous year, in models with the same
control variables as the first stage wage models.
17
The coefficient of the fraction of immigrant dropouts in a model for the log relative supply of high
school versus dropout labor is -6.10, with a standard error of 0.20 (F-statistic = 902). The fraction of
immigrant dropouts explains 74 percent of the variation in the relative supply variable across the 325 MSA’s
in the 2000 Census.
18
In my 2001 study I presented estimates for 6 occupation groups in 175 cities. Interestingly, the
estimated relative supply effects were typically smaller when the sample was restricted to low-skilled
occupations.

11
This pattern has persisted despite steady inflows of relatively unskilled immigrants that have created

ever greater differences across cities in the relative supply of dropouts over the past two decades.

d. Explaining the Absence of Local Labor Market Impacts

A variety of explanations have been offered for the finding that wages of less skilled natives

are insensitive to the relative supply pressure created by unskilled immigrants. The first is

unobserved relative demand shocks, which enter the relative wage and employment equations and

are potentially correlated with the relative share of low skilled workers.19 The leading solution to this

problem is to instrument relative supply (or the relative number of low skilled immigrants) with

information on historical immigration patterns. Immigrants from a given source country tend to go

to the same places they went many years ago, and relative skill levels of the immigrants from a

country are highly correlated over time, so instruments based on historical immigration patterns

have reasonable predictive power. My reading of the evidence is that instrumenting sometimes

moves the coefficients in the “right direction”, but does not change the conclusion that immigrant

impacts are small (see for example, Lewis (2003) who analyzes changes in relative wages in major

cities between 1980 and 1990 using OLS and IV methods).

A second explanation is that, as predicted by a Hecksher-Olin (HO) model, variation in the

relative supply of unskilled labor across local labor markets is absorbed by changing industry

structure. As pointed out by Lewis (2003), the magnitude of any HO-style adjustments can be

calculated by looking at data on industry shares across cities. To illustrate this point, start with an

identity that expresses the overall fraction of dropouts employed in a given city, sd(c), as a weighted

19
For example, using the model outlined in footnote 14, the residual in the relative wage equation is
u1 = (F!1)/(F+,) log (ed/eH) ! (Nd!NH )/(F+,).

12
sum of the industry shares in the city, times the dropout intensity in each industry:

(2) sd(c) = 1/N(c) 3i Ndi(c)

= 3i Ni(c)/N(c) Ndi(c)/Ni(c )

= 3i 8i(c) sdi(c) ,

where N(c) is total employment in city c, Ndi(c) is the number of dropouts employed in industry i in

city c, Ni(c) is total employment in industry i in city c, 8i(c) /Ni(c)/N(c) is the employment share of

industry i in city c, and sdi(c) = Ndi(c)/Ni(c ) is the share of dropout workers in industry i in city c. It

follows that the gap between sd(c) and the national average fraction of dropouts, sd, can be written as

the sum of a “between industry component” B representing shifts in the relative fractions of

different industries in the city, a “within industry component” W, representing shifts in the relative

fraction of dropout workers in each industry, and an interaction component I:

(3) sd(c) - sd = B(c) + W(c) + I(c),

where

B(c) = 3i sdi [ 8i(c) ! 8i ]

W(c) = 3i 8i [ sdi(c) ! sdi ]

I(c) = 3i [ 8i(c) ! 8i ] × [ sdi(c) ! sdi ] .

The HO theorem states that under certain conditions all of the variation in the share of dropout

labor across cities can be absorbed by expansion or contraction of high-dropout-intensity industries

(i.e., via the B(c) term), with no city-level variation in relative wages or the dropout intensity of any

particular industry.20

In Card and Lewis (2005), we used data on employment classified by 3 digit industry from

20
These conditions would include infinitely elastic supplies of capital, perfectly integrated product
markets, and the existence of at least one industry that produces a tradeable good or service that has a
dropout intensity that exceeds the maximum dropout share in any city.

13
the 2000 Census to compute the terms in equation (2) for each of 150 larger cities. We then

performed a series of regressions:

(4a) B(c) = aB + bB [ sd(c) - sd] + eB(c)

(4b) W(c) = aw + bw [ sd(c) - sd] + ew(c)

(4c) I(c) = aI + bI [ sd(c) - sd] + eI(c) .

Since (3) is an identity, bw + bB + bI =1. A strict version of HO implies bB = 1.

Figure 4 plots the between-industry component B(c) against the excess fraction of dropouts

in each of the 150 larger MSA’s. For reference, the figure also shows the 45 degree line: if changing

industry structure accounted for the absorption of immigrants the points would lie along this line.

Although the points suggest an upward-sloping relationship, the slope is relatively modest,

suggesting that changing industry structure accounts for only a small share of the absorption of

dropouts. Indeed, the OLS estimate, reported in the first column of Table 4, is 0.22, and is

significantly different from 1. By contrast, Figure 5 plots the within-industry component W(C)

against the excess fraction of dropouts in each city. This component is more highly correlated with

the dropout share: as shown in column 2 of Table 4, the estimate of bw is 0.76. Though not shown

in a figure, the interaction terms are relatively small, and essentially uncorrelated with differences

across cities in the share of dropout workers. The estimate of bI in column 3 of Table 4 is 0.02

(with a very small R-squared =0.03).

The MSA’s with relatively high dropout shares are labeled in Figures 4 and 5. Interestingly,

most of these MSA’s are comprised of counties in California and Texas with substantial agricultural

employment. Since agriculture relies on the availability of land resources, it is debatable whether the

high employment shares of agriculture in these MSA’s represents a reaction to abundant supplies of

less-educated labor. Rather, it seems more likely that the relative supplies of less-educated labor in

14
these MSA’s are driven by the availability of farm jobs.

The framework of equation (3) can be used to examine the contribution of the changing

scale of specific industries to the absorption of local supplies of dropout labor. For example, the

contribution of industry i is sdi [ 8i(c) ! 8i ], which is excess employment share of the industry in

city c relative to its national average share, multiplied by the average dropout intensity of the

industry. Columns 4-6 of Table 4 show estimates of models similar to (4a), focusing on the

absorption contributions of agriculture, textiles apparel and footwear industries, and a set of low-

skilled service industries.21 These 3 industry clusters together account for most of the total between

industry effect: agriculture alone accounts for nearly one-half. Overall, though there is some

evidence that textiles and apparel manufacturing tends to cluster in cities with high dropout shares,

these results suggest that most of the absorption of unskilled labor across cities occurs within very

narrow industries. Apart from a few small sectors it is difficult to find much evidence of HO-style

industry adjustment across cities.

Similar conclusions were reached by Lewis (2003), who examined changes in the relative

absorption of 4 education groups over the 1980-1990 period. Lewis used Census data to estimate

first-differenced versions of equation (4b) for each skill group.22 He also compared OLS estimates

to IV estimates that used immigrant inflows based on historical immigration patterns as instruments

for the changes in the relative shares of each skill group. As in the 2000 cross-section, the industry

composition effects over the 1980-1990 period are only weakly related to local skill-group-specific

21
We include textiles, apparel, knitting mills, footwear, and leather industries as apparel, and the
following as “low skilled services”: building services, landscaping services, carwashes, landscaping, dry
cleaning and laundry services, private household services, and other personal services.
22
One difference is that Lewis regresses the between-industry effects on the population share of the
skill group in the local labor market, rather than the employment share. An advantage of a first differenced
approach is that it eliminates the confounding caused by permanent factors like differences in the amount of
agricultural land in an MSA.

15
population growth. Lewis’ estimates of bB for manufacturing industries (which are arguably best

able to respond to local factor availability) are very close to 0, while his estimates for all industries

range from 0 to 0.08. He also reports parallel specifications in which the dependent variable is the

within-industry relative employment term. These are much more strongly correlated with relative

population growth, accounting for 90 percent of the adjustment to skill-group specific relative

supply shocks.

The evidence suggests that HO-style changes in industry structure play relatively little role in

explaining how cities like Los Angeles were able to absorb massive inflows of relatively uneducated

immigrant workers over the past two decades. Instead, most of the less-educated labor was

absorbed by city-specific within-industry increases in dropout intensity, which took place despite any

corresponding changes in the relative wages of dropout workers.

One possible explanation for this pattern is that local relative demand shocks for dropout

workers are “caused” by the presence of low skilled immigrants. For example, Acemoglu’s (1998)

model of endogenous technological change suggests that firms will innovate in a direction to take

advantage of more readily available factors, even in the absence of relative wage changes. Beaudry

and Green’s (2003) model of technological adoption has a similar flavor. Lewis (2004) presents

some of the first direct evidence for this mechanism, using data on the number of advanced

technologies adopted by manufacturing plants in the late 1980s and early 1990s. He finds that

controlling for very detailed (4 digit) industry effects, the adoption of advanced technologies by

individual plants is significantly reduced by the presence of a greater relative supply of unskilled

labor in the local labor market. These results are potentially consistent with the evidence on within-

industry absorption in Table 4. More work is clearly needed to better understand how firms choose

which technologies to use, and whether the choice is influenced by the relative availability of

16
different skill groups.

e. Aggregate Evidence on Relative Wages of Dropouts

My reading of the evidence is that the two main mechanisms that economists have proposed

to explain the adjustment of local labor markets to immigration-based supply shocks – selective

mobility and HO-style realignment of local industry structure – are relatively unimportant, and that

the bulk of the absorption occurs within industries. In view of the weak correlation between local

wages and local immigrant supplies, some researchers – notably Borjas, Katz and Freeman (1996,

1997) and Borjas (2003) – have argued that aggregate time series analyses are required to measure

the full impacts of immigration on native wages. A complete analysis of aggregate trends is beyond

the scope of this paper. However, in light of the data in Table 1 showing the relative education

distribution of immigrants, it is useful to briefly examine trends in the relative wages of high school

dropouts.

Figure 6 plots two measures of the wage gap between high school dropouts and high school

graduates: the mean log wage differential between the groups, and the average return per year of

schooling among those with 12 or fewer years of schooling, multiplied by 4. These wage gaps refer

to the hourly earnings of men age 18-64 in the 1980-2002 March Current Populations Survey (CPS),

and are estimated from models that include controls for a cubic in potential experience and

dummies for black race and Hispanic ethnicity. For reference Figure 6 also plots the college-high

school wage premium, estimated from samples of men with 12 or 16 years of schooling. Since 1979

the wage premium for high school graduates relative to dropouts has fluctuated in the range of 25 to

30 percent, with a modest rise in the early 1980s and more or less steady declines since then. The

return per year of schooling for those with 0-12 years of school has fluctuated between 7 and 8

17
percent, and also increased slightly in the early 1980s. In contrast, the college-high school wage

premium has varied a lot more, rising by about 12 log points in the early 1980s, and nearly 22 log

points over the past two decades.

Although immigration presumably exerts downward pressure on the relative wages of

dropouts, the wage gap between dropouts and high school graduates has been nearly constant since

1980, and has fallen by more than 50 percent relative to the gap between high school graduates and

holders of bachelor’s degrees.23 The absence of an aggregate trend in the relative wages of high

school dropouts is consistent with the remarkable stability of the relative wage of dropouts across

different local labor markets. Of course, even taking account of unskilled immigrant inflows the

relative supply of dropouts has declined over the past two decades, so depending on what is

assumed about the rate of growth of relative demand for dropouts versus high school graduates, one

can argue that immigration lowered the wages of the least educated natives relative to the

counterfactual trend.24 Without knowing the trend in relative demand for dropouts, however, the

aggregate data are uninformative, so estimates of the effect on native wages amount to simply

multiplying the relative share of dropouts attributable to immigration by some estimate of the

elasticity of substitution (Johnson, 1980; Borjas, 2003).

23
According to the data in Table 1, the presence of immigrants increased the relative supply of
dropouts in 2000 by about 21 percent, reduced the relative supply of high school graduates by about 5
percent, and had no net effect on the relative supply of people with a college degree or more. Assuming that
the elasticity of substitution between education groups is -1.4 (Borjas, 2003; Katz and Murphy, 1992) and
ignoring labor supply effects, the presence of immigrants in the U.S. labor market should have raised the
wage premium for high school graduates relative to dropouts by about 26/1.4 = 18 log points, in the absence
of other factors.
24
In the 1980 Census, 26.3 percent of the population age 18-64 were dropouts, 39.2 percent had
exactly 12 years of schooling, 19 percent had some college, and 15.6 had a college degree or more.
Comparing these numbers to those in Table 1 there was a 35 percent decline in the log relative supply of
dropout versus high school labor between 1980 and 2000.

18
II. Assimilation of Immigrants

While immigrant men in the 1970 Census earned about as much an natives, a wage gap

opened up over the 1970s and has persisted. Currently, immigrant men’s hourly wages are about 20

percent lower than natives’, while immigrant women’s wages are about 10 percent lower.25 Given

the gap in education between immigrants and natives, and the importance of education in the U.S.

wage structure, this is not too surprising. Moreover, the quality of education in many of the major

immigrant sending countries is arguably below the quality in the U.S. (Bratsberg and Terrell, 2002),

and many immigrants have limited English skills, implying that immigrant human capital is even

lower than observed education would suggest.

Following Chiswick (1978) there is an extensive literature on the question of whether the

immigrant-native earnings gap narrows with time in the U.S. Such “earnings assimilation” could be

due to formal or informal training, acquisition of language skills, or a variety of other processes.26

Borjas (1985, 1995) noted that a synthetic cohort analysis like Chiswick’s will overstate earnings

growth if more recent immigrant arrival cohorts have lower unmeasured skill characteristics than

earlier arrivals, as seems to have been true in 1980 and 1990. Moreover, many immigrants return to

their home country within a few years, and others move back and forth, further complicating

inferences from cross sectional data. Limited evidence from true longitudinal data (Lubotsky, 2000)

suggests that immigrant earnings rise with time in the U.S., though the gains (about 10-15 percent in

the first 20 years in the U.S.) are not enough to offset the 35-40 percent immigrant-native earnings

25
These numbers come from an analysis of March CPS data from 1995 to 2002.
26
Cortes (2004) shows that recent immigrant arrivals have relatively high rates of participation in
schooling. She finds that 1975-80 immigrant arrivals show a gain in English proficiency between 1980 and
1990. Manning (2003, chapter 6) notes that some fraction of life cycle earnings growth is due to accumulated
“search capital”. Immigrants may start off with less efficient search and gradually catch up to natives.

19
gap at arrival.

Although the precise magnitude of immigrant earnings assimilation will probably be debated

for many years (see Duleep and Regets, 2002 for a recent analysis), few of the 40 percent of

immigrants who arrive in the U.S. as adults without a high school credential will ever earn as much

as average natives. Likewise, the 22 percent of immigrants with a college degree or more will earn

more than average natives.27 In my opinion, a more interesting question is how well the U.S.-born

children of immigrants are doing. Focusing on the status of immigrants’ children is important for a

number of reasons. Second generation immigrants are a growing fraction of the population,

accounting for 10 percent of teenagers nationwide.28 Nearly all of them will spend their entire lives

in the U.S., and will pay taxes and receive income support payments. Thus, the success of

immigrant children is an important component of the long run costs and benefits of immigration.

For these and other reasons the relative success of the second generation provides a key gauge of

the extent to which their parents assimilated into the U.S.

Table 5 presents some simple descriptive regression models showing the relative status of

immigrants and second generation immigrants in the 1995-2002 CPS. (I define second generation

immigrants as people born in the U.S. with at least one foreign-born parent). The upper panel of

the table shows results for men, while the lower panel shows results for women. The first two

columns shows models for years of schooling and the probability of working in the previous year, fit

over the entire population of 21-64 year olds, while columns 3-6 show models for log hourly wages,

27
Using March CPS data for 1995-2002, I estimate that immigrants with at least a college degree earn
about 30 percent more than average natives.
28
Since 1994, the CPS has asked individuals where their parents were born. Using March 1995-2002
CPS files, I estimate that about 11 percent of people age 16-19 were born in the U.S. with at least one
immigrant parent.

20
fit to workers only.

Looking first at the education models, immigrants have about 1.2 - 1.4 fewer years of

education then natives, standardizing for their age. (The raw gaps are a little smaller). On the other

hand, second generation immigrants have 0.3-0.4 years more education than people whose parents

were born in the U.S. (the “third and higher” generation). Among men, immigrants are only slightly

less likely to work than members of third and higher generation, while second generation men are a

little more likely to work. Among women there is a larger immigrant gap in the probability of

working (a 13.6 percent lower annual employment rate than natives) but again second generation

women are a little more likely to work than third and higher generation natives.

The models in columns 3 and 4 present wage models that control for age and geographic

location, but not for education. Among men, immigrants have 18-23 percent lower wages than third

and higher generation natives, while second generation immigrants have 4-8 percent higher wages.

The wage gaps for second generation women are about the same as for second generation men, but

for immigrant women the wage gap is smaller than for immigrant men, perhaps reflecting the

relative selectivity of labor force participation among immigrant women. When controls are added

for education, the wage gap for immigrant men falls to about 11 percent while the gap for immigrant

women falls to 7 percent. The wage gaps for second generation men and women also fall, to under

2 percent in each case. Thus, the higher wages of second generation immigrants are largely

explained by their geographic location and their higher education.

The models in the final column of Table 5 add two additional controls for black race and

Hispanic ethnicity. Reflecting the fact that many immigrants are Hispanic, and that third and higher

generation Hispanics earn a little less than non-Hispanics, these added controls reduce the

immigrant wage gaps slightly, to 8 percent for men and 5 percent for women, and slightly increase

21
the wage advantage of second generation immigrants.

The results in Table 5 suggest a couple of conclusions. First, immigrant workers in the U.S.

labor market today (over 90 percent of whom arrived after 1965) earn less than natives, but the

magnitude of the wage gaps are not enormous. After controlling for education, which explains

about an 11 percent gap in immigrant earnings for both men and women, the gaps are under 10

percent – comparable to the wage gaps for blacks or native Hispanics. Second, the children of

immigrants do well, on average, with most of their wage advantage relative to natives attributable to

higher education. Despite the lower education of their parents, children born to immigrant parents

seem to catch up and even surpass the levels of children born to U.S. natives.

An interesting perspective on this catch-up phenomenon is provided by examining

differences across parental source countries (Borjas, 1993; Card Dinardo and Estes, 2000). Looking

in the recent CPS data, one can identify second generation men and women whose parents were

from different countries, and compare the earnings or educational attainment of each second

generation group against the corresponding outcomes for their parents. This idea is illustrated in

Figures 7a and 7b, which show mean education levels for second generation younger adults (age 21-

40) in the 1995-2002 March CPS by country of origin of their father, plotted against mean levels of

education for fathers of children age 0-15 from the same source country in the 1980 Census. For

reference, I have also shown the point representing members of the third and higher generation, and

the estimated regression line across the 39 country of origin groups shown in the figures.29

The figures suggest that there is a strong intergenerational correlation in education that is

29
I selected countries of origin with at least 50 observations for second generation sons and
daughters. The largest group is Mexico (4998 second generation children). Italy, Canada, Cuba, Germany
and the Philippines also have at least 500 second generation children. The smallest origin groups are Panama
(54 observations), Austria (53 observations) and Israel (51 observations).

22
similar for sons and daughters. Indeed, the coefficient estimates and R-squared statistics are nearly

identical for sons and daughters (slope=0.30 for men, standard error=0.03, R-squared=0.77;

slope=0.29 for women, standard error=0.03, R-squared=0.77). Interestingly, the coefficient of 0.3

for the effect of fathers education on either sons or daughters is almost identical to the estimates

obtained in a micro level regression using samples of men and women from the General Social

Survey.30 Thus, the intergenerational transmission of education is about the same for families of

immigrants as for other families in the US. In particular, there is no evidence that second generation

immigrants’ education outcomes regress toward the mean more slowly than other children.

Even more interestingly, in both Figure 7a and 7b the fitted line for the second generation

group over-predicts the outcomes for natives: by 0.71 years for men and by 0.77 years for women.

This means, for example, that second generation sons whose fathers had as little as 10.4 years of

schooling (2.3 years below the average for native fathers) ended up ahead of their third generation

peers. Even sons of Mexican immigrants, whose fathers had 5.5 years of schooling less than native-

born fathers in 1980 (7.3 years versus 12.8 years for native-born fathers) ended up with 12.2 years of

schooling, closing 80 percent of the education gap faced by their fathers.31

Finally, it is interesting to compare the results in Figures 7a and 7b with similar results from

an earlier generation of immigrant children. Card, DiNardo and Estes (2000) conduct a parallel

analysis using 1970 Census data for second generation immigrants, and 1940 Census data for their

parents. The estimated intergenerational coefficients in education are 0.41 (standard error 0.10) for

30
I used the 1972-1996 GSS. The sample has 6667 men and 7745 women between the ages of 21 and
45 with observed father’s education. In a regression controlling for age and age-squared, the effect of father’s
education is 0.32 for men (standard error 0.01) and 0.30 for women (standard error 0.01).
31
The mean level of education of third and higher generation sons is 13.29 (14.4 for daughters), while
the mean level of education of second generation Mexican sons is 12.19 (12.41 for daughters).

23
men and 0.47 (standard error 0.08) for women. These point estimates are a little higher than the

ones for more recent cohorts, though relatively imprecise. If anything, however, they suggest that

the rate of assimilation (which is 1 minus the intergenerational correlation) is slightly faster for more

recent cohorts than for older ones.

These results paint a relatively optimistic picture of the success of post-1965 immigrants.

Conditional on their parents’ human capital, the U.S.-born children of these immigrants have done

remarkably well. Indeed, of the 39 largest country-of-origin groups, sons from 33 groups and

daughters from 32 groups have higher average educational attainment than the children of natives.

III. Conclusions

Immigration is a major policy concern in many countries around the world. Two important

questions that economic research can answer concern the impact of immigrants on the labor market

opportunities of natives, and the relative success of immigrants in integrating into the domestic

economy. Economists have struggled with both questions for the past couple of decades, with

varying degrees of success, and the lessons from the U.S. literature provide potentially valuable

lessons to researchers in other contexts.

On the question of immigrant competition the U.S. has a structural advantage, since there

are many large U.S. cities, with widely varying levels of immigration, and samples from the

Decenniel Censuses can be used to develop detailed models of local labor market outcomes. New

evidence from the 2000 Census re-confirms the main lesson of earlier studies: Although immigration

has a strong effect on relative supplies of different skill groups, local labor market outcomes of low

skilled natives are not much affected by these relative supply shocks. Recent evidence on the

24
response of local industry structure to immigration-induced supply shocks shows that the absorption

of unskilled immigrants takes place within industries in high-immigrant cities, rather than between

industries, as implied by simple trade models. It remains a fascinating question how firms in a given

industry can adapt their production technology so closely to local supplies of different types of labor

without substantial changes in relative wages.

As the evidence has accumulated over the past two decades that local labor market outcomes

are only weakly correlated with immigrant densities, some analysts have argued that the cross-city

research design is inherently compromised by intercity mobility of people, goods, and services.

Underlying this argument is the belief that labor market competition posed by immigration has to

affect native opportunities, so if we don’t find an impact, the research design must be flawed. The

leading alternative to a local labor market approach is a time series analysis of aggregate relative

wages. Surprisingly, such an analysis shows that the wages of native dropouts (people with less than

a high school diploma) relative to native high school graduates have remained nearly constant since

1980, despite pressures from immigrant inflows that have increased the relative supply of dropout

labor, and despite the rise in the wage gap between other education groups in the U.S. economy.

While the counterfactual is unknown, it is hard to argue that the aggregate time series evidence

points to a negative impact of immigration unless one starts from that position a priori.

On the question of immigrant assimilation, a major constraint in the U.S. literature has been

the absence of true longitudinal data. Nevertheless, I believe that a narrow focus on immigrant

earnings is misplaced. Few of the 40 percent of immigrants who come to the U.S. without

completed high school education will ever catch up with the average earnings of natives. Most of

their U.S.-born children, however, will catch up with the children of natives. Evidence on the

intergenerational progress of immigrants’ children is now becoming available, and points to above-

25
average levels of educational attainment, even for children whose fathers had much lower schooling

than native-born fathers. The relatively strong educational progress of second generation

immigrants, together with the limited evidence of adverse effects on less skilled natives, suggest that

the new immigration may not be so bad after all.

26
 References

Acemoglu, Daron. “Why Do New Technologies Complement Skills? Directed Technical Change
and Wage Inequality.” Quarterly Journal of Economics 114 (November 1998): 1055-1089.

Altonji, Joseph G., and David Card “The effects of immigration on the labor market outcomes of
less-skilled natives”. In John M. Abowd and Richard B. Freeman, eds., Immigration, Trade, and the
Labor Market. Chicago: The University of Chicago Press, 1991.

Beaudry, Paul and David Green. “Wages and Employment in the United States and Germany: What
Explains the Differences?” American Economic Review 93 (June 2003): 573-603.

Borjas, George. “Assimilation, Changes in Cohort Quality, and the Earnings of Immigrants". Journal
of Labor Economics 3 (1985): 463-489.

Borjas, George, “Immigrants, Minorities, and Labor Market Competition” Industrial and Labor
Relations Review 40 (1987): 382-392.

Borjas, George. “The Intergenerational Mobility of Immigrants.” Journal of Labor Economics 11

(1993): 113-135.

Borjas, George. “The Economics of Immigration.” Journal of Economic Literature 32 (1994): 1667–717.

Borjas, George. “Assimilation and Changes in Cohort Quality Revisited: What Happened to
Immigrant Earnings in the 1980s? Journal of Labor Economics 13 (1995): 201-245

Borjas, George, Richard B. Freeman and Lawrence Katz. “Searching for the Effect of Immigration
on the Labor Market.” American Economic Review 86 (May 1996): 246-251.

Borjas, George, Richard B. Freeman and Lawrence Katz. “How Much Do Immigration and Trade
Affect Labor Market Outcomes?” Brookings Papers on Economic Activity 1 (1997): 1-67.

Borjas, George. Heaven’s Door. Princeton, NJ: Princeton University Press, 1999.

Borjas, George. “The Labor Demand Curve is Downward Sloping: Reexamining the Impacts of
Immigration on the Labor Market.” Quarterly Journal of Economics 118 (November 2003): 1335-1374.

Bratsberg, Bernt, and Dek Terrell. “School Quality and Returns to Education of U.S. Immigrants.”
Economic Inquiry 40 (April 2002): 177-198.

Brigham, Carl C. A Study of American Intelligence. Princeton NJ: Princeton University Press, 1923.

Card, David. "Immigrant Inflows, Native Outflows and the Local Labor Market Impacts of Higher
Immigration." Journal of Labor Economics, 19 (January 2001).

27
Card, David and John E. DiNardo. “Do Immigrant Inflows Lead to Native Outflows?” American
Economic Review 90 (May 2000).

Card, David and John E. DiNardo. “Skill Biased Technical Change and Rising Wage Inequality:
Some Problems and Puzzles.” Journal of Labor Economics 20 (October 2002).

Card, David, John E. DiNardo, and Eugena Estes. “The More Things Change: Immigrants and the
Children of Immigrants in the 1940s, the 1970s, and the 1990s”. In G. Borjas, editor, Issues in the
Economics of Immigration. University of Chicago Press for NBER: Chicago, 2000.

Card, David and Ethan G. Lewis. “The Diffusion of Mexican Immigrants in the 1990s: Patterns
and Impacts.” UC Berkeley Center for Labor Economics Unpublished Working Paper, February
2005.

Card, David and Thomas Lemieux. “Dropout and Enrollment Trends in the Post-War Period:
What Went Wrong in the 1970s?” In Jonathan Gruber, editor, Risky Behavior Among Youth: An
Economic Analysis. Chicago: University of Chicago Press, 2000.

Card, David and Thomas Lemieux. “Can Falling Supply Explain the Rising Return to College for
Younger Men? A Cohort-Based Analysis.” Quarterly Journal of Economics 116 (May 2001).

Chiswick, Barry R. “The Effects of Americanization on the Earnings of Foreign-Born Men.”

Journal of Political Economy 86 (1978): 897-921.

Cortes, Kalena E. “Are Refugees Different from Economic Immigrants? Some Empirical Evidence
on the Heterogeneity pf Immigrant Groups in the United States.” Review of Economics and Statistics,
forthcoming 2004.

Dahl, Gordon B. “Mobility and the Return to Education: Testing a Roy Model with Multiple
Markets.” Econometrica 70 (November 2002): 2367-2420.

Duleep, Harriet and Mark C. Regets. “The Elusive Concept of Immigrant ‘Quality’: Evidence from
1970-1990.” IZA Working Paper No. 631, November 2002.

Friedberg, Rachel M. and Jennifer Hunt. “The Impact of Immigration on Host Country Wages,
Employment, and Growth.” Journal of Economic Perspectives 9 (Spring 1995): 23-44.

Frey, William H. “Immigration and Internal Migration `Flight' from US Metropolitan Areas:
Toward a New Demographic Balkanisation.” Urban Studies 32 (1995) 733--757.

Frey, William H. “Immigration, Domestic Migration, and Demographic Balkanization in America:

New Evidence for the 1990s.” Population and Development Review 22 (December 1996): 741-763.

Grossman, Jean B. “The Substitutability of Natives and Immigrants in Production.” Review of

Economic and Statistics 64 (1982): 596-603.

28
Heckman, James J. and Jose Scheinkman. “The Importance of Bundling in a Gorman-Lancaster
Model of Earnings.” Review of Economic Studies 54 (1987): 243-255.

Jaeger, David. “Skill Differences and the Effect of Immigrants on the Wages of Natives.” U.S.
Bureau of Labor Statistics Working Paper No. 273. March 1996.

Johnson, George E. “The Labor Market Effects of Immigration.” Industrial and Labor Relations Review
33, (April 1980): 331-41.

Katz, Lawrence and Kevin M. Murphy. “Changes in Relative Wages, 1963-1987: Supply and
Demand Factors.” Quarterly Journal of Economics 107 (February 1992): 35-78.

Kuhn, Peter and Ian Wooten. “Immigration, Trade, and the Wages of Natives.” In John M. Abowd
and Richard B. Freeman, eds., Immigration, Trade, and the Labor Market. Chicago: The University of
Chicago Press, 1991.

LaLonde, Robert J. and Robert H. Topel. “Labor Market Adjustments to Increased Immigration.”
In John M. Abowd and Richard B. Freeman, eds., Immigration, Trade, and The Labor Market, Chicago:
University of Chicago Press, 1991.

Lazear, Edward P. “Culture and Language.” Journal of Political Economy 107: 6 (December 1999):
S95-S126.

Lewis, Ethan G. “Local Open Economies within the U.S. How do Industries Respond to
Immigration?” Federal Reserve Bank of Philadelphia Working Paper (December 2003).

Lewis, Ethan G. “ The Effect of Local Skill Mix on New Technology Adoption in U.S.
Manufacturing.” Federal Reserve Bank of Philadelphia Unpublished Working Paper (December
2004).

Lubotsky, Darren. “Chutes or Ladders? A Longitudinal Analysis of Immigrant Earnings.”

Princeton University Industrial Relations Section Working Paper No. 445, August 2000.

Manning, Alan. Monopsony in Motion. Princeton NJ: Princeton University Press, 2003.

Sjaastad, Larry A. “The Costs and Returns of Human Migration.” Journal of Political Economy 70
(October 1962, Part 2): S80-S93.

Topel, Robert H. “Local Labor Markets.” Journal of Political Economy 94 (June 1986, Part 2): S111-
S143.

Welch, Finis. "Linear Synthesis of Skill Distribution." Journal of Human Resources 4 (Summer 1969):
311-327.

Wright, Richard, Mark Ellis, and M. Reibel. “The linkage between immigration and internal
migration in large metropolitan areas in the United States.” Economic Geography 73 (1997): 234-253.

29
Table 1: Educational Attainment of Natives and Immigrants in 2000 Census

Immigrants:
In US In US
Highest Education: All Natives All 5+ Yrs. < 5 Yrs.

Dropouts 17.8 14.7 38.2 37.6 40.3

Of Which:
1-8 Years Completed Schooling 5.3 2.8 21.5 21.3 22.2

9-11 Years Completed Schooling 12.6 11.9 16.7 16.3 18.1

High School Diploma 37.2 39.2 24.0 24.3 23.1

Some College (including 22.6 23.7 15.5 16.3 12.5

Associates Degree)
Bachelors Degree 14.8 15.0 13.2 12.9 14.0

Advanced Degree 7.7 7.5 9.2 8.8 10.1

Of Which:
Masters Degree 5.2 5.2 5.4 5.1 6.3

Professional Degree 1.7 1.6 2.3 2.2 2.4

Doctorate 0.8 0.7 1.5 1.5 1.4

Note: Based on tabulations of individuals age 18-64 in 2000 Census. High school diploma group
includes people with less than 1 year of college (8 percent of the overall sample).
Table 2: Immigrant Densities and the Relative Fractions of Less Educated Workers, Selected Cities 1980 and 2000

1980: 2000:
Percent Percent Dropouts: Percent Percent Percent Dropouts: Percent
Immigrants Among Among Dropouts Immigrants Among Among Dropouts
In City Immigrants Natives In City In City Immigrants Natives In City

All Cities 9.5 38.9 23.0 24.3 18.0 37.8 13.0 17.7

New York 23.2 39.6 26.4 29.5 41.8 32.0 17.5 23.6
Los Angeles 25.3 49.2 19.5 27.0 47.8 47.2 14.4 30.1
Chicago 11.8 44.0 23.7 26.1 21.2 37.7 11.8 17.3
Philadelphia 4.9 31.1 25.2 25.5 8.3 21.9 13.3 14.0
Detroit 6.3 34.3 25.8 26.4 8.6 26.2 14.4 15.5
Houston 9.4 46.1 25.1 27.1 26.0 51.6 15.5 24.9
Dallas 5.1 43.7 24.3 25.3 19.7 54.2 13.6 21.6
Washington DC 9.6 18.3 16.8 16.9 20.6 25.8 9.9 13.2
Boston 10.3 35.6 15.6 17.6 17.8 24.0 7.9 10.7
San Fancisco 17.0 28.4 14.3 16.7 36.4 26.6 6.9 14.0
Miami 41.1 38.5 23.3 29.6 61.2 33.3 18.6 27.6
Atlanta 3.1 14.8 24.9 24.6 12.1 34.0 13.6 16.1
Pittsburgh 2.6 28.1 21.5 21.7 2.6 12.5 10.4 10.5
Cleveland 5.8 34.5 24.0 24.6 5.6 19.7 14.2 14.5

Note: Based on tabulations of 1980 and 2000 Census public use files. "All cities" includes 272 Standard Metropolitan
Areas in 1980 and 325 Metropolitan Statistical Areas in 2000. Boundaries of some cities change between 1980 and
2000. Samples include individuals age 18-64 only.
Table 3: Effects of Relative Supply on the Relative Wages and Employment of Native Male Dropouts

Relative Outcomes of Native Male Dropouts:

Fraction Employed Mean Log
Last Year Hourly Wage
OLS IV OLS IV

Log Relative Supply of -0.013 -0.012 0.006 0.010

Dropouts vs. High School (0.003) (0.003) (0.009) (0.010)
Graduates

R-squared 0.056 0.035 0.001 0.003

Note: Standard errors in parentheses. All models fit to sample of 325 Metropolitan Statistical Areas
using weighted least squares. City data are derived from the 2000 Census public use files and
pertain individuals age 18-64. Outcomes are adjusted differences in employment-population
or mean log wages between high school dropouts and high school graduates -- see text.
Instrument is fraction of low education immigrants in city.
Table 4: Regression Models Measuring Cross-City Absorption of Excess Dropout Workers

Effect Across All Industries: Industry-Specific Expansion:

Between Within Apparel Low Skill
Industry Industry Interaction Agric. & Textiles Services
(1) (2) (3) (4) (5) (6)
Excess Fraction of 0.22 0.76 0.02 0.02 0.02 0.02
Dropout Employment (0.02) (0.02) (0.01) (0.01) (0.01) (0.01)
in City
R-squared 0.37 0.84 0.03 0.17 0.24 0.33

Note: All models estimated on sample of 150 larger MSA's, using 264 industry cells per city in
columns 1-3. Regressions are weighted by city size. See text for definitions of industries used
in columns 4-6.
Table 5: Education and Earnings Gaps Between Immigrants, Second Generation, and Others

Fit to All Individuals: Fit to Workers Only:

Years of Percent Models for Log Hourly Wage
Education Working (coefficients x 100)
(1) (2) (3) (4) (5) (6)
Estimates for Men:
Immigrant -1.24 -0.6 -18.3 -23.4 -11.1 -8.0
(0.02) (0.2) (0.4) (0.4) (0.4) (0.4)

Second Generation 0.45 0.8 8.0 3.6 1.5 2.3

(0.02) (0.2) (0.5) (0.5) (0.5) (0.5)

Controls for Age yes yes yes yes yes yes

Controls for Region/Urban no no no yes yes yes

Control for Education no no no no yes yes

Control for Race/Ethnicity no no no no no yes

Estimates for Women:

Immigrant -1.37 -13.6 -11.8 -18.6 -7.1 -5.4
(0.01) (0.2) (0.4) (0.4) (0.4) (0.4)

Second Generation 0.31 0.5 8.3 3.0 1.2 1.9

(0.02) (0.3) (0.5) (0.5) (0.5) (0.5)

Controls for Age yes yes yes yes yes yes

Controls for Region/Urban no no no yes yes yes

Control for Education no no no no yes yes

Control for Race/Ethnicity no no no no no yes

Notes: Standard errors in parentheses. Models estimated on pooled sample of 1995-2002 March Current
Population Surveys. Samples include individuals age 21-64 only. Hourly wage is estimated from data on
wage and salary earnings last year, weeks worked last year, and usual hours per week last year. Wages are
censored below at $2/hour (in 2002 dollars) are set to $2 and above at $200 per hour in 2002 dollars.
Controls for Region/urban are 8 region dummies and dummy for living in Metropolitan Area. Control
for education is linear term in years of education. Controls for race/ethnicity are dummies for black race
and Hispanic ethnicity.
Figure 1: Fraction of Immigrant Dropouts and Overall Fraction of Dropouts

0.5
Fraction Dropouts in Adult Population

0.4

0.3 Los Angeles

Miami

NYC

0.2
Chicago

0.1
Note: line with slope = 1 shown

0.0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Fraction of Immigrant Dropouts in Local Adult Population
Figure 2: Relative Wage of High School Grads and Dropouts vs. Fraction Low
Education Immigrants

0.5
Mean Log Wage High School Grads - Dropouts

0.4

0.3 NYC
Chicago
Miami

0.2
Los Angeles

0.1

national average=0.22

0.0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Fraction of Immigrant Dropouts in Local Adult Population
Figure 3: Relative Employment of High School Grads and Dropouts vs. Fraction Low
Education Dropouts

0.28
Log Employment Rate HS Grads/Dropouts

0.24

Los Angeles
NYC
0.20
Chicago
Miami

0.16

national average = 0.20

0.12
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Fraction of Immigrant Dropouts in Local Adult Population
Figure 4: Contribution of Between-Industry Component to Absorption of Dropouts

0.25
Excess of Dropout-Intensive Industries in MSA

0.20

0.15

0.10 Visalia Ca

Hickory NC Bakersfield Ca
0.05 Fresno Ca

Salinas Ca McAllen Tx

0.00
Los Angeles

-0.05

-0.10
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
Excess Fraction of Dropouts in MSA
Figure 5: Contribution of Within-Industry Component to Absorption of Dropouts

0.25

0.20
Excess Dropout Intensity in MSA

McAllen Tx
0.15
Los Angeles

El Paso
0.10 Bakersfield Ca
Miami Visalia Ca

Fresno Ca
0.05 Salinas Ca

Hickory NC
0.00 Stockton Ca

Modesto CA
-0.05

-0.10
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
Excess Fraction of Dropouts in MSA
Figure 6: College/High School and High-School/DropoutWage Gaps

0.60

0.50 College-High School Gap

Log Wage Differentials

0.40

Return per year x 4 (Education <=12)

0.30

High School Grad/Dropout Gap

0.20

0.10
1979 1982 1985 1988 1991 1994 1997 2000
Figure 7a: Father-Son Intergenerational Correlation in Education

17

16 India
Average Education of Sons (1995-2002)

15

14

13

Natives
Mexico
12

11

10
6 8 10 12 14 16 18
Average Education of Fathers (1980)
Figure 7b: Father-Daughter Intergenerational Correlation in Education

17
Average Education of Daughters (1995-2002)

16 India

15

14

13 Natives

Mexico
12

11

10
6 8 10 12 14 16 18
Average Education of Fathers (1980)