Illegal Migration From Mexico to the US (Dec 2006)

8/3/2019 Illegal Migration From Mexico to the US (Dec 2006)

1/57

869

Journal of Economic LiteratureVol. XLIV (December 2006), pp. 869924

Illegal Migration from Mexico tothe United States

GORDON H. HANSON

In this paper, I selectively review recent literature on illegal migration from Mexico tothe United States. I begin by discussing methods for estimating stocks and flows ofillegal migrants. While there is uncertainty about the size of the unauthorized popu-lation, new data sources make it possible to examine the composition of legal and ille-

gal populations and the time-series covariates of illegal labor flows. I then considerthe supply of and demand for illegal migrants. Wage differentials between the UnitedStates and Mexico are hardly a new phenomenon, yet illegal migration from Mexico

did not reach high levels until recently. An increase in the relative size of Mexicos working-age population, greater volatility in U.S.Mexico relative wages, andchanges in U.S. immigration policies are all candidate explanations for increasinglabor flows from Mexico. Finally, I consider policies that regulate the cross-border

flow of illegal migrants. While U.S. laws mandate that authorities prevent illegal entry

and punish firms that hire unauthorized immigrants, these laws are imperfectlyenforced. Lax enforcement may reflect political pressure by employers and other

interests that favor open borders.

University of California, San Diego, and NationalBureau of Economic Research. I thank Roger Gordon andseminar participants at the 2006 AEA meetings, BostonUniversity, MIT, Syracuse University, and Tufts Universityfor helpful comments and Jeff Lin and Maribel Pichardofor excellent research assistance.

1. Introduction

There is increasing interest by academ-ics and policymakers in Mexicanmigration to the United States. Mexico isthe most important source country forU.S. immigration, accounting for 34 per-cent of total immigrant arrivals since 1990.In 2004, the 10.5 million Mexican immi-grants living in the United States were 31

1 Figure 1 reports the total number of Mexican immi-grants in the United States (legal and illegal) as a share ofMexicos population, the total U.S. population, and the for-eign-born U.S. population.

2 Over the period 1841 to 1860, Ireland accounted for39 percent of U.S. immigration and over the period 1841to 1890, Germany accounted for 30 percent of U.S. immi-gration (U.S. Department of Homeland Security 2004).

percent of the U.S. foreign-born populationand equivalent to 10 percent of the totalpopulation of Mexico (see figure 1).1 TheUnited States has not experienced a concen-trated immigration wave of this magnitude

since the influxes from Germany and Irelandin the mid-nineteenth century.2 For Mex-ico, the continuing population outflow is


2/57

870 Journal of Economic Literature, Vol. XLIV (December 2006)

Figure 1. Mexican Immigrants in the U.S.

1960 1970 1980 1990 2000

0

.1

.2

.3

0

.1

.2

.3Share of All U.S. Immigrants

Share of U.S. PopulationShare of Mexican Population

3 The one episode that approaches the current outflowis the Mexican Revolution (191120). Between 1911 and1925, 680,000 legal immigrants from Mexico (or 5 percentof Mexicos 1910 population) entered the United States(and were joined by a large number of illegal immigrants).By the late 1920s, many of these individuals had returnedto Mexico. See Lawrence Cardoso (1980) and Fernando

Alanis Enciso (1999).

unprecedented.3 In both countries, the

cross-border flow of labor appears to haveaffected the structure of wages, the intrana-tional distribution of population, and thepattern of industrial specialization.

Beyond its scale, the distinguishing fea-ture of Mexican immigration is that mostnew arrivals enter the United States illegal-ly. In 2004, there were an estimated 5.9 mil-lion unauthorized Mexican immigrants inthe United States, among a total unautho-rized population of 10.3 million (Jeffrey S.

Passel 2005). Thus, 56 percent of Mexicanimmigrants appear to lack permission to bein the country, compared to 17 percent ofall other immigrants. Large-scale illegal

immigration in the United States is a rela-

tively new phenomenon. It has provokedpolitical debate about whether to providepublic services to illegal immigrants, grantthem status as legal residents, or militarizeU.S. borders to prevent further illegalinflows. In Mexico, migration abroad hashelped ease the countrys adjustment torapid growth in its working-age populationand to macroeconomic shocks, though not

without disrupting the families and commu-nities whose members have moved to the

United States.There is an emerging body of economic

research on illegal migration from Mexico tothe United States. This literature has beenmade possible by the recent availability ofdata sources on the cross-border move-ments of legal and illegal Mexican migrants.Also prompting attention is the realizationthat, with unauthorized entrants accountingfor over half of Mexican immigrants and

over three tenths of all U.S. immigrants, any


3/57

Hanson: Illegal Migration from Mexico to the United States 871

4 For reviews of this literature, see Douglas S. Masseyet al. (1994) and Thomas J. Espenshade (1995).

5 The current U.S. immigration quota system wasestablished by the HartCellar Act of 1965 and amendedby the Immigration Act of 1990. The 1990 law set a flexi-ble cap for U.S. legal admissions at 675,000 individuals ofwhich 480,000 are to be family-based, 140,000 are to beemployment-based, and 55,000 are to be diversity immi-grants. Immediate relatives of U.S. citizens are not subject

to immigration quotas.

discussion of international migration in theUnited States or Mexico ends up confrontingthe issue of illegality either explicitly orimplicitly.

Much of the initial research on unautho-

rized labor flows was done by nonecono-mists.4 The principal themes of this body ofwork resemble those in the economics liter-ature on internal migration in developingcountries (see Robert E. B. Lucas 1997;Hillel Rapoport and Frdric Docquierforthcoming). Early waves of illegal migra-tion from Mexico appear to have originatedin rural areas of the country (Wayne A.Cornelius 1992; Jorge Durand, Massey, andRen Zenteno 2001), involved households

financing the migration of one or moremembers in return for remittances from themigrants (Durand, Emilio A. Parrado, andMassey 1996; Durand et al. 1996), anddepended on family and community net-

works that helped migrants enter and findemployment in the United States (Massey etal. 1994; Massey and Kristin E. Espinosa1997).

Yet, internal migration and illegal interna-

tional migration differ in important respects.While policy barriers that restrict within-country regional labor flows are rare, coun-tries actively regulate the inflow of laborfrom abroad. The United States determinesthe level of legal immigration through quo-tas on entry visas, which change infrequent-ly over time.5 The country implicitly sets thelevel of illegal immigration through select-ing the intensity with which it enforces bor-ders against illegal entry. Key issues for the

study of illegal migration are how countrieschoose their border regulation policies and

6 For surveys of this literature, see James P. Smith andBarry Edmonston (1998) and George J. Borjas (1999a,1999b) and for recent work in the area see Borjas (2003)

and David Card (2005). For analysis of these issues in thecontext of Mexican immigration, see Smith (2003),Stephen J. Trejo (2003), Francise D. Blau and LawrenceM. Kahn (forthcoming), Borjas and Lawrence F. Katz(forthcoming), Card and Ethan Lewis (forthcoming), andBrian Duncan and Trejo (forthcoming). There is a smallerliterature on the consequences of emigration for Mexico.See Prachi Mishra (forthcoming) and Gordon H. Hanson(forthcoming).

7 Exceptions include Timothy J. Hatton and Jeffrey G. Williamson (2004) and Anna Maria Mayda (2005), whoexamine the correlates of international migration flows. Inresearch on internal migration, there is considerable workon the incentive to migrate. See Michael J. Greenwood

(1997) and Lucas (1997) for reviews of the literature.

how prospective migrants respond to thesepolicies.

There is a large literature on U.S. immi-gration, which tends to focus on the labor-market consequences of immigrant inflows

and the economic performance of immi-grants.6 This body of work examines, amongother questions, whether immigrationreduces wages for U.S. native workers;

whether immigrants make relatively greateruse of means-tested entitlement programs;and whether earnings, education, fertility, orother outcomes for immigrants converge tonative levels over time.

Largely taken for granted in the U.S. liter-ature iswhy foreign residents migrate to the

United States. One obvious reason is thatU.S. real wages far exceed those in manyother countries. Large wage differentials,coupled with binding and slowly changingquotas on U.S. legal immigration, createqueues to enter the United States. Givenextended delays in clearing such queues,annual variation in the level of legal immi-gration appears to be more or less insensitiveto contemporaneous annual fluctuations in

U.S. or foreign econ-omies. Perhaps as aresult, the quantity of literature on the con-sequences of U.S. legal immigration vastlyexceeds that on its causes.7

With illegal immigration, the determi-nants of migrant flows and the high-frequency variation of these flows have


4/57


attracted more attention. Geographic prox-imity allows unauthorized migrants fromMexico to move to the United Statesrelatively quickly. The existence of well-established migration networks enables U.S.

employers to communicate changes in theirdemand for labor to prospective migrants inMexico. Migrants use these same networksto find jobs and housing in the UnitedStates. Shocks to either the Mexican or U.S.economies may be transmitted into changesin cross-border labor flows with relativelyshort time lags, making illegal migrationpotentially quite responsive to changes inbinational business-cycle conditions.

Another feature that distinguishes legal

and illegal migrant flows is their composi-tion. While legal migrants face entry costsassociated with queues in obtaining visas,illegal migrants face costs associated withevading immigration authorities. Once in thereceiving country, the risk of detection maymake some employers unwilling to hire ille-gal migrants, limiting their occupationalprospects and reducing the returns to skillthey perceive. Variation in migration costs

and in receiving-country wage profilesbetween legal and illegal migration suggestthe characteristics of illegal migrants maydiffer from those of legal migrants.Observed changes in the composition ofU.S. immigrants, which has received muchattention in research on consequences ofU.S. immigration, could be partly a by-product of the rising share of unauthorizedentrants in immigrant inflows. For Mexico,the composition of migrant outflows matters

not just for the labor-market effects of emi-gration but also for the ties that migrantsretain with the country. Illegal migrantsappear to be more likely than legal migrantsto send remittances to family members inMexico. Rising illegal migration fromMexico may be partly responsible for therecent surge in remittances in the country,

which rose from 0.1 percent of GDP in 1990to 2.2 percent of GDP in 2004 and now gen-

erate more foreign exchange in the country

than tourism or foreign direct investment(Inter-American Development Bank 2004).

In this paper, I selectively review recentliterature on illegal migration from Mexico tothe United States. In section 2, I discuss

methods for estimating stocks and flows ofillegal migrants. While there is uncertaintyabout the size of the unauthorized popula-tion, new data sources make it possible toexamine the composition of legal and illegalpopulations and the time-series covariates ofillegal labor flows. In section 3, I consider thesupply of and demand for illegal migrants.

Wage differentials between the United Statesand Mexico are hardly a new phenomenon,

yet illegal migration from Mexico did not

reach high levels until recently. An increasein the relative size of Mexicos working-agepopulation, greater volatility in U.S.Mexicorelative wages, and changes in U.S. immigra-tion policies are all candidate explanations forincreasing labor flows from Mexico. In sec-tion 4, I consider policies that regulate thecross-border flow of illegal migrants. WhileU.S. laws mandate that authorities preventillegal entry and punish firms that hire unau-

thorized immigrants, these laws are imper-fectly enforced. Lax enforcement may reflectpolitical pressure by employers and otherinterests that favor open borders. In section5, I discuss directions for further research.

My goal in this paper is not to conduct anexhaustive survey of work on illegal migra-tion but rather to highlight major findings inthe literature, assess the state of importantdebates, and identify unresolved issues, withan eye toward advancing questions to help

guide future work. As much of the literatureis empirical, I will focus on this strand ofresearch, with occasional forays into theory.One topic I will not discuss at much length isthe economic consequences of illegal migra-tion, in part because literature on the conse-quences of immigration has been subject toseveral recent surveys (see note 6) and inpart because there is little research on thespecific aspects of these consequences that

are attributable to illegal immigration.


5/57


8 This also appears to be the case in Western Europe

(Tito Boeri, Barry McCormick, and Hanson 2002).

Though my focus is on the United Statesand Mexico, insights from the literature arerelevant for other regions, as well.Unauthorized migration has become a glob-al phenomenon. In the last two decades,

there have been sizable flows of illegalmigrants from North Africa and EasternEurope to Western Europe, from Indonesiato elsewhere in Southeast Asia, and fromneighboring countries to South Africa. TheU.S.Mexico experience may be instructivefor these and other cases regarding how tomeasure unauthorized migration, estimatethe causes and consequences of migrationflows, and gauge the potential impacts ofpolicy interventions.

2. Stocks and Flows of Illegal Migrants

Illegal immigrants account for a large andgrowing fraction of the U.S. foreign-bornpopulation.8 One may imagine that, as partof the underground economy, unauthorizedmigrants are not easily subject to measure-ment. However, there is now abundant evi-dence that illegal immigrants are representedin official household surveys, including the

U.S. Census of Population and Housing andthe U.S. Current Population Survey. Givenknown levels of U.S. legal immigration, thenumber of foreign-born individuals enumer-ated in these sources is far too large for themall to be legal.

The most common method to estimatethe number of illegal immigrants is to takethe difference between the measured immi-grant population and the sum of past legalimmigrant inflows. Estimates using thisresidual approach suggest that stocks of ille-gal immigrants have risen sharply over time.However, there is considerable variability inthe estimates, associated with differingassumptions about the magnitude of errorsin enumerating legal and illegal immigrantsin official data sources.

Knowing the overall share of immigrants who are unauthorized is not the same asknowing which specific immigrants in agiven data source are unauthorized.Fortunately, there are now several micro-

level surveys that provide information onindividual migration status. Either bydesign or default, migrants from Mexicoaccount for a large fraction of those repre-sented in these data sources. While eachsurvey has limitations, their use in conjunc-tion with large public data sets fromMexico and the United States providesconsiderable detail on the population oflegal and illegal migrants from Mexico whoare living or have lived in the United

States.A third data source on unauthorized

migration is a by-product of U.S. immigra-tion policy. To prevent illegal immigration,the U.S. Border Patrol polices U.S. bordersand ports of entry, attempting to apprehendthose seeking to enter the country illegally.The Border Patrol compiles high-frequencydata on apprehensions and enforcement, the

vast majority of which occur along the

U.S.Mexico border. Data on border appre-hensions and enforcement allow one toexamine how attempts at illegal entry vary

with economic conditions in the UnitedStates and Mexico and to see which factorsare associated with the intensity of U.S.enforcement activities.

2.1 The Residual Foreign-Born Population

In the United States, there are two classesof legal immigrants who appear in official

data sources (i.e., are surveyed by the U.S.Census Bureau, the U.S. Bureau of LaborStatistics (BLS), or other official entities).One is permanent legal immigrants, whohave the right to reside in the country indef-initely. Another is temporary legal immi-grants, who have the right to reside in thecountry for a defined time period, as speci-fied by an entry visa (such as those for stu-dents, specific skill classes of workers,

diplomats, and family members of temporary


6/57


9 The Census Bureau does not survey individuals ontourist or business visas or other short-term visitors.

10 See Bean et al. (1998) for a survey of the literature on

estimating illegal immigrant populations.

legal immigrants).9 Since government sur-veys do not screen individuals based on theirimmigration status, illegal immigrants alsoappear in official data, to the extent theymake themselves available to be surveyed.

The standard method to estimate the num-ber of illegal immigrants is to assume it isequal to the residual foreign-born population,

which is given by

(1) Ut = Ft t

s=1Ls(1 ms es) Tt,

where Ut is the unauthorized foreign-bornpopulation in year t, Ft is the total foreign-born population in yeart,Ls is the number ofpermanent legal immigrants that entered in

year s t, ms is the mortality rate betweenyears and yeart for legal immigrants enter-ing in year s, es is the emigration ratebetween year s and year t for legal immi-grants entering in years, and Tt is temporarylegal immigrants present in yeart.

Table 1 reports estimates of the unautho-rized foreign-born population by the U.S.Bureau of the Census (Joe Costanzo et al.2001), Bean et al. (2001a, 2001b), Passel(2005), and the U.S. Immigration andNaturalization Service (INS) (2001).10 In themost recent projection, Passel (2005) esti-mates that between 2002 and 2004 the ille-gal immigrant population rose from 9.3 to10.3 million, for an average annual netinflow of 500,000 migrants, with 57 percentof these individuals coming from Mexico.This compares to an average annual net ille-gal inflow during the 19902000 period of581,000 migrants (with 58 percent of net

new immigrants coming from Mexico),based on U.S. Census Bureau estimates(Costanzo et al. 2001), and 350,000 migrants(with 79 percent coming from Mexico),based on INS (2001) estimates.

Obviously, there are a host of assumptionsinvolved in estimating (1). One must assign

mortality and emigration rates to each entrycohort and determine the fraction of tempo-rary immigrants admitted in previous yearsthat are still in the country. In practice, mostdiscrepancies evident in table 1 appear unre-

lated to differences in these assumptions. Ofgreater importance are assumptions aboutmeasurement error in Ft.

To simplify matters, I reexpress equation(1) in contemporaneous values as Ft =Lt+ Ut,

whereLt is the total legal foreign-born pop-ulation in year t. The U.S. Census Bureau(and other entities that conduct householdsurveys) tends to undercount the total popu-lation (with undercount rates for low-incomehouseholds, which would include many

Mexican immigrants, believed to be relative-ly high), in which case the measured foreign-born population, Ftt, is less than the actualforeign-born population, Ft. The total legal-immigrant population, Lt, in contrast,appears to be measured with greater accura-cy, since immigration authorities haverecords on how many entry visas they award.Suppose the legal-immigrant population thatis enumerated in the census is,

(2) Ltt =Lt(1 t) + Lt,

where t is the fraction of legal immigrantsthat go uncounted, and the unauthorized-immigrant population enumerated in thecensus is,

(3) Utt = Ut(1 t) + Ut,

where t is the fraction of unauthorizedimmigrants that go uncounted and Lt and Utare mean zero iid errors. Lt t and Utt are notobserved individually but are observed interms of their sum, Ftt. While the U.S.Census Bureau estimates that it under-counts the Hispanic population by 5 per-cent, Bean et al. (2001b) put the undercountrate for legal Mexican immigrants at 2 per-cent to 4 percent. Similarly, while the U.S.Census Bureau estimates that it under-counts the illegal immigrant population by

15 percent, the INS assumes an undercount


7/57


TABLE 1ESTIMATES OF THE U.S. ILLEGAL-IMMIGRANT POPULATION, 19902004

Total Population of Illegal Immigrants in the United States (millions)

Costanzo et al. (2003) Bean et al. (2001a, 2001b)

INS Undercount Rate Undercount Rate PasselYear (2001) 10% 15% 20% 15% Median 25% (2005)

1990 3.500 3.766 4.430 4.707 -- -- -- --1991 4.025 -- -- -- -- -- -- --1992 4,204 -- -- -- -- -- -- --1993 4.492 -- -- -- -- -- -- --1994 4.750 -- -- -- -- -- -- --1995 5.146 -- -- -- -- -- -- --1996 5.581 -- -- -- -- -- -- --1997 5.862 -- -- -- -- -- -- --1998 6.098 -- -- -- -- -- -- --1999 6.488 -- -- -- -- -- -- --2000 7.000 8.705 10.242 10.882 -- -- -- --2001 -- -- -- -- 5.918 7.751 9.864 --2002 -- -- -- -- -- -- -- 9.3002003 -- -- -- -- -- -- -- --2004 -- -- -- -- -- -- -- 10.300

Population of Illegal Immigrants from Mexico in the United States (millions)

1990 2.040 1.008 -- -- -- -- -- --1996 -- -- -- -- 1.524 2.543 3.706 --2000 4.808 3.872 -- -- -- -- -- --2001 -- -- -- -- 3.462 4.510 5.765 --2002 -- -- -- -- -- -- -- 5.3002003 -- -- -- -- -- -- -- --

2004 -- -- -- -- -- -- -- 5.900

Notes: Costanzo et al. (2003) report estimates of the total U.S. unauthorized population using three alternativeundercount rates for illegal immigrants (10 percent, 15 percent, or 20 percent). The INS (2001) uses a constantundercount rate for illegal immigrants of 10 percent, as does Passel (2005). The 1996 Bean et al. (2001a) esti-mates are based on undercount rates of either (a) 15 percent for illegal immigrants and 3 percent for legal immi-grants or (b) 25 percent for illegal immigrants and 5 percent for legal immigrants. The 2001 Bean et al. (2001b)estimates are based on undercount rates of either (a) 15 percent for illegal immigrants and 0.5 percent for alllegal immigrants (2 percent for legal Mexican immigrants) or (b) 25 percent for illegal immigrants and 2 percentfor all legal immigrants (4 percent for legal Mexican immigrants). The median estimates for Bean et al. (2001a,2001b) are for the full set of reported of undercount rates, whch vary depending on the years and sample.

of 10 percent, and Bean et al. (2001) useundercount rates of 15 percent to 25 per-cent. It is primarily these differences inassumptions that account for variation in theestimates in table 1.

Assumptions about undercount rates arebased on comparisons of the enumeratedpopulation with the population measured inpostenumeration surveys conducted in spe-cific localities. This requires one to assume

that underenumeration in these localities is

representative of the total U.S. population(see Bean et al. 1998). The U.S. CensusBureau justifies its undercount assumptionsbased on results from its own postenumera-tion surveys, from which standard errors forpopulation estimates are derived (Costanzoet al. 2001). The INS (2001) justifies a 10percent undercount rate based on a casestudy of Mexican immigrants in Los AngelesCounty. Bean et al. (2001a, 2001b) justify

their range of undercount rates based on


8/57


11 Ignoring interactions in undercount rates,t ttUt t [Ftt (t t t)Lt Lt Ut ].

(1 t)(1 tt)

When the assumed undercount rate for illegal immigrantsis low, this expression will be negative (there will beupward bias in the estimated number of illegal immi-grants). This effect will be exaggerated if the assumedundercount rate for legal immigrants is also low.

12 In this exercise, I assume the undercount rate for

legal migrants is held constant at 2 percent.

results in Bean and Jennifer Van Hook(1998).

Given observed values of Ftt and Lt andassumed values for t or t, which I indicateby a tilde, the estimated value of the illegal-

immigrant population in year t can beapproximated as

Ftt Lt(1 t t)(4) t = .1 tt

If true undercount rates fall over time andwe fail to account for this, estimates of theillegal-migration population will be biasedupwards (t > Ut).11 Consider values for Fttand Lt in 1996 and 2001 in Bean et al.(2001a, 2001b). If we change the assumption

for the 19962001 period from a constantillegal undercount rate of 25 percent (whichis at the upper end of the Bean et al. rates)to a reduction in the illegal undercount ratefrom 25 percent to 15 percent (to the lowerend of the Bean et al. rates),12 the estimatedannual illegal net inflow from Mexico wouldfall by 112,000 migrants (from 432,000 to320,000), which is 30 percent of the averageannual estimated illegal inflow over the peri-

od. Since different authors tend to use dif-ferent postenumeration surveys as the basisfor selecting undercount rates, there is littleconsensus in the literature about what hashappened to the undercount of illegal immi-grants over time, other than it exhibiting adownward trend.

In table 1, the span of estimates for theillegal-immigrant population in the UnitedStates is wide. Between 1990 and 2004, theestimated average annual net inflow of unau-

thorized immigrants ranges from 350,000 to580,000 individuals, with 55 percent to 80

percent of these individuals coming fromMexico. Missing in the literature are attemptsto gauge how the validity of different assump-tions about undercount rates vary acrosstime. The postenumeration surveys (typically

conducted by or on behalf of the U.S. CensusBureau or BLS) that are the basis for theundercount assumptions are carried out in asmall number of communities around thetime of the main survey. Since samples forpostenumeration surveys differ across surveyperiods, there is little basis for making time-series comparisons in undercount rates. Untillarge-scale public surveys ask about individ-ual migration status directly, estimating thesize of the illegal-immigrant population will

remain a speculative enterprise.

2.2 Churning in Legal and IllegalImmigrant Populations

Estimates of the stock of illegal immi-grants give little indication of how longunauthorized migrants are likely to remain

without a legal resident visa. Each year,there appears to be a large flow of individu-als from the pool of illegal migrants to the

pool of permanent legal immigrants. Manyimmigrants who obtain visas for legal perma-nent residence (green cards) are at the timethey obtain their visas residing in the UnitedStates illegally. Figure 2 shows the numberof Mexican immigrants awarded legal per-manent residence and the fraction of theseindividuals who are adjusting status. Overthe period 19922002, status adjustersaccounted for 56 percent of new legal per-manent immigrants from Mexico. Some of

those adjusting their visa status are tempo-rary legal immigrants who have succeeded inobtaining permanent entry visas. However,for Mexican immigrants, the majority ofthose adjusting status to permanent legalresidence appear to have been living in theUnited States as illegal immigrants (U.S.Department of Homeland Security 2004).

Further evidence of churning in the pop-ulation of legal and illegal immigrants is

available in the New Immigrant Survey


9/57


Figure 2. U.S. Legal Immigration from Mexico

LegalImmigrants

ShareAdjustingStatus

Legal ImmigrantsShare Adjusting Status

1992 1994 1996 1998 2000 2002 2004

100000

150000

200000

250000

0

.2

.4

.6

.8

(NIS), which in its pilot form includes a ran-

dom sample of 1,134 immigrants whoreceived U.S. legal permanent resident visasin 1996 (Guillermina Jasso, Massey, Mark R.Rosenzweig, and Smith 2000). Based on theNIS data, Massey and Nolan J. Malone(2002) find that 54 percent of Mexicannationals who obtained a green card in 1996reported having entered the United Statesillegally at an earlier date in time, either bycrossing the U.S. border (41 percent) oroverstaying a temporary entry visa (13 per-

cent). Overall, 21 percent of U.S. green-cardrecipients in 1996 reported having crossedthe U.S. border illegally and 11 percentreported having overstayed a temporaryentry visa.

Transitions from illegal to legal residencestatus indicate that many individuals queu-ing for U.S. green cards choose to do so asillegal immigrants, rather than waiting outthe process as residents of their home coun-

tries. Between 1992 and 2004, 90 percent

13 Though immediate relatives of U.S. citizens are notsubject to immigration quotas, to obtain a green card theystill must screened by immigration authorities, a processthat can take as long as two years. The screening process ismore protracted for individuals who meet the qualifica-tions for a green card but whose preference category issubject to immigration quotas (e.g., more distant relativesof U.S. citizens and relatives of U.S. legal residents) and

can take five years or more (David A. Martin 2005).

of Mexican immigrants who obtained

U.S. green cards qualified under family-reunification provisions of U.S. immigrationlaw. Since 1965, the United States has grant-ed unrestricted legal entry to the immediaterelatives of U.S. citizens and restricted legalentry, subject to annual immigration quotas,to more distant relatives of U.S. citizens andrelatives of U.S. permanent legal residents.Most applicants take several years or moreto clear the queue for a green card.13Whilethe United States periodically attempts to

limit the granting of green cards to thoseapplicants who either have valid temporary


10/57


entry visas or are residing abroad, the sheervolume of applications has made this provi-sion difficult to enforce (Jessica Vaughan2003). Consequently, the U.S. governmentroutinely grants green cards to individuals

who currently are residing in the countryillegally. Massey and Malone (2002) reportthat prior illegality is more common amongthose who receive green cards under family-based immigration provisions than underemployment-based immigration provisions.The latter category requires sponsorship bya U.S. employer and applies mostly to high-ly skilled individuals.

Churning in the illegal immigrant popula-tion suggests some unauthorized migrants

may view their visa status as mutable. Inmaking the decision to migrate to the UnitedStates, individuals in Mexico who have rela-tives that are U.S. legal residents may inter-nalize the prospect of obtaining a U.S. greencard in the future. They may consider beingan illegal immigrant as simply an intermedi-ate step in becoming a legal permanent resi-dent. The possibility of transitioning fromillegal to legal status may blur differences in

the expectations and behavior of legal andillegal migrants.A second means by which prospective

migrants might expect to obtain a U.S. greencard is through a future amnesty for illegalaliens. In 1986, the U.S. Immigration Reformand Control Act (IRCA) awarded permanentlegal residence visas to illegal immigrants

who could demonstrate either (1) continuousU.S. residence since 1982 or (2) sixty days ofemployment in U.S. agriculture since 1985.

Over the next eight years, 1.6 million illegalimmigrants received green cards under thefirst provision and 1.1 million illegal agricul-tural workers received green cards underthe second provision, with Mexican nationalsaccounting for 2 million of the 2.7 millionIRCA legalizations (U.S. Immigration andNaturalization Service 2001). While there ispolitical opposition in the United States toanother amnesty, there have been numerous

recent legislative proposals to legalize at least

14 These nine states are Aguascalientes, Colima,Durango, Guanajuato, Jalisco, Michoacn, Nayarit, SanLuis Potos, and Zacatecas. During the period 1944 to1964, this group of states accounted for 55 percent ofmigration from Mexico to the United States (Durand,

Massey, and Zenteno 2001).

some unauthorized migrants (Hanson 2005).The prospect of a future amnesty is anotherfactor that helps diminish distinctions betweenlegal and illegal migrants.

2.3 Composition of the Legal and Illegal

Immigrant PopulationsA longstanding conception of Mexico-to-

U.S. migration is that it is driven by theneeds and rhythms of agriculture.According to this view, most migrants fromMexico are from the countryside, come tothe United States to work as farm laborersduring peak agricultural months, and returnto their families in Mexico for the winter off-season. Migrants would tend to be male,

rural in origin, relatively uneducated, andresiding in the United States on an itinerantbasis. While there is little doubt that at onetime this view of Mexican migration wasaccurate, the Mexican immigrant populationin the United States has since become moreheterogeneous and more permanent.

Large-scale emigration from Mexicobegan in the early twentieth century.Railroad construction in the late 1800s

linked interior Mexico to the U.S. border,giving U.S. employers improved access toMexican labor (Cardoso 1980). In the early1900s, Texas farmers began to recruit labor-ers in Mexico. They followed the main railline into the country, which ran southwestthrough agricultural states in Mexicos cen-tral and western regions. Early migrantsfrom Mexico came primarily from nineMexican states in this area (Durand, Massey,and Zenteno 2001).14 Migration expanded

further in the 1940s, after the U.S. Congressenacted the Bracero Program (194264),

which allowed U.S. employers to bring inworkers from Mexico (and the Caribbean) tofulfill short-term labor contracts (of less than


11/57


15 On illegal immigration and guest-worker programs,

see Gil S. Epstein, Arye L. Hillman, and Avi Weiss (1999).

a year in length). At the end of their con-tracts, workers were required to return totheir home countries. The vast majority ofbraceros worked on U.S. farms (KittyCalavita 1992). At its peak, from 1954 to

1960, 300,000 to 450,000 migrant workersfrom Mexico entered the United Statesannually. The end of the Bracero Programmarked the beginning of large-scale illegalimmigration from Mexico.15

After working in the United States, manybraceros returned to Mexico where theyassisted later generations in migratingabroad. They helped establish informal net-

works through which earlier migrants helpnew migrants enter the United States, find

housing in U.S. cities, and obtain jobs withU.S. employers. Networks are often embed-ded in relationships involving family, kin, orcommunity of birth, which gives them aregional component. Partly as a result, thereis strong historical persistence in migrationrates to the United States across Mexicanregions. Figure 3 plots emigration rates inthe 1950s against those in the 1990s acrossMexicos thirty-two states. Data for the

1950s are from Woodruff and Zenteno(2001), who calculate the fraction of eachMexican states population that migrated tothe United States between 1955 and 1959under the U.S. Bracero Program. Data forthe 1990s, taken from the 2000 MexicoCensus of Population and Housing, reportthe fraction of households in a state having amember migrate to the United Statesbetween 1995 and 2000. The correlationbetween state migration rates in the

19952000 and the 195559 periods is 0.73.Figure 4 shows that most high-migrationstates are located in central Mexico, which isneither close to the United States nor hometo Mexicos poorest households. States onthe U.S.Mexico border tend to have lowemigration rates, as do states in low-incomesouthern Mexico.

Perhaps as a result of migration networks,current generations of Mexican immigrantsin the United States tend to live near indi-

viduals from their home regions in Mexico.For instance, Kaivan Munshi (2003) finds

that immigrants from the state of Jalisco aremuch more likely to live in Los Angeles orSan Diego than immigrants from the stateof Guanajuato, who prefer Chicago orDallas. Migrants reinforce networks by cre-ating home-town associations that helpmembers of their communities in Mexicomake the transition to living north of theborder. Of 218 home-town associationsformed by Mexican immigrants enumeratedin a 2002 survey of such organizations in

southern California, 87 percent were associ-ated with one of the nine west-central statesthat dominated migration to the UnitedStates under the Bracero Program (GustavoCano 2004).

While migration networks are a consistentfeature of cross-border labor flows, the com-position of these flows is not. Since the 1960s,Mexico has urbanized, become a more edu-cated nation, and incorporated women into

the labor force in greater numbers. Cornelius(1992) and Cornelius and Enrico A. Marcelli(2001) suggest these changes have shifted thecomposition of Mexican migrants in theUnited States from sojourners, who followthe harvest season through the rural UnitedStates and then return to Mexico at the endof the year, to settlers, who have a permanentpresence in U.S. communities. Resisting thisnotion, Durand, Massey, and Zenteno (2001)suggest instead that migration to the United

States remains dominated by men from tra-ditionally agricultural states in WesternMexico. While migrants have become bettereducated and more urban, they retain strongties to Mexico, returning often and tending toavoid permanent U.S. settlement.

Until recently, it would have been difficultto muster much more than case-study evi-dence to evaluate these claims. There arenow several data sources on migrants from

Mexico that give details on an individuals


12/57


Figure 3. Migration Rates to the U.S. by Mexican States, 1950s and 1990s

ShareofHouseholdsSending

0 .02 .04 .06

0

.05

.1

.15

Aguascal

Baja Cal

Baja CalCampeche

Coahuila

Colima

Chiapas

Chihuahu

DF

Durango

Guanajua

Guerrero

HidalgoJalisco

Mexico

Michoaca

MorelosNayarit

Nuevo Le

Oaxaca

Puebla

Queretar

Quintana

San Luis

Sinaloa

Sonora

Tabasco

Tamaulip

Tlaxcala

Veracruz

Yucatan

Zacateca

MigrantstotheU.S.,

1995

2000

Share of Residents Migrating to U.S., 19551959

Figure 4. Migration Rates to the U.S. by Mexican State and Distance to the U.S.

Kilometers to U.S. Border

0 500 1000 1500 2000 2500

0

.05

.1

.15

Aguascal

Baja Cal

Baja CalCampeche

Coahuila

Colima

Chiapas

Chihuahu

DF

Durango

Guanajua

Guerrero

HidalgoJalisco

Mexico

Michoaca

MorelosNayarit

Nuevo Le

Oaxaca

Puebla

Queretar

Quintana

San Luis

Sinaloa

Sonora

Tabasco

TamaulipTlaxcala

Veracruz

Yucatan

Zacateca

Householdsw/

MigrantinU.S.,

2000


13/57


17 Recent papers using the MMP include Massey andEspinosa (1997), Belinda I. Reyes (1997), Munshi (2003),Christina Gathmann (2004), Pia M. Orrenius andMadeline Zavodny (2005), and McKenzie and Rapoport

(2004).

16 A similar source is the National Survey of

Demographic Dynamics (ENADID), conducted by thegovernment of Mexico in 1992 and 1997. The ENADIDasks households in Mexico whether any of its membershave ever worked in or looked for work in the UnitedStates (and the year in which this occurred). In the 1997ENADID, 9 percent of individuals report having been tothe United States and 21 percent of households reporthaving a member in the United States, up from 8 percentand 17 percent in 1992. As with the MMP, the ENADIDonly includes households with at least one memberremaining in Mexico. See Durand, Massey, and Zenteno(2001) and David McKenzie and Rapoport (2004) for workusing the ENADID and Cornelius and Marcelli (2001)and Durand, Massey, and Zenteno (2001) for discussions

of other surveys.

migration status.16 Perhaps the best knownand most utilized source is the Mexican

Migration Project (MMP).17 The MMP is ahousehold survey conducted in wintermonths (when seasonal migrants tend toreturn to Mexico) in 1982 and over the

period 1987 to 1997 in several dozen ruralcommunities in western Mexico, chosen forhaving high rates of migration to the UnitedStates (Massey et al. 1994; Durand et al.1996). In each community and in each year,the MMP surveyed a random sample of

TABLE 2MIGRANT AND NONMIGRANT MEXICAN NATIONALS, MALES

Mexico Migration Project, 19891991

In Mexico at Time of Survey

1990 1990 Currently In U.S. atU.S. Mexico All on Last Time of 1989

Census Census Respondents Migration Survey LPS

Percent Male 56.2 48.0 48.7 64.7 55.9 57.2

Age 18 to 27 38.5 37.6 34.1 38.4 43.5 33.028 to 37 31.9 25.8 29.7 35.7 35.7 42.538 to 47 17.1 17.4 19.0 18.1 11.7 16.548 to 57 8.2 11.7 10.7 6.7 5.9 5.858 to 67 4.4 7.5 6.4 1.1 3.2 2.3

Males Aged 18 to 47Years of None 9.6 9.2 5.2 3.8 2.1 4.0

Schooling 1 to 4 10.3 17.3 23.4 24.3 15.3 19.75 to 8 28.0 30.1 32.5 41.4 34.3 39.1

9 to 11 17.2 24.7 18.5 19.4 23.8 18.412 to 15 31.7 11.1 12.9 9.3 19.1 18.316 plus 3.2 7.7 7.5 1.8 5.6 0.5

Live in Urban Area 91.9 74.8 80.6 74.5 89.2 --In Labor Force 91.0 85.2 95.8 98.6 94.1 95.2

Work in Agriculture 15.5 23.9 28.9 31.2 9.1 11.9Has Migrated to U.S. -- -- 50.3 100.0 100.0 --

Migrate U.S. Last Year -- -- 2.5 0.1 0.1 --

Years 0 to 5 28.8 -- 65.6 45.2 36.9 13.9

in U.S. 6 to 10 23.4 -- 14.8 17.3 20.2 19.411 to 20 35.7 -- 15.2 28.2 36.3 6.720 plus 12.2 -- 4.5 9.3 6.6 59.9

Sample size, 1867 96,487 196,729 5,370 722 375 1,670Sample size, 1847 83,703 158,917 4,448 666 341 1,535


14/57


several hundred households, collecting infor-mation on past migration behavior of eachhousehold member.18 An advantage of theMMP is that it allows one to construct retro-spective migration histories on a reasonably

large sample of individuals.

19

Among malehousehold heads, 23 percent report havingmigrated to the United States within three

years of being surveyed (during the period1984 to 1996). Of those reporting hav-ing migrated over the period 1970 to 1990, 89percent state that on their first trip tothe United States they entered withoutdocuments (McKenzie and Rapoport 2004).

The MMP is subject to several potentialproblems associated with how migrants and

communities are selected into the sample.Since communities included in the MMPare chosen on the basis of being rural andhaving residents with high migrationpropensities, they are unlikely to be repre-sentative of Mexico as a whole (McKenzieand Rapoport 2004). Within communities,the households surveyed are those with atleast one member remaining in Mexico, thusexcluding households that have migrated to

the United States in their entirety. And,within households, the migrants surveyeddirectly are those who have returned toMexico, for at least part of the year. Thereare no direct observations on individualsresiding in the United States.

One way to evaluate the issue of sampleselection in the MMP is to compare individ-uals in the survey with individuals in MexicosCensus of Population and Housing and withMexican immigrants in the U.S. Census of

Population and Housing. Tables 2 and 3 provide summary statistics for working-ageadults in the 1990 U.S. and Mexico censusesand in the 1989, 1990, and 1991 MMP sur-

veys. I consider three MMP subsamples: (a)

all respondents; (b) those who report resid-ing in the United States but who are presentin Mexico at the time of the MMP survey(seasonal migrants); and (c) those in theUnited States at the time of the MMP sur-

vey (permanent migrants), whose responsesare provided by other members of theirhousehold in Mexico.

While Mexican immigrants in the UnitedStates (census immigrants) and MMP per-manent migrants have relatively similarcharacteristics, they differ considerably fromMMP seasonal migrants. Males account for65 percent of MMP seasonal migrants, butonly 56 percent of census immigrants andMMP permanent immigrants. And, while

age profiles are similar among the threegroups, educational attainment is not. Males

with nine or more years of schooling accountfor 52 percent of census immigrants and 48percent of MMP permanent migrants, butonly 31 percent of MMP seasonal migrants.Employment patterns also differ acrossgroups. Among males, 16 percent of censusimmigrants and 9 percent of MMP perma-nent migrants work in agriculture, compared

to 31 percent of MMP seasonal migrants.Seasonal migrants also appear to be less estab-lished in the United States. For males, 55 per-cent of MMP seasonal migrants have spentmore than five years in the United States,compared to 71 percent of census immigrantsand 63 percent of MMP permanent migrants.For each of these comparisons, results aresimilar for females.

Over time, Mexican immigrants haveshifted out of agriculture as a main industry

of U.S. employment. Using data from theU.S. census, Card and Lewis (forthcoming)show that between 1990 and 2000, amongrecent Mexican immigrants (05 years in theUnited States), the share working in agricul-ture fell from 23 percent to 15 percent formen and from 13 percent to 7 percent for

women. Among men, construction account-ed for the largest growth in employmentshares, while among women retail trade

showed the largest increase.

18 Different households are surveyed in differentyears, such that the MMP is a repeated cross-sections ofhouseholds and not a true panel.

19 On measurement error in retrospective data, seeMegan Beckett et al. (2001) and Smith and Duncan

Thomas (2003).


15/57


Other surveys of illegal immigrants fromMexico also suggest their characteristics aremore similar to permanent migrants (whetherin the U.S. census or the MMP) than season-al migrants. The Legalized Persons Survey

(LPS) covered illegal immigrants who were

granted permanent legal residence in theUnited States under the amnesty provisionof IRCA (eligibility for which required proofof U.S. residence from 1982 forward). TheLPS consisted of an initial survey in 1989 and

a follow-up survey in 1992 of immigrants

TABLE 3MIGRANT AND NONMIGRANT MEXICAN NATIONALS, FEMALES

Mexico Migration Project, 19891991

In Mexico at Time of Survey

1990 1990 Currently In U.S. atU.S. Mexico All on Last Time of 1989Census Census Respondents Migration Survey LPS

Percent Female 43.8 52.0 51.3 35.3 44.1 42.8

Age 18 to 27 33.0 38.0 35.0 35.5 43.2 31.928 to 37 31.4 26.1 29.3 37.1 33.5 42.038 to 47 18.9 17.0 18.6 21.3 16.6 16.848 to 57 10.4 11.5 10.7 5.6 4.4 6.658 to 67 6.3 7.6 6.5 0.5 2.4 2.7

Females Aged 18 to 47Years of None 8.5 12.7 6.6 3.0 5.1 4.6

Schooling 1 to 4 9.8 18.8 25.9 18.7 13.8 18.75 to 8 28.3 31.4 35.4 56.2 35.5 43.09 to 11 16.2 24.8 16.8 12.4 23.8 18.4

12 to 15 34.0 8.0 11.9 7.6 28.3 15.916 plus 3.3 4.4 3.5 2.2 2.2 0.6

Live in Urban Area 92.7 75.8 80.6 70.8 88.8 --In Labor Force 56.7 26.6 28.8 35.7 47.5 61.3

Work in Agriculture 7.9 2.3 2.4 2.2 6.6 4.1Children Ever Born 2.6 3.3 -- -- -- --

Own Children in HH 1.7 1.6 1.8 1.0 2.2 2.1Has Migrated to US -- -- 18.6 100.0 100.0 --

Migrated U.S. Last Year -- -- 0.8 0.0 0.0 --

Years in U.S. 0 to 5 26.2 -- 58.1 44.3 41.5 10.76 to 10 20.9 -- 17.5 17.6 16.5 11.1

11 to 20 37.3 -- 18.7 28.9 29.2 4.720 plus 15.6 -- 5.7 9.2 12.7 73.5

Sample size, 1867 76,518 212,912 5,658 394 296 1,248Sample size, 1847 63,278 172,458 4,688 370 276 1,132

Note: Table 2 gives summary statistics on working-age adults, either 1847 or 1867 years old. The samples areresidents of Mexico (1990 Mexico Census of Population and Housing); Mexico-born residents of the UnitedStates (1990 U.S. Census of Population and Housing); respondents of the 19891991 Mexican Migration ProjectSurvey (all respondents, those residing in United States but in Mexico at the time of the survey, or those residingin United States at the time of the survey whose responses were given by household members in Mexico); and

illegal immigrants from Mexico in the United States who qualified for amnesty under the U.S. ImmigrationReform and Control Act (1989 Legalized Persons Survey). The definition of an urban area is a locality with morethan 2,500 inhabitants.


16/57


17/57


22 Individuals apprehended in the U.S. interior, in con-trast, could have crossed the border at a much earlier date,making interior apprehensions less strongly correlatedwith current attempts at illegal entry.

23 Between 1990 and 2003, 95 percent of those theBorder Patrol apprehended agreed to depart voluntarily.

24 A further issue is that the majority of those attempt-ing illegal entry do not appear to be apprehended on anygiven attempt. Using MMP data, Massey and AudreySinger (1995) find that for trips to the United States in the1970s and 1980s the average probability of apprehension

was 35 percent.

correlated with the contemporaneous levelof attempted illegal immigration.22 However,

apprehensions are likely to be a poor indica-tor of the actual level of illegal inflows(Espenshade 1995). Within a single month,one individual may be apprehended multipletimes. Those apprehended who agree to bedeported voluntarily are not processed bythe U.S. justice system. For Mexican nation-als, voluntary deportation often involves littlemore than a bus ride across the border, leav-ing them in position to attempt illegal entry

again in the near future.

23, 24

To gauge how apprehensions might berelated to illegal immigration, consider the

level of apprehensions as a function ofthe average probability of apprehensionand the number of attempts at illegal en-try. Extending Wilfred J. Ethiers (1986)model, let

(5) At = P(Ht,Mt)Mt,

whereAt is the level of apprehensions, Mt isthe number of attempts at illegal entry, andP(Ht, Mt) is the average probability an indi-

vidual is apprehended on any given attemptto cross the border. The apprehensionsprobability is a function ofHt, the intensity

with which authorities police the border,and the number of entry attempts. Greaterenforcement is likely to raise the apprehen-sion probability, making P(.) increasing inHt. For a given level of enforcement, moretotal attempts are likely to reduce the prob-ability any single attempt results in capture

(since enforcement resources are spread

Apprehensions(000s)

Year

OfficerHours(000s)

Apprehensions (000s)Officer Hours (000s)

1965 1970 1975 1980 1985 1990 1995 2000 2005

0

500

1000

1500

0

2500

5000

7500

10000

Figure 5. Linewatch Apprehensions and Enforcement by the U.S. Border Patrol


18/57


27 See Mary Jordan, Mexicans Caught at Border inFalling Numbers, Washington Post, May 24, 2002.

28 One limitation of this exercise is that I assumeborder-crossing technology and border-apprehensionstechnology have been stable over time. There is anec-dotal evidence that both may have changed considerably,especially since September, 2001. However, Hanson andSpilimbergo (1999) find no evidence of a structuralbreak in the apprehensions function for the 1968 to 1996

period.

more thinly across those attempting entry),making P(.) decreasing in Mt. 25

Suppose P(Ht,Mt) = cH1t M

2t , where c, 1,

and 2 are positive constants, in which caselog apprehensions can be expressed as

(6) lnAt = 0 + 1lnHt + 1 2)lnMt

Hanson and Antonio Spilimbergo (1999) usemonthly data over the period 196896 to esti-mate equation (6), modeling attempted ille-gal entry as a reduced-form function of real

wages in Mexico, real wages in the UnitedStates, other indicators of economic condi-tions in the two countries, a time trend, andmonthly dummy variables.26 With an esti-

mate of

1, we can solve for a function that isan affine transformation of lnMt, given by

(7) 0 + (1 2)lnMt = lnAt 1lnHt

The expression on the right of (7) will posi-tively covary with lnMt as long as 2 < 1.Approximated attempts at illegal entry in (7)do not give an estimate of the level of illegalimmigration. However, they may indicatethe variation across time and the magnitude

of log changes in attempted illegal entry.Figure 6 shows estimates of (7), based oninstrumental-variables estimates of equation(6) (see note 26). Approximated attempts atillegal entry rise from the 1960s to the mid-1980s, are stable from the mid-1980s to themid-1990s, and then decline somewhat in2000 and 2001. Part of the 2001 decline mayreflect a change in border-crossing activityafter the events of September 11th in the

United States. In late 2001, the U.S. BorderPatrol increased its vigilance at border cross-ings, which may have dissuaded somemigrants from crossing as frequently as theyhad in the past.27

The trend in figure 6 is roughly consistentwith table 1 and results from previous esti-mates of the U.S. unauthorized population.Illegal immigration appears to have risensteadily after the end of the BraceroProgram in 1965 and has been relatively sta-ble at high levels for the past two decades.28

I will return to data on border apprehen-sions and enforcement when evaluating fac-tors that affect the level of illegalimmigration and the political economy of

U.S. policy on illegal immigration.

2.5 Summary

Currently, no data source gives preciseestimates of the size of the U.S. illegal-immigrant population over medium or longtime spans. Government data-gatheringagencies have been wary of asking questionsabout an individuals immigration status,

perhaps out of fear of dissuading illegalmigrants from participating in surveys. Theresult is gaps in our knowledge about unau-thorized migrants, which the literature hasbeen able to partially fill in through otherdata sources.

The perspective that emerges from thedata that are available is that Mexico-to-U.S.illegal migration increased in the 1970s and1980s and averaged around 200,000 to300,000 net unauthorized entries per year in

the 1990s and early 2000s. The population ofillegal immigrants from Mexico in the

25 For earlier work using apprehensions data, see Beanet al. (1990) and Borjas, Richard B. Freeman, and KevinLang (1991).

26 To deal with the possible correlation betweenenforcement and unobserved shocks to apprehensions,Hanson and Spilimbergo (1999) instrument for enforce-ment using U.S. government spending on nationaldefense and indicators for whether there is an upcomingU.S. presidential, congressional, or border-state guberna-torial election. Border enforcement tends to follow a polit-ical cycle, dropping during election years (Hanson,Raymond Robertson, and Spilimbergo 2002). The report-ed coefficient estimate for 1 is the long-run elasticity of

apprehensions with respect to enforcement.


19/57


Year1970 1975 1980 1985 1990 1995 2000 2005

0

1

ln(Apprehensions)0.8ln(Officer Hours)

1

2

3

Figure 6. Estimated Attempts at Illegal Entry along the U.S.Mexico Border

United States includes a substantial fractionof women, is predominantly employed innonagricultural jobs, and has schooling lev-els that are comparable to or higher thannonmigrating individuals in Mexico.Though many migrants maintain ties withfamily members in their origin communi-ties, a majority appear to have settled in theUnited States on a medium or long-termbasis.

3. The Supply of and Demand for MexicanImmigrants

Beginning with Larry A. Sjaastad (1962),economists have viewed migration as aninvestment decision. An individual migratesif the expected discounted difference in thestream of income between the new and oldlocation exceeds moving costs. The incentiveto migrate will vary across individuals

according to differences in their expectations

of future earnings, discount rates, and per-ceived cost of migrating. The cost of unau-thorized migration includes transport tothe border, the physical risks and monetarycharges incurred in crossing the borderillegally, the psychic penalty from leavingones friends and family behind, and thetime and monetary expense of settling inanother country. To uncover sources of

variation in the demand for and the supplyof illegal migrants, recent work estimates

the sensitivity of migrant outflows fromMexico to variation in U.S. and Mexican

wages, border-crossing costs, and access tomigration networks.

Also beginning with Sjaastad, economiststend to model the migration decision asirreversible (Greenwood 1997; Lucas 1997).In many contexts, this assumption may bereasonable. U.S. legal immigrants, if they

wish to keep their green cards valid, must

make the United States their permanent


20/57


residence.29 Those migrating to the UnitedStates legally would thus tend to be individ-uals expecting to stay in the country for anextended period of time. In the context ofillegal migration from Mexico, individuals

may operate on shorter time horizons. Thesubstantial round-trip migration document-ed by the MMP indicates that for at leastsome Mexican nationals the cost of movingback and forth across the border is suffi-ciently low to warrant making the trip annu-ally.

Of interest to both Mexico and the UnitedStates is not just the volume of migrant flowsbut their composition. Widening differencesin earnings between immigrants and natives

in the United States is cited as evidence thatrecent U.S. immigrants are negatively select-ed in terms of skill (Borjas 1999a). Followingthis line of thought, one might expect nega-tive selection to be especially strong amongillegal immigrants. Recent work examinesmigrant selection by comparing the charac-teristics of Mexican immigrants in theUnited States with those of nonmigratingindividuals in Mexico.

3.1 The Incentive to Migrate from Mexicoto the United States

A long line of research applies the generalframework in Sjaastad (1962) to examine thesensitivity of migration flows to the observedcosts and benefits of migrating. For prospec-tive illegal migrants in Mexico, the costsinclude the four components identifiedabove and the benefits include gains in realincome associated with moving from low-

wage Mexico to the high-wage UnitedStates.

30 Coyotes also help migrants navigate unfamiliar ter-rain. Recent changes in U.S. enforcement strategy havemade it more difficult to cross the border in urban areas,encouraging migrants to enter the United States throughthe mountainous desert regions of Arizona and EasternCalifornia, where temperature extremes expose migrantsto physical risks. Annual deaths of border crossers have

increased from an average of 100 during the mid 1990s to410 during the period 20002004 (Cornelius 2005). Withnet annual illegal immigration from Mexico since 2000 atapproximately 300,000 individuals (see table 1), thereappear to be around 1.4 deaths per 1,000 successful netunauthorized migrants.

31 For the more expensive complete package of smug-gling services, the coyote typically receives a portion of thefee up front and the remaining portion once the migrant issafely delivered to friends or family members in theUnited States. See Charlie LeDuff, The Crossing: ASpecial Report; A Perilous 4,000 Mile Passage to Work,New York Times, May 29, 2001, p. 1.

32 To reduce recall bias, I only show data for years

within four of years of an MMP survey.

First, consider the components of illegalmigration costs. Of these, transport costs tothe border are likely to be small and psychiccosts difficult to evaluate. Border-crossingcosts and settlement costs are in principle

measurable, though it is only the formerthat has been subject to much in the way offormal research. To avoid capture by theU.S. Border Patrol, migrants often pur-chase the services of a smuggler, known as acoyote.30 Coyotes offer a range of services,from simply guiding migrants across theborder to more complete packages thatinclude transport to an interior U.S. city,such as Houston, Los Angeles, orPhoenix.31 The MMP is one of the few data

sources that asks migrants from Mexicoabout their border-crossing behavior. UsingMMP data, Orrenius (2001) documentsthat, during the period 1978 to 1996, 69percent of migrants reported hiring a coy-ote, as shown in figure 7.32 During this peri-od, the average price paid for coyoteservices varied between $385 and $715 ( in2000 dollars). Since 2001, when the BorderPatrol became more vigilant in monitoring

U.S. border crossing points, coyote priceshave risen. Based on a 2005 survey of returnmigrants in rural areas of two Mexican

29 A green-card holder may lose U.S. permanent resi-dent status by taking permanent residence abroad, remain-ing abroad without obtaining a reentry visa, or by filing aforeign tax return as a nonimmigrant. Once a legal immi-grant becomes U.S. citizens, he or she is free to enter and

leave the country at will.


21/57


33 The increase in prices is based on respondents recol-lections in 2005 of prices they paid in previous years andthus may be subject to recall bias. In 2005, 90 percent ofrespondents in Corneliuss data (all of whom are from thestates of Jalisco or Zacatecas) report using a coyote on their

previous trip to the United States.

Figure 7. Use of Smugglers by Migrants in the Mexican Migration Project

MigrantsUsingCoyotes

Year

CoyotePrices(2000USD)

Migrants Using CoyotesCoyote Prices (2000 USD)

1980 1984 1988 1992 1996

.6

.7

.8

.9

1

400

500

600

700

states (that are also in the MMP), Cornelius(2005) finds that between the 199698 and200204 periods average coyote prices roseby 37 percent from $1180 to $1680.33

One might be concerned that coyoteprices based on MMP data are biased down-

wards. As mentioned in the last section, theMMP surveys households in communities

with high rates of migration to the UnitedStates. Individuals in these communities

may have relatively good access to migrationnetworks, making them less dependent oncoyote services or better able to negotiatelower prices from smugglers. Cornelius(2005) reports that 65 percent of migrants

34 Recent articles in the popular press give a pricerange for coyote services of $1,500$2,000 (Man Accusedof Smuggling Immigrants, The Oregonian, September 17,2005, p. B2; US Immigration, The Economist, May 19,2005; Between Here and There, The Economist, July 5,2001; Illegal Immigration; Border Agents Understand aComplex Issue, The San Diego Union-Tribune, May 26,

2005, p. B12).

hiring a smuggler in 2004 report usingfriends or family to find a coyote. Resultsusing the MMP itself are consistent withconcerns about bias in observed coyoteprices. Gathmann (2004) finds that migrants

with family members in the United Statesare less likely to use coyote services and,among migrants who do hire a smuggler,those with family members abroad pay lowerprices. Yet, even if one accepts the high coy-

ote prices quoted in the press,34

border-crossing costs since 2001 appear to be nomore than $2,000, which is 35 percent ofMexicos 2003 per capita GDP.


22/57


35 In earlier work, J. Edward Taylor (1987) examinesmigration behavior in a single rural community in Mexico.

36 See also Oded Stark and Taylor (1989, 1991) andMassey and Espinosa (1997).

37 McKenzie and Rapoport (2004), who also use MMP

data, obtain similar results.

Next, consider the benefits to migration. Tocalculate the gross return to migration, one

would need to account for the expectedlength of stay in the United States, the expect-ed path of future earnings, and the discount

rate applied to these earnings. In the absenceof data on these components, much of theresearch on the decision to migrate takes areduced-form approach to modeling how thereturns to migration affect behavior.35 Usingretrospective data from the MMP, Orreniusand Zavodny (2005) estimate the hazard that

young males migrate to the United States as afunction of individual and household charac-teristics and economic conditions in the twocountries.36 In the MMP, most migrants

appear to be entering the United States ille-gally. The migration hazard is nonmonotonicin schooling, increasing at low and mediumschooling levels (primary and secondaryschool) and decreasing at higher schoolinglevels (preparatory school and beyond).37 Thehazard is decreasing in per capita GDP inMexico and increasing in U.S. average wages.These results suggest that migrants tend to bedrawn from the middle of the schooling dis-

tribution and that migration is more likelyduring periods when U.S. income is expand-ing relative to income in Mexico. The migra-tion hazard is also higher for individuals

whose fathers have migrated or whose sib-lings have migrated. This result could indicatethe presence of family migration networks, orit could indicate the presence of unobservedhousehold characteristics (e.g., unmeasured

wealth) that influence migration behavior.By treating migration as a function of age,

education, and other individual characteris-tics, Orrenius and Zavodny pick up variationacross individuals in the incentive tomigrate. By also including macroeconomicconditions, they pick up time-series variation

in the migration incentive. Yet, since theirspecification is a reduced form, the coeffi-cient estimates do not allow one to recoverthe elasticity of migration with respect to

wage differentials between the United

States and Mexico. Further, since the MMPis restricted to communities with historicallyhigh propensities to migrate to the UnitedStates, the results may not be informativeabout how prospective migrants in otherregions of Mexico respond to changes inbinational economic conditions.

To examine how the gross level of attempted illegal migration responds tochanges in U.S.Mexican wages, Hansonand Spilimbergo (1999) estimate an appre-

hensions function, similar to that in equation(6). Using monthly data, they regress appre-hensions at the U.S.Mexico border onlagged apprehensions, current and laggedlinewatch enforcement hours, the real peso

wage for production workers in Mexicanindustry, real peso and real dollar U.S. wages(measured as the weighted average of wagesin U.S. industries that employ recent Mexicanimmigrants in large numbers), and other con-

trols. They instrument for enforcement usingU.S. government spending on nationaldefense and indicators for whether there is anupcoming U.S. presidential, congressional, orborder-state gubernatorial election (see note26). Figure 8 shows a partial regression plotof log apprehensions on log Mexican realaverage hourly earnings, based on Hansonand Spilimbergos estimates. Border appre-hensions appear to be very responsive tochanges in Mexican wages. A 10 percent

decline in Mexican real wages is associatedwith a 68 percent percent increase in borderapprehensions. Moreover, this effect is almostfully realized within three months following a

wage change, suggesting that shocks to theMexican economy are rapidly transmitted tochanges in attempted illegal migration.

Over the past three decades, Mexico hasexperienced wide variation in real income, asperiodic devaluations of the peso have lead

to bursts of inflation, which have caused


23/57


Figure 8. Border Apprehensions and Average Wages in Mexico

LogBorderApprehensions

Log Mexico Hourly Wage

0 .25

0

.6

Coeff.= 0.60, Std. Error = 0.10

.6

.25

Figure 9. Border Apprehensions and U.S.Mexico Relative Per Capita Income

Ln

Apprehensions

Year

Ln ApprehensionsP

1975 1980 1985 1990 1995 2000 2005

.25

.75

1.25

0

.1

.2

Ln U.S.Mexico Per Capita GD

LnU.

.

S

MexicoPerCapitaGDP

.25

.1

.2


24/57


38 For Mexico, average hourly wages are calculated asmonthly labor income/(4.5 hours worked last week); forthe United States, average hourly wages are calculated asannual labor income/(weeks worked last year usual hoursworked per week). For Mexico, I need to assume individ-uals work all weeks of a month, which could bias wage esti-mates downwards. To avoid measurement error associated with implausibly low wage values or with top coding ofearnings, I drop the largest and smallest 0.5 percent ofwage values.

39 In 2000, Mexicos PPP-adjusted price level was 61

percent of the U.S. price level.

incomes to fall sharply. Figure 9, which plotsrelative per capita GDP in the United Statesand Mexico, gives evidence of this volatility.During three separate episodes in the lastthirty years, Mexicos per capita GDP

declined by five log points or more (relativeto the United States) within the space ofthree years. Each of these episodes was fol-lowed by an increase in border apprehen-sions. Much of the research on interregionalmigration in the United States and othercountries finds that it is labor earnings in thereceiving region, and not the sendingregion, that appear to drive migration flows(Greenwood 1997). However, this does notappear to be the case in Mexico, where

income volatility appears to be a strong pushfactor for illegal migration.

While apprehensions are the only avail-able high-frequency measure of grossattempts at illegal migration, these datahave important limitations. Since individualmigrants may be apprehended multipletimes in a given time period, the number ofapprehensions may far exceed the grossnumber of migrants (Espenshade 1995).

Controlling for border enforcementaddresses this problem, but only if one has valid instruments for enforcement (sinceenforcement is likely to be endogenous toshocks to attempted illegal migration) and ifthe impact of enforcement on apprehen-sions (controlling for the incentive tomigrate) is stable over time (see notes 26and 28).

3.2 U.S.Mexico Wage Differences

The reduced-form results of Orrenius andZavodny (2005) and Hanson andSpilimbergo (1999) suggest that illegalmigration flows are highly responsive tochanges in the return to migration. However,these results give no indication of the magni-tude of the returns themselves. As a crudeapproximation of the short-run gross returnto migration, I examine differences in hourly

wages for men in Mexico and Mexican immi-

grant men in the United States in 2000. I ask

how long a migrant from Mexico would haveto work in the United States in order torecoup border-crossing costs, as approximat-ed by the price of coyote services. I focus onmales, since, as tables 2 and 3 show, there are

large differences in labor-force participationrates between women in Mexico andMexican immigrant women in the UnitedStates, which complicates comparing female

wage outcomes across national borders. Bylimiting the analysis to current wage differ-ences and a single component of migrationcosts, this exercise falls well short of a com-plete costbenefit accounting of the migra-tion decision. Still, given large back and forthflows of labor across the U.S.Mexico bor-

der, the current U.S.Mexico wage differen-tial is likely to be the relevant gross return tomigration for at least some prospectivemigrants.

Table 4 reports average hourly earningsby age and schooling categories for males inMexico (based on the 2000 Mexico Censusof Population and Housing) and for immi-grant males from Mexico in the UnitedStates (based on the 2000 U.S. Census of

Population and Housing).38

To increase theshare of illegal immigrants among Mexicanimmigrant men, I limit the sample to veryrecent immigrants (individuals residing inthe United States for 03 years). To adjustfor cost of living differences between thecountries, I scale up Mexican hourly wagesto achieve purchasing power parity with theUnited States, using the 2000 PPP adjust-ment factor for Mexico in the Penn WorldTables.39


25/57


TABLE 4AVERAGE HOURLYWAGES FOR MEXICAN MALES, 2000

Years of Schooling Completed

Age 4 5 to 8 9 to 11 12 13 to 15 16+

Mexican 18 to 22 7.83 7.60 7.45 8.07 8.76 8.44Immigrants 23 to 27 8.44 8.19 8.21 9.06 9.53 13.02

in U.S. 28 to 32 8.27 8.56 8.70 9.66 9.56 15.6933 to 37 9.46 9.25 9.34 10.07 11.36 16.8438 to 42 9.19 9.39 9.33 11.01 12.11 16.2643 to 47 9.75 8.90 9.35 10.68 12.80 15.8848 to 52 9.57 9.37 9.42 9.31 11.65 17.78

Residents 18 to 22 1.36 1.56 1.76 2.06 2.61 3.91of Mexico 23 to 27 1.43 1.80 2.10 2.79 3.77 5.20

28 to 32 1.56 1.93 2.42 3.22 4.80 6.6333 to 37 1.65 2.08 2.56 3.45 5.25 7.07

38 to 42 1.64 2.14 2.88 3.74 5.62 7.4243 to 47 1.69 2.30 3.00 4.40 5.86 8.0548 to 52 1.66 2.30 3.15 4.21 6.11 8.71

Note: Table shows average hourly wages in 2000 U.S. dollars for Mexican immigrant males in the United Statesor male residents of Mexico who report working 20 to 84 hours a week (in either sample, the highest and lowest0.5 percent of wage values are excluded). Data for Mexico are a 10 percent random sample of the 10 percentmicrosample of theXIII Censo General de Poblacion y Vivienda, 2000; data for the United States are from the 5percent U.S. PUMS in 2000. Mexican immigrants are restricted to individuals who have resided in the UnitedStates for three years or less. See the text on the calculation of wage values.

40 These states are Aguascalientes, Colima, Durango,Guanajuato, Guerrero, Hidalgo, Jalisco, Michoacan,

Morelos, Nayarit, Oaxaca, San Luis Potosi, and Zacatecas.

Not surprisingly, wages are substantiallyhigher among Mexican immigrants in theUnited States than among residents ofMexico. For 2327 year-old males, the PPP-adjusted hourly wage differential varies from$7.01 for those with 04 years of schooling to$5.76 for those with 1315 years of schoolingand to $7.82 for those with 16 or more yearsof schooling. Given that migration propensi-ties vary widely across regions of Mexico, onemight think that the average hourly wage forthe country as a whole may not be the rele-

vant alternative wage for most prospectivemigrants. Table 5 reports wage differentialsbetween the United States and high-migra-tion states in Mexico, defined to be states

with above-average emigration rates in2000.40 U.S.Mexico wage differentials for 41 On the returns to education in Mexico, see also

Michael Ian Cragg and Mario Epelbaum (1996) and JamesAirola and Chinhui Juhn (2003). If unobserved ability andschooling are correlated, estimates of returns to schoolingmay be biased. Also, self-selection into the labor force or

into migration may introduce further biases.

high-migration states are similar to those forMexico as a whole.

In table 5, the log binational wage differ-ence declines with schooling, which is in line

with empirical research that has suggestedestimated returns to schooling are higher inMexico. In the 1990s in the United States,an additional year of schooling is associated

with an increase in wages for Mexican immi-grant men of 2.5 to 3.2 log points (Borjas1996; Trejo 1997; Jeffrey Grogger and Trejo2002). In the 1990s in Mexico, in contrast,an additional year of schooling is associated

with an increase in wages for men of 7.6 to9.7 log points (Daniel Chiquiar 2003).41

Returns to experience also appear to be


26/57


higher for Mexican residents than forMexican immigrants. Higher returns to edu-cation in Mexico are consistent with thecountry having a low relative supply ofskilled labor compared to the United States.Higher average wages in the United States(within schooling groups) are consistent withthe country having a higher relative level ofTFP and higher relative supplies of physical

capital.

In 2000, a 2327 year-old recent Mexicanmigrant with 5 to 8 years of schooling (thecategory just below the national mean level ofschooling for Mexico) would recoup border-crossing costs of $2,000 in 313 hours, or 7.8

weeks based on a forty-hour work week. Thespeed with which migrants would seem torecover border-crossing costs suggests thatother costs to migration (including psychic

costs and financing costs associated with

TABLE 5LOG U.S.MEXICO HOURLYWAGE DIFFERENTIAL, MALES IN 2000

Years of Schooling Completed

Age 4 5 to 8 9 to 11 12 13 to 15 16+

Log of U.S.Mexico Wage Differential, All Males in Mexico

18 to 22 1.263 1.098 0.985 0.895 0.750 0.361(0.022) (0.013) (0.015) (0.018) (0.042) (0.099)

23 to 27 1.281 1.043 0.904 0.746 0.507 0.387(0.024) (0.015) (0.018) (0.018) (0.038) (0.048)

28 to 32 1.225 1.036 0.821 0.652 0.306 0.357(0.029) (0.018) (0.024) (0.023) (0.047) (0.052)

33 to 37 1.258 1.025 0.823 0.613 0.339 0.294(0.034) (0.023) (0.034) (0.031) (0.057) (0.059)

38 to 42 1.287 1.019 0.775 0.575 0.288 0.216(0.035) (0.029) (0.045) (0.043) (0.068) (0.075)

43 to 47 1.265 0.905 0.708 0.498 0.275 0.118(0.040) (0.037) (0.063) (0.063) (0.103) (0.116)

48 to 52 1.264 0.987 0.715 0.418 0.164 0.122(0.050) (0.046) (0.096) (0.076) (0.122) (0.142)

Log of U.S.Mexico Wage Differential, Males in High-Migration States

18 to 22 1.188 1.061 0.989 0.912 0.676 0.381(0.026) (0.016) (0.017) (0.025) (0.052) (0.128)

23 to 27 1.211 1.013 0.895 0.722 0.481 0.425(0.027) (0.017) (0.019) (0.024) (0.050) (0.052)

28 to 32 1.144 1.011 0.824 0.657 0.361 0.415(0.031) (0.020) (0.025) (0.028) (0.057) (0.053)

33 to 37 1.151 0.988 0.821 0.652 0.317 0.411

(0.035) (0.025) (0.034) (0.034) (0.061) (0.060)38 to 42 1.206 0.966 0.792 0.601 0.185 0.291

(0.036) (0.030) (0.044) (0.046) (0.070) (0.073)43 to 47 1.194 0.867 0.671 0.528 0.237 0.181

(0.041) (0.037) (0.061) (0.067) (0.101) (0.112)48 to 52 1.193 0.961 0.727 0.292 0.284 0.227

(0.049) (0.045) (0.092) (0.080) (0.125) (0.137)

Notes: Table shows the log difference in average hourly wages (and standard errors) forMexican immigrant males and male residents of Mexico. The top half of the tableincludes all regions of Mexico; the bottom half includes only high-migration states. Seetable 4 and the text for details.


27/57


42 Francisco L. Rivera-Batiz (1999), who also uses theLPS, estimates larger wage differences between legal andillegal immigrants. For other work on the wage consequences

of IRCA, see Cynthia Bansak and Steven Raphael (2001).

credit constraints) are large. In a simple stat-ic model, the equilibrium wage differentialbetween two countries is fixed by migrationcosts. To reconcile persistent U.S.Mexico

wage differences with small border-crossing

costs, there would need to be positive unob-served migration costs (otherwise migrationflows would be larger) and heterogeneityacross individuals in these costs (otherwisemigration flows would be lumpier, witheither everyone or no one in a skill group

willing to migrate).There are reasons to be skeptical about how

well the observed U.S.Mexico wage differ-ential captures the incentive to migrate. Thisdifferential reflects the binational difference

in returns to both observed and unobservedcharacteristics. If migrants are positivelyselected in terms of unobserved skill,observed wage differences will tend to over-state true wage differences. In the next sec-tion, we address the issue of which types ofindividuals in Mexico select into migration.

Another issue is that, since average wagesfor Mexican immigrants in the United Statesare a composite of wages for legal and illegal

immigrants, they may overstate wages anillegal immigrant would expect to earn.Sherrie A. Kossoudji and Deborah A. Cobb-Clark (2002) use the LPS to compare wagesfor illegal immigrants before and after theyobtained green cards under the amnesty pro-

vision of the 1986 IRCA. Between 1989 and1992, average hourly earnings for newlylegalized immigrant men (64 percent of

whom were Mexican nationals) rose by 6 logpoints relative to earnings for Latino men in

the NLSY, controlling for observable charac-teristics. Also, prior to legalization, illegal-immigrant men had relatively slow wagegrowth. Not surprisingly, illegal status is asso-ciated with lower wages and less opportunityfor occupational advancement (Kossoudjiand Cobb-Clark 1996, 2000).42

What is surprising, perhaps, is that theestimated wage premium for legalization isso small. One might expect that being con-fined to the underground economy wouldsubstantially limit workers employment

prospects. Once legalized, they would enjoystrong wage growth. By this reasoning, a 6-log-point gain over three years is unimpres-sive. One possibility is that the three yearscovered by the LPS is not long enough forindividuals to realize the wage benefits asso-ciated with gaining legal status. Legalizationmay open up new opportunities to movebetween occupations or between regions,

which migrants need more time to exploit.Another possibility is that Kossoudji and

Cobb-Clarks estimate is biased downwards.Their control group includes legal and illegalimmigrants from Latin America, as well assecond and later generation Latin Americanimmigrants. Suppose illegal immigrants aremore negatively selected in terms of unob-served skill than legal immigrants (whereoverall legal immigrants may be positively ornegatively selected in terms of skill). In aneconomy where the return to skill is rising

(as in the United States during the 1990s),stronger negative selection of illegal immi-grants in terms of unobserved skill wouldtend to make their unobserved wage growthrelatively small and lead one to underesti-mate the wage premium due to legalization.

The differential in U.S. and Mexican wages in table 5 may also miss importantsources of variation across individuals in theincentive to migrate. For instance, average

wages for Mexican immigrants in the United

States may hide heterogeneity in expecta-tions about U.S. wages among prospectivemigrants. Individuals with previous U.S.labor-market experience or who speakEnglish well might have relatively strong

wage expectations due to the ease withwhich they expect to find a U.S. job or thehigh productivity they anticipate having onthe job. Though not specific to Mexico, HoytBleakley and Aimee Chin (2004) find that,

for immigrants from non-English-speaking


28/57


43 Bleakley and Chin (2004) instrument for English-lan-guage ability using immigrant age at arrival, exploiting thefact that younger children appear to learn new languagesmore easily than older children.

44 For other work on migration networks in Mexico, seePaul Winters, Alain de Janvry, and Elisabeth Sadoulet

(2001).

countries that arrived in the United States aschildren, wages are higher for those withstronger English-language skills.43

Another source of heterogeneity in wageexpectations among prospective migrants is

variation in access to migration networks.Individuals with friends or family that havemigrated abroad may have better informa-tion about how to find a job in the UnitedStates. Munshi (2003), using data from theMMP, finds that Mexican migrants in theUnited States are more likely to beemployed and more likely to be employed ina (higher-paying) nonagricultural job thelarger is the U.S. population of residentsfrom their origin community in Mexico.44

He instruments for the time-varying size ofthe U.S. population from a migrants origincommunity in Mexico using lagged rainfallin the migrants origin commun

Documents

Illegal Migration From Mexico to the US (Dec 2006)