the economics of immigration · pdf filemigration occurred between 1881 and ... mexico 22.3...

Journal of Economic Literature, Vol. XXXII (December 1994)Borjas: The Economics of ImmigrationJournal of Economic LiteratureVol. XXXII (December 1994), pp. 1667–1717

The Economics of Immigration

By GEORGE J. BORJAS

University of California at San Diegoand National Bureau of Economic Research

I am grateful to Julian Betts, Daniel Hamermesh, James Rauch,and Stephen Trejo for useful comments, and to the National Sci-ence Foundation for research support.

1. Introduction

THERE HAS BEEN a resurgence of im-migration in the United States and in

many other countries. The United Na-tions estimates that over 60 million peo-ple, or 1.2 percent of the world’spopulation, now reside in a countrywhere they were not born (United Na-tions 1989, p. 61). Although most immi-grants choose a “traditional” destination(over half typically go to the UnitedStates, Canada, or Australia), many othercountries are receiving relatively largeimmigrant flows. Nearly 11 percent ofthe population in France, 17 percent inSwitzerland, and 9 percent in the UnitedKingdom is foreign-born. Even Japan,which is thought of as being very homo-geneous and geographically immune toimmigrants, now reports major problemswith illegal immigration.

As a result of these changes in the “im-migration market,” the impact of immi-gration on the host economy is now be-ing debated heatedly in many countries.The political discussion is centeredaround three substantive questions.First, how do immigrants perform in thehost country’s economy? Second, whatimpact do immigrants have on the em-

ployment opportunities of natives? Fi-nally, which immigration policy mostbenefits the host country?

The policy significance of these ques-tions is evident. For example, immi-grants who have high levels of productiv-ity and who adapt rapidly to conditionsin the host country’s labor market canmake a significant contribution to eco-nomic growth. Natives need not be con-cerned about the possibility that theseimmigrants will increase expenditures onsocial assistance programs. Conversely, ifimmigrants lack the skills that employersdemand and find it difficult to adapt, im-migration may significantly increase thecosts associated with income mainte-nance programs as well as exacerbate theethnic wage differentials already in exist-ence in the host country.

Similarly, the debate over immigrationpolicy has long been fueled by the wide-spread perception that “immigranthordes” have an adverse effect on theemployment opportunities of natives.Which native workers are most adverselyaffected by immigration, and how largeis the decline in the native wage?

Finally, there is great diversity in im-migration policies across countries. Somecountries, such as the United States,

1667

award entry visas mainly to applicantswho have relatives already residing inthe country. Other countries, such asAustralia and Canada, award visas to per-sons who have a desirable set of socio-economic characteristics, and still othercountries, such as Germany, encouragedthe migration of “temporary” guest work-ers in the 1960s, only to find that thetemporary migrants became a permanentpart of the German population. Thechoice of the “right” immigration policycan obviously have a significant impacton economic activity both in the shortrun and in the long run.

The past decade witnessed an explo-sion in research on many aspects of theeconomics of immigration. This litera-ture is motivated mainly by the variouspolicy concerns and provides valuable in-sights into all these issues. This paperdoes not attempt to provide an encyclo-pedic summary of the empirical resultsin the literature; instead, it surveys thethemes and lessons suggested by the on-going research. Perhaps the most impor-tant theme is that an assessment of theeconomic impact of immigration re-

quires an understanding of the factorsthat motivate persons in the sourcecountries to emigrate and of the eco-nomic consequences of pursuing particu-lar immigration policies. As a result, themost important lesson is that the eco-nomic impact of immigration will vary bytime and by place, and can be eitherbeneficial or harmful. Although the dis-cussion focuses on the experience of theUnited States (simply because most stud-ies in the literature use data drawn fromthe U.S. decennial Censuses), we willsee that much can be learned by compar-ing the U.S. experience to that of otherhost countries.

2. Immigration to the United States: ABrief History

As Table 1 shows, the size of the immi-grant flow has fluctuated dramaticallyduring the past century. The First GreatMigration occurred between 1881 and1924, when 25.8 million persons enteredthe country. Reacting to the increase inimmigration and to the widespread per-ception that the “new” immigrants dif-

TABLE 1LEGAL IMMIGRANT FLOW TO THE UNITED STATES 1881–1990

DecadeImmigrant Flow

(in 1000s)

Immigrant Flow asPercentage of Change

in Population

Percentage of Populationthat is Foreign-Born at

End of Decade

1881–1890 5,246.6 41.0 14.71891–1900 3,687.6 28.3 13.61901–1910 8,795.4 53.9 14.61911–1920 5,735.8 40.8 13.21921–1930 4,107.2 24.6 11.61931–1940 528.4 5.9 8.81941–1950 1,035.0 5.3 6.91951–1960 2,515.5 8.7 5.41961–1970 3,321.7 13.7 4.71971–1980 4,493.3 20.7 6.21981–1990 7,338.1 33.1 7.9

Sources: U.S. Department of Justice. Immigration and Naturalization Service (1993, p. 25); U.S. Department ofCommerce. Bureau of the Census (1975, pp. 8, 14; 1993b, p. 50).

1668 Journal of Economic Literature, Vol. XXXII (December 1994)

fered from the old, Congress closed thefloodgates in the 1920s by enacting thenational-origins quota system. This sys-tem restricted the annual flow fromEastern Hemisphere countries to150,000 immigrants, and allocated the vi-sas according to the ethnic compositionof the U.S. population in 1920. As aresult, 60 percent of all available visaswere awarded to applicants from twocountries, Germany and the UnitedKingdom.

During the 1930s, only .5 million im-migrants entered the United States.Since then, the number of legal immi-grants has increased at the rate of aboutone million per decade, and is now near-ing the historic levels reached in theearly 1900s. By 1993, nearly 800,000 per-sons were being admitted annually.There has also been a steady increase inthe number of illegal aliens. Demo-graphic studies conclude that around twoto three million persons were illegallypresent in the United States in the late1980s, and that the net flow of illegalaliens is on the order of 200,000 to300,000 persons per year (U.S. GeneralAccounting Office 1993).

Table 1 also illustrates that the size ofthe immigrant flow has increased notonly in absolute terms, but also as a per-centage of population growth. In fact,the contribution of the Second Great Mi-gration to population growth is fast ap-proaching the level reached during theFirst Great Migration, when immigrationaccounted for 40 to 50 percent of thechange in population. As a result of thesetrends, the fraction of the populationthat is foreign-born rose from 4.7 to 7.9percent between 1970 and 1990.

The huge increase in immigration inrecent decades can be attributable partlyto changes in U.S. immigration policy.Prior to 1965, immigration was guided bythe national-origins quota system. The1965 Amendments to the Immigration

and Nationality Act (and subsequent re-visions) repealed the national origin re-strictions, increased the number of avail-able visas, and made family ties to U.S.residents the key factor that determineswhether an applicant is admitted into thecountry. As a consequence of both the1965 Amendments and of major changesin economic and political conditions inthe source countries relative to theUnited States, the national origin mix ofthe immigrant flow changed substantiallyin the past few decades. As Table 2shows, over two-thirds of the legal immi-grants admitted during the 1950s origi-nated in Europe or Canada, 25 percentoriginated in Western Hemisphere coun-tries other than Canada, and only 6 per-cent originated in Asia. By the 1980s,only 13 percent of the immigrants origi-nated in Europe or Canada, 47 percentin Western Hemisphere countries otherthan Canada, and an additional 37 per-cent originated in Asia.

In recent years, the debate over immi-gration policy led to the enactment oftwo major pieces of legislation. Fueledby charges that illegal aliens were over-running the country, Congress enactedthe 1986 Immigration Reform and Con-trol Act (IRCA). This legislation gaveamnesty to three million illegal aliensand introduced a system of employersanctions designed to stem the flow ofadditional illegal workers.1 The 1990 Im-migration Act permits the entry of an ad-ditional 150,000 legal immigrants annu-ally. The legislated increase in the size ofthe immigrant flow makes it likely thatthe United States will admit a recordnumber of immigrants during the 1990s.

1 In 1986, the Border Patrol apprehended 1.8million illegal aliens. Although the number of an-nual apprehensions declined to about one millionfollowing the enactment of IRCA, they are nowback up to about 1.3 million, or 2.5 apprehensionsper minute (U.S. Department of Justice. Immigra-tion and Naturalization Service 1993, p. 156).

Borjas: The Economics of Immigration 1669

TABLE 2NATIONAL ORIGIN COMPOSITION OF LEGAL IMMIGRANT FLOW TO UNITED STATES, 1931–1990

1931–40 1941–50 1951–60 1961–70 1971–80 1981–90

Number of Immigrants (in 1000s)

All Countries 528.4 1035.0 2515.5 3321.7 4493.3 7338.1

Europe 347.6 621.1 1325.7 1123.5 800.4 761.6 Germany 114.1 226.6 477.8 190.8 74.4 92.0 Greece 9.1 9.0 47.6 86.0 92.4 38.4 Ireland 11.0 19.8 48.4 33.0 11.5 32.0 Italy 68.0 57.7 185.5 214.1 129.4 67.3 Poland 17.0 7.6 10.0 53.5 37.2 83.3 United Kingdom 31.6 139.3 202.8 213.8 137.4 159.2

Asia 16.6 37.0 153.2 427.6 1588.2 2738.2 China 4.9 16.7 9.7 34.8 124.3 346.7 India 0.5 1.4 3.4 10.3 164.1 250.8 Iran 0 0.5 25.5 29.6 45.1 116.2 Japan 1.9 1.6 46.3 40.0 49.8 47.1 Korea 0 0.1 6.2 34.5 267.6 333.7 Philippines 0.5 4.7 19.3 98.4 355.0 548.8 Vietnam 0 0 0.3 4.3 172.8 280.8

America 160.0 354.8 996.9 1716.4 1982.7 3615.2 Canada 108.5 171.7 378.0 413.3 169.9 156.9 Mexico 22.3 60.6 299.8 453.9 640.3 1655.8 Cuba 9.6 26.3 78.9 208.5 264.9 144.6 Dominican Republic 1.2 5.6 9.9 93.3 148.1 252.0 Haiti 0.2 0.9 4.4 34.5 56.3 138.4

Africa 1.8 7.4 14.1 29.0 80.8 176.9

Oceania 2.5 14.6 13.0 25.1 41.2 45.2Percentage Distribution

Europe 65.8 60.0 52.7 33.8 17.8 10.4 Germany 21.6 21.9 19.0 5.7 1.7 1.3 Greece 1.7 .9 1.9 2.6 2.1 .5 Ireland 2.1 1.9 1.9 1.0 .3 .4 Italy 12.9 5.6 7.4 6.4 2.9 .9 Poland 3.2 .7 .4 1.6 .8 1.1 United Kingdom 6.0 13.5 8.1 6.4 3.1 2.2

Asia 3.1 3.6 6.1 12.9 35.3 37.3 China .9 1.6 .4 1.0 2.8 4.7 India .1 .1 .1 .3 3.7 3.4 Iran .0 .0 1.0 .9 1.0 1.6 Japan .4 .2 1.8 1.2 1.1 .6 Korea .0 .0 .2 1.0 6.0 4.5 Philippines .1 .5 .8 3.0 7.9 7.5 Vietnam .0 .0 .0 .1 3.8 3.8

America 30.3 34.3 39.6 51.7 44.1 49.3 Canada 20.5 16.6 15.0 12.4 3.8 2.1 Mexico 4.2 5.9 11.9 13.7 14.3 22.6 Cuba 1.8 2.5 3.1 6.3 5.9 2.0 Dominican Republic .2 .5 .4 2.8 3.3 3.4 Haiti .0 .1 .2 1.0 1.3 1.9

Africa .3 .7 .6 .9 1.8 2.4

Oceania .5 1.4 .5 .8 .9 .6

Source: U.S. Department of Justice. Immigration and Naturalization Service (1993, pp. 27–28).

3. How Do Immigrants Perform in theHost Country?

Many studies in the modern economicliterature on immigration focus on deter-mining the trends in the skill level andearnings of the immigrant population inthe host country.2 These studies view thelabor market performance of immigrantsin the host country as a measure of theimmigrant contribution to the economy’sskill endowment and productivity. In ad-dition, the trends in immigrant skillshelp determine the impact of immigra-tion on the employment opportunities ofnative-born workers and on expendituresin social insurance programs.

A. Aging and Cohort Effects

The pioneering work of BarryChiswick (1978) and Geoffrey Carliner(1980) analyzed how immigrant skillsadapted to the host country’s labor mar-ket by estimating the cross-section re-gression model:

log wi = Xiφ + δAi + γoIi + γ1yi + εi, (1)

where wi is worker i’s wage rate; Xi is avector of socioeconomic characteristicswhich might include education and re-gion of residence; Ai gives the worker’sage or potential labor market experience;Ii is a dummy variable indicating if theworker is an immigrant; and yi gives thenumber of years an immigrant workerhas resided in the United States (and isset to zero for native-born workers). Inpractice, the model typically includeshigher-order polynomials in age andyears-since-migration, and the coeffi-cient vector (φ,δ) is allowed to vary be-tween immigrants and natives. For sim-

plicity, we restrict the discussion to thesimpler specification.

The coefficient γ0 gives the percentagewage differential between immigrantsand natives at the time of arrival, whilethe coefficient γ1 gives the rate at whichthe earnings of immigrants rise relativeto the earnings of natives. The earlystudies of wage determination among im-migrant and native men in the UnitedStates reached a quick consensus: the co-efficient γ0 was negative and the coeffi-cient γ1 was positive.3 The essence of theresults is summarized in Figure 1, whichillustrates the predicted immigrant andnative age-earnings profiles implied byChiswick’s analysis of the 1970 Census.At the time of arrival, immigrants earnabout 17 percent less than natives. Be-cause immigrants experience faster wagegrowth, immigrant earnings “overtake”native earnings within 15 years after arri-val. After 30 years in the United States,the typical immigrant earns about 11percent more than a comparable nativeworker.

Two distinct arguments were used toexplain these results. At the time of arri-val, immigrants earn less than natives be-cause they lack the U.S.-specific skillsthat are rewarded in the American labormarket (such as English proficiency). Asthese skills are acquired, the humancapital stock of immigrants grows rela-tive to that of natives, and immigrantsexperience faster wage growth. The hu-

2 These questions are not restricted to the mod-ern literature. Paul Douglas (1919), for example,analyzed the occupational distribution of immi-grants who arrived during the First Great Migra-tion to determine if the newer immigrants were asskilled as the old.

3 There is a widespread, though erroneous, per-ception that studies based on cross-section datafrom other countries and other time periods reachsimilar conclusions. However, Chiswick’s (1980)study of immigrants in Britain reports that years-since-migration has no impact on immigrant earn-ings. Similarly, both Francine Blau (1979) andBarry Eichengreen and Henry Gemery (1986) ana-lyze the economic mobility of immigrants who en-tered the United States at the turn of the 20thcentury, but reach conflicting conclusions. Blaufinds wage convergence between immigrants andnatives, while Eichengreen and Gemery find littlewage convergence between the two groups.


man capital investment hypothesis, how-ever, does not by itself generate an over-taking point. After all, why would immi-grants accumulate more human capitalthan natives? The overtaking point wasinstead interpreted in terms of a selec-tion argument: immigrants are “moreable and more highly motivated” thannatives (Chiswick 1978, p. 900), and im-migrants “choose to work longer andharder than nonmigrants” (Carliner1980, p. 89). This assumption was typi-cally justified by arguing that only themost driven and most able persons havethe ambition and wherewithal to packup, move, and start life anew in a foreigncountry.

The optimistic appraisal of immigrantadjustment implied by the results sum-marized in Figure 1 was challenged byBorjas (1985), who argued that the posi-tive cross-section correlation betweenthe relative wage of immigrants andyears-since-migration need not indicatethat the wage of immigrants converges tothat of natives. The basic problem withthe “assimilationist” interpretation of the

regression in (1) is that it draws infer-ences about how the earnings of immi-grant workers evolve over time from asingle snapshot of the immigrant popula-tion. It might be the case, however, thatnewly arrived immigrants are inherentlydifferent from those who migratedtwenty years ago. Hence we cannot usethe current labor market experiences ofthose who arrived twenty years ago toforecast the future earnings of newly ar-rived immigrants.

Figure 2 illustrates the implications ofthis alternative hypothesis. For concrete-ness, consider a situation where thereare three separate immigrant waves, onewave arrived in 1950, the second in1970, and the last in 1990. Assume thatimmigrants enter the United States atage 20. The earliest cohort is assumed tohave the highest productivity level of anygroup in the population, including U.S.-born workers. If we could observe theirearnings in every year after they arrive inthe United States, their age-earningsprofile would be given by the line PP inthe figure. Let’s also assume that the lastimmigrant wave (i.e., the 1990 arrivals) isthe least productive of any group in thepopulation. Their age-earnings profile isgiven by the line RR in the figure. Fi-

Log Earnings in 19709.1

9.0

8.9

8.8

8.7

8.6

8.5

8.4

of Immigrants and Natives in the United States, 1970

Source: Chiswick (1978, Table 2, Column 3). All the variables in the regression are evaluated at the means of the immigrant sample, and immigrants are assumed to enter the United States at age 20.

Figure 1. The Cross-Section Age-Earnings Profiles

20 25 30 35 40 45 50 55 60 65

Natives

Immigrants

Age

Wage

P

Age-Earnings Profile of ImmigrantsFigure 2. Cohort Effects and the Cross-Section

20 40 60

Q

RC

1950 Cohort

1970 Cohortand Natives

1990 Cohort

Age

P

Q

R

C


nally, suppose that the 1970 wave has thesame skills as natives; the age-earningsprofiles of the two groups is given byQQ. There is no wage convergence be-tween immigrants and natives in this hy-pothetical example.

Suppose we have access to data drawnfrom the 1990 Census cross-section.These data allow us to identify only onepoint on each of the immigrant age-earn-ings profiles. In particular, we can ob-serve the earnings of the immigrants whoarrived in 1990 when they are 20 yearsold; the earnings of the 1970 arrivals atage 40; and the earnings of the 1950 arri-vals at age 60. The age-earnings profilegenerated by the cross-section data,therefore, is given by the line CC in Fig-ure 2. The cross-section regression lineis steeper than the native age-earningsprofile, making it seem as if there iswage convergence between immigrantsand natives, when in fact there is none.Moreover, the cross-section regressionline crosses the native age-earnings pro-file at age 40, making it seem as if immi-grant earnings overtake native earningsafter 20 years in the United States, whenin fact no immigrant group experiencedsuch an overtaking.

Figure 2 shows how a cross-section re-gression can yield erroneous insightsabout the adaptation process experi-enced by immigrants if there are intrin-sic differences in productivity across im-migrant cohorts (or “cohort effects”).Cohort effects can arise as a result ofchanges in immigration policy. For ex-ample, the 1965 Amendments de-empha-sized the role of skills in allocating entryvisas, and instead makes these awardsbased almost entirely on whether the ap-plicant has family ties with current U.S.residents. If this policy shift generated aless-skilled immigrant flow, the cross-section finding that more recent immi-grants earn less than earlier immigrantssays little about wage convergence, but

instead may reflect innate differences inability or skills across cohorts.

Cohort effects may also arise as a re-sult of changes in economic or politicalconditions in the source countries and inthe United States. Even if the UnitedStates had not adopted the 1965 Amend-ments, improving economic conditions inWestern Europe would have reduced thenumber of immigrants from these “tradi-tional” source countries. The changingnational origin mix of the immigrant flowgenerates cohort effects if skill levelsvary across countries or if skills from dif-ferent countries are not equally transfer-able to the United States. Finally, cohortdifferences in average productivity willbe observed in a cross-section whenthere is nonrandom return migration. Iflow-wage immigrant workers return totheir source countries, the earlier waveshave been “weeded out” and will haverelatively higher earnings than more re-cent waves.

It is evident that both the immigrantand native populations must be “tracked”over time to correctly measure wage con-vergence between immigrants and na-tives. Most longitudinal data sets eithercontain very few immigrants or providenonrandom samples of the foreign-bornpopulation. As a result, the literature haspursued the alternative of creating syn-thetic cohorts of immigrants by trackingspecific immigrant waves across the de-cennial Censuses or across the CurrentPopulation Surveys (CPS). The empiricalevidence typically found in these studiesis summarized in Table 3, which reportsthe unadjusted percentage wage differ-ential between immigrant and nativemen in each of the decennial Censusesbetween 1970 and 1990.4

4 The calculations use a 1/500 random sample ofnative workers and a 5/100 random sample of im-migrant workers in each Census (except in 1970when the immigrant extract forms a 2/100 randomsample). The resulting data set contains 920,700


Each Census cross-section shows thatimmigrants who have been in the UnitedStates for several decades have higherwages than natives, while more recentarrivals have lower wages. In 1990, forexample, immigrants who arrived in theUnited States between 1950 and 1960earned 19.6 percent more than natives,while immigrants who arrived between1985 and 1989 earned 31.7 percent less.The data, however, also support the hy-pothesis that there exist cohort effects inthe foreign-born population, with morerecent immigrant cohorts having rela-tively lower wage rates. For example, themost recent cohort enumerated in the1970 Census (i.e., the 1965–1969 arri-vals) earned only 16.6 percent less thannatives in 1970; the wage gap betweenthe most recent arrivals and natives grew

to 27.6 percent by 1980, and to 31.7 per-cent by 1990.5

Because of these cohort effects, thecross-section relationship between therelative wage of immigrants and years-since-migration overestimates the wagegrowth actually experienced by a particu-lar cohort. The 1990 cross-section sug-gests that over a 20-year period (1970 to1990), the relative earnings of immi-grants grow by about 33 percentagepoints.6 In fact, the relative wage of the1965-1969 wave increased by only 18percentage points over the 20-year pe-riod, or about half of the cross-sectionrate of convergence.

The implications of the data summa-rized in Table 3 are clear and provoca-tive. If we interpret the difference inwages between immigrants and nativesas a measure of relative skills, more re-cent immigrant waves are relatively lessskilled than earlier waves. Moreover, im-migrant wage growth is more sluggishthan suggested by the early cross-sectionstudies. It is extremely unlikely that theearnings of more recent cohorts will everreach parity with (let alone overtake) theearnings of natives.

observations. The percent wage differential be-tween immigrants and natives equals 100(ex − 1),where x is the difference in average log wages be-tween the groups. See Borjas (forthcoming) for amore detailed discussion of the data and of thetrends in immigrant earnings.

TABLE 3 PERCENTAGE WAGE DIFFERENTIAL BETWEEN

IMMIGRANT AND NATIVE MEN, 1970–1990

Group: 1970 1980 1990

All Immigrants .9 −9.2 −15.2Cohort: 1985–1989 Arrivals — — −31.7 1980–1984 Arrivals — — −27.8 1975–1979 Arrivals — −27.6 −17.8 1970–1974 Arrivals — −18.9 −9.3 1965–1969 Arrivals −16.6 -7.8 1.1 1960–1964 Arrivals −4.4 .1 9.0 1950–1959 Arrivals 5.6 5.7 19.6 Pre-1950 Arrivals 10.3 10.6 26.2

Source: Author’s tabulations from the 1970, 1980, and1990 Public Use Samples of the U.S. Census. Thestatistics are calculated in the subsample of menaged 25–64 who work in the civilian sector, who arenot self-employed, and who do not reside in groupquarters.

5 These results differ slightly from those re-ported by Edward Funkhouser and Trejo (forth-coming), who use CPS data from various supple-ments to describe the trend in immigrant skillsduring the 1980s. The CPS data indicate that thedecline in relative skills was reversed somewhat bythe late 1980s. The CPS, however, contains rela-tively small samples of immigrants. In addition,the national origin composition of immigrant co-horts is extremely unstable across CPS surveys.For instance, 21 percent of the cohort that immi-grated between 1982 and 1984 in the June 1988CPS is of Mexican origin, while the respective sta-tistic for the same cohort in the November 1989CPS is 37 percent. These statistics suggest that thechange in the relative immigrant wage across theCurrent Population Surveys provides unreliablemeasures of both cohort effects and of the rate ofwage convergence.

6 This statistic is calculated by comparing therelative wage of the immigrants who arrived in thelate 1980s with the relative wage of the immi-grants who arrived in the late 1960s.


Needless to say, these findings havegenerated a great deal of controversyand debate. (See, for example, Chiswick1986; Harriet Orcutt Duleep and MarkRegets 1992b; Robert LaLonde andRobert Topel 1992; and AndrewYuengert 1994.) Many of these studies(including the original work of Borjas,1985) point out that interpreting the in-tercensal trend in the relative wage ofimmigrants as a measure of relativechanges in skills implicitly assumes thatperiod effects influence the wage of im-migrants and natives by the same relativeamount. To see this point formally, con-sider the following generic model thatcharacterizes the analytical frameworknow used in the literature. Suppose wepool all the data in two cross-sections(such as the 1980 and 1990 Censuses)and estimate the regression equations:

log wij = Xjφi + δiAj + αyj + βCj + γiπj + εi j, (2)

log wnl = Xlφn + δnAl + γnπl + εnl, (3)

where wij gives the wage of immigrant j;wnl gives the wage of native l; X gives avector of standardizing socioeconomiccharacteristics; A gives the worker’s ageat the time of the Census; y gives thenumber of years that the immigrant hasresided in the United States; C is the cal-endar year of arrival in the UnitedStates; and π is a dummy variable in-dicating if the observation was drawnfrom the 1990 Census. To easily illus-trate the identification problem, theage, years-since-migration, and calendaryear-of-arrival variables are entered lin-early.

The coefficients γi and γn give the pe-riod effects for immigrants and natives,respectively. The coefficient δn gives theaging effect for natives; the rate at whichnative earnings increase over the life cy-cle. The respective aging effect for im-migrants is given by δi + α. The age-

earnings profiles of immigrants and na-tives converge if (δi + α) > δn (assumingimmigrants earn less than natives at thetime of arrival).7 Finally, the coefficientβ measures the cohort effect, the rate ofchange in the entry wage across immi-grant cohorts.8

It is well known that the key parame-ters of the regression model in equations(2) and (3) are not identified. The years-since-migration variable is a linear com-bination of the period effect and the co-hort variable:

yi ≡ πi(1990 − Ci) + (1 − πi) (1980 − Ci) = 1980 − Ci + 10π. (4)

In order to identify the period effects,the aging effects, and the cohort effect,therefore, a restriction must be imposedon the model. One possible restriction isthat the period effects are the same forimmigrants and natives, or:

γi = γn. (5)

Equation (5) implies that the relativewage of immigrants and natives is inde-pendent of secular changes in the wagelevel. We implicitly imposed this restric-tion on the data when we interpreted theintercensal trends in Table 3 as changesin the relative skills of immigrants. Bynetting out the secular trend in the na-tive wage (i.e., by using a difference-in-differences estimator), we are simply leftwith the trend in immigrant productivity.Note, however, that the wage is theproduct of the rate of return to skillstimes the worker’s human capital stock.

7 Although the regression model in (2) assumesthat the aging effect is the same for all immigrantcohorts, many of the empirical studies in the lit-erature relax this assumption.

8 The model assumes that there are no cohorteffects in the native population (perhaps due tochanges in the quality of education). Even thoughthis is a standard assumption in the literature, theestimated cohort effects in the immigrant popula-tion may be sensitive to the existence of cohorteffects among native workers.


If period effects influence the price ofskills differently for immigrants and na-tives, the intercensal change in relativewages could be reflecting differences inprices rather than differences in humancapital.

There were historic changes in theU.S. wage structure during the 1980sand these changes did not affect all skillgroups equally (Frank Levy and RichardMurnane 1992). In particular, there wasa sizable increase in the wage gap be-tween highly educated and less educatedworkers; and among workers within nar-rowly defined occupation and industrycells. It is unlikely that these changes inthe wage structure affected the earningsof immigrant and native workers by thesame percentage amount. The immigrantpopulation in the United States is rela-tively unskilled (at least in terms of edu-cational attainment). Because the rate ofreturn to skills increased during the1980s, the relative wage of immigrants

would have fallen between 1980 and1990 even if immigrant skills had re-mained constant. In other words, thechanges in the wage structure could ac-count for both the observed decline inthe relative wage of successive immi-grant cohorts and for the sluggish wagegrowth experienced by a particular co-hort as it entered the 1980s.

It is unlikely, however, that changes inthe wage structure account for the down-ward trend in relative wages across suc-cessive immigrant cohorts or for the slowwage convergence between immigrantsand natives. Consider the trends in im-migrant educational attainment, a skillmeasure that is invariant to changes inthe wage structure. Table 4 documentsthe changes in the schooling distributionof immigrants and natives in the past twodecades. In 1970, 39.6 percent of nativeswere high school dropouts; by 1990, only14.8 percent of natives lacked a highschool diploma. Among immigrants, 48.2

TABLE 4 EDUCATIONAL ATTAINMENT OF IMMIGRANT AND NATIVE MEN, 1970–1990

1970 1980 1990

Group

PercentHigh School Dropouts

PercentCollege

Graduates


PercentCollege

Graduates


PercentCollege

Graduates

Natives 39.6 15.4 23.1 22.9 14.8 26.6Immigrants 48.2 18.9 37.4 25.3 36.9 26.6

Cohort: 1985–89 Arrivals — — — — 35.2 31.5 1980–84 Arrivals — — — — 40.4 24.1 1975–79 Arrivals — — 36.2 30.4 42.2 24.8 1970–74 Arrivals — — 44.0 24.9 42.7 24.1 1965–69 Arrivals 45.2 28.3 41.6 24.7 34.1 26.2 1960–64 Arrivals 44.8 21.1 34.7 24.8 27.5 27.9 1950–59 Arrivals 47.4 17.1 31.4 23.7 25.9 27.8 Pre-1950 Arrivals 51.7 15.0 35.3 21.6 25.2 31.8

Source: Author’s tabulations from the 1970, 1980, and 1990 Public Use Samples of the U.S. Census. The statisticsare calculated in the subsample of men aged 25–64 who work in the civilian sector, who are not self-employed, andwho do not reside in group quarters.

1 LINE SHORT


percent were dropouts in 1970, 37.4 per-cent in 1980, and 36.9 percent in 1990.Relative to natives, immigrants wereabout 21.7 percent more likely to behigh school dropouts in 1970, but arenow more than twice as likely to be highschool dropouts.

Moreover, even though the percentageof immigrant workers who are collegegraduates rose during the period, thepercentage of natives who are collegegraduates rose even faster. Immigrantswere more likely to be college gradu-ates in 1970 (18.9% for immigrants ascompared to 15.4% for natives). By1990, both groups had exactly the sameprobability of being college graduates(26.6%). Therefore, changes in the“quantity” of immigrants’ human capitalare partly responsible for the decline inthe relative immigrant wage.

It is also easy to show that changes inthe U.S. wage structure were not suffi-ciently large to account for a sizable partof the declining relative wage of immi-grants across successive waves. For ex-ample, we know that the wage structurechanged in different ways for variousage-education groups, with groups withmore education and experience havinglarger wage growth between 1970 and1990. We can then use the wage growthobserved in 56 age-education cellsamong native workers to “deflate” thewage growth of immigrants in the sameage-education cells.9 To take into ac-count changes in wage inequality even

within age and education cells, LaLondeand Topel (1992) suggest using a defla-tor based on an immigrant’s ranking inthe native wage distribution. If all work-ers who fall in the pth percentile of thewage distribution are equally skilled,then we can use the wage growth experi-enced by natives in the pth percentile todeflate the wage growth of immigrantswho fall in the same percentile in the1970–1990 period.10

Table 5 reports the changes in the de-flated relative wage of immigrants be-tween 1970 and 1990. Regardless ofwhich deflator is used, more recent im-migrant cohorts have substantially lowerrelative wages than earlier cohorts. Themost recent cohort in 1970 earned 16.6percent less than natives at the time ofarrival. The most recent cohort in 1990earned 29.5 percent less than natives ifwe use the deflator based on age-educa-tion cells, and 29.4 percent less if we usethe percentile deflator. The change inthe wage structure, therefore, accountsfor only 15 percent of the drop in therelative immigrant wage between 1970and 1990.

The cohort and aging effects calcu-lated from the synthetic cohorts in theCensus data may be biased because thesample composition of a particular immi-grant cohort changes systematicallyacross Censuses. Perhaps one-third ofimmigrants in the United States eventu-

9 The eight age categories are: 25–29 years old;30–34; 35–39; 40–44; 45–49; 50–54; 55–59; and60–64. The seven education categories are: lessthan 8 years of schooling; 9 years; 10–11 years; 12years; 13–15 years; 16 years; and more than 16years. Define Drs(t) to be the wage growth experi-enced by the typical native worker in age group rand education group s between 1970 and year t(t =1980, 1990). The deflated wage is then given bylog wl,rs(t) = log wl,rs(t) − ∆rs(t), where log wl,rs(t) isthe log wage of person l in skill group rs in Censusyear t.

10 Neither deflator fully solves the problem ofaccounting for changes in the wage structure. Theage-education deflator, for example, ignores theincrease in inequality that occurred within age-education cells. The percentile deflator assumesthat immigrants and natives in the pth percentileare perfect substitutes. This is unlikely to be true.Newly arrived immigrants might place badly in thenative wage ranking not because they are un-skilled, but because they are going through aninitial “testing” period. In the end, therefore, animmigrant who initially places in the pth percentilemay have skills that are comparable to those ofnatives in the (p + q)th percentile, where q > 0.

1 LINE SHORT


ally return to their countries of origin(Robert Warren and Jennifer Peck1980). Suppose that the return migrantsare mainly composed of workers withlower than average wages (i.e., the “fail-ures”). The intercensal tracking of a par-ticular cohort would reveal an improve-ment in relative wages even if no wageconvergence is taking place. Alterna-tively, if the return migrants are “suc-cesses,” the rate of wage convergencewould be underestimated. Because dataon the size and composition of the returnmigration flow is scarce, few studies sys-tematically analyze the selection mecha-nism generating the return migrationflow (the limited available evidence isdiscussed in the next section). As a re-sult, the bias introduced by nonrandomreturn migration is typically ignored.

Even if there were no return migra-tion, Rachel Friedberg (1992) and JamesSmith (1992) have shown that the samplecomposition of a particular immigrantcohort changes over time because thesample of working-aged immigrants in

later Censuses includes a larger numberof persons who migrated as children.11

The economic experiences of “immigrantchildren” may resemble those faced bynative workers. The inclusion of the im-migrant children in later Censuses thusbiases the estimated rate of wage conver-gence upward. A better measure of wageconvergence, therefore, is obtained bytracking a specific immigrant cohort, de-fined in terms of both year-of-migrationand age-at-arrival, across the variousCensuses.

Table 6 summarizes the trend in thepercent wage differential between a par-ticular group of immigrants and similarlyaged natives, so that immigrants who ar-rived when they were between 25 and 34years old in the late 1960s are compared

TABLE 5PERCENTAGE WAGE DIFFERENTIAL BETWEEN IMMIGRANT, AND NATIVE, MEN, 1970–1990, DEFLATED BY

CHANGES IN WAGE STRUCTURE

Using Age-EducationDeflator Using Percentile Deflator

Group: 1970 1980 1990 1980 1990

All Immigrants .9 −9.4 −14.4 −8.6 −13.9

Cohort: 1985–1989 Arrivals — — −29.5 — −29.4 1980–1984 Arrivals — — −25.0 — −25.4 1975–1979 Arrivals — −25.2 −15.8 −26.2 −16.0 1970–1974 Arrivals — −17.5 −8.8 −17.9 -8.3 1965–1969 Arrivals −16.6 −8.2 −.2 −7.2 1.1 1960–1964 Arrivals −4.4 −1.0 6.0 .2 7.9 1950–1959 Arrivals 5.6 3.9 13.1 5.4 17.1 Pre-1950 Arrivals 10.3 4.7 16.0 10.2 23.2


11 An earlier study by Sherrie Kossoudji (1989)used the 1976 Survey of Income and Educationcross-section to estimate models of occupationalmobility which differentiate between persons whomigrated as children and those who migrated asadults. She finds that controlling for age-at-migra-tion leads to flatter occupational mobility profilesamong immigrants than among natives.


to natives aged 25–34 in 1970, to natives35–44 in 1980, and to natives aged 45–54in 1990. About half of the wage conver-gence implied by the statistics presentedin Table 5 disappears after controllingfor age-at-migration. Consider, for exam-ple, the group of immigrants who arrivedbetween 1965 and 1969 and who were25–34 years old in 1970. They earned12.0 percent less than natives in 1970and 2.5 percent less in 1990. Over a 20-

year period, therefore, the relative wageof this immigrant cohort increased by 10percentage points, in contrast to the 18percent growth suggested by the inter-censal comparison that does not controlfor age-at-migration and to the 33 per-cent growth implied by the 1990 cross-section.

Table 6 reveals that practically all im-migrants, regardless of when they ar-rived in the country, experience the

TABLE 6PERCENTAGE WAGE DIFFERENTIAL BETWEEN IMMIGRANTS AND NATIVES, BY AGE GROUP

AND YEAR OF ARRIVAL

Actual Wage Actual WageUsing Age-Education

Deflator

Cohort/Age Group: 1970 1980 1990 1980 1990

1960–1964 Arrivals: 15–24 in 1970 — 1.1 4.2 .9 4.5 25–34 in 1970 3.1 −.3 −.2 .0 .1 35–44 in 1970 −6.0 −6.7 1.1 −6.7 1.4 45–54 in 1970 11.1 −10.8 — −10.9 —

1965–1969 Arrivals: 15–24 in 1970 — −4.6 −6.9 −6.2 −5.5 25–34 in 1970 −12.0 −5.9 −2.5 −5.4 −2.3 35–44 in 1970 −15.9 −15.3 −8.8 −15.5 −8.3 45–54 in 1970 −22.5 −21.1 — −21.6 —

1970–1974 Arrivals: 25–34 in 1980 — −11.4 −11.8 −12.5 −10.4 35–44 in 1980 — −17.7 −16.4 −17.1 −15.6 45–54 in 1980 — −26.0 −20.7 −26.4 −20.0

1975–1979 Arrivals: 25–34 in 1980 — −21.3 −15.5 −21.2 −14.8 35–44 in 1980 — −24.9 −24.1 −24.2 −23.4 45–54 in 1980 — −29.8 −26.3 −29.8 −26.1

1980–1984 Arrivals: 25–34 in 1990 — — −18.6 — −18.2 35–44 in 1990 — — −25.3 — −24.5 45–54 in 1990 — — −34.0 — −33.0

1985–1989 Arrivals: 25–34 in 1990 — — −23.0 — −23.5 35–44 in 1990 — — −28.6 — −28.3 45–54 in 1990 — — −36.2 — −35.7



same sluggish relative wage growth. Thisresult is significant because it suggeststhat more recent immigrant cohorts havenot had faster wage growth despite theirlower starting positions.12 In fact, immi-grants who arrived during the 1970s ex-perienced the same wage growth asthose who arrived during the 1960s dur-ing their first decade in the UnitedStates. Immigrants who arrived between1975 and 1979 and were around age 30at the time of arrival earned 21.3 percentless than natives in 1980 and 15.5 per-cent less than natives in 1990, an in-crease of only 5.8 percentage points.This wage growth is similar to that expe-rienced by similarly aged immigrantswho arrived between 1965 and 1969;they earned 12.0 percent less than na-tives in 1970 and 5.9 percent less in1980.

Many studies have confirmed thatthere has been an overall decline in therelative skills of successive immigrant co-horts. For example, Yuengert (1994, p.86) finds that the relative wage of theimmigrants who migrated in the late1960s was about 9 percentage pointslower than the relative wage of thosewho arrived in the 1950s; LaLonde andTopel (1992, p. 89) report a 22 percent-age point drop in the relative wage ofimmigrants cohorts between the late1960s and the late 1970s; and Funk-houser and Trejo (forthcoming, Table 6)report a 10 percentage point drop duringthe same period. There is also a consen-

sus that much of the decline is due tochanges in “observables.” Both Funk-houser and Trejo (forthcoming, Table 6)and LaLonde and Topel (1992, p. 89)conclude that at least two-thirds of thedecline can be attributed to changes inthe educational attainment of immi-grants relative to natives. Some studiesalso show that the changing national ori-gin mix of the immigrant flow (which ob-viously implies changes in the observableskills of immigrants) accounts for muchof the decline in skills across successivecohorts. This result will be discussed indetail below.

B. Wage Convergence Between Immigrants and Ethnically Similar Natives

The data summarized in the previoussection describe how the immigrantwage adjusts relative to that of the typi-cal native worker. Because recent immi-grant waves start off at such a disadvan-tage, it is not too surprising that theirearnings fail to reach parity with theearnings of the average U.S.-born worker(who is typically a white person of Euro-pean ancestry). A number of studies thusinvestigate if immigrant earnings con-verge to the earnings of U.S.-born work-ers who share the same ethnic back-ground. These intra-ethnic comparisonscan help assess if the “new immigration”will exacerbate the ethnic differences al-ready prevalent in the U.S. labor market.

There is, however, little consensus onwhether the relative skills of immigrantsdeclined within specific ethnic groups,or on whether the wage of immigrantsconverges to that of ethnically similar na-tives. Most studies typically focus onfour large ethnic groups: Mexican immi-grants, other Hispanic immigrants, Asianimmigrants (excluding the Middle East),and “white” immigrants (defined as per-

12 Duleep and Regets (1992b) use the 1970 and1980 Censuses to estimate correlations betweenwage growth and entry wages across national ori-gin groups. These correlations tend to be negative,leading them to conclude that the low entry wageof the immigrants who arrived in the late 1970sdid not represent their true “quality” because theywould have faster wage growth than earlier immi-grants. The additional data provided by the 1990Census indicates that the less-skilled cohorts whomigrated in the 1970s did not, in fact, experiencefaster wage growth than earlier waves.

1 LINE SHORT


sons originating in Europe or Canada).13

The four native base groups are: Mexi-can-American natives (i.e., U.S.-bornpersons of Mexican ancestry); other His-panic-American natives (all other U.S.-born persons who report being of His-

panic ancestry); Asian-American natives(non-Hispanic persons whose race isAsian); and white natives (non-Hispanicwhites).

Table 7 summarizes the trends in thewage of immigrants in particular cohortsand age groups relative to ethnicallysimilar natives in the same age group, sothat Mexican immigrants aged 25–34 in1970 are contrasted with Mexican-American natives aged 25–34 in 1970,with Mexican-American natives aged 35–

1 LINE SHORT

13 Kristin Butcher (1994) describes the processof wage convergence between black immigrantsand U.S.-born black workers, and finds that thelabor market experience of black immigrants re-sembles that of black natives who had moved outof their state of birth.

TABLE 7PERCENTAGE WAGE DIFFERENTIAL BETWEEN IMMIGRANTS AND NATIVES OF SAME ETHNIC BACKGROUND

(Using Age/Education Deflator)

Mexican Other Hispanics Asian White

Cohort/Age Group 1970 1980 1990 1970 1980 1990 1970 1980 1990 1970 1980 1990

1960–64 Arrivals 15–24 in 1970 — −1.8 −5.1 — 20.3 29.2 — 1.9 .2 — 2.7 10.3 25–34 in 1970 −5.8 −9.6 −16.0 9.0 16.9 19.1 6.9 15.0 11.7 9.5 7.9 10.3 35–44 in 1970 −22.4 −19.7 −14.2 8.0 8.5 13.7 −15.3 4.1 17.1 4.5 3.9 14.3

1965–69 Arrivals 15–24 in 1970 — −11.7 −13.0 — .6 3.1 — 9.0 3.0 — .5 7.1 25–34 in 1970 −26.5 −16.5 −19.5 −15.8 −3.4 .1 −17.6 9.1 8.5 .3 2.9 12.3 35–44 in 1970 −32.5 −23.0 -29.2 −15.9 −9.8 −6.5 −15.2 −13.1 −5.6 −5.4 −6.2 9.4

1970–74 Arrivals 25–34 in 1980 — −19.5 −21.2 — −7.1 −1.0 — 2.7 5.3 — −2.9 8.1 35–44 in 1980 — −23.8 −29.3 — −11.7 −6.8 — −9.9 −3.9 — −6.9 3.0

1975–79 Arrivals 25–34 in 1980 — −33.8 −29.5 — −21.5 −16.7 — −19.7 −10.2 — −.6 11.7 35–44 in 1980 — −38.3 −36.7 — −22.4 −15.2 — −28.1 −25.8 — −1.8 4.2

1980–84 Arrivals 25–34 in 1990 — — −25.0 — — −19.7 — — −14.9 — — 12.4 35–44 in 1990 — — −39.6 — — −27.3 — — −28.8 — — 10.1

1985–89 Arrivals 25–34 in 1990 — — −33.9 — — −28.2 — — −24.3 — — 4.0 35–44 in 1990 — — −45.1 — — −36.2 — — −30.6 — — −1.2

Percent of ImmigrantPopulation Belongingto Particular EthnicGroup 9.7 18.5 26.2 11.4 13.1 16.1 8.6 16.4 21.7 62.4 36.8 21.5



44 in 1980, and with Mexican-Americannatives aged 45–54 in 1990. There areinteresting differences in the directionand magnitude of cohort effects acrossthe various groups. The relative wage ofsuccessive waves of Mexican immigrantsdeclined during the past two decades. In1970, the typical Mexican immigrantaged 25–34 who had just arrived in theUnited States earned 26.5 percent lessthan the typical Mexican-American na-tive; by 1990, the latest wave of Mexicanimmigrants earned 33.9 percent less thantheir native counterparts. Note, how-ever, that the wage gap between Mexicanimmigrants and Mexican-American na-tives underestimates the “true” economicstatus of Mexican immigrants in theUnited States. After all, Mexican-Ameri-can natives are themselves a relativelydisadvantaged group, earning 16 percentless than the typical U.S.-born worker in1990.

The relative wage of other Hispanicimmigrants and Asian immigrants alsofell across successive cohorts. In contrastto these groups, the relative wage of suc-cessive waves of European and Canadianimmigrants rose slightly between 1970and 1990. The most recent “white” arri-vals (aged 25–34 at the time of arrival)earned .3 percent more than white na-tives in 1970, but by 1990 they earned 4percent more.

The raw data thus suggest that somegroups experienced a decline in relativewages across successive cohorts, whilewhite immigrants experienced an in-crease. The data also indicate that mostnon-Asian cohorts experienced a 5 to 10percentage point increase in their rela-tive wage between 1970 and 1990. Forinstance, the Mexican immigrant whohad just arrived in the United States in1970 and was between 25 and 34 yearsold earned 26.5 percent less than thetypical Mexican-American native. By1990, the wage gap had narrowed by only

7 percentage points. Similarly, the typi-cal white immigrant in the same agegroup who had just arrived in the UnitedStates in 1970 earned .3 percent morethan white natives, and this wage gapgrew to 12.3 percent by 1990. This rateof wage convergence allows white immi-grants to substantially outperform whitenatives after 20 years in the UnitedStates, but prevents Mexican immigrantsfrom reaching wage parity with Mexican-American natives.

Finally, there seems to be a structuralshift in the rate of wage convergence forAsian immigrants who migrated after1970. Asian immigrants who migrated inthe 1960s experienced a very high rate ofwage convergence. The typical Asian im-migrant who arrived in the late 1960s(and was 25–34 years old at the time ofarrival) earned 17.6 percent less thanAsian-American natives in 1970, andabout 9 percent more in both 1980 and1990. In contrast, a similarly aged Asianimmigrant who arrived in the late 1970searned 19.7 percent less than Asian-American natives in 1980, and 10.2 per-cent less in 1990. In effect, this later co-hort of Asian immigrants has a rate ofwage convergence which is half of thatexperienced by earlier immigrant waves.

Table 7 thus suggests that there is agreat deal of diversity in the economicexperiences of various immigrant groupsin the United States. In view of this di-versity, it is not surprising that there is agreat deal of disagreement in the litera-ture (which is mostly based on compari-sons of the 1970 and 1980 Censuses) asto whether there has been a decline inthe average skill level of successive im-migrant waves within ethnic group, andon whether there is wage convergencewith ethnically similar natives. For exam-ple, Smith (1992, p. 79) concludes thatthere is “very little within-cohort wageassimilation for [Mexican] immigrantsacross their labor market careers,” and


that there is “strong evidence of declin-ing labor market quality.” In contrast,LaLonde and Topel (1992, p. 82) con-clude that Mexican immigrants “showsubstantial assimilation” with “no signifi-cant evidence of a decline in immigrantquality.” Even more striking, Yuengert’s(1994, p. 86) examination of the samedata reveals an increase in the skill levelof Mexican immigrant cohorts over time,with Mexicans who arrived between 1965and 1969 having 13 percent higher rela-tive earnings than those who arrived inthe 1950s.

There are many differences acrossthese studies which can potentially ex-plain the disparity in results. Smith, forexample, stresses the importance of con-trolling for age-at-migration when esti-mating the rate of wage convergence, avariable that LaLonde and Topel andYuengert ignore. In contrast, LaLondeand Topel stress the importance of con-trolling for the impact of changes in thewage structure on the wage of differentskill groups, a factor that Smith ignores.The data summarized in Table 7 con-trols both for age-at-migration and forchanges in the wage structure, as well asextends the span of time studied by an-other decade (using the 1990 Census).

Although intra-ethnic comparisons arecommon in the literature, there are anumber of conceptual problems in thesestudies that have not been sufficientlyappreciated. Most obvious is the aggre-gation bias introduced by pooling immi-grants from different countries into aparticular “ethnicity” (such as creatingthe Asian group by combining personsfrom countries as diverse as India, Ja-pan, and Vietnam). Because immigrantgroups from different countries differsubstantially, it is doubtful that the com-posite “other Hispanic” or “Asian” re-sembles the average person in any of thenational origin groups making up theethnic category. Moreover, there are siz-

able changes in the national origin mix ofthe immigrant flow over very short timeperiods even within a particular ethnicgroup. As a result, we do not know howto interpret the cohort effects or thechanges in the rate of wage convergenceamong Asians or other Hispanics unlesswe deal directly with a more primitivedefinition of ethnicity (i.e., the one thatcoincides with national origin).

Moreover, the composition of the na-tive base in these broadly defined ethnicgroups is changing systematically overtime. In 1970, for example, there werefew adult Cubans in the other Hispanic-American native sample. By 1990, as theU.S.-born children of the early Cubanwaves enter the labor market, the wageof the other Hispanic native base ispartly determined by the skill endow-ment of immigrant flows that arrived ageneration earlier. The comparison ofHispanic immigrants to Hispanic-Ameri-can natives in 1970 thus differs funda-mentally from the comparison of His-panic immigrants to Hispanic-Americannatives in 1990.

Most importantly, there is a sense inwhich these intra-group comparisonsmiss the point. What would we concludeif the relative wage of Mexican immi-grants converged to that of Mexican-American natives, or the relative wage ofAsian immigrants converged to that ofAsian-American natives? The fact re-mains that the wage of Mexican-Ameri-can natives is itself 16 percent below thatof the typical U.S.-born worker, whilethe wage of Asian-American natives is 12percent above. Intra-group convergenceis not an interesting phenomenon if wewant to identify the groups of nativeworkers who are most likely to be ad-versely affected by immigration, or if weare concerned about the impact of immi-gration policy on poverty rates, on thecosts of welfare programs, and on thecontribution of immigrants to the econ-


omy’s skill endowment. The costs andbenefits of immigration are more closelyrelated to how immigrants perform rela-tive to the population average than tohow immigrants perform relative to anonrandom subset of the population.

C. Language and the Process of Wage Convergence

Although there are many estimates ofthe rate of wage convergence betweenimmigrants and natives, we do not yetunderstand why some wage convergencetakes place. For the most part, the stud-ies investigating the differential accumu-lation of human capital by immigrantsand natives focus on one single factor,the acquisition of “language capital” inthe host country.

The early work of Gilles Grenier(1984), and Walter McManus, WilliamGould, and Finis Welch (1983) con-cluded that U.S. immigrants who areproficient in the English language havehigher earnings than immigrants who arenot.14 Grenier reports that Hispanic im-migrants who do not speak English pay a17 percent wage penalty, even after ad-justing for differences in education andother socioeconomic characteristics. Thiswage differential implies a $96,600 (in1993 dollars) increase in lifetime earn-ings for a Hispanic immigrant who be-comes proficient in the English language(McManus 1985). Presumably, profi-ciency in the host country’s language in-creases immigrant earnings because bi-lingualism opens up many employmentopportunities.

There also seems to be a link betweenEnglish language proficiency and therate of wage convergence between immi-grants and natives. Chiswick (1991), for

example, documents that an additionalyear of residence in the United States in-creases the probability of English profi-ciency by about 3 percentage points in asmall sample of illegal aliens appre-hended in Los Angeles. Moreover, add-ing variables measuring the worker’sEnglish skills to a cross-section earningsfunction reduces the coefficient of years-since-migration by 10 to 20 percent(Evelina Tainer 1988; Chiswick 1991).

In 1990, 47.0 percent of the immigrantstock in the United States did not speakEnglish very well (U.S. Department ofCommerce 1993a, p. 129). Given the ap-parent high returns to English languageproficiency, it is worth asking why moreimmigrants do not pursue this humancapital investment. The rate of return tolanguage capital, however, may havelittle to do with the wage differentialbetween immigrants who are English-proficient and immigrants who are not.English proficiency and earnings mightbe correlated simply because more ableworkers are likely to speak English andto earn more. Some studies correct forthe endogeneity of the language variableby using instrumental variable estima-tors, but these attempts are not convinc-ing. For example, Chiswick and Miller(1992, p. 265) use such instruments asthe worker’s veteran status, number ofchildren, and the fraction of persons inthe state who speak the same language.It is doubtful that this set of identifyinginstruments is correlated with Englishproficiency, but is not correlated withthe worker’s earnings capacity.

Even if language proficiency were ex-ogenous, the returns to language capitalare affected by the clustering of immi-grants in ethnic enclaves, such as the Cu-bans in Miami’s Little Havana and theMexicans in East Los Angeles. Immi-grants residing in these enclaves mightface low returns to language capital be-cause most of their economic exchanges

14 Similar findings are reported in Carliner’s(1981) study of immigrants in Canada. DavidBloom and Grenier (1992) and Chiswick and PaulMiller (1992) compare the returns to languagecapital in the United States and Canada.


are with persons of the same ethnic (andlinguistic) background. For example, al-most half of the Cubans who arrived inthe Mariel boatlift in 1980 worked forCuban employers in 1986 (AlejandroPortes 1987).15 McManus (1990) findsthat the wage gap between Hispanicswho are English proficient and Hispanicswho are not is 26 percent for workerswho live in a county that is only 10 per-cent Hispanic, but falls to 11 percent forworkers who live in a county that is 75percent Hispanic.

Although many studies measure thecorrelation between language capital andwage convergence, there are many othervariables which influence the assimila-tion process, such as the acquisition offormal education or on-the-job trainingin the post-migration period, invest-ments in geographic mobility within thehost country, and differences in jobsearch activities. Few studies, however,investigate how natives and immigrantsdiffer in these human capital invest-ments.16

4. National Origin and theSelf-selection of Immigrants

Why did the relative wages of succes-sive immigrant cohorts arriving in theUnited States decline? The empiricalevidence suggests that one single factor,the changing national origin mix of theimmigrant flow, can explain much of thedecline (Borjas 1992b; LaLonde andTopel 1992).

A. National Origin and the Decline in Immigrant Skills

Table 8 illustrates the huge differ-ences in educational attainment andearnings across national origin groups in1990. Mean years of schooling rangefrom eight years for immigrants originat-ing in Mexico or Portugal, to about 15years for immigrants originating in suchdiverse countries as Austria, India, Ja-pan, and the United Kingdom. Similarly,immigrants from El Salvador or Mexicoearn 40 percent less than natives, whileimmigrants from Australia or South Af-rica earn 30 to 40 percent more than na-tives. These differences cannot be attrib-uted to the fact that some national origingroups have lived in the United Statesfor longer periods. There is substantialdispersion in both educational attain-ment and relative wages even among im-migrants who have been in the countrymore than 10 years.

In view of the post-1950 changes inthe national origin mix of immigrantflows, it is not surprising that thesechanges “explain” the decline in relativewages across successive immigrantwaves. Borjas (1992b, p. 41) decomposesthe skill decline into a portion due tochanges in the national origin mix andinto a portion due to the changing skilllevel of immigrants from specific coun-tries. The changing national origin mixexplains over 90 percent of the declinein educational attainment and relativewages across successive waves between1960 and 1980.

To some extent, the inter-group vari-ation in skills documented in Table 8mirrors the dispersion in skills across thepopulations of the various source coun-tries. There is, for example, a great dealof dispersion in educational attainmentacross countries (Robert Barro and Jong-Wha Lee 1993). Even if the immigrantflow was randomly drawn from the popu-

15 Borjas (1990, ch. 10) and Ivan Light andEdna Bonacich (1988) provide detailed studies ofself-employment in the immigrant population.

16 An exception is given by Ann Bartel’s (1989)analysis of the internal migration decisions of for-eign-born workers in the United States. Bartelfinds that immigrants choose to reside in areaswhere there are other immigrants, and that theirinternal migration decision are much less sensitiveto regional wage differentials than those of na-tives.


TABLE 8 EDUCATIONAL ATTAINMENT AND WAGES OF IMMIGRANT MEN IN 1990, BY NATIONAL ORIGIN GROUP

Educational Attainment

Percentage WageDifferential Between

Immigrants and Natives

Country of BirthAll

ImmigrantsPre-1980Arrivals

AllImmigrants

Pre-1980Arrivals

Europe: Austria 14.68 14.50 38.4 40.9 Czechoslovakia 14.46 14.49 25.9 37.4 France 14.76 14.03 25.7 27.8 Germany 13.88 13.69 24.5 25.1 Greece 11.83 11.59 −.9 2.4 Hungary 13.59 13.37 27.3 31.9 Italy 10.90 10.71 16.1 17.4 Poland 12.77 12.36 −.3 19.8 Portugal 8.29 8.40 −3.1 −.1 U.S.S.R. 14.23 14.17 6.2 20.2 United Kingdom 14.60 14.35 37.2 37.9 Yugoslavia 11.75 11.47 11.5 17.5

Asia: Cambodia 10.22 11.71 −30.8 −14.6 China 12.82 13.20 −21.3 1.9 India 15.94 16.61 17.6 56.2 Iran 15.52 15.90 6.8 18.6 Japan 15.18 14.67 49.3 27.5 Korea 14.25 14.87 −12.0 10.8 Laos 9.98 10.49 −32.4 −28.3 Lebanon 14.16 13.90 −2.0 10.2 Philippines 14.05 14.09 −5.9 9.7 Taiwan 16.32 17.18 13.9 50.7 Vietnam 12.26 13.25 −18.9 −2.4

, North and South America: , Argentina 13.35 13.17 4.7 17.2 Canada 13.79 13.56 24.0 23.9 Colombia 12.08 12.31 −19.1 −5.5 Cuba 11.74 12.26 −15.3 −5.3 Dominican Republic 10.28 10.46 −29.2 −21.7 Ecuador 11.55 11.88 −20.6 −9.6 El Salvador 8.61 9.60 −39.7 −27.5 Guatemala 9.23 10.27 −38.2 −21.8 Haiti 11.22 12.22 −30.2 −13.6 Jamaica 11.97 12.35 −11.2 −3.1 Mexico 7.61 7.56 −39.5 −32.3 Nicaragua 11.73 12.32 −34.8 −11.3 Panama 13.41 13.44 1.9 11.3 Peru 12.99 13.13 −20.6 .3

Africa: Egypt 15.62 15.71 12.2 41.9 Ethiopia 13.97 15.43 −21.0 6.5 Nigeria 15.80 16.52 −18.9 −3.9 South Africa 15.91 15.93 43.6 58.4

Australia 15.21 15.10 33.0 30.5

Source: See Table 3. The educational attainment of native men in 1990 is 13.2 years.

lation of the source countries, the educa-tional attainment of immigrants who en-tered the United States in the 1980swould differ from that of earlier immi-grant waves.

To illustrate the importance of thiscompositional effect, Borjas (1992b) cal-culated the average schooling level ofthe country represented by the typicalimmigrant for a number of immigrantwaves. The typical immigrant who ar-rived between 1955 and 1960 originatedin a country where the average personhad 9.5 years of schooling. This statisticdeclined to 7.7 years for the 1975–1980flow. If immigrants were randomlydrawn from the source country’s popula-tion and if the rate of return to schoolingis on the order of 7 percent, the declin-ing educational attainment of the typicalsource country would alone be responsi-ble for a 14 percent decline in relativewages across immigrant cohorts.

There is also a great deal of variationin other types of work-related skillsacross the various source countries, andthese skills are not equally transferableto the United States. Clearly, the kindsof skills workers acquire in highly devel-oped economies differ from those ac-quired in less-developed countries. Itseems likely that skills acquired in ad-vanced economies are more easily trans-ferable to the U.S. labor market. In fact,there is a strong positive correlation be-tween immigrant earnings in the UnitedStates and the level of economic devel-opment in the country of origin, as mea-sured by the country’s per capita GNP(Guillermina Jasso and Mark Rosenzweig1986).

There has been a dramatic drop in theper capita income of the country repre-sented by the typical immigrant enteringthe United States (Borjas 1992b). Theaverage person who immigrated between1955 and 1960 originated in a countrywhich had a 1980 per capita GNP of

$6,823 (in 1980 dollars). By contrast, therespective statistic for the typical immi-grant who arrived in the late 1970s is$3,828. Because the elasticity of theearnings of immigrants in the UnitedStates with respect to per capita GNP inthe source country is on the order of .04,immigrants who arrived in the late 1950swill earn about 4 percent more thanthose who arrived in the late 1970s, evenif the immigrant flow were randomly se-lected from the source countries.

B. The Self-Selection of the Immigrant Flow

The immigrant flow, however, is notrandomly selected from the populationof the source countries. Borjas (1987) ar-gues that the self-selection of the immi-grant flow generates some of the nationalorigin differentials documented in Table8. Suppose that residents of country 0(the source country) consider migratingto country 1 (the host country). Assumealso that migration decisions are irre-versible so that no return migration oc-curs. If they choose to remain in thesource country, residents of the sourcecountry have an earnings distributiongiven by:

log w0 = µ0 + ε0, (6)

where w0 gives the worker’s earnings inthe source country; µ0 is the mean logearnings in the source country; and therandom variable ε0 measures deviationsfrom mean earnings, and is assumed tobe normally distributed with mean zeroand variance σ0

2.If the entire population of the source

country were to migrate to the hostcountry, they would face the earningsdistribution:

log w1 = µ1 + ε1, (7)

where µ1 is the mean log earnings inthe host country, and the random vari-


able ε1 measures deviations from meanearnings, and is normally distributedwith mean zero and variance σ1

2. The cor-relation coefficient between the randomvariables ε0 and ε1 equals ρ.

The population mean µ1 need notequal the mean earnings of native work-ers in the host country. The averageworker in the source country, for in-stance, might be less skilled than the av-erage worker in the United States. Forconvenience, it is useful to assume thatthe typical person in both countries isequally skilled, so that µ1 also gives themean earnings of natives in the hostcountry. This assumption helps isolatethe impact of the selection process onthe skill composition of the immigrantflow.

Equations (6) and (7) summarize theearnings opportunities available to po-tential migrants in the source and hostcountries. The migration decision is de-termined by a comparison of earningsopportunities across countries, net of mi-gration costs (C). Define the index func-tion:

I = log

w1

w0 + C ≈ (µ1 −µ0 −π) + (ε1 −ε0), (8)

where π = C/w0 gives a “time-equivalent”measure of migration costs. A workermigrates to the host country if I > 0and remains in the source country other-wise.

Migration costs C will differ amongworkers. For instance, newly arrived im-migrants may be unemployed while theylook for employment, suggesting thathigh-wage migrants might have highermigration costs. High-wage migrants,however, are more likely to have priorjob connections and better informationabout job opportunities, suggesting anegative correlation between migrationcosts C and wages. The immigrant alsoincurs transportation costs. It is instruc-tive to assume initially that the time-

equivalent migration costs, π, are con-stant in the population (so that migrationcosts are proportional to wages). Theprobability that a person migrates to thehost country can then be written as:

P = Pr{υ > (µ0 + π − µ1)} = 1 − Φ(z), (9)

where υ = ε1 – ε0, z = (µ0 + π – µ1)/σv,and Φ is the standard normal distribu-tion function. It is easy to show that:

∂P∂µ0

< 0, ∂P∂µ1

> 0, and ∂P∂π

< 0. (10)

The emigration rate is negatively cor-related with mean earnings in the sourcecountry and with migration costs, and ispositively correlated with mean earningsin the host country. Although most stud-ies analyzing internal migration flows fo-cus on the determinants of the size anddirection of migration flows, there areother equally important questions whichcan be analyzed in the context of the in-come maximization model. For instance,which persons find it worthwhile to mi-grate to the host country?

This question is at the heart of AndrewRoy’s (1951) well-known model of self-selection, which describes how workerssort themselves among employment op-portunities (Michael Sattinger 1993).The implications of the income-maximi-zation hypothesis for the selection of theimmigrant flow are easily grasped byevaluating the conditional means E(logw0 | I > 0), which gives the earnings ofimmigrants prior to their migration, andE(log w1 | I > 0), which gives immigrantearnings in the host country. Because ofthe normality assumption, these condi-tional means are given by:17

17 To derive equation (11), note that: E(log w0 I > 0) = µ0 + σ0E(ε0

∗ v∗ > z),where ε0

∗ = ε0/σ0, v∗ = v/σv. Because the conditionalexpectation of a normal density is linear, we can write ε0

∗ =ρ0v v∗ + ξ, where ρ0v is the correlation between ε0 and v,


E(log w0 I > 0)

= µ0 + σ0σ1

σv ρ −

σ0

σ1 λ, (11)

E(log w1 I > 0)

= µ1 + σ0σ1

σv

σ1

σ0 − ρ

λ, (12)

where λ = φ(z)/(1 − Φ(z)), and φ is thedensity of the standard normal. The vari-able λ is inversely related to the emigra-tion rate and is positive as long as somepersons find it profitable to remain inthe source country (P < 1).

Let Q0 = E(e0 | I > 0) and Q1 = E(e1 |I > 0). Inspection of equations (11) and(12) indicates that there are three possi-ble types of selection characterizing theimmigrant flow:

Q0 > 0 and Q1 > 0

if and only if ρ > σ0

σ1 and

σ1

σ0 > 1. (13)

Q0 < 0 and Q1 < 0

if and only if ρ > σ1

σ0 and

σ0

σ1 > 1. (14)

Q0 < 0 and Q1 > 0

if and only if ρ < min

σ1

σ0,

σ0

σ1

. (15)

Equation (13) shows that immigrantsare positively selected (i.e., have above-average earnings in both the source andhost countries) when the correlation be-tween skills in the two countries is suffi-ciently high and when the host countryhas more dispersion in its earnings distri-

bution. The strong positive correlationbetween earnings in the source and hostcountries ensures that skills are portableacross countries. The immigrant popula-tion is then drawn from the upper tail ofthe earnings distribution because thesource country, in a sense, “taxes” high-ability workers and “insures” less ableworkers against poor labor market out-comes.

Equation (14) indicates that immi-grants are negatively selected (i.e., havebelow-average earnings in both thesource and host countries) when the cor-relation coefficient ρ is sufficiently highand when the earnings distribution inthe source country has a larger variancethan the earnings distribution in the hostcountry. The immigration flow is nega-tively selected, therefore, when the hostcountry taxes high-income workers andprovides better insurance for low-incomeworkers.

Finally, equation (15) describes thecharacteristics of a “refugee sorting,”where immigrants have below-averageearnings in the source country but endup in the upper tail of the earnings dis-tribution of the host country. This sort-ing occurs when ρ is small or negative.The correlation ρ might be negative aftera source country experiences a Commu-nist takeover. This political system (atleast in its initial stages) redistributes in-comes by confiscating the assets of rela-tively successful persons. The model sug-gests that immigrants from such systemswill be in the lower tail of the “revolu-tionary” earnings distribution, but willperform well in the host country’s mar-ket economy.

Note that the type of selection charac-terizing the immigrant flow depends onthe second moments of the earnings dis-tributions. Put differently, because theunderlying distribution of skills is beingheld constant, the variance of the earn-ings distribution proxies for the price of

and ξ is independent of v*. The mean earnings ofimmigrants in the source country are then givenby:

E(log w0 I > 0) = µ0 + σ0 ρ0v E(v∗ v∗ > z).Equation (11) follows directly by noting thatρ0v = (ρσ0σ1 − σ0

2 )/σ0σv and λ = E(v∗ v∗ > z). Equa-tion (12) can be derived in an analogous manner. It isworth noting that the random variables ε0 and ε1 can bedecomposed into observable and unobservable compo-nents so that the framework applies to selection in bothtypes of skill characteristics.


skills, and thus serves an allocative rolein the sorting of persons across coun-tries. It is worth noting that neither thedifference in mean earnings nor the levelof migration costs determines the type ofselection that characterizes the immi-grant flow.18 Although the first momentsdetermine the size and direction of theflow, they do not determine if immi-grants are drawn mainly from the upperor lower tails of the earnings distribu-tion.

The empirical evidence provides somesupport for the equilibrium skill sortingimplied by the model. Borjas (1990, ch.7) reports that measures of income in-equality in the source country, which area rough proxy for the rate of return toskills, are negatively correlated with theearnings of immigrant men in the UnitedStates.19 Holding constant a vector of ob-servable socioeconomic characteristics(including educational attainment andage), the point estimates suggest thatMexican immigrant men earn about 4percent less than British immigrantssimply because of the selectivity effectresulting from Mexico having a higherrate of return to skills than the United

Kingdom. Deborah Cobb-Clark (1993)finds a similar negative correlation be-tween the earnings of immigrant womenin the United States and measures of therate of return to schooling in the sourcecountries. Finally, Edward Taylor’s(1987) case study of migration in a ruralMexican village concludes that Mexicanswho migrated illegally to the UnitedStates are less skilled, on average, thanthe typical person residing in the village.This type of selection is consistent withthe fact that Mexico has a relatively highrate of return to skills.

The discussion provides an interestingexplanation of the decline in the relativeskills of immigrant cohorts admitted tothe United States in the postwar era.Prior to the 1965 Amendments, the allo-cation of visas was guided by the ethniccomposition of the U.S. population in1920, and thus favored immigration froma small number of Western Europeancountries. The 1965 Amendments re-pealed the national-origins quota systemand greatly increased the number of im-migrants originating in Asian and LatinAmerican countries. The new immigra-tion, therefore, is more likely to origi-nate in countries where the populationtends to be less skilled, where skills areless easily transferable to the UnitedStates, and where the rate of return toskills is relatively high. All these factorscontribute to a decline in the relativeskills of successive immigrant waves.20

The self-selection model can be ex-

18 Although the discussion assumed that migra-tion costs (in time-equivalent terms) are constant,it is not difficult to incorporate liquidity con-straints or variable migration costs into the model.For instance, economic conditions might motivatethe least-skilled to migrate, but liquidity con-straints prevent the migration of these workers.The “best of the worst” will then move if the flowis negatively selected. Similarly, if migration costsare correlated with earnings, the selection charac-terizing the immigrant flow may change in eitherdirection. If, for example, migration costs are posi-tively correlated with earnings, the immigrant flowis more likely to be negatively selected. It is easyto show that the correlation between migrationcosts and earnings can change the type of selec-tion only if the variance in migration costs is suffi-ciently high relative to the variance in skills.

19 Alan Barrett (1993) shows that immigrantswho enter the United States using a family reunifi-cation visa have relatively lower earnings whenthey originate in countries where the income dis-tribution has a large variance.

20 Part of the national origin wage differentialsmay also arise from discrimination against particu-lar groups. The literature has not investigated thishypothesis seriously because the evidence on theHispanic/non-Hispanic or the Asian/white wagedifferential among native workers does not lenditself to a simple discrimination interpretation.Cordelia Reimers (1983) finds that much of theHispanic/non-Hispanic wage differential is attrib-utable to differences in observable characteristics,while Chiswick (1983) shows that Asian groups ac-tually have higher wages than white workers, evenafter controlling for observable characteristics.


tended to incorporate the fact that mi-gration decisions are reversible. Returnmigration can arise for two distinct rea-sons. First, it may be the optimal resi-dential location plan over the life cycle.In other words, workers reside in thehost country for a few years and then re-turn to their home countries after accu-mulating sufficiently large levels of hu-man capital or wealth. This mobilitypattern allows some workers to attainhigher utility or wealth than if the migra-tion decision was permanent. Alterna-tively, return migration flows may resultfrom mistakes in the initial migration de-cision.21 Potential immigrants are uncer-tain about the economic conditions avail-able to them in the destination. As aresult, the actual outcomes experiencedin the host country’s labor market differfrom the expected outcomes that guidedthe immigration decision. As long as re-turn migration costs are relatively low,immigrants who experience worse-than-expected outcomes will return to theirhome country.

Borjas and Bernt Bratsberg (forthcom-ing) argue that regardless of which ofthese two factors generates return migra-tion, the implications for the skill com-position of the “surviving” immigrantstock are the same: return migration ac-centuates the selection that characterizesthe initial migration flow. The intuitionis illustrated in Figure 3 for the specialcase where earnings are perfectly corre-lated across countries.22 Suppose that

the immigrant flow is positively selectedso that all workers with skill level ex-ceeding vH emigrate. The worker withskill level vH is the “marginal” immi-grant; he is indifferent between migrat-ing and not migrating. As a result, theimmigrants with skill level in the neigh-borhood of vH are most susceptible toimproved opportunities in the sourcecountry or to adverse random shocks inthe host country’s labor market. The re-turn migrants are the “worst of the best.”If the return migration flow is negativelyselected, the immigrants have skills be-low vL. The persons in the neighborhoodof vL are the marginal immigrants, andthe return migrants are the “best of theworst.”

The limited empirical evidence sup-ports these theoretical implications. Fer-nando Ramos (1992) analyzes the returnmigration decisions of Puerto Ricans liv-ing in the United States. The joint studyof the Puerto Rican and U.S. Censusesprovides valuable information on thecharacteristics of Puerto Ricans living inthe United States, of Puerto Ricans whoremained in their homeland, and ofPuerto Ricans who returned to PuertoRico after living in the United States fora brief period. The data indicate thatPuerto Rican “immigrants” in the UnitedStates are relatively unskilled, but that

21 Eliakim Katz and Oded Stark (1987) presenta model of how the immigrant flow is selectedbased on the assumption that there is asymmetricinformation in the migration decision (workersknow their skills and earnings in the source coun-try, but not in the host country).

22 The assumption that earnings are perfectlycorrelated across countries implies that we canwrite the wage structure for country i (i = 0,1) aslog wi = µi + hi s, where s is a random variabledescribing a worker’s skills; and hi is the rate ofreturn to skills. Ignoring migration costs, a resi-dent of the source country migrates when µ1 +

h1s > µ0 + h0s. We can rewrite this decision rule as(h1 − h0) s > (µ0 − µ1). Thus, there exists a thresh-old level of skills that separates out the migrantsfrom the nonmigrants. Note that this result doesnot depend on the distribution of the random vari-able s.

Figure 3. The Self-selection of Return MigrantsnL

Negatively-SelectedImmigrant Flow

SkillsnH

Positively-SelectedImmigrant Flow


the return migrants are relatively moreskilled than the typical immigrant. Thetypical Puerto Rican who migrated to theUnited States prior to 1975 had 9.4 yearsof schooling, as compared to 10.8 yearsfor a Puerto Rican who never left PuertoRico. The Puerto Rican migration flow,therefore, is negatively selected. In con-trast, the typical Puerto Rican who re-turned to Puerto Rico after a stint in theUnited States had 9.8 years of schooling.

Borjas and Bratsberg (forthcoming)find a relationship between the rate ofreturn migration for a particular nationalorigin group and the average earnings ofthe surviving stock of immigrants in theUnited States. In particular, a high rateof return migration for the national ori-gin group increases the average earningsof the surviving immigrants when the im-migrant flow is positively selected (i.e.,originates in a country with a low rate ofreturn to skills), and reduces the averageearnings of the surviving stock when theimmigrant flow is negatively selected(i.e., originates in a country with a highrate of return to skills). Even though thereturn migration rates in the Borjas-Bratsberg study are measured with agreat deal of error, the empirical evi-dence suggests that return migrationdoes accentuate the selection of immi-grants at either tail of the skill distribu-tion. Bratsberg (1993) shows that the re-turn migration rate of foreign students inthe United States differs substantiallyacross source countries. For example,only about 3 percent of students origi-nating in Mexico or Germany choose toremain in the United States, as opposedto nearly 30 percent of students originat-ing in Israel, Poland, and Kenya. Thedata indicate that foreign students aremore likely to return to wealthier coun-tries and to countries which offer highrates of return to schooling.

Roy’s framework has also been ex-panded to incorporate the idea that im-

migration decisions are made in a familycontext (Cobb-Clark 1990; Borjas andStephen Bronars 1991). The maximiza-tion of family income implies that theimmigrant flow contains some tied mov-ers, persons who would not have mi-grated on their own but who migrate aspart of the household. This approach willlikely play a crucial role in under-standing skill trends among immigrantwomen, both in terms of cohort effectsand wage convergence. The early workof James Long (1980), based on the 1970Census cross-section, suggests that thelabor market experiences of immigrantwomen in the United States differ sub-stantially from those of men. For exam-ple, the earnings of immigrant womenare negatively correlated with years-since-migration. Remarkably, there hasbeen little empirical research document-ing the skill trends among immigrantwomen since that early study.

C. The Host Country’s Demand for Immigrants

Even though Roy’s self-selectionmodel has influenced our thinking abouthow the immigrant flow is chosen fromthe source country’s population, it is im-portant to stress that the model onlygives the “supply side” of the immigra-tion market. Workers who wish to mi-grate to a particular host country can doso only if the host country’s governmentallows it. The immigration market ishighly regulated. Most countries havestrict policies describing the demo-graphic characteristics of persons whoare allowed to enter the country (such asskills, national origin, or family ties withcurrent residents). The size and skillcomposition of the immigrant flow,therefore, are jointly determined by thesupply-side considerations stressed inthe self-selection model as well as by fac-tors which influence the host country’s


demand for immigrants (or, equivalently,the supply of visas).

In general, the supply of visas is deter-mined by the host country’s political andeconomic gains from immigration. Forinstance, the returns to immigration willdepend partly on the benefits accruingfrom admitting workers who can special-ize in particular industries and occupa-tions, and will also be determined by theimpact of immigrant flows on the em-ployment opportunities of natives as wellas on the social fabric of the host coun-try. It is also clear that there will be dif-ferential benefits from admitting skilledor unskilled immigrant flows, dependingon the skill composition of the nativework force and on the generosity of so-cial insurance programs.

Unfortunately, the literature does notyet provide a systematic analysis of thefactors that generate the host country’sdemand function for immigrants. Recentwork by Jess Benhabib (1993) constructsa demand curve by noting that nativesdiffer in their wealth, so that there willbe both winners and losers from thechoice of a particular immigration policy.The demand function for immigrants isthen an exercise in political economy,and depends on the extent to which thewinners can compensate the losers. Rich-ard Freeman (1993) conjectures that thedemand curve for immigrants might bemostly determined by discriminationagainst some national origin groups.

A promising exploration of the factorsthat shift the U.S. demand for immi-grants is given by Claudia Goldin’s(1994) study of the origins of the na-tional-origins quota system. In 1915,Congress enacted legislation requiringimmigrants to pass a literacy test, effec-tively reducing the demand for unskilledimmigrants. President Woodrow Wilsonvetoed the legislation. Legislators repre-senting districts with large immigrantpopulations voted not to override Wil-

son’s veto (suggesting that their immi-grant constituents did not support a re-strictionist policy towards unskilledworkers).23 In contrast, legislators repre-senting districts where wages were stag-nant voted to override the veto (implyingthat their constituents had little to gain,and perhaps much to lose, from admit-ting more immigrants). Therefore, itseems as if further research on the politi-cal economy of immigration policy mightgreatly improve our understanding of theproperties of equilibrium in the immi-gration market.

5. International Differences inImmigrant Performance

The performance of immigrants in thehost country’s labor market has beendocumented in a number of other coun-tries, including Australia (John Beggsand Bruce Chapman 1991); Britain(Chiswick 1980); Germany (ChristianDustmann 1993; Jörn-Steffen Pischke1993); and Israel (Friedberg 1993).These international comparisons help as-sess the impact of differences in immi-gration policy. The most extensive re-search has been conducted on theimmigrant experience in Canada, whichby the early 1990s had an annual immi-grant flow on the order of one percent ofits population (Michael Baker andDwayne Benjamin 1994; Bloom,Grenier, and Morley Gunderson forth-coming; and Robert Wright and PaulMaxim 1993).

Until 1961, Canadian immigration pol-icy, like that of the United States, per-mitted the entry of persons originating inonly a few countries, such as the United

23 Lindsay Lowell, Frank Bean, and Rodolfo DeLa Garza (1986) report that Congressmen repre-senting districts with large Hispanic populationswere more likely to oppose enactment of an earlyversion of the 1986 Immigration Reform and Con-trol Act (which made it illegal for employers tohire illegal aliens).


Kingdom, or of persons who were de-pendents of Canadian residents. Majorpolicy changes in 1962 and 1967 re-pealed the national origin restrictions,and shifted the emphasis towards skillsrequirements. Under current regula-tions, applicants for entry into Canadaare classified into three classes: the fam-ily class (which includes close relatives ofCanadian residents), assisted relatives(which includes more distant relatives ofCanadian residents), and independentimmigrants. Visa applicants in the lasttwo classes are screened by means of a“point system.” Points are awarded ac-cording to such factors as the applicant’seducation, age, and occupation. Appli-cants who get a passing score areawarded an entry visa.

As Table 9 shows, the point systemseems to have had a major impact on theskill level of immigrants in Canada. Inthe early 1960s, the typical immigrantentering Canada had about half-a-yearless schooling than the typical immigrantentering the United States. By the late1970s, the typical immigrant enteringCanada had almost one more year of

schooling than the typical immigrant en-tering the United States. In addition, thetypical immigrant entering Canada in thelate 1970s earned 16 percent less thanCanadian-born workers, while the typicalimmigrant entering the United Statesearned about 28 percent less than U.S.-born workers.24

A number of recent studies attempt todetermine why Canada “attracts” rela-tively more skilled immigrants than theUnited States. Surprisingly, there is littledifference in average skills between im-migrants in Canada and in the UnitedStates for given national origin groups(Borjas 1993b; Duleep and Regets1992a). In other words, the typical Ital-ian immigrant in Canada has about asmuch schooling and does about as well in

TABLE 9EDUCATIONAL ATTAINMENT AND WAGES OF IMMIGRANTS IN CANADA AND THE UNITED STATES

Host Country

Cohort/Census Year Canada United States

Educational Attainment: 1960–64 Cohort as of 1970 10.5 10.9 1975–80 Cohort as of 1980 12.6 11.8 Natives as of 1970 9.9 11.5 Natives as of 1980 11.3 12.7

Percentage Wage Differential Between Immigrants and Natives in Host Country: 1960–64 Cohort as of 1970 −.8 −4.4 1975–80 Cohort as of 1980 −15.8 −27.6

Source: Borjas (1993b, p. 28). The statistics are calculated in the subsample of men aged 25–64 who work in thecivilian sector, who are not self-employed, and who do not reside in group quarters. The Canadian data are drawnfrom the 1971 and 1981 Public Use Samples of the Canadian Census, while the U.S. data are drawn from the 1970and 1980 Public Use Samples of the U.S. Census.

24 The evidence also indicates that Canada ex-perienced a decline in relative wages across suc-cessive immigrant waves, although not as steep asthe decline observed in the United States. Wrightand Maxim (1993) suggest that the Canadian de-cline occurred both because of a change in thenational origin mix of immigrants and because of adecrease in the share of independent class immi-grants (so that the point system became less rele-vant over time).


the labor market as the typical Italianimmigrant in the United States. The na-tional origin mix of immigrants in Can-ada and the United States, however, dif-fers substantially, with a larger fractionof the Canadian immigrant flow originat-ing in European countries. During the1980s, 27 percent of the immigrant flowentering Canada originated in Europe, ascompared to only 10.4 percent for theUnited States. Therefore, part of the dif-ference between the average skill levelof immigrants in Canada and the UnitedStates is attributable to the different na-tional origin mix of immigrants in thetwo host countries.

This finding raises important questionsabout how a Canadian-style point systemworks. It would be a mistake to claimthat the point system is ineffective be-cause it seems to have little impact onthe education level or relative wages ofspecific national origin groups. An alter-native, though little discussed, effect ofthe point system is to reallocate visasacross source countries. Consider, for in-stance, the implications of how educa-tion is rewarded in the point system. Inthe late 1960s, a visa applicant was given1 point per year of education, and only50 out of 100 points were needed to“pass the test.” Persons originating incountries that have a high level of educa-tional attainment are more likely to qual-ify for entry than persons originating incountries where the typical person haslittle schooling. It is likely, therefore,that the point system plays an importantrole in determining the national originmix of the immigrant flow.25

The experience of immigrants in Can-ada differs from the experience of theirU.S. counterparts in one other notableway. Baker and Benjamin (1994) docu-ment that over a 10 year period, the rela-tive wage of immigrants in Canada mightincrease by at most 3 percentage points,less than half of the wage growth experi-enced by immigrants in the UnitedStates. This finding suggests that a fruit-ful avenue for future research might beto investigate why the adjustment pro-cess of immigrants differs across hostcountries.

As noted above, a number of studiesanalyze immigrant labor market perform-ance in other host countries. There is agreat deal of diversity in the results ofthese studies, but the evidence generallysuggests that countries which have skillfilters “attract” a relatively more skilledimmigrant flow. Australia, for example,has a point system similar to the Cana-dian one. Beggs and Chapman (1991) re-port that immigrants in Australia havehigh relative wages. In contrast, Pischke(1993) finds that immigrants in Germany(who for the most part were Turkishguest workers admitted in the 1960s)have lower wages than native Germansand do not experience any wage conver-gence over the life cycle.

6. The Impact of Immigrants on NativeEarnings and Employment

Do immigrants have an adverse impacton native earnings and employment op-portunities? If so, how large is the loss inthe economic welfare of native workers?Are all native groups equally affected bythe entry of immigrants into the labormarket? A rapidly growing literature nowpurports to document the impact of im-migrants on the native labor market in a

25 Allan Green and David Green (1994) arguethat the point system affects the skill level of im-migrants because it influences the occupationaldistribution of the immigrant flow, with occupa-tions that are “in demand” being much more rep-resented in the immigrant flow. The point systemthus alters the national origin mix because workersoriginating in countries where the occupational

distribution is more compatible with the one de-sired by Canada have a larger probability of ob-taining a visa.


number of host countries. As we will seebelow, however, a number of conceptualproblems plague this literature. As a re-sult, the accumulated empirical evidencehas little to say about the underlyingquestions.

To understand the impact of immigra-tion on native employment opportuni-ties, suppose we view a labor market as aclosed economy where a single competi-tive industry uses a linear homogeneousproduction function to produce Q unitsof a good (Joseph Altonji and David Card1991). The production process uses bothskilled and unskilled workers. The wagerates of skilled and unskilled workers arews and wu, respectively. The cost func-tion in this industry is then given byQc(ws, wu), where c(ws, wu) is the unitcost function. Perfect competition im-plies that the price of the output, p,equals the unit cost of production, sothat p = c(ws, wu).

Both skilled and unskilled workerspurchase the good. Each type-i worker(i = s, u) has an output demand functiongiven by Di(wi, p). There are Ns skilledworkers and Nu unskilled workers, andthe fraction of unskilled workers in thepopulation is b. Product market equilib-rium requires:

Q = Ns Ds (ws, p) + Nu Du (wu, p). (16)

To close the model, suppose the laborsupply function of each type-i worker isLi(wi, p). Labor market equilibrium im-plies:

Ns Ls (ws, p) = Q cs(ws,wu) (17)

Nu Lu (wu, p) = Q cu(ws,wu), (18)

where ci = ]c/]wi.Consider now what would happen if

∆N immigrants enter the labor marketexogenously. Suppose that the fraction ofunskilled workers in the immigrant flowequals β. Under some simplifying condi-tions, Altonji and Card (1991, pp. 204–

05) show that the resulting change in thewage of skilled and unskilled workers isgiven by:26

∆log ws = λ

εs − δs ⋅ β − b

b(1 − b) ⋅ ∆N

N

= αs∆NN

, (19)

∆log wu = 1 − λεu − δu

⋅ b − βb(1 − b)

⋅∆NN

= αu∆NN

, (20)

where λ = Nu Du(wu, p)/Q; εi $ 0 is the la-bor supply elasticity of type-i workers;and δi < 0 is the labor demand elasticityfor type-i workers.

Equations (19) and (20) give the re-duced-form impact of immigration onthe skilled and unskilled wage. Supposethat the fraction of unskilled workers inthe immigrant flow (β) equals the frac-tion of unskilled workers in the nativepopulation (b). The linear homogeneityof the production function then impliesthat neither the skilled nor the unskilledwage changes as a result of immigration.Alternatively, if the fraction of unskilledworkers in the immigrant flow exceedsthe respective fraction among natives(β > b), immigration increases the skilledwage and decreases the unskilled wage.

This conceptual experiment, there-fore, indicates precisely how the impactof immigration on native employmentopportunities can be measured. If wecould observe a number of closed labormarkets which immigrants penetrate ran-domly, we can then relate the change inthe wage of skilled and unskilled workers

26 To derive these equations, differentiate thelabor market equilibrium conditions, the productmarket equilibrium condition, and the zero profitcondition, assuming that ]Di/]wi = 0, ]Li/]p = 0,and that the cross-elasticities of factor demand arezero, so that the demand for skilled (unskilled)workers is independent of the unskilled (skilled)wage.


to the proportion of immigrants in thepopulation (after adjusting for the skillcomposition of both the native popula-tion and the immigrant flow). The esti-mated parameters would summarize theimpact of immigrants on native employ-ment opportunities.

Practically all empirical studies in theliterature, beginning with Jean Grossman(1982), attempt to replicate this experi-ment by treating a city or metropolitanarea as the empirical counterpart of theclosed labor market in the theoreticalanalysis. The typical study then regressesa measure of the native wage in the lo-cality on the relative quantity of immi-grants in that locality (or the change inthe wage in the locality over a specifiedtime period on the change in the numberof immigrants in the locality). Equations(19) and (20) show that the impact of im-migration will also depend on the skilldistribution of immigrants relative tothat of natives. The empirical studies,however, typically ignore the skill differ-entials that exist in both the native andimmigrant populations across metropoli-tan areas, and simply calculate the corre-lation between the immigrant share andthe native wage.

Table 10 summarizes the results ofrepresentative studies in this literature.The across-city correlations in theUnited States generally indicate that theaverage native wage is slightly lower inlabor markets where immigrants tend toreside.27 The point estimates of the elas-ticity of the native wage with respect tothe number of immigrants clusteraround −.01 to −.02, so that if one cityhas 10 percent more immigrants than an-other, the native wage in the city withmore immigrants is only about .2 percentlower. The evidence also indicates thatthe numerically weak relationship be-tween native wages and immigration isobserved across all types of native work-ers, white or black, skilled or unskilled,male or female.28 In terms of the pa-

TABLE 10 ELASTICITY OF NATIVE WAGES WITH RESPECT TO THE NUMBER OF IMMIGRANTS IN LOCALITY

StudyImpact of

Immigrants on: Dependent VariableElasticityEstimate

Altonji and Card (1991, p. 220)

Less Skilled Natives Weekly wages +.01

Bean, Lowell, and Taylor (1988, p. 44)

Native Mexican MenBlack Men

Annual earningsAnnual earnings

−.005 to + .05−.003 to + .06

Borjas (1990, p. 87) White Native Men Annual earnings -.01Black Native Men Annual earnings −.02

Grossman (1982, p. 600)

All Natives Factor share of native workers −.02

LaLonde and Topel Young Black Natives Annual earnings −.06 (1991, p. 186) Young Hispanic Natives Annual earnings −.01

27 Many of these studies also find a significantnegative correlation between immigration and theimmigrant wage. For instance, Grossman (1982)reports that a 10 percent increase in the numberof immigrants reduces the immigrant wage by 2percent, while Altonji and Card (1991) concludethat a 10 percent increase in the number of immi-grants reduces the immigrant wage by at least 4percent.

28 An exception to this result is given by Altonjiand Card (1991), who relate the wage change ex-perienced by natives in a particular metropolitanarea between 1970 and 1980 to the change in the


rameters of equations (19) and (20),therefore, the literature concludes thatαs ≈ αu ≈ −.02.

Though most of the studies focus onthe relationship between native earningsand the immigrant share in the local la-bor market, some studies also estimatethe correlation between immigration andnative labor force participation rates,hours worked, and unemployment rates.Table 11 summarizes representative re-sults in the literature. It is evident thatimmigration has a weak effect on the em-ployment of natives.

Studies of specific labor markets con-firm the finding that immigration seemsto have little impact even when the mar-ket receives very large immigrant flows.On April 20, 1980, Fidel Castro declaredthat Cuban nationals wishing to move to

the United States could leave freely fromthe port of Mariel. By September 1980,about 125,000 Cubans, mostly unskilledworkers, had chosen to undertake thejourney. Almost overnight, Miami’s laborforce had unexpectedly grown by 7 per-cent. Card’s (1990) influential analysis ofthe data indicates that the time-seriestrend in wages and employment opportu-nities for Miami’s workers, including itsblack population, was barely nudged bythe Mariel flow. The trend in the wageand unemployment rates of Miami’sworkers between 1980 and 1985 wassimilar to that experienced by workers insuch cities as Los Angeles, Houston, andAtlanta, cities which did not experiencethe Mariel flow.

In short, the estimated correlationsbetween native wages and the immigrantshare in local labor markets do not sup-port the hypothesis that the employmentopportunities of U.S.-born workers arestrongly and adversely affected by immi-gration. Moreover, the evidence forother host countries is similar. Pischkeand Johannes Velling’s (1994) study ofthe German labor market relies on the

share of immigrant workers in that locality. Whenthey instrument the change in the locality’s immi-grant share with a second-order polynomial in thefraction of the work force that was foreign-born in1970, the estimated elasticity is −.8. It is doubtful,however, that the immigrant share in 1970 is avalid instrument for the growth in the immigrantshare.

TABLE 11ELASTICITY OF NATIVE EMPLOYMENT WITH RESPECT TO THE NUMBER OF IMMIGRANTS IN LOCALITY

StudyImpact of

Immigrants on: Dependent Variable/RemarksElasticityEstimate

Altonji and Less Skilled Natives Employment-population ratio −.038 Card (1991, p. 220) Weeks worked −.062

Borjas (1990, p. 92) White Native Men Labor force participation rate -.01Black Native Men Labor force participation rate +.04

Thomas Muller and Thomas Espenshade (1985, p. 100) Black Natives Unemployment rate −.01

Julian Simon, Stephen Moore, and Richard Sullivan (1993)

Natives Unemployment rate +.001

C. Winegarden and Lay Young White Natives Unemployment rate .01 Khor (1991, p. 109) Young Black Natives Unemployment rate −.003


same across-city comparisons that domi-nate the U.S. literature, and finds a weaknegative correlation between the nativewage and the fraction of immigrants inthe work force; and Jennifer Hunt (1992)reports that, even though 900,000 per-sons returned to France within one yearafter the 1962 independence of Algeria(increasing the French labor force by 2percent), there was little impact on theaffected localities.29

The correlations estimated in this ex-tensive literature, however, misspecifythe theoretical experiment describedearlier and hence do not answer thequestion of whether native workers areadversely affected by immigration. Inparticular, the comparison of economicconditions in different metropolitan ar-eas, as well as the pre- and post-immi-gration comparison in a particular metro-politan area, presumes that the labormarkets are closed (once immigrationtakes place) and that the migration flowis exogenous.

Metropolitan areas in the UnitedStates (and abroad) are not closed econo-mies; labor, capital, and goods flowfreely across localities and tend to equal-ize factor prices in the process. As longas native workers and firms respond tothe entry of immigrants by moving to ar-eas offering better opportunities, there isno reason to expect a correlation be-tween the wage of natives and the pres-ence of immigrants. As a result, the com-parison of local labor markets may bemasking the “macro” effect of immigra-tion. Moreover, immigrants do not sim-ply land in a randomly chosen metropoli-tan area; presumably they choose areas

which provide them the best opportuni-ties. Therefore, the correlations typicallyestimated in the literature have no struc-tural interpretation; they do not estimatethe demand function for native workers,nor do they estimate the reduced-formimpact of immigrants on native employ-ment opportunities.30

A recent study of time-series datadrawn from the CPS by Borjas, Free-man, and L. Katz (1992) provides indi-rect evidence of the macro impact of im-migration. As noted earlier, the 1980switnessed a substantial increase in thewage gap between workers who do nothave a high school diploma and workerswith more education. The decade alsowitnessed the entry of large numbers ofless skilled immigrants. Given reasonableestimates of labor demand elasticities,Borjas, Freeman, and L. Katz concludethat perhaps a third of the 10 percentagepoint decline in the relative wage of highschool dropouts between 1980 and 1988can be attributed to the less skilled im-migration flow.31

To reconcile the finding that local la-bor markets do not seem to be affectedby immigration with the possible exist-ence of an economy-wide impact, Ran-

29 William Carrington and Pedro de Lima(1994) report inconclusive results when they ana-lyze the impact of the 600,000 refugees who en-tered Portugal after the country lost the Africancolonies of Mozambique and Angola in the mid-1970s, increasing Portugal’s population by almost7 percent.

30 Some studies use the industry, rather thanthe local labor market, as the unit of observationand analyze native employment and wages as im-migrants penetrate a particular industry (ThomasBailey 1987; John DeNew and Klaus Zimmermann1994; Roger Waldinger 1993). The correlations aresometimes interpreted in terms of a displacementeffect. As with studies of local labor markets, thesecorrelations have no structural interpretation aslong as workers and firms can move across indus-tries.

31 Using CPS data, Topel (1994) also finds thatthe relative decline in the wage of less skilledworkers during the 1980s was steepest in labormarkets which had a sizable immigrant presence.It is important to stress, however, that the CPSdata do not identify persons by nativity status, sothat the decline in the relative wage of unskilledworkers could be attributable to the fact that theunskilled wage fell because the new immigrantsearn even lower wages than the unskilled nativepopulation.


dall Filer (1992) and Michael White andLori Hunter (1993) analyze how the in-ternal migration flows of U.S.-bornworkers respond to immigration. Using1980 Census data, they find that metro-politan areas where immigrants clusterexperienced lower rates of native in-mi-gration and somewhat higher rates of na-tives out-migration. This pattern of na-tive mobility, of course, dissipates theimpact of immigration over the entireeconomy. The evidence for more recenttime periods, however, seems to bemixed. Using various CPS supplementsfrom the 1980s, Butcher and Card (1991)and White and Zai Liang (1993) estimatea positive correlation between immigra-tion flows and the in-migration rates ofnatives to particular cities, while WilliamFrey’s (1994) study of the 1990 Censusreveals that less skilled native workersresiding in states which received largeimmigrant flows in the late 1980s hadrelatively high probabilities of out-migra-tion.

Although native workers and firmsprobably “vote with their feet” and at-tenuate the negative or positive impactof immigration on local labor markets,this argument does not fully explain whyimmigration has little impact on local la-bor markets. Card’s Mariel study, inparticular, raises a number of puzzlingquestions. The Marielitos had no im-pact on Miami’s labor market even inthe year when the large migration tookplace. As a result, the internal flows oflabor, capital, and goods can explain theapparent lack of correlation between na-tive earnings and the presence of immi-grants only if markets adjust instanta-neously.

There exists a great deal of regionalvariation in many labor market charac-teristics and these differences are oftenviewed as the result of equilibrium pro-cesses that are specific to the locality andthat do not disappear quickly. Olivier

Blanchard and L. Katz’s (1992) empiricalstudy of regional labor market adjust-ments in the United States is quite in-structive. They find that a one-time ad-verse economic shock to a state (on theorder of a −1 percent demand shock onemployment) reduces the state’s realwage for up to 10 years before the inter-nal migration of workers reequilibratesthe wage across regions. The unresolvedpuzzle facing those who interpret thelack of correlation between immigrationand native wages in the local labor mar-ket in terms of an economy-wide equilib-rium process is clear: Why should it bethat many other regional variations per-sist over time, but that the impact of im-migration on native workers is arbitragedaway immediately?

A fair appraisal of the literature thussuggests that we still do not fully under-stand how immigrants affect the employ-ment opportunities of natives in local la-bor markets; nor do we understand thedynamic process through which nativesrespond to these supply shocks and rees-tablish labor market equilibrium.

7. Immigration and Welfare

Historically, the debate over immigra-tion policy in the United States has re-volved around the questions of whetherimmigrants assimilate in the UnitedStates and whether they take jobs awayfrom natives. The rapid growth of enti-tlement programs in the past three de-cades introduces an additional explosivequestion into the political arena: Do im-migrants pay their way in the welfarestate?

A. Trends in Immigrant Welfare Participation

The early work of Blau (1984) used the1976 Survey of Income and Education toassess if immigrants and natives had thesame propensity for receiving public as-


sistance (see also Leif Jensen 1988). Blauconcluded that immigrant householdshad roughly the same probability of par-ticipating in public assistance programsas native households, but that immi-grants had lower participation rateswhen compared to natives who had thesame socioeconomic characteristics (suchas household composition and educa-tional attainment of the householdhead).

As with the early studies analyzing theevolution of immigrant earnings, thesefindings were based on studies of cross-section data sets. Beginning with Borjasand Trejo (1991), recent work analyzesthe trends in immigrant welfare partici-pation using synthetic cohorts created bypooling Census cross-sections. The re-sults of this type of research are summa-rized in Table 12, which uses the 1970,1980, and 1990 U.S. Censuses to docu-

ment how immigrant participation incash-benefit welfare programs changedover the past twenty years.32 Immigrantswere slightly less likely than natives toreceive cash benefits in 1970. By 1990,the fraction of immigrant householdson welfare was 1.7 percentage pointshigher than the fraction of native house-holds.

Two distinct factors account for thedisproportionate increase in welfare par-ticipation among immigrant households.Recent immigrant waves are more likelyto use welfare than earlier waves, bothrelative to natives and in absolute terms.In 1970, only 5.5 percent of the most re-cent immigrant households (i.e., house-holds that have been in the UnitedStates fewer than five years) receivedwelfare as compared to 6.0 percent fornative households. By 1990, 8.3 percentof the newly arrived immigrant house-holds received public assistance as com-pared to 7.4 percent of native house-holds. There are, therefore, significantcohort effects in welfare participationrates among immigrants.

In addition, the welfare participationrate for a specific immigrant wave in-creases over time. Even though only 5.5percent of the households that migratedbetween 1965 and 1969 received publicassistance in 1970, the welfare participa-tion rate of this group increased to about10 percent in both 1980 and 1990. Immi-grant households, therefore, assimilate

TABLE 12WELFARE PARTICIPATION RATES OF NATIVE AND

IMMIGRANT HOUSEHOLDS, 1970–1990(Percentage of Households Receiving Public

Assistance)

All Households

Group 1970 1980 1990

Natives 6.0 7.9 7.4All Immigrants 5.9 8.7 9.1

Cohort: 1985–1989 Arrivals — — 8.3 1980–1984 Arrivals — — 10.7 1975–1979 Arrivals — 8.3 10.0 1970–1974 Arrivals — 8.4 9.7 1965–1969 Arrivals 5.5 10.1 9.8 1960–1964 Arrivals 6.5 9.2 8.4 1950–1959 Arrivals 4.9 7.1 6.7 Pre-1950 Arrivals 6.2 9.3 8.1

Source: Author’s tabulations from the 1970, 1980, and1990 Public Use Samples of the U.S. Census. The sta-tistics are calculated in the subsample of householdswhere the household head is at least 18 years of ageand does not reside in group quarters.

32 The Census data report participation only incash benefits programs, such as Aid to Familieswith Dependent Children (AFDC) and Supple-mental Security Income (SSI). The data do notcontain any information on participation in non-cash programs such as Food Stamps and Medicaid.The statistics are calculated using a 1/1000 ran-dom sample of native households in each of theCensuses and a 5/100 random sample of immi-grant households (except in 1970 when the immi-grant extract forms a 2/100 random sample, and in1990 when the native extract forms a 5/1000 ran-dom sample). The resulting data set contains1,296,699 observations.


into welfare.33 This perverse pattern ofadaptation might arise because newly ar-rived immigrants fear that they jeopard-ize their chances for naturalization ifthey receive public assistance, or be-cause immigrants learn about welfareprograms the longer they reside in theUnited States.34

Not surprisingly, there are huge differ-ences in welfare participation propensi-ties among national origin groups. Table13 reports the welfare participation ratesfor selected groups. Only about 2 to 4percent of the households originating inSouth Africa, Taiwan, or the UnitedKingdom receive public assistance, asopposed to 11 to 12 percent of thehouseholds originating in Ecuador orMexico, and nearly 50 percent of thehouseholds originating in Laos or Cam-bodia.

The statistics presented in Table 13suggest a major “structural shift” be-tween two types of national origingroups. In particular, refugee groupstend to exhibit much higher rates ofwelfare participation than non-refugeegroups. As noted earlier, householdsoriginating in Cambodia or Laos had awelfare participation rate of near 50 per-cent; those originating in Vietnam have awelfare participation rate of 25.8 per-cent; while those originating in Cuba orthe Soviet Union have a participationrate of 16 percent. Moreover, the partici-pation rate of refugee groups remainshigh even after a decade in the United

States. Refugee groups that are typicallythought of as being economically suc-cessful, such as the pre-1980 Cubans(who migrated prior to the Mariel flow),have a welfare participation rate of over15 percent in 1990.

The high propensity of refugee house-holds to enter and stay in the welfaresystem may be the result of governmentpolicies which supposedly ease the tran-sition of refugees into the United States.Persons who enter the country as refu-gees have immediate access to a wide ar-ray of social services that neither otherlegal immigrants nor natives qualify for.The early introduction of refugees topublic assistance programs seems to havea profound and long-term impact.

The Census data indicate that not onlyare the recipiency rates of immigranthouseholds rising over time, but that thedollar costs of immigrant welfare partici-pation are also rising. Table 14 docu-ments that the typical native householdon welfare received roughly $4,000 incash benefits (in 1989 dollars) in allthree Census years under study. In con-trast, the typical immigrant household onwelfare received about $3,800 in 1970,nearly $4,700 in 1980, and about $5,400in 1990. There are sizable cohort effectsin the welfare income received by immi-grant households. In 1970, householdswho had just entered the country andwere on welfare received an average of$3,800 in cash benefits. By 1990, thenewly arrived immigrant households onwelfare received an average of $6,400.

Unfortunately, few studies documentimmigrant participation in public assis-tance programs for other host countries.An important exception is the work ofBaker and Benjamin (1993), who findthat the typical immigrant in Canada hada lower probability of participating inwelfare programs than the typical native.In 1991, the typical native household inCanada had a 9.4 percent welfare partici-

33 Borjas and Trejo (1991) show that immigranthouseholds assimilate into welfare programs evenwhen particular age groups are tracked acrossCensuses.

34 Even in 1990, the gap in welfare participationrates between immigrants and natives can be at-tributed to differences in observable socioeco-nomic characteristics, such as educational attain-ment and household composition. In other words,it is not “immigrant-ness” that generates high wel-fare participation rates in the immigrant popula-tion. Rather, it is the socioeconomic charac-teristics of the immigrant population.


TABLE 13WELFARE PARTICIPATION RATES IN 1990, BY NATIONAL ORIGIN GROUP

Country of BirthAll

ImmigrantsPre-1980Arrivals

Europe: Austria 4.3 4.5 Czechoslovakia 4.9 4.9 France 4.8 5.9 Germany 4.1 4.2 Greece 5.5 5.6 Hungary 5.1 5.1 Italy 5.4 5.6 Poland 5.7 5.9 Portugal 7.1 7.6 U.S.S.R. 16.3 10.1 United Kingdom 3.7 4.1 Yugoslavia 5.3 5.7

Asia: Cambodia 48.8 24.4 China 10.4 11.1 India 3.4 4.2 Iran 7.5 4.1 Japan 2.3 3.7 Korea 8.1 8.6 Laos 46.3 34.1 Lebanon 7.3 8.8 Philippines 9.8 10.5 Taiwan 3.3 4.2 Vietnam 25.8 15.9

North and South America: Argentina 4.8 5.7 Canada 4.8 5.1 Colombia 7.5 8.9 Cuba 16.0 15.3 Dominican Republic 27.9 29.9 Ecuador 11.9 13.8 El Salvador 7.3 10.2 Guatemala 8.7 11.4 Haiti 9.1 9.7 Jamaica 7.5 8.7 Mexico 11.3 12.8 Nicaragua 7.8 11.8 Panama 9.0 8.7 Peru 5.9 7.8

Africa: ,

Egypt 5.5 6.7 Ethiopia 5.9 3.0 Nigeria 3.2 3.3 South Africa 1.6 1.6

Australia 3.7 3.8

Source: Author’s tabulations from the 1990 Public Use Sample of the U.S.Census. The statistics are calculated in the subsample of households wherethe household head is at least 18 years of age and does not reside in groupquarters.

pation rate, as compared to only 7.4 per-cent for the typical immigrant household(Baker and Benjamin 1993, Table 1).The lower propensities of immigrants inCanada to enter the welfare system maybe the result of the screen filters whichhinder relatively unskilled immigrantsfrom entering Canada (although Bakerand Benjamin do not provide any directevidence to indicate that the point sys-tem reduces expenditures in welfare pro-grams).35

B. Do Immigrants Pay Their Way?

There has been a great deal of discus-sion in recent years about whether immi-grants take more out of the social wel-

fare system than they put in. JeffreyPassel and Rebecca Clark (1994), for ex-ample, conclude that immigrants pay $27billion more in taxes than they take outof the system, while Donald Huddle(1993) claims that immigration increasesthe native tax burden by about $40 bil-lion annually.

As with all accounting exercises, thesestudies make many disputable assump-tions which effectively determine the an-swer to the question. There are, how-ever, a few facts that are directlyrelevant to the debate and that do notdepend on accounting assumptions. Ta-ble 15 summarizes the data for the 1970–1990 period. The first row of the tablereports the fraction of households in theUnited States that have an immigranthead. This fraction rose from 6.8 percentin 1970 to 8.4 percent in 1990. UsingCensus data it is easy to calculate thefraction of immigrant households in thepopulation of “welfare households” (i.e.,households that receive public assis-tance). In 1970, 6.7 percent of welfarehouseholds had an immigrant head, sothat immigrants were slightly under-rep-resented among welfare households. By1990, the situation had changed dramati-cally: 10.4 percent of welfare householdshad a foreign-born head, so that immi-grants were substantially over-repre-sented among welfare households.

The Census data also indicate that in1970, a total of $14.6 billion in cashbenefits was distributed to households;by 1980, this expenditure had risen to$26.8 billion; and by 1990, to $28.6 bil-lion (all in 1989 dollars). The third rowof Table 15 reports the fraction of “wel-fare income” that was distributed to for-eign-born households. In 1970, 6.7 per-cent of cash benefits were distributed toimmigrant households, again indicatingthat immigrant were slightly under-rep-resented in the distribution of welfarebenefits. By 1990, the situation had

TABLE 14MEAN WELFARE INCOME OF NATIVE AND

IMMIGRANT HOUSEHOLDS, 1970–1990(Calculated in Subsample of Households Receiving

Public Assistance, 1989 dollars)

Group 1970 1980 1990

Natives 3,837 4,248 4,017All Immigrants 3,806 4,662 5,363

Cohort:1985–1989 Arrivals — — 6,3851980–1984 Arrivals — — 6,5711975–1979 Arrivals — 5,228 5,6521970–1974 Arrivals — 5,220 4,8841965–1969 Arrivals 3,830 5,044 4,7961960–1964 Arrivals 4,144 5,050 4,4801950–1959 Arrivals 4,402 4,680 4,514Pre-1950 Arrivals 3,629 4,022 4,262

Source: Author’s tabulations from the 1970, 1980, and1990 Public Use Samples of the U.S. Census. The sta-tistics are calculated in the subsample of householdswhere the household head is at least 18 years of ageand does not reside in group quarters.

35 The evidence also indicates that immigrantsin Canada, like their counterparts in the UnitedStates, assimilate into the welfare system. Over a10-year period, the probability of participating inpublic assistance programs for the typical immi-grant in Canada rises by about 5 percentage points(relative to natives).


changed drastically: 13.1 percent of allcash benefits were distributed to immi-grant households, indicating a substantialover-representation of immigrants inwelfare expenditures. Put differently,the total amount of cash benefits re-ceived by immigrant households was 56percent higher than would have been thecase if immigrants used the welfare sys-tem to the same extent as natives.

Immigrants, therefore, now receive adisproportionately high share of cashbenefits. Moreover, they do not receive adisproportionately high share of non-welfare income. In 1990, immigrants re-ceived 8.3 percent of all non-welfareincome (about the same as their popula-tion proportion).36 Because immigrantsdo not receive a disproportionately highshare of income, they also do not pay adisproportionately high share of taxes.

As noted earlier, accounting exercisesthat assign a dollar figure to the tax bur-

den imposed by immigration inevitablyincorporate a number of hidden andquestionable assumptions. Table 16 illus-trates the problem by presenting a back-of-the-envelope calculation of the taxburden in 1990. The first row reportsthat immigrants received $3.7 billiondollars in cash welfare benefits in 1990,or as noted earlier, 13.1 percent of ex-penditures in cash benefit programs. Atthat time, expenditures on all means-tested entitlement programs was $181.3(U.S. Bureau of the Census, 1992, p.357).37 If we assume that immigrants alsoaccount for 13.1 percent of these expen-ditures, immigrants increase expendi-tures on all means-tested entitlementprograms by $23.8 billion.

The next step in the calculation is tocompute the taxes that immigrants pay.According to the 1990 Census, immi-

TABLE 15IMMIGRANT CONTRIBUTION TO WELFARE EXPENDITURES

1970 1980 1990

1. Percentage of Households with Immigrant Heads

6.8 7.6 8.4

2. Percentage of Households with Immigrant Heads in Population of Households Receiving Public Assistance

6.7 8.3 10.1

3. Percentage of Public Assistance Income Distributed to Households with Immigrant Heads

6.7 9.1 13.1

4. Percentage of Non−Welfare Income Received by Households with Immigrant Heads

6.3 7.0 8.3

Source: Author’s tabulations from the 1970, 1980, and 1990 Public Use Samples of the U.S.Census.

36 Even though the typical immigrant workerearns less than the typical native, immigrants as agroup do not have a disproportionately low shareof non-welfare income. This discrepancy is ex-plained by the fact that immigrants have larger la-bor force participation rates than natives.

37 The means-tested entitlement programs in-clude such programs as Food Stamps, Medicaid,low-income housing assistance, and Head Start.Expenditures on means-tested entitlement pro-grams totaled $186.4 billion. The figure reportedin the text nets out expenditures on Indian HealthServices and on pensions for needy veterans fromthe total because few immigrants are likely toqualify for these programs.


grant households received a total income(net of welfare payments) of $284.7 bil-lion. Richard Kasten, Frank Sammartino,and Eric Toder (1993) have recently cal-culated the federal tax burden for U.S.households at various points in the in-come distribution. Applying their esti-mated tax rates to the immigrant incomedistribution suggests that the federal taxburden for immigrants is on the order of22 percent. If the total tax rate (includ-ing state and local taxes) is 30 percent,immigrant households then pay about$85.4 billion in taxes.

The calculations thus indicate that im-migrants pay more in taxes ($85.4 bil-lion) than they take out of the system($23.8 billion). But this comparison ismisleading. It is, in effect, saying thatimmigrant taxes are only used to fundtheir use of means-tested entitlementprograms. This assumption is justifiableif all other government programs providepure public goods, so that expendituresin these programs are unaffected by im-migration. It is likely, however, that im-migrants increase the congestion associ-

ated with the provision of many of thesepublic goods (e.g., more crowded parks,schools, and roads). In other words, themarginal cost of providing these publicgoods to the immigrant population is notzero. Immigrants, therefore, should becharged a user fee for the various gov-ernment services.

It is obviously very difficult to deter-mine the “correct” user fee schedule forthe services provided to immigrants. Wedo not even know, for instance, if themarginal cost of providing many of thepublic services to immigrants (such as anexpansion of the public school system orthe construction of additional roads) isless than or greater than the averagecost. Obviously, different assumptionsabout the marginal cost of providingservices will lead to very different con-clusions about whether immigrants paytheir way in the welfare state. For exam-ple, a “revenue-neutral” immigrationpolicy (i.e., one that would neither subsi-dize nor penalize natives for the provi-sion of government services to the immi-grant population) requires that the

TABLE 16ACCOUNTING OF WELFARE EXPENDITURES AND TAXES PAID BY IMMIGRANT

HOUSEHOLDS IN 1990 (in billions of dollars)

Tax Rate

30% 40%

1. Cash Benefits Received by Immigrant Households (= 698, 071 Households x $5,363) $3.7 $3.72. Dollar Value of Benefits from Means-Tested Programs Received by Immigrant Households (13.1% of $181.3 Billion)

$23.8

$23.83. Non-Welfare Income Received by Immigrant Households $284.7 $284.74. Taxes Paid by Immigrant Households $85.4 $113.95. Taxes Allocated to Means-Tested Entitlement Programs (8.9% of Taxes Paid)

$7.6 $10.1

6. Fiscal Burden on Native Taxpayers Imposed by Immigrant Households $16.2

$13.7

Source: Author’s tabulations from the 1990 Public Use Sample of the U.S. Census.


average tax rate for immigrants be setequal to:

ti = ∆GiYi

, (21)

where ∆Gi gives the increase in govern-ment expenditures attributable to immi-gration, and Yi is the income received byimmigrants. This tax rate depends notonly on the increase in expenditures, butalso on the mean income and labor forceparticipation rate of immigrant workersbecause less-skilled immigrant popula-tions with low rates of work attachmentwould have to be taxed at a higher rate(for a given increase in government ex-penditures).

Suppose that the marginal cost of pro-viding services to immigrants equals theaverage cost and that per capita incomein the immigrant population equals thatof natives. These assumptions imply thatimmigrants should be charged for thecosts of the various government pro-grams as if they were natives. In otherwords, if x percent of a native worker’staxes pay for defense, then x percent ofan immigrants’ taxes should also be allo-cated to pay for defense. In 1990, 91.1percent of taxes were used to pay forprograms other than means-tested enti-tlement programs. If we charge immi-grants 91.1 percent of their tax paymentsfor using these other programs, thenonly 8.9 percent of immigrants’ taxes areleft to fund their use of means-tested en-titlement programs. As reported in row 5of Table 16, immigrants would then con-tribute only $7.6 billion to the funding ofthe entitlement programs. The annualloss associated with immigration is onthe order of $16 billion.38 As this back-

of-the-envelope calculation suggests,therefore, accounting exercises can leadto radically different conclusions aboutwhether immigrants pay their way.

The calculation reported in Table 16illustrates why the studies of Passel andClark (1994) and Huddle (1993) reachsuch different conclusions. Passel andClark estimate that immigrants payabout $70 billion in taxes, but increaseexpenditures in such programs as welfareand education by $43 billion, thus gener-ating a “net surplus” of $27 billion. Thiscalculation, of course, assumes that themarginal cost of providing all other pro-grams to immigrants is zero. In contrast,Huddle simply concludes that immi-grants pay less in total taxes than theytake out of the system. Huddle’s calcula-tions, however, assume that immigrantspay only 7 percent of their income intaxes (net of payments to the Social Se-curity system), and overestimate thecosts of immigration by claiming that forevery six immigrants who enter the coun-try, one native is displaced from his joband joins the welfare rolls.

The cost-benefit calculation presentedhere focuses exclusively on immigrantparticipation in means-tested entitle-ment programs. Adding other govern-ment programs, such as Social Security,could change the results substantially.For example, it is often argued that im-migrants make a net contribution to theSocial Security system because many im-migrants leave the United States prior toretiring and do not collect benefits, de-spite their having contributed to the sys-tem. It is important to realize, however,that the median age of immigration is 30,so that many immigrants pay into the So-cial Security system for a much shortertime span than natives, yet collectroughly the same benefits (the benefits

38 To determine if there is a net benefit fromimmigration, these fiscal costs must be contrastedwith an estimate of the benefits from immigration.Consumers, for example, may be able to buycheaper goods and employers can hire some work-ers at lower wage rates. The literature, however,

does not provide a systematic accounting of thesebenefits.


of a person who works for 30 years arenot much greater than those of a personwho works for only 15 years). Despitethe potential importance of the SocialSecurity program for any cost-benefitcalculation, no studies exist which incor-porate the long-run impact of immigra-tion on the Social Security system.

The calculation also ignores the costsand benefits of providing schooling toimmigrant children. The expendituresassociated with putting the childrenthrough the public school system aresubstantial. In California alone, it is esti-mated that roughly $1.7 billion was spenton educating the children of illegalaliens in 1993 (California Department ofFinance 1994). These costs, however,must be contrasted with the benefits ofhaving a more educated work force lateron in the life cycle. It is also the casethat immigrants who enter the UnitedStates after they complete their educa-tion import “free” human capital into theUnited States. To the extent that immi-grants do not receive their entire prod-uct as wages, substantial benefits mightaccrue from this infusion of human capi-tal.39

8. The Second Generation

In 1990, 9.7 percent of the U.S. popu-lation was native-born with foreign par-entage (or “second-generation”). By theyear 2050, the share of second-genera-tion persons will increase to 13.9 per-cent, and an additional 8.5 percent willbe composed of the grandchildren ofcurrent immigrants (Barry Edmonstonand Passel 1992, p. 471). The economic

impact of immigration obviously dependsnot only on how immigrants adapt to thelabor market, but also on the adjustmentprocess experienced by their offspring.

The traditional view of this intergen-erational adjustment is vividly depictedby the melting pot metaphor. Over thecourse of two or three generations, im-migrants are transformed from a collec-tion of diverse national origin groupsinto a homogeneous native population.Beginning with Nathan Glazer andDaniel Moynihan (1963), modern socio-logical research argues that this meta-phor does not correctly portray the eth-nic experience in the United States. Infact, Glazer and Moynihan (1963, p.xcvii) conclude that “the American ethosis nowhere better perceived than in thedisinclination of the third and fourthgeneration of newcomers to blend into astandard, uniform national type.” The re-visionist literature suggests that many ofthe cultural and economic differencesamong immigrant groups are transmittedto their children, so that the heterogene-ity found among today’s immigrants be-comes the heterogeneity found amongtomorrow’s ethnic groups.

The modern economic literature onthe intergenerational mobility experi-enced by immigrant households is domi-nated by two questions. First, is there asignificant improvement in economicstatus between the first and second gen-erations? Second, do the national origindifferentials in skills and earnings evi-dent in the immigrant generation disap-pear over time?

In contrast to the voluminous litera-ture analyzing the economic status of im-migrants, few studies document the skillsand labor market performance of theirAmerican-born children. Early work byChiswick (1977) and Carliner (1980)used the 1970 Census cross-section tocalculate the relative wage of variousgenerations of Americans. The 1970

39 Calculations conducted in particular locali-ties, such as Los Angeles and San Diego, suggestthat the costs imposed by immigration on thecriminal justice system can also be substantial.There is, however, little systematic study of theextent to which immigrants participate in criminalactivities. John Tanton and Wayne Lutton (1993)report that 20 percent of federal inmates in theUnited States are non-citizens.


Census data allow the precise identifica-tion of first- and second-generationworkers.40 The generation of the remain-ing workers, who had both parents bornin the United States, cannot be deter-mined, but for convenience we will callthem “third-generation” Americans. Thelast column of Table 17 summarizes theevidence provided by the 1970 Censuscross-section. Second-generation work-ers earn about 12 percent more than im-migrants and 16 percent more thanthird-generation workers. Both Chiswickand Carliner concluded that there was agreat deal of economic mobility amongthe ambitious children of immigrants,but that the “hunger” disappeared by thethird generation.

This conclusion, however, is prema-ture (Borjas 1993a). In any cross section,the family ties among the three genera-tions identifiable in the data are tenuous.At the time of the survey, many mem-bers of the first generation have just ar-rived in the United States and cannot

have any native descendants employed inthe U.S. labor market. Second-genera-tion workers can only be descendants ofimmigrants who have been in the coun-try for at least two or three decades.41

Therefore, if there are cohort differ-ences among immigrants and if these dif-ferences are partially transmitted to theirchildren, the labor market performanceof second-generation workers now par-ticipating in the labor market (who arethe offspring of the immigrant wavesthat arrived 30 or 40 years ago) cannotbe used to forecast the future perform-ance of the children of newly arrived im-migrants.

We showed earlier how tracking spe-cific immigrant waves across Censusesyields a rate of wage convergence in theimmigrant generation. The intercensaltracking of immigrants and their off-spring also yields an estimate of the rateof economic mobility across generations.

TABLE 17RELATIVE WAGES OF FIRST AND SECOND GENERATIONS IN 1940 AND 1970

(Percentage Wage Differential Relative to Third Generation)

Variable/Group 1940 1970

Log Wage First Generation 20.3 4.0 Second Generation 26.4 16.3Log Wage, Adjusting for Education and Age First Generation 20.6 7.8 Second Generation 26.0 11.7

Source: Borjas (1993a, p. 119). The statistics are calculated in the subsampleof men aged 25–64 who work in the civilian sector, who are not self-employed,and who do not reside in group quarters.

40 The first-generation includes persons bornabroad, while the second generation includes per-sons who had at least one foreign-born parent.Since 1970, the Census does not contain any infor-mation on the birthplace of the parents, but in-stead reports information on a person’s ancestry,regardless of parental birthplace.

41 It is also extremely unlikely that the so-calledthird-generation workers are direct descendants ofthe immigrants enumerated in the Census cross-section. The persons identified as members of thethird generation form a diverse collection of work-ers whose presence in the United States may date40 or 400 years. Moreover, the cross-section com-parison requires that working-age immigrants haveAmerican-born grandchildren who are also ofworking age.


It is likely, for example, that the childrenof the immigrant stock present in theUnited States in 1940 show up as sec-ond-generation workers in the 1970Census. The data reported in the firstcolumn of Table 17 indicate that immi-grants in 1940 earned about 20 percentmore than third-generation workers. Aswe saw earlier, the children of these im-migrants, presumably the second-genera-tion workers enumerated in the 1970Census, earn only about 16 percent morethan the third generation. The intercen-sal tracking thus contradicts the percep-tion that second-generation workershave, on average, higher earnings thanthe first. There is, instead, a slight re-

gression towards the mean. Because theimmigrant stock present in the UnitedStates in 1940 had very high wages, thesecond-generation enumerated in 1970had relatively lower wages than their im-migrant parents.

Table 18 uses the 1940 and 1970 Cen-suses to document the huge wage differ-entials across national origin groups inboth the first and second generations. In1970, second-generation Americans ofBritish ancestry earned about 23.1 per-cent more than third-generation Ameri-cans, while second-generation Mexicansearned 14.9 percent less. To measureboth the “shift” in economic fortunes be-tween the first and second generations as

TABLE 18NATIONAL ORIGIN WAGE DIFFERENTIALS AMONG FIRST GENERATION WORKERS IN 1940,

AND SECOND GENERATION WORKERS IN 1970(Percentage Wage Differential Relative to Third Generation)

Country of Origin: First Generation in 1940 Second Generation in 1970

Austria 32.2 23.5Canada 28.7 12.1Cuba −.4 −2.5Czechoslovakia 31.9 14.7Denmark 33.8 12.6France 25.7 24.6Germany 21.9 13.7Greece −9.8 20.8Hungary 28.0 22.6Ireland 23.2 21.7Italy 17.2 14.7Mexico −39.1 −14.9Netherlands 11.1 17.5Norway 31.0 15.5Poland 24.9 16.6Portugal 5.2 −.3Romania 34.3 39.1Spain 6.7 11.2Sweden 30.0 19.5Switzerland 21.9 12.4United Kingdom 37.3 23.1USSR 31.8 37.7Yugoslavia 34.9 18.9

Source: Borjas (1993a, p. 124). The statistics are calculated in the subsample of men aged 25–64who work in the civilian sector, who are not self-employed, and who do not reside in groupquarters.


well as the correlation in the mean earn-ings of an ethnic group across genera-tions, it is useful to contrast second-gen-eration Americans in 1970 with theearnings of immigrants in 1940. Let w2jbe the average log wage (in 1970) of sec-ond-generation Americans in group jrelative to that of third generationAmericans; and w1j be the average logwage (in 1940) of immigrants in group jrelative to that of third generationAmericans. Borjas (1993a, p. 125) re-ports that the regression line relating therelative log wage of these two genera-tions is given by:

w2j = .070 + .447 w1j,(.017)

(.065) (22)

where the standard errors are reportedin parentheses and the regression usesthe 23 national origin groups listed inTable 18.42 The intercept reveals a 7 per-cent increase in earnings potential be-tween the first and second generationsthat is common to all national origingroups. The empirical evidence, there-fore, indicates that second-generationworkers do experience a “jump” in theirearnings capacity. The data, however,also reveal a strong correlation betweenthe economic status of national origingroups in the first and second genera-tions. The slope estimate of .45 impliesthat roughly half of the wage differentialbetween any two national origin groupsin the first generation persists into thesecond. There is some regression towardthe mean, but national origin is still animportant determinant of the earnings ofsecond generation Americans. In fact, ifthe intergenerational correlation is onthe order of .5 and is constant across

generations, the evidence suggests thatthe ethnic skill differentials introducedby immigration will persist into the thirdgeneration and perhaps even into thefourth.

The long-run persistence of ethnic dif-ferences is evident in a recent analysis ofthe children and grandchildren of theimmigrants who entered the UnitedStates at the turn of the 20th Century.Using data drawn from the 1910 Census,Borjas (1994) finds sizable differences inthe skills and earnings of the nationalorigin groups that made up the FirstGreat Migration. Using data drawn fromthe 1940 and 1980 Censuses, and theGeneral Social Surveys, Borjas thenshows that there are sizable differencesin the skills and earnings of the childrenand grandchildren of these immigrants.A 20 percentage point difference in theliteracy rate between any two groups inthe first generation implies a 1-year dif-ference in educational attainment amongsecond-generation workers, and a .5-yeardifference among third generation work-ers. Similarly, a 20 percent wage differ-ential between the two groups in thefirst generation implies roughly a 12 per-cent differential in the second genera-tion, and a 5 percent differential in thethird. Ethnicity matters, and it seems tomatter for a very long time.

This conclusion is not consistent withthe widespread perception that the cor-relation between parental skills and chil-dren’s skills is small and might be on theorder of .2 (Gary Becker and NigelTomes 1986). Recent work by GarySolon (1992) and David Zimmerman(1992), however, suggests that measure-ment error in parental background leadsto a substantial underestimate of the cor-relation in earnings across generations.Correcting for this measurement errorincreases the intergenerational correla-tion to between .3 to .4. Because the re-gression reported in (22) uses the aver-

42 The regression reported in (22) is based onthe data summarized in Table 18. The regression,however, uses a generalized least squares estima-tor to account for the heteroscedasticity intro-duced by the sampling error in the dependentvariable


age earnings of workers in each group,the relatively high intergenerational cor-relation between the first and secondgenerations is partly due to the fact thatthe data net out a substantial amount ofmeasurement error.

Recent work, however, suggests thatmeasurement error alone does not ex-plain the very high correlation in themean earnings of ethnic groups overtime. These studies argue that there areracial or ethnic externalities in the labormarket which influence the human capi-tal accumulation of persons belonging toparticular racial or ethnic groups (GlennLoury 1977). Put differently, the rate ofintergenerational mobility between im-migrants and their children is influencednot only by parental background, butalso by the “quality” of the ethnic envi-ronment where the children grow up.These ethnic effects increase the correla-tion in earnings across generations andcan substantially delay the convergenceof ethnic skill differentials.

A simple formulation of this idea isgiven by Borjas (1992a), who argues thatthe average human capital stock in theparental generation for ethnic group j, k

_j,

which he calls “ethnic capital,” acts as anexternality in the production of the hu-man capital of children. The productionfunction for child quality is given by:

Child Quality = f (parental inputs, k_

j). (23)

The hypothesis that ethnicity has ex-ternal effects on human capital accumu-lation has been used widely in the sociol-ogy literature. For instance, JamesColeman (1988) stresses that the culturein which the individual is raised (whichhe calls “social capital”) can be thoughtof as a form of human capital common toall members of that group. He arguesthat social capital alters the opportunityset of workers and has significant effectson behavior, human capital formation,and labor market outcomes. Similarly, in

his influential study of the underclass,William Wilson (1987) argues that thepresence of mainstream role models inpoor neighborhoods serves an importantsocial and economic function.43

To determine the relative importanceof parental inputs and ethnic spilloverson the intergenerational transmissionprocess, Borjas (1992a) estimated thefollowing regression model in the Na-tional Longitudinal Surveys of Youth andthe General Social Surveys:

yij(t) = β1yij(t − 1) + β2y_

j (t − 1) + εi j(t), (24)

where yij(t) measures the skills (such aseducation or wage) of person i in ethnicgroup j in generation t; yij(t − 1) givesthe skills of his father; and y

_j(t − 1) gives

the average skills of the ethnic group inthe father’s generation. All variables aremeasured in deviations from the mean.

It is easy to show the link between themicro model in (24) and the regressionusing the mean earnings of ethnic groupsin the first and second generations re-ported in equation (22). Aggregating (24)within an ethnic group yields:

y_

j(t) = (β1 + β2) y_

j(t − 1) + ε_

j (t). (25)

The regression estimated in aggregateCensus data, therefore, estimates β1 +β2. This sum yields precisely the inter-generational transmission coefficientrelevant for determining the rate atwhich the mean skills of ethnic groupsconverge across generations, or “mean-convergence.” If the sum is less thanone, ethnic differences converge overtime; if not, ethnic differences diverge.The empirical evidence indicates that β1is on the order of .2 to .3, and that β2 isalso on the order of .2 to .3, so that therate of mean convergence is around .4 to

43 The rapidly growing literature on the deter-minants of endogenous economic growth alsostresses the hypothesis that human capital has ex-ternal effects in production; see Robert Lucas(1988) and Paul Romer (1986).


.6. There is, therefore, a great deal ofpersistence in ethnic skill differentialsover time, and about half of the persist-ence is due to the effects of ethnic spill-overs on intergenerational mobility.

We cannot yet determine if the ethnicexternalities model provides a useful ap-proach for analyzing the long-run eco-nomic impact of immigration. Future re-search will have to specify the precisemechanism through which ethnic and ra-cial spillovers operate, such as neighbor-hoods, schools, and religious institutions,as well as document the extent to whichintra-ethnic contacts influence job searchactivities, occupational choice, and otherlabor supply and labor demand deci-sions.

9. Conclusion

The literature investigating the eco-nomic impact of immigration on theUnited States and on other host coun-tries grew rapidly in the past decade.This explosion of research substantiallysharpened our understanding of the eco-nomics of immigration. The stylized factsthat long dominated the discussion overthe costs and benefits of immigrationwere radically altered during the 1980s,and a number of new questions, issues,and perceptions replaced them.

To appreciate the magnitude of thisupheaval, consider the perceived wisdomas of ten years ago. The available studiessuggested that even though immigrantsgenerally arrived with an economic dis-advantage, their economic opportunitiesimproved rapidly over time. Within adecade or two after arrival, immigrantearnings would approach, reach paritywith, and overtake the earnings of na-tives of comparable socioeconomic back-ground. Moreover, there was little evi-dence to suggest that immigrants had anadverse impact on native employmentopportunities. Overall, the empirical evi-

dence painted a very optimistic pictureof the contribution of immigrants to theAmerican economy.

In the past ten years, many morebrushstrokes were applied to the canvas,and the theme and shape of the picturechanged. The new research established anumber of new stylized facts: The rela-tive skills of successive immigrant wavesdeclined over much of the postwar pe-riod; it is unlikely that recent immigrantswill reach parity with the earnings of na-tives during their working lives; althoughthere is only a weak negative correlationbetween the presence of immigrants in alocal labor market and the earnings ofnatives in that labor market, immigrationmay have been partly responsible for thedecline in the earnings of unskilled na-tive workers that occurred during the1980s; the new immigration may have anadverse fiscal impact because recentwaves participate in welfare programsmore intensively than earlier waves; im-migration policy matters, so that hostcountries which filter applicants in termsof observable skills “attract” immigrantswho are more skilled, have higher earn-ings, and are less likely to participate inpublic assistance programs; and, finally,there exists a strong correlation betweenthe skills of immigrants and the skills ofsecond-generation Americans, so thatthe huge skill differentials observedamong today’s foreign-born groups be-come tomorrow’s differences amongAmerican-born ethnic groups.

An important lesson of the recent re-search is that immigration has a far-reaching and long-lasting impact. In asense, we are only beginning to observethe economic consequences of the his-toric changes in the size, national originmix, and skill composition of immigrantsadmitted to the United States during thepast three decades. The Second GreatMigration surely will alter the skill en-dowment of the labor force, the employ-


ment opportunities of native workers,and the costs of social insurance pro-grams not only in our generation, but forour children and grandchildren as well.In addition, current immigration in theUnited States and in many other hostcountries is setting the stage for the eth-nic differences in economic outcomesthat are likely to be a dominant featureof labor markets in these countriesthroughout the next century.

In view of the economic, cultural, andpolitical significance of the issues raisedby immigration, it is not surprising thatimmigration policy is now a central in-gredient in the debate over social policyin many countries. For the most part,this debate focuses on economic issuesand uses the evidence provided by eco-nomic research to frame and formulatethe discussion. Because the economicimpact of current immigration will befelt for many decades to come and be-cause the immigrant flow to many hostcountries continues unabated, the explo-sion of research that we witnessed in thepast decade is sure to continue.

REFERENCES

ABOWD, JOHN M. AND FREEMAN, RICHARD B.,eds. Immigration, trade, and the labor market.Chicago: U. of Chicago Press, 1991.

ALTONJI, JOSEPH G. AND CARD, DAVID. “The Ef-fects of Immigration on the Labor Market Out-comes of Less-Skilled Natives,” in JOHN M.ABOWD AND RICHARD B. FREEMAN, eds. 1991,pp. 201–34.

BAILEY, THOMAS R. Immigrant and native work-ers: Contrasts and competition. Boulder, CO:Westview, 1987.

BAKER, MICHAEL AND BENJAMIN, DWAYNE. “TheReceipt of Transfer Payments by Immigrants inCanada.” U. of Toronto, 1993.

———. “The Performance of Immigrants in theCanadian Labor Market,” J. Lab. Econ., July1994, 12(3), pp. 369–405.

BARRETT, ALAN M. “Three Essays on the LaborMarket Characteristics of Immigrants.” Ph.D.Dissertation, Michigan State U., 1993.

BARRO, ROBERT J. AND LEE, JONG-WHA. “Inter-national Comparisons of Educational Attain-ment.” NBER Working Paper No. 4349, Apr.1993.

BARTEL, ANN P. “Where Do the New U.S. Immi-

grants Live?” J. Lab. Econ., Oct. 1989, 7(4), pp.371–91.

BEAN, FRANK D.; LOWELL, B. LINDSAY AND TAY-LOR, LOWELL J. “Undocumented Mexican Im-migrants and the Earnings of Other Workers inthe United States,” Demography, Feb. 1988,25(1), pp. 35–52.

BECKER, GARY S. AND TOMES, NIGEL. “HumanCapital and the Rise and Fall of Families,” J.Lab. Econ., July 1986, 4(3, Part 2), pp. S1–39.

BEGGS, JOHN J. AND CHAPMAN, BRUCE J. “MaleImmigrant Wage and Unemployment Experi-ence in Australia,” in JOHN M. ABOWD ANDRICHARD B. FREEMAN, eds. 1991, pp. 369–84.

BENHABIB, JESS. “A Note on the Political Econ-omy of Immigration.” New York U., 1993.

BLANCHARD, OLIVIER JEAN AND KATZ,LAWRENCE F. “Regional Evolutions,” Brook-ings Pap. Econ. Act., 1992, 1, pp. 1–61.

BLAU, FRANCINE D. “Immigration and LaborEarnings in Early Twentieth Century America,”Research in population economics. 2. Green-wich, CT: JAI Press, 1979, pp. 21–41.

———. “The Use of Transfer Payments by Immi-grants,” Ind. Lab. Relat. Rev., Jan. 1984, 37(2),pp. 222–39.

BLOOM, DAVID E. AND GRENIER, GILLES. “Earn-ings of the French Minority in Canada and theSpanish Minority in the United States,” inBARRY R. CHISWICK, ed. 1992, pp. 373–409.

BLOOM, DAVID E.; GRENIER, GILLES AND GUN-DERSON, MORLEY. “The Changing Labor Mar-ket Position of Canadian Immigrants,” Can. J.Econ., forthcoming.

BORJAS, GEORGE J. “Assimilation, Changes in Co-hort Quality, and the Earnings of Immigrants,”J. Lab. Econ., Oct. 1985, 3(4), pp. 463–89.

———. “Self-Selection and the Earnings of Immi-grants,” Amer. Econ. Rev., Sept. 1987, 77(4),pp. 531–53.

———. Friends or strangers: The impact of immi-grants on the U.S. economy. New York: BasicBooks, 1990.

———. “Ethnic Capital and IntergenerationalMobility,” Quart. J. Econ., Feb. 1992a, 107(1),pp. 123–50.

———. “National Origin and the Skills of Immi-grants in the Postwar Period,” 1992b, inGEORGE J. BORJAS AND RICHARD B. FREE-MAN, eds. 1992, pp. 17–47.

———. “The Intergenerational Mobility of Immi-grants,” J. Lab. Econ., Jan. 1993a, 11(1), pp.113–35.

———. “Immigration Policy, National Origin, andImmigrant Skills: A Comparison of Canada andthe United States,” in Small differences thatmatter: Labor markets and income maintenancein Canada and the United States. Eds.: DAVIDCARD AND RICHARD B. FREEMAN. Chicago,IL: U. of Chicago Press, 1993b, pp. 21–43.

———. “Long-Run Convergence of Ethnic SkillDifferentials: The Children and Grandchildrenof the Great Migration,” Ind. Lab. Rel. Rev.,July 1994, 47(4), pp. 553–73.


———. “Assimilation and Changes in CohortQuality Revisited: What Happened to Immi-grant Earnings in the 1980s?” J. Lab. Econ.,forthcoming.

BORJAS, GEORGE J. AND BRATSBERG, BERNT.“Who Leaves? The Outmigration of the For-eign-Born,” Rev. Econ. Statist., forthcoming.

BORJAS, GEORGE J. AND BRONARS, STEPHEN G.“Immigration and the Family,” J. Lab. Econ.,Apr. 1991, 9(2), pp. 123–48.

BORJAS, GEORGE J. AND FREEMAN, RICHARD B.,eds. Immigration and the work force: Economicconsequences for the United States and sourceareas. Chicago: U. of Chicago Press, 1992.

BORJAS, GEORGE J.; FREEMAN, RICHARD B. ANDKATZ, LAWRENCE F. “On the Labor Market Ef-fects of Immigration and Trade,” in GEORGE J.BORJAS AND RICHARD B. FREEMAN, eds. 1992,pp. 213–44.

BORJAS, GEORGE J. AND TREJO, STEPHEN J. “Im-migrant Participation in the Welfare System,”Ind. Lab. Rel. Rev., Jan. 1991, 44(2), pp. 195–211.

BRATSBERG, BERNT. “The Incidence of Non-Re-turn Among Foreign Students in the UnitedStates.” Kansas State U., Sept. 1993.

BUTCHER, KRISTIN F. “Black Immigrants in theUnited States: A Comparison with Native Blacksand Other Immigrants,” Ind. Lab. Rel., Jan.1994, 47(2), pp. 265–84.

BUTCHER, KRISTIN F. AND CARD, DAVID. “Immi-gration and Wages: Evidence from the 1980’s,”Amer. Econ. Rev., May 1991, 81(2), pp. 292–96.

CALIFORNIA DEPARTMENT OF FINANCE. “Edu-cating Illegal Immigrants.” Technical Notes toGovernor’s Budget 1994–95, Jan. 19, 1994.

CARD, DAVID. “The Impact of the Mariel Boatlifton the Miami Labor Market,” Ind. Lab. Rel.Rev., Jan. 1990, 43(2), pp. 245–57.

CARLINER, GEOFFREY. “Wages, Earnings andHours of First, Second and Third GenerationAmerican Males,” Econ. Inquiry, Jan. 1980,18(1), pp. 87–102.

———. “Wage Differences by Language Groupand the Market for Language Skills in Canada,”J. Human Res., Summer 1981, 16(3), pp. 384–99.

CARRINGTON, WILLIAM J. AND DE LIMA, PEDRO.“Large-Scale Immigration and Labor Markets:An Analysis of the Retornados and Their Impacton Portugal.” Johns Hopkins U., Jan. 1994.

CHISWICK, BARRY R. “Sons of Immigrants: AreThey at an Earnings Disadvantage?” Amer.Econ. Rev., Feb. 1977, 67(1), pp. 376–80.

———. “The Effect of Americanization on theEarnings of Foreign-Born Men,” J. Polit. Econ.,Oct. 1978, 86(5), pp. 897–921.

———. “The Earnings of White and ColouredMale Immigrants in Britain,” Economica, Feb.1980, 47(185), pp. 81–87.

———. “An Analysis of the Earnings and Employ-ment of Asian-American Men,” J. Lab. Econ.,Apr. 1983, 1(2), pp. 197–214.

———. “Is the New Immigration Less Skilled

Than the Old?” J. Lab. Econ., Apr. 1986, 4(2),pp. 168–92.

———. “Speaking, Reading, and Earnings AmongLow-Skilled Immigrants,” J. Lab. Econ., Apr.1991, 9(2), pp. 149–70.

———–, ed. Immigration, language, and ethnic-ity: Canada and the United States. Washington:American Enterprise Institute, 1992.

CHISWICK, BARRY R. AND MILLER, PAUL W.“Language in the Immigrant Labor Market,” inBARRY R. CHISWICK, ed. 1992, pp. 229–96.

COBB-CLARK, DEBORAH A. “Immigrant Selectiv-ity: The Roles of Household Structure and U.S.Immigration Policy.” Ph.D. Dissertation, U. ofMichigan, 1990.

———. “Immigrant Selectivity and Wages: TheEvidence for Women,” Amer. Econ. Rev., Sept.1993, 83(4), pp. 986–93.

COLEMAN, JAMES S. “Social Capital in the Crea-tion of Human Capital,” Amer. J. Sociology,1988 Supplement, 94, pp. S95–120.

DENEW, JOHN P. AND ZIMMERMANN, KLAUS F.“Native Wage Impacts of Foreign Labor: A Ran-dom Effects Panel Analysis,” J. PopulationEcon., June 1994, 7(2), pp. 177–92,

DOUGLAS, PAUL. “Is the New Immigration MoreUnskilled than the Old?” J. Amer. Statist. As-soc., June 1919, 16, pp. 393–403.

DULEEP, HARRIET ORCUTT AND REGETS, MARKC. “Some Evidence on the Effect of AdmissionCriteria on Immigrant Assimilation,” in 1992aBARRY R. CHISWICK, ed. 1992, pp. 410–39.

———. “The Elusive Concept of Immigrant Qual-ity.” Urban Institute, 1992b.

DUSTMANN, CHRISTIAN. “Earnings Adjustment ofTemporary Immigrants,” J. Population Econ.,May 1993, 6(2), pp. 153–68.

EDMONSTON, BARRY AND PASSEL, JEFFREY S.“Immigration and Immigrant Generations inPopulation Projections,” Int. J. Forecasting,Nov. 1992, 8(3), pp. 459–76.

EICHENGREEN, BARRY AND GEMERY, HENRY A.“The Earnings of Skilled and Unskilled Immi-grants at the End of the Nineteenth Cen-tury,” J. Econ. Hist., June 1986, 46(2), pp. 441–54.

FILER, RANDALL K. “The Effect of Immigrant Ar-rivals on Migratory Patterns of Native Workers,”in GEORGE J. BORJAS AND RICHARD B. FREE-MAN, eds. 1992, pp. 245–69.

FREEMAN, RICHARD B. “Immigration from Poorto Wealthy Countries: Experience of the UnitedStates,” Europ. Econ. Rev., Apr. 1993, 37(2/3),pp. 443–51.

FREY, WILLIAM H. “The New White Flight,”American Demographics, Apr. 1994, 16(4), pp.40–48.

FRIEDBERG, RACHEL M. “The Labor Market As-similation of Immigrants in the United States:The Role of Age at Arrival.” Brown U., 1992.

———. “You Can’t Take It With You: ImmigrantAssimilation and the Portability of Human Capi-tal, Evidence from Israel.” Brown U., Dec.1993.


FUNKHOUSER, EDWARD AND TREJO, STEPHEN J.“The Labor Market Skills of Recent Male Immi-grants: Evidence from the Current PopulationSurveys.” Ind. Lab. Rel Rev., forthcoming.

GLAZER, NATHAN AND MOYNIHAN, DANIEL P.Beyond the melting pot: The Negroes, Puerto Ri-cans, Jews, Italians, and Irish of New York City.Cambridge, MA: M.I.T. Press, 1963.

GOLDIN, CLAUDIA. “The Political Economy of Im-migration Restrictions in the United States,1890–1921,” in The regulated economy. Eds.:CLAUDIA GOLDIN AND GARY LIBECAP. Chi-cago: U. of Chicago Press, 1994, pp. 223–58.

GREEN, ALLAN G. AND GREEN, DAVID A. “Cana-dian Immigration Policy: The Effectiveness ofthe Point System and Other Instruments.” TheU. of British Columbia, Jan. 1994.

GRENIER, GILLES. “The Effects of LanguageCharacteristics on the Wages of Hispanic-American Males,” J. Hum. Res., Winter 1984,19(1), pp. 35–52.

GROSSMAN, JEAN BALDWIN. “The Substitutabilityof Natives and Immigrants in Production,” Rev.Econ. Statist., Nov. 1982, 64(4), pp. 596–603.

HUDDLE, DONALD. “The Costs of Immigration.”Rice U., July 1993.

HUNT, JENNIFER. “The Impact of the 1962 Repa-triates from Algeria on the French Labor Mar-ket,” Ind. Lab. Rel. Rev., Apr. 1992, 45(3), pp.556–72.

JASSO, GUILLERMINA AND ROSENZWEIG, MARKR. “What’s in a Name? Country-of–Origin Influ-ences on the Earnings of Immigrants in theUnited States,” in Research in human capitaland development. Vol. 4. Ed.: ODED STARK.Greenwich, CT: JAI Press, 1986, pp. 75–106.

JENSEN, LEIF. “Patterns of Immigration and Pub-lic Assistance Utilization, 1970–1980,” Interna-tional Migration Rev., Spring 1988, 22(1), pp.51–83.

KASTEN, RICHARD; SAMMARTINO, FRANK ANDTODER, ERIC. “Trends in Federal Tax Progres-sivity: 1980–1993.” U.S. Congressional BudgetOffice, Feb. 1993.

KATZ, ELIAKIM AND STARK, ODED. “InternationalMigration under Asymmetric Information,”Econ. J., Sept. 1987, 97(387), pp. 718–26.

KOSSOUDJI, SHERRIE A. “Immigrant Worker As-similation: Is It a Labor Market Phenomenon?”J. Human Res., Summer 1989, 24(3), pp. 494–527.

LALONDE, ROBERT J. AND TOPEL, ROBERT H.“Labor Market Adjustments to Increased Immi-gration,” in JOHN M. ABOWD AND RICHARD B.FREEMAN, eds. 1991, pp. 167–99.

———. “The Assimilation of Immigrants in theU.S. Labor Market,” in GEORGE J. BORJAS ANDRICHARD B. FREEMAN, eds. 1992, pp. 67–92.

LEVY, FRANK AND MURNANE, RICHARD J. “U.S.Earnings Levels and Earnings Inequality: A Re-view of Recent Trends and Proposed Explana-tions,” J. Econ. Lit., Sept. 1992, 30(3), pp.1333–81.

LIGHT, IVAN AND BONACICH, EDNA. Immigrant

entrepreneurship: Koreans in Los Angeles,1965–1982. Berkeley: U. of California Press,1988.

LONG, JAMES E. “The Effect of Americanizationon Earnings: Some Evidence for Women,” J.Polit. Econ., June 1980, 88(3), pp. 620–29.

LOURY, GLENN C. “A Dynamic Theory of RacialIncome Differences,” in Women, minorities,and employment discrimination. Eds.: PHYLLISA. WALLACE AND ANNETTE M. LAMOND. Lex-ington, MA: Lexington Books, 1977, pp. 153–86.

LOWELL, B. LINDSAY; BEAN, FRANK D. AND DELA GARZA, RODOLFO O. “Undocumented Im-migration: An Analysis of the 1984 Simpson-Mazzoli Vote,” Soc. Sci. Quart., Mar. 1986,67(1), pp. 118–27.

LUCAS, ROBERT E., JR. “On the Mechanics ofEconomic Development,” J. Monet. Econ., July1988, 22(1), pp. 3–42.

MCMANUS, WALTER S. “Labor Market Costs ofLanguage Disparity: An Interpretation of His-panic Earnings Differences,” Amer. Econ. Rev.,Sept. 1985, 75(4), pp. 818–27.

———. “Labor Market Effects of Language En-claves: Hispanic Men in the United States,”J. Human Res., Spring 1990, 25(2), pp. 228–52.

MCMANUS, WALTER; GOULD, WILLIAM ANDWELCH, FINIS. “Earnings of Hispanic Men:The Role of English Language Proficiency,” J.Lab. Econ., Apr. 1983, 1(2), pp. 101–30.

MULLER, THOMAS AND ESPENSHADE, THOMAS J.The fourth wave. Washington, DC: Urban Insti-tute Press, 1985.

PASSEL JEFFREY S. AND CLARK, REBECCA L.“How Much Do Immigrants Really Cost? A Re-appraisal of Huddle’s ‘The Costs of Immi-grants’.” Urban Institute, Feb. 1994.

PISCHKE, JÖRN-STEFFEN. “Assimilation and theEarnings of Guestworkers in Germany.” Massa-chusetts Institute of Technology, 1993.

PISCHKE, JÖRN-STEFFEN AND VELLING, JOHAN-NES. “Wage and Employment Effects of Immi-gration to Germany: An Analysis Based on LocalLabor Markets.” Massachusetts Institute ofTechnology, Jan. 1994.

PORTES, ALEJANDRO. “The Social Origins of theCuban Enclave in Miami,” Sociological Perspec-tives, Fall 1987, 30(3), pp. 340–72.

RAMOS, FERNANDO A. “Out-Migration and Re-turn Migration of Puerto Ricans,” in GEORGE J.BORJAS AND RICHARD B. FREEMAN, eds. 1992,pp. 49–66.

REIMERS, CORDELIA W. “Labor Market Discrimi-nation against Hispanic and Black Men,” Rev.Econ. Statist., Nov. 1983, 65(4), pp. 570–79.

ROMER, PAUL M. “Increasing Returns and Long-run Growth,” J. Polit. Econ., Oct. 1986, 94(5),pp. 1002–37.

ROY, ANDREW D. “Some Thoughts on the Distri-bution of Earnings,” Oxford Econ. Pap., N.S.,June 1951, 3, pp. 135–46.

SATTINGER, MICHAEL. “Assignment Models ofthe Distribution of Earnings,” J. Econ. Lit.,June 1993, 31(2), pp. 831–80.


SIMON, JULIAN L.; MOORE, STEPHEN AND SULLI-VAN, RICHARD. “The Effect of Immigration onAggregate Native Unemployment: An Across-City Estimation,” J. Lab. Res., Summer 1993,14(3), pp. 299–316.

SMITH, JAMES. “Hispanics and the AmericanDream: An Analysis of Hispanic Male LaborMarket Wages, 1940–1980.” Rand Corporation,1992.

SOLON, GARY R. “Intergenerational Income Mo-bility in the United States,” Amer. Econ. Rev.,June 1992, 82(3), pp. 393–408.

TAINER, EVELINA. “English Language Proficiencyand the Determination of Earnings among For-eign-Born Men,” J. Human Res., Winter 1988,23(1), pp. 108–22.

TANTON, JOHN AND LUTTON, WAYNE. “Immigra-tion and Criminality in the U.S.A.,” J. Social,Polit. Econ. Studies, Summer 1993, 18(2), pp.217–34.

TAYLOR, J. EDWARD. “Undocumented Mexico-U.S. Migration and the Returns to Householdsin Rural Mexico,” Amer. J. Agr. Econ., Aug.1987, 69(3), pp. 626–38.

TOPEL, ROBERT H. “Regional Labor Markets andthe Determinants of Wage Inequality,” Amer.Econ. Rev., May 1994, 84(2), pp. 17–22.

UNITED NATIONS. World population at theturn of the century. New York: United Nations,1989.

U.S. DEPARTMENT OF COMMERCE. BUREAU OFTHE CENSUS. Historical statistics of the UnitedStates, colonial times to 1970. Washington, DC:U.S. GPO, 1975.

———. The foreign-born population in the UnitedStates. Washington, DC: Bureau of the CensusReport 1990 CP-3–1, 1993a.

———. Statistical abstract of the United States,1993. Washington, DC: U.S. GPO, 1993b.

U.S. DEPARTMENT OF JUSTICE. IMMIGRATIONAND NATURALIZATION SERVICE. Statisticalyearbook of the immigration and naturalization

service, 1992. Washington, DC: U.S. GPO,1993.

U.S. GENERAL ACCOUNTING OFFICE. Illegalaliens: Despite data limitations, current meth-ods provide better population estimates. ReportNumber GAO/PEMD-93–25, August 1993.

WALDINGER, ROGER. “Who Gets the ‘LousyJobs’? New Immigrants and African-Americansin New York, 1940–1990.” U. of California, LosAngeles, Oct. 1993.

WARREN, ROBERT AND PECK, JENNIFER MARKS.“Foreign-Born Emigration from the UnitedStates: 1960–1970,” Demography, Feb. 1980,17(1), pp. 71–84.

WHITE, MICHAEL J. AND HUNTER, LORI. “TheMigratory Response of Native-Born Workers tothe Presence of Immigrants in the Labor Mar-ket.” Brown U., July 1993.

WHITE, MICHAEL J. AND LIANG, ZAI. “The Effectof Immigration on the Internal Migration of theNative-Born Population, 1981–90.” Brown U.,Aug. 1993.

WILSON, WILLIAM JULIUS. The truly disadvan-taged: The inner city, the underclass, and pub-lic policy. Chicago: U. of Chicago Press, 1987.

WINEGARDEN, C. R. AND KHOR, LAY BOON. “Un-documented Immigration and Unemploymentof U.S. Youth and Minority Workers:Econometric Evidence,” Rev. Econ. Statist.,Feb. 1991, 73(1), pp. 105–12.

WRIGHT, ROBERT E. AND MAXIM, PAUL S. “Im-migration Policy and Immigrant Quality: Em-pirical Evidence from Canada,” J. PopulationEcon., 1993, 6(4), pp. 337–52.

YUENGERT, ANDREW. “Immigrant Earnings, Rela-tive to What? The Importance of EarningsFunction Specification and Comparison Points,”J. Applied Econometrics, Jan.-Mar. 1994, 9(1),pp. 71–90.

ZIMMERMAN, DAVID J. “Regression Toward Medi-ocrity in Economic Stature,” Amer. Econ. Rev.,June 1992, 82(3), pp. 409–29.


the economics of immigration · pdf filemigration occurred between 1881 and ... mexico 22.3...

Documents