SEARCH

SEARCH BY CITATION

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES

A common perception is that immigrants, including illegal immigrants, use disproportionate public aid and select locations based on characteristics of services offered. This paper asks to what extent geographic clustering of undocumented immigrant agricultural laborers in the U.S. is correlated with take-up of public aid broadly defined. Evidence from a nationally representative farmworker survey does not support welfare migration for undocumented immigrants, who have been previously unidentifiable in the literature. The paper, therefore, challenges existing notions of welfare migration by illegal immigrants that have inspired state-level public policy initiatives.

1. INTRODUCTION

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES

A common perception is that immigrants are costly in terms of public expenditures, especially at the state level. When states face budget trade-offs, immigrants (and particularly illegal immigrants) are often among the proposed scapegoats. As public aid and education participation imposes direct costs on state governments, states are concerned about attracting a disproportionate number of program participants. Budget shortfalls in recent years and tense public policy discussions concerning racial and ethnic profiling, such as the Arizona immigration bill that received media attention in 2010, make these issues even more salient.

Migrants are said to engage in welfare migration if they choose residence in response to public aid differentials across locations. Analogously, a state is a “welfare magnet” if it attracts disproportionate numbers of these migrants. Studies have demonstrated the presence of welfare migration and Tiebout-style “voting with one's feet” within low-income native and legal immigrant populations.1 Despite common public perceptions, little statistical evidence exists concerning how these mechanisms compare across illegal and legal immigrant categories.

The hypothesis of welfare migration within the undocumented population may seem irrelevant given that extensive eligibility requirements for many U.S. welfare programs exclude those without documents. Empirical evidence, however, shows that public aid participation rates among undocumented immigrants are significant. Undocumented immigrants may collect benefits on behalf of legal children or by using false documents. In addition, public medical services and education programs often are not subject to legal status verification.

Evidence of geographic clustering of immigrants in general is well-documented. Official statistics on residential choices of legal immigrants—naturalized citizens and legal permanent residents—reveal concentrations of individuals in California, New York, Florida, and Texas. Estimates of the undocumented immigrant population also show patterns consistent with purposeful locational clustering, particularly in places with established ethnic, racial, and linguistic networks. The question here, therefore, is to what extent this clustering is related to public aid program participation.

California, known for welfare generosity, is usually suspected to be a welfare magnet for immigrants. California has traditionally offered more generous benefits than other U.S. states and therefore is a common candidate for a welfare magnet in the literature. In 2009, the maximum monthly combined Temporary Assistance for Needy Families (TANF) and Food Stamp Program (FSP) benefit for a family of four in California was $1,055 compared with $870 in the median state.2 These patterns have been evident over time, and persist after cost of living adjustments are made. Furthermore, in addition to its higher absolute benefits, California uses state funds to provide cash welfare, food stamps, Medicaid, and Supplemental Security Income to immigrants otherwise ineligible due to the five-year residency requirement introduced during the 1996 welfare reform (Kaushal, 2005).

Given that provision of welfare, medical, and education programs is costly, several states have taken action to limit the use of these programs by undocumented immigrants. A well-publicized state position in California was Proposition 187 in 1994 that proposed an almost complete restriction of public aid and services including public education, health care, and welfare to undocumented residents of the state.3 Proposition 187 passed by a margin of 59 to 41 percent in 1994, but was ruled unconstitutional in federal court in 1995 based on exceeding state authority on immigration policy. Notably, this historical example is just one instance of state-level responses to undocumented immigration. In 2011, for example, state legislators presented more than 1,500 immigration-related bills and resolutions, of which 109 in 36 states focused on public benefits.4 If welfare migration is not significant for all legal status groups alike, then some of these efforts may be misdirected.

Welfare migration may exist to different degrees across legal status groups. Immigrants may respond not only to generosity differences across potential locations, but also to differences in availability that are correlated with legal status. Undocumented immigrants, for example, may be less likely than documented immigrants to respond to public aid incentives both due to eligibility issues and temporary migration (especially if application processing times discourage take-up among transitory persons). Furthermore, illegal immigrants may have greater dependence on social networks that may assist with employment and housing, and therefore these networks may substitute to some extent for public benefits. Almost 21 percent of undocumented immigrants in comparison to 17 percent of documented immigrants in the sample in this paper, for example, report free housing as part of their work compensation. On the other hand, undocumented immigrants may be more responsive to welfare if networks also assist in understanding and applying for public aid (especially when language is an issue). Furthermore, differences in family structure between undocumented and documented immigrants may promote differences in take-up given program characteristics. Almost 32 percent of the undocumented sample used here reports living without other family members versus only about 13 percent of documented immigrants. Thus, while there are many reasons to believe that undocumented and documented immigrants will respond differently to public aid differentials, the magnitude and direction of these differences are empirical questions not answered using data on legal immigrants alone.

Primary data in this paper come from the National Agricultural Workers Survey (NAWS), a nationally representative data set conducted by the U.S. Department of Labor. While restricting to one industry, the sample design of the NAWS, unlike traditional micro-level data sources, specifically accounts for migratory behavior and directly asks about both usage of public aid and, most importantly, about legal status. Due to data restrictions, previous studies of welfare migration by immigrants have only considered legal permanent residents and naturalized citizens. The NAWS affords the opportunity to extend this area of research to illegal immigrants and to those with other types of work authorization.

Raw tabulations from the NAWS are consistent with lesser behavioral response to welfare differences by undocumented immigrants than by documented immigrants. Of illegal immigrants between 1989 and 2009, approximately 21 percent report some form of public aid use with higher percentages among legal immigrants (60.5 percent) by a broad definition of public aid usage. Figure 1 shows aggregate welfare participation rates for agricultural workers by legal status in the sample over time where public aid is defined to include Aid for Families with Dependent Children (AFDC)/Temporary Aid for Needy Family (TANF), FSP, General Assistance, low-income housing, government health clinics, Medicaid, and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).5

image

Figure 1. Welfare Participation Rates of U.S. Farmworkers, by Legal Status.Source: National Agricultural Workers Survey, sample-weighted and pooled cross-sections, 1989–2009.

Download figure to PowerPoint

A primary contribution of this paper is to test for the presence of welfare migration by refined legal status groups: undocumented immigrants, naturalized citizens, green card holders, immigrants with other work authorization, and U.S.-born citizens. The analysis is presented as a case study from U.S. agricultural labor markets. While agricultural workers as a whole may be generalized to the greater population imperfectly at best, it is more likely that undocumented workers in agriculture are similar to undocumented immigrants overall. It, therefore, is argued here that illegal immigrant behavior may follow common patterns across industries. Thus, documenting the existence or absence of welfare migration for different groups contributes to a better understanding of the effect of state and local public finance on the locational distribution of migrants, and the budgetary effects of illegal immigrant clusters beyond the context of the case study.

2. IMMIGRANT LOCATIONAL CLUSTERING

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES

The literature concerning legal immigrant welfare migration is characterized by debates over data sources, appropriate econometric methods, and ultimate conclusions. Buckley (1996) uses Immigration and Naturalization Service (INS) admissions data from 1985 to 1991, and regresses the annual number of legal permanent residents in a state divided by its population on a measure of state-specific welfare levels (total AFDC monthly payments times a cost of living deflator divided by total recipients) and other state-level socioeconomic regressors. Consistent with a welfare migration story, Buckley finds a strongly positive, significant relationship between legal immigration flows and AFDC levels. He further finds refugees and asylees more responsive and employment category immigrants less responsive to welfare generosity than those gaining admission for family reasons. Dodson (2001), like Buckley (1996), provides support for the existence of immigrant welfare migration after regressing the number of immigrants by admission category from a given country who locate in a specific state on the maximum combined AFDC and FSP benefit for a family of three using cross-sectional INS data.

Zavodny (1999) challenges these conclusions using INS legalization data from 1989 to 1994 supplemented with refugee data. She regresses the log number of persons immigrating to a state in a given year on state-level variables, including real combined AFDC and FSP benefits for a family of three. She controls for state fixed-effects and for country-specific immigrant stock and finds welfare levels to have a significant positive effect only on the location choices of refugees and asylees. She concludes that welfare is not an important determinant of locational choice overall. Finally, Kaushal (2005) offers additional challenges by creating a state-level policy dummy variable for whether or not new immigrants are eligible for means-tested programs using 1995–1996 and 1998–1999 INS data. She uses the proportion of newly arrived immigrants in a year who locate in a given state as her dependent variable and concludes that means-tested programs have “at best a weak effect on the location choices of newly arrived immigrants” (p. 79).

In broader literature on welfare migration, Brueckner (2000) argues that strategic response by policymakers is evidence that a state's benefit choice is a function of neighboring states' levels, and therefore policy effects relating to welfare reform may be greater than migration effects alone. Elsewhere, Kaestner, Kaushal, and Van Ryzinc (2003) conclude that welfare reform was associated with increased employment-related intrastate migration (and decreased nonemployment-related migration) suggesting that welfare migration as defined in the literature may be less of a concern than migration for work-related reasons caused by changes in welfare availability requirements.

The differences in conclusions evident in the literature may be at least partially due to differences in behavior across legal status groups as none of these authors distinguish between documented and undocumented individuals. Furthermore, it is useful to extend the analysis through the current period while controlling for year, and thus for differential policies in effect. These objectives are addressed in what follows.

3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES

The existence of various welfare programs and other public services may differently influence the locational choices of U.S.-born citizens, legal immigrants, and illegal immigrants. Sorting may be a function of differential eligibility requirements, permanence of migration, social network availability and usage, and family structure characteristics across legal status groups. These differences may affect the expected benefits and costs of choosing one location over another and therefore may result in different patterns than those illustrated in literature focused solely on documented immigrants.

Equilibrium Sorting

Borjas, Bronars, and Trejo (1992) present a multiregion extension of the Roy model applied to internal U.S. migrants. Their model predicts that regions paying high returns to skills attract higher skilled workers while lower return regions attract those with fewer skills. Workers then self-select into the location that gives them their highest expected earnings. Borjas (1999) extends the model to immigrants and to welfare participation. His extension predicts that foreign-born welfare recipients will cluster in locations offering the highest welfare benefits more so than natives will. The intuition behind the general result is that immigrants face similar costs of migrating to any U.S. state while current native residents of a state have zero cost of staying put but positive cost of moving. This indicates that natives should be less responsive to small differences in state-level public aid than immigrants should be. Borjas confirms this pattern with U.S. Census data.

It can be argued theoretically that agents prefer to locate in the area for which their expected net benefits from making a migration is highest. The expected net benefits from migrating to a particular location is a function of a worker's expected wage earnings in that location, his or her expected cash value of public aid programs in that location (broadly defined to include welfare, education, medical, and any other aid and allowed to depend on earnings to account for the means-tested aspects of many programs), and his or her expected migration (and opportunity) cost from moving. Because variables are expected values, they represent interactions between the probability of having access to welfare, the choice to participate, and state-level generosity.6 Differences in earnings across regions can be attributable to differing endowments of factors of production and varying socioeconomic and technological conditions. An assumption of a skill measure determining worker earnings in each state allows characterization of equilibrium sorting.

Legal Status Considerations

If mobility costs and expected welfare benefits differ between illegal and legal immigrants, then the extent of welfare clustering by one group or another may be heterogeneous. If immigrants are distinguished by legal status and regions with high returns to skills attract higher skilled workers (and lower return regions attract those with lower skills), specific skill-level cutoffs should depend on the differing migration cost and public aid benefit schedules faced by each legal status group. Expected costs of migration to certain areas may differ by legal status, and in the context of differential availability and generosity of public aid by legal status group, expected benefits of migration also may differ. Undocumented immigrants, for example, may employ “coyotes”—border smugglers—or take longer routes to their destination to elude border patrol (see Gathmann, 2008) and thus may face higher migration costs than do legal immigrants and natives. They may have fewer opportunities to receive supplemental sources of income from public aid programs, and therefore may have lower propensities to participate than those in other legal status groups, and lower benefit levels if they do participate.

Asymmetric migration costs alter the locational distribution of migrants, and asymmetries may exist not only between natives and immigrants but also between specific groups of immigrants that are distinguishable by their legal status characteristics. Furthermore, while legal immigrants may be argued to face the same wage-skills relationship as natives, the earnings of undocumented immigrants are likely to be distributed differently. This may be true because illegal immigrants may have less bargaining power (all else equal) in negotiating wage contracts with employers who face penalties if caught with undocumented members in their workforces. Observable positive wage differentials between legal and illegal immigrants in the data are consistent with this story. Furthermore, illegal immigrants, unlike their legal immigrant or U.S.-born counterparts who are more likely to pay taxes, may consider gross wages as opposed to net wages when deciding whether to migrate.

In summary, when expected migration costs are allowed to vary across origin country/destination pairs, it is theoretically ambiguous which legal status group will cluster to the greatest extent, and a multiplicity of possible equilibria exist in contrast to the results of the earlier literature. Pena (2011) finds that Borjas (1999)'s immigrant welfare clustering result in the 1990 Census data continues to hold for Census year 2000, but that the result is sensitive to occupation. Occupations for which the result reverses are specifically those thought to have higher illegal immigrant percentages (e.g., farmwork, construction, and nonconstruction laborers). Census data, however, does not allow for analysis separating illegal from legal workers in order to confirm this hypothesis. The question, therefore, remains an empirical one, and one that can only be addressed with data that distinguishes individuals by legal status group.

4. THE NAWS AND PUBLIC AID PARTICIPATION

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES

To date, the empirical literature on welfare migration has primarily used two data sources: INS administrative record data or the U.S. Census. As INS data are primarily available in count form, authors using it generally do not control for individual demographic characteristics important to locational decision making. In addition, these researchers only consider legal permanent residents and are unable to characterize broader groups of immigrants, such as naturalized citizens and illegal immigrants. Legal status also cannot be fully controlled for in Census data, and the Census may underrepresent certain immigrant groups. This paper, however, extends this literature by presenting a case study using an underused but representative survey of undocumented and documented, immigrant and native, U.S. agricultural workers in order to draw further conclusions.7

The NAWS is a representative data set of employed farmworkers conducted by the U.S. Department of Labor. Advantages of the NAWS include that its sample design, unlike traditional micro-level data sources, specifically accounts for migratory behavior, and that it contains information relating to the legal status of its respondents.8 Survey respondents are employed by growers and farm labor contractors in crop agriculture. The NAWS is a cross-sectional sample from work sites three times per year (fall, winter/spring, summer) since fall of 1988. This paper uses the public use data covering fiscal years 1989 through 2009.

Of the 52,479 workers in the sample, 22,413 indicate undocumented immigration status. U.S.-born workers total 10,412. In addition, 2,284 naturalized citizens, 12,818 green cards holders, and 3,779 individuals with other work authorization (e.g., those with special work permits) are identifiable. Mexican workers total 38,228 (73 percent), and 20,644 (55 percent) of Mexican workers are undocumented. Respondents are guaranteed confidentiality and less than 2 percent decline to answer legal status questions. Furthermore, the survey has a longstanding history having been conducted multiple times a year since 1988 and never having resulted in an enforcement raid. Legal status questions are multistep and cross-checked for consistency as per Department of Labor procedures. Despite relatively smaller sample sizes on a per year and per region basis, the NAWS is nationally and regionally representative of agricultural workers within 12 spatial divisions (with sampling weights, which are used here in all tables). The 12 regions in the proprietary data set are collapsed into six regions for the public use data. California is one of these “regions,” but no other states are identifiable alone.

Table 1 shows key demographic and employment variables by legal status after pooling the data. Immigrants working in agriculture are more likely to be male than are U.S.-born citizens. Legal immigrants are older on average than natives, and undocumented immigrants are younger. In comparison with natives, immigrants have fewer years of education and are less likely to report English language proficiency. Undocumented immigrants report fewer years of U.S. experience than do documented immigrants. In terms of locational distributions across U.S. regions, immigrants are more likely to reside in California than are their native counterparts. More than 95 percent of all immigrant groups in U.S. agriculture are Hispanic, and more than 90 percent of nonnaturalized immigrants are from Mexico. Approximately 60 percent of those with work authorization other than green cards and almost 62 percent of undocumented workers report being migrant, and 51 and 41 percent of these groups, respectively, report working in a harvest task.

Table 1. Means of Key Demographic and Employment Variables, by Legal Status
  Nat.GreenOther 
 Nativecitizencardauthor.Illegal
  1. Source: National Agricultural Workers Survey, sample-weighted and pooled cross-sections, 1989–2009.

Female (%)34.1020.2023.8815.3315.73
Age (years)33.9940.1939.3831.3628.14
Married, spouse in U.S. (%)0.430.520.600.360.21
Children in U.S. (#)0.701.121.380.950.43
None (%)67.0155.0446.6864.4180.63
One (%)11.7011.4512.3510.497.05
More than one (%)21.2933.5240.9725.1112.33
Education (years)11.107.505.875.656.32
U.S. farmwork experience (years)12.4716.2715.588.644.84
Speaks English (%)96.3841.7022.1516.527.38
Reads English (%)94.8734.5917.0211.435.52
Hispanic (%)29.3596.0496.9298.2298.81
Has work network (%)54.2559.6659.9562.9378.39
In California (%)5.3628.3849.3234.4936.70
In Eastern region (%)18.8630.036.2312.9218.07
In Southeast (%)15.1711.727.6518.6915.60
In Midwest (%)40.679.5712.5910.8312.07
In Southwest (%)9.789.4712.269.104.64
In Northwest (%)10.1610.8511.9513.9712.94
From Mexico (%)0.0056.5594.9894.4093.99
In field crops (%)26.6612.6711.877.5514.12
In fruit crops (%)12.2630.0043.7140.7238.35
In horticulture (%)28.5916.8711.6913.9112.71
In vegetables (%)21.8334.5028.1032.0829.84
In misc. crops (%)10.475.954.635.564.91
Preharvest task (%)21.8517.3615.3715.5320.37
Harvest task (%)14.7329.8231.8850.5740.57
Postharvest task (%)17.9214.9312.728.119.48
Semiskill (%)21.4922.6327.3519.4617.24
Supervisor (%)0.350.660.460.300.04
Other task (%)23.6614.6012.216.0412.31
Migrant (%)16.0038.7639.1459.9161.66
Observations7,9402,04911,8342,73021,671

Welfare participation rates by region are presented in Table 2. It is assumed for the analysis here that respondents work and reside in the same region. Compared with the national average within legal status group, California residents have higher participation rates than those elsewhere. More than 70 percent of U.S.-born Californian workers, for example, indicate that they (or their families) participate in some sort of welfare program compared with the national average of 42 percent. In fact, all legal status groups display this qualitative pattern though some differences are not as statistically and economically significant in the raw statistics (e.g., for undocumented workers).

Table 2. Welfare Program Participation, by Legal Status and Location (% of Total)
  Nat.GreenOther 
 Nativecitizencardauthor.Illegal
  1. Source: National Agricultural Workers Survey, sample-weighted and pooled cross-sections, 1989–2009.

California70.5880.0378.1452.2520.37
East36.8768.2351.3026.9911.89
Southeast51.8847.2846.8214.0113.01
Midwest38.0859.8151.4219.5120.72
Southwest49.7359.7157.6432.9922.34
Northwest39.1968.7760.7848.7633.94
United States42.3167.4566.1636.4919.59

The primary empirical analysis in this paper focuses on interrelationships between locational choice and the extent of participation in public aid programs by undocumented and documented immigrants relative to a base category of U.S.-born workers. One previous paper using the NAWS is directly related to this one. Moretti and Perloff (2000) examine farmworkers' welfare program and private charity take-up decisions. The authors find that undocumented immigrant families are more likely to use public medical assistance and less likely to use other public aid programs when compared with documented immigrants and U.S.-born citizens. In addition, they show a positive correlation between public aid participation and U.S.-born children in households headed by illegal immigrants. Although they do not consider geographic clustering, their paper has implications for this study. If welfare migration does exist for the undocumented group, it may be driven by specific programs such as medical service. Second, indicators of family structures should be among the controls.

Public Aid Participation in the NAWS

As documented in Table 2, participation rates vary both across legal status and across locations. In the absence of welfare clustering, roughly equal percentages of welfare participants and nonparticipants would be expected to reside in each state. For example, if 30 percent of U.S. welfare participants live in California, then 30 percent of nonparticipants also should live there. Likewise, equal percentages of participants and nonparticipants in each legal status/location category are expected. The difference between the percentage of participant households living in a region and the percentage of nonparticipant households in that region is an unconditional estimate of the “welfare clustering gap” associated with that location. This type of formulation would not control for socioeconomic characteristics, and therefore might reflect differences in the distributions of characteristics associated with participation across areas instead of behavioral clustering. An empirical test for the existence of a “welfare clustering gap” using multivariate regression analysis, therefore, is developed following Borjas (1999). Specifically, a differences-in-differences model can be used to create conditional estimates of this gap. A positive and statistically significant coefficient on the clustering gap variable would be consistent with a welfare migration story.

Empirical Test of Welfare Clustering

The descriptive empirical model allows for multiple treatment groups by legal status in a linear probability model. Each immigrant legal status group is compared with a U.S.-born control group. The specification takes the form:

  • display math

where inline image is a binary variable for whether or not individual i is observed in a particular state or region; inline image is a dummy variable indicating whether or not a migrant farmworker is of illegal status; inline image, inline image, and inline image are binary variables for naturalized citizen, green card, and other work authorization, respectively; and inline image is defined as whether or not the respondent indicates receiving public aid. The coefficients β6 through β9 are estimators of the clustering gaps in the state or region of interest between public aid participants and nonparticipants for each legal status group relative to the estimate of this gap for the native population. The specification follows Borjas' (1999) empirical model in which an interaction term between public aid usage and immigrant status overall was used to measure this gap for immigrants as a total group relative to natives in the U.S. Census.

The characteristics in inline image control for demographic and work-related factors associated with program eligibility (e.g., family structure, crop type, migrant status, etc.) and for systematic differences in the averages of these factors across legal status groups.9 National origin effects are included in reported specifications. Previous papers (e.g., Borjas, 1999; Zavodny, 1999; Dodson, 2001) use immigrant country of origin to control for linguistic and cultural networks. Survey year fixed effects are further included in specifications to account for policy differences over time as indicated by the literature.

If actual public aid usage is correlated with better labor market opportunities or the locational choices of earlier migrants, migration networks may matter. The NAWS allows for the construction of an individual work network indicator variable, which is included as a regressor in order to avoid spurious effects pertaining to the labor market. Specifically, the network variable equals one if the worker was referred to his or her job by a relative, friend, or workmate. A large literature cites the importance of networks for the immigrant population. Scott, Coomes, and Izyumov (2005), for example, find that employment-based immigrants specifically tend to locate in cities where there are other immigrants like themselves. However, they are not similarly attracted to communities with large immigrant populations originating from sources other than their own.10 Munshi (2003) establishes the importance of networks for Mexican–U.S. migration particularly. Notably, more than 78 percent of the undocumented sample reports using a personal network to find employment (Table 1) in comparison to percentages around 50–60 percent for other legal status groups.

Table 3 presents estimates of the clustering gaps between welfare participants and nonparticipants for the four immigrant legal status groups relative to natives for California. The California results are presented separately for two reasons. First, as described above, California is the only state that is identifiable separately due to the regional sampling methodology underlying the survey and the construction of the public use data. Second, however, this structure is fortuitous for this paper as the California case specifically was highlighted in Borjas' original model given relatively high benefit levels in that location along with the observation of immigrant clustering. Focus on California, therefore, facilitates a comparison between earlier results in the literature for immigrants in general and for the refined immigrant groups identifiable here.

Table 3. Welfare Clustering Gap Hypothesis Test-California
Dependent Variable: Probability of California(1)(2)(3)(4)(5)
  1. Source: National Agricultural Workers Survey, sample-weighted and pooled cross-sections, 1989–2009. Notes: Robust standard errors in parentheses. Base category (excluded legal status group) is native, U.S.-born farmworkers. *inline image, **inline image, ***inline image.

Naturalized citizen0.143***0.126***0.280***0.296***0.201***
 (0.0186)(0.0194)(0.0324)(0.0324)(0.0329)
Green card0.298***0.266***0.265***0.284***0.185***
 (0.0108)(0.0126)(0.0273)(0.0277)(0.0269)
Other work author.0.244***0.225***0.221***0.252***0.157***
 (0.0135)(0.0152)(0.0286)(0.0299)(0.0296)
Illegal0.337***0.342***0.341***0.349***0.228***
 (0.00636)(0.00902)(0.0262)(0.0265)(0.0262)
Used public aid0.0514***0.0481***0.0511***0.0478***0.0456***
 (0.00539)(0.00607)(0.00604)(0.00621)(0.00707)
Naturalized*used public aid0.103***0.129***0.118***0.121***0.0689***
 (0.0261)(0.0268)(0.0258)(0.0257)(0.0265)
Green card*used public aid0.214***0.234***0.234***0.233***0.148***
 (0.0154)(0.0156)(0.0157)(0.0160)(0.0150)
Other author.*used public aid0.190***0.210***0.215***0.217***0.134***
 (0.0267)(0.0282)(0.0282)(0.0286)(0.0290)
Illegal*used public aid−0.0336**−0.0424***−0.0407***−0.0401***−0.0814***
 (0.0141)(0.0145)(0.0146)(0.0145)(0.0138)
Demographic characteristics?NoYesYesYesYes
Country of origin controls?NoNoYesYesYes
Year fixed-effects?NoNoNoYesYes
Crop, task, and migrant characteristics?NoNoNoNoYes
Observations51,52549,52849,52749,52748,984
R20.1500.1580.1710.1780.312

The results in the table indicate that all immigrant groups are statistically significantly more likely to be observed in California than are natives (the base category). Those who use public aid also are more likely (on the order of approximately 5 percent) to reside in California than in the other agricultural regions of the NAWS. The four interaction terms between aid usage and immigrant legal status groupings all are highly statistically significant. Those corresponding to naturalized citizens, green card holders, and those with other work authorization all are positive, consistent with a welfare migration story where legal immigrants are more responsive to public aid differentials than are natives. Coefficients on the interaction term for undocumented immigrants, however, display an opposite pattern indicating that illegal immigrants are less responsive to public aid differentials in comparison to U.S.-born workers. Magnitudes of these identified effects are decreased for legal immigrants and increased (in the negative direction) for illegal immigrants once controls for crop, agricultural task, and migrant status are incorporated (column 5).11 By these results, the welfare clustering gap for California between naturalized citizens and natives is 6.9 percent, while these gaps for green card holders and those with other work authorizations are 14.8 and 13.4, respectively, in comparison to the U.S.-born population. The lower magnitude for naturalized citizens is expected if more permanent immigrants are more similar to natives than are temporary immigrants. The welfare clustering gap for undocumented immigrants on the other hand is negative 8.1 percent in comparison to natives for the California case, indicating that the difference in propensities to locate in high-benefit California between undocumented immigrant participants and nonparticipants is less than this difference between native participants and nonparticipants. Results, therefore, are inconsistent with the hypothesis that California is a welfare magnet for undocumented agricultural workers.

Table 4 presents estimates of the clustering gaps between welfare participants and nonparticipants for the four immigrant legal status groups relative to natives in separate regressions for each of the other five regions identifiable in the NAWS public use data set. Columns represent separate linear probability regressions for these regions relative to everywhere else. California therefore is included in each of these bases. All else equal, welfare participants overall have an increased propensity to locate in the southern parts of the country (both Southeast and Southwest regions in Table 4), as was also evident for the California case. The opposite is true for the East and Midwest regions.12

Table 4. Welfare Clustering Gap Hypothesis Test-Other Regions
 (1)(2)(3)(4)(5)
Dependent Variable: Probability of Region NotedEastSoutheastMidwestSouthwestNorthwest
  1. Source: National Agricultural Workers Survey, sample-weighted and pooled cross-sections, 1989–2009. Notes: Robust standard errors in parentheses. Base category (excluded legal status group) is native, U.S.-born farmworkers. *inline image, **inline image, ***inline image.

Naturalized citizen−0.01040.114***−0.294***0.0246−0.0354
 (0.0393)(0.0283)(0.0328)(0.0245)(0.0297)
Green card0.02220.110***−0.277***0.00945−0.0494**
 (0.0306)(0.0236)(0.0271)(0.0114)(0.0227)
Other work author.0.0587*0.239***−0.353***−0.0358***−0.0656**
 (0.0354)(0.0300)(0.0343)(0.0132)(0.0258)
Illegal0.120***0.174***−0.374***−0.0829***−0.0646***
 (0.0309)(0.0237)(0.0275)(0.0106)(0.0229)
Used public aid−0.0336**0.0468***−0.0670***0.0220**−0.0138
 (0.0141)(0.0116)(0.0193)(0.00958)(0.00980)
Naturalized*used public aid0.0844***−0.131***0.0384−0.0507**−0.0102
 (0.0317)(0.0258)(0.0321)(0.0246)(0.0253)
Green card*used public aid0.0474***−0.0937***0.0109−0.0646***−0.0477***
 (0.0168)(0.0159)(0.0233)(0.0139)(0.0143)
Other author.*used public aid0.0308−0.212***0.0193−0.0331**0.0605**
 (0.0263)(0.0233)(0.0313)(0.0156)(0.0263)
Illegal*used public aid0.0127−0.104***0.114***−0.0237**0.0824***
 (0.0174)(0.0145)(0.0225)(0.0114)(0.0148)
Demographic characteristics?YesYesYesYesYes
Country of origin controls?YesYesYesYesYes
Year fixed effects?YesYesYesYesYes
Crop, task, and migrant characteristics?YesYesYesYesYes
Observations48,98448,98448,98448,98448,984
R20.1530.0700.1880.0460.039

Other regional patterns also emerge. Particularly, naturalized citizens and green card holders who use public aid are found more likely to locate in the eastern region than are native workers though those with other work authorization and undocumented workers display similar rates to natives conditional on the full set of control variables. The results for naturalized citizens and green card holders, therefore, also are consistent with a welfare migration story for legal immigrants relative to natives, but not for illegal immigrants, which is similar to what is found for California. Clustering gap estimators are statistically significant and negative for both the Southwest and Southeast regions across legal status groups indicating that the differences in propensity to locate in these regions between all types of immigrants who use aid and their counterparts who do not are less than the corresponding difference for the native population. This is despite the increased propensity for program participants overall to locate in these areas, and is in contrast to the California case where results are consistent with welfare clustering among the legal immigrant population but not among undocumented immigrants. For the Southeast and Southwest cases, natives who participate in programs are the overrepresented group.

The Midwest and Northwest columns in the table display mixed results. For each of these cases, differences between illegal immigrants who use aid versus those who do not are greater than this difference for the native population. For the Midwest case, other legal status groups do not show significant differences from natives. For the Northwest case, the welfare clustering gap for green card holders is negative and that for those with other work authorization is positive. These results can relate to the distribution of welfare recipients across regions. If the difference in the propensity to locate in the Midwest between natives who use welfare and those who do not is small, for example, then it is more likely that this difference for undocumented immigrants who use versus who do not use benefits will be larger in comparison to this base.13 This is what is observed for the Midwest and Northwest regions, but is not observed for the Southeast, the Southwest, or California. As opposed to a causal interpretation, results are indicative of equilibrium relationships between program participation and regional locational choice by legal status subgroup in the U.S. farmworker population. Results overall therefore can be seen as painting a picture of the geographic distribution of legal and illegal immigrant versus native welfare recipients throughout the relatively low-wage agricultural sector.

Self-Selection and Locational Choice

Questions arise as to how these results can be reconciled with those of the previous literature and whether the results here are generalizable beyond the population of farmworkers. Given that previous results on welfare migration come from data predating welfare reform, differences may be attributable to time frame of study. Second, the NAWS and Census are based on sampling methodologies of different philosophies that could potentially generate discrepancies. Particularly, unemployed persons who may have higher probabilities of using welfare are not represented in the NAWS, but are included in the U.S. Census.14 Buckley (1996), for example, finds employment category immigrants less responsive to welfare generosity than those gaining admission for family reasons. Third, because the NAWS is a survey of agricultural workers, differences may be due to characteristics of this specific occupation. Agricultural work is often physically demanding and workers in this industry may be selectively healthy (and likewise less likely to participate in public aid programs) in comparison to the general population. These explanations are examined in the context of occupation in Pena (2011) using year 1990 and 2000 Census data. There it is shown that the welfare clustering results for immigrants in general hold across Census years, thus discounting the study year explanation for differences. Furthermore, results also hold for the overall population and across employed and unemployed subgroups (though to differing magnitudes). This sheds doubt on employment differentials as an explanation for differences between the results in this study and those in previous literature. Results in Pena (2011), however, are found to be occupation-dependent reversing for those in agriculture and other fields with higher illegal immigrant concentrations, such as construction and nonconstruction laborers. Undocumented immigrant prevalence, therefore, is argued to be a primary determinant of difference.

The discussion in Section 'DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS' suggests that immigrants within each legal status group should sort based on their skill levels in response to differences in the wage-skill relationship across states. Therefore, selection might shed further light on the agricultural worker results if undocumented and documented immigrants within a state differ from those elsewhere in the country. For example, welfare clustering in California for some groups but not others in the NAWS might be associated with selection by higher skilled immigrant farmworkers to California farms as opposed to elsewhere.

To examine this hypothesis, again consider a linear probability model with variables defined as before:

  • display math

The vectors δ5 through δ8 capture differential effects of socioeconomic characteristics on the probability that workers in the various legal status groups reside in a particular state.

Using education as a proxy for skill, Table 5 suggests that self-selection differences between agricultural workers and the rest of the population of immigrants may drive some of the results. The significant coefficients on the legal status interactions with education for naturalized citizens, and with experience for naturalized citizens and green card holders, suggests positive selection by these measures into California particularly as legal status/education interactions are positive and significant for all legal status groups for this location.15 For illegal workers to California, it is also specifically notable that positive selection exists on the dimension of work networks all else equal, which is expected given the literature (e.g., Munshi, 2003).

Table 5. Selection in Observable Characteristics (Dependent Variable: Probability of Region Noted)
  (1)(2)(3)(4)(5)(6)
  CaliforniaEastSoutheastMidwestSouthwestNorthwest
  1. Sources: National Agricultural Workers Survey, sample-weighted and pooled cross-sections, 1989–2009. Regressions also include survey year, country of origin, crop, task, and migrant indicators, and noninteracted socioeconomic characteristics. Base category (excluded legal status group) is native, U.S. born farmworkers.

  2. Note: Robust standard errors in parentheses. *inline image, **inline image, ***inline image.

Nat.Female0.0259−0.114***0.0762***0.003910.003440.00447
Citizen* (0.0410)(0.0296)(0.0236)(0.0391)(0.0312)(0.0437)
 Age0.000621−0.00673***−1.10e-050.00608***0.00131−0.00128
  (0.00129)(0.00149)(0.00121)(0.00184)(0.000946)(0.00107)
 Spouse0.0191−0.0656−0.01170.0215−0.02210.0587*
  (0.0347)(0.0459)(0.0304)(0.0488)(0.0300)(0.0318)
 Children0.001210.007080.00116−0.00948−0.005730.00577
  (0.00970)(0.0112)(0.00939)(0.0123)(0.00830)(0.0100)
 Education0.0120***0.001480.000988−0.0183***0.0157***−0.0118***
  (0.00435)(0.00411)(0.00346)(0.00434)(0.00328)(0.00419)
 U.S.0.00329**0.00313*−0.00290*−0.00318*0.00127−0.00161
 farmwork experience(0.00142)(0.00174)(0.00149)(0.00189)(0.00147)(0.00123)
 Work−0.03230.0368−0.0671***0.004750.02740.0305
 network(0.0261)(0.0293)(0.0240)(0.0312)(0.0214)(0.0255)
GreenFemale0.0628***−0.0777***0.0857***−0.0206−0.0102−0.0400**
card* (0.0186)(0.0185)(0.0214)(0.0280)(0.0173)(0.0177)
 Age−0.00260***−0.00258***−0.00232**0.00374***0.00413***−0.000372
  (0.000803)(0.000931)(0.000908)(0.00123)(0.000868)(0.000670)
 Spouse0.0234−0.0401**0.01290.00335−0.007150.00759
  (0.0179)(0.0204)(0.0208)(0.0268)(0.0155)(0.0176)
 Children−0.0217***0.0255***−0.00424−0.006370.001100.00566
  (0.00552)(0.00725)(0.00761)(0.0115)(0.00586)(0.00570)
 Education−0.00130−0.0002850.00352−0.0165***0.0225***−0.00793***
  (0.00234)(0.00311)(0.00282)(0.00376)(0.00267)(0.00243)
 U.S.0.00349***0.00266**−0.00402***−0.000593−0.00228*0.000738
 farmwork experience(0.000995)(0.00109)(0.00104)(0.00137)(0.00117)(0.000829)
 Work0.000437−0.0116−0.0379**0.0492**−0.0372***0.0370***
 network(0.0140)(0.0172)(0.0156)(0.0228)(0.0141)(0.0139)
OtherFemale0.0758−0.137***0.04360.01480.0293−0.0263
auth.* (0.0466)(0.0243)(0.0287)(0.0472)(0.0239)(0.0437)
 Age0.00455***−0.00392**0.0008260.0003310.000650−0.00244*
  (0.00141)(0.00156)(0.00159)(0.00171)(0.000896)(0.00126)
 Spouse−0.00380−0.0656**0.0251−0.02480.02080.0483
  (0.0314)(0.0278)(0.0290)(0.0348)(0.0205)(0.0297)
 Children−0.01110.0268***−0.0215***−0.0198*−0.004640.0303**
  (0.0104)(0.00828)(0.00796)(0.0112)(0.00636)(0.0143)
 Education0.004800.0114**0.00159−0.0202***0.0139***−0.0115***
  (0.00366)(0.00473)(0.00381)(0.00545)(0.00282)(0.00337)
 U.S.0.00231−9.19e-05−0.0106***0.00382*0.00227*0.00230
 farmwork experience(0.00200)(0.00168)(0.00214)(0.00206)(0.00135)(0.00168)
 Work−0.0430*−0.0817***0.01260.0898***−0.01990.0423**
 network(0.0234)(0.0280)(0.0249)(0.0280)(0.0155)(0.0207)
Illegal*Female0.0373**−0.0938***0.0739***−0.0009170.0242*−0.0407**
  (0.0170)(0.0194)(0.0171)(0.0269)(0.0134)(0.0172)
 Age−0.00141**−0.00243***−0.00278***0.00411***0.0008450.00166**
  (0.000652)(0.000925)(0.000692)(0.00109)(0.000548)(0.000732)
 Spouse−0.00408−0.0486**−0.006980.0463*−0.01950.0328
  (0.0171)(0.0209)(0.0186)(0.0278)(0.0132)(0.0203)
 Children−0.0242***0.0162**−0.00985−0.0115-0.0005450.0299***
  (0.00682)(0.00771)(0.00774)(0.0110)(0.00634)(0.00910)
 Education−0.002820.002270.00168−0.0187***0.0177***−0.000110
  (0.00196)(0.00306)(0.00259)(0.00358)(0.00214)(0.00257)
 U.S.−0.001320.000167−0.00408***0.00311**0.00197**0.000151
 farmwork experience(0.00104)(0.00115)(0.000995)(0.00140)(0.000915)(0.000979)
 Work0.0825***−0.0101−0.0717***0.0140−0.0263**0.0117
 network(0.0123)(0.0181)(0.0146)(0.0217)(0.0110)(0.0139)
 Observations49,14049,14049,14049,14049,14049,692
 R20.3060.1600.0760.1950.0560.043

5. CONCLUSIONS

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES

This paper examines whether individual states or regions are welfare magnets for immigrants in various legal status groups. Previous studies of welfare migration have excluded undocumented immigrants, yet ongoing legislative initiatives suggest that welfare migration by this population is of concern to specific U.S. locations. A focus of this paper is to extend the literature to this group using case study data drawn from agriculture which is distinctive in that it allows estimation of differentials in behavior across legal status groups unlike what has been possible in previous literature. While agricultural workers may not be representative of the population as a whole and therefore generalizability may be limited for legal workers, it is arguable that undocumented workers in agriculture may share commonalities with undocumented immigrants overall as costs and benefits of welfare take-up to these populations are likely similar. Furthermore, local areas may face the same predicaments of weighing costs of supporting illegal immigrants versus benefits of supporting the workforce needed by local industries. While there is strong evidence that undocumented immigrants cluster in certain states, such as the common immigrant destination of California, this study does not find that these patterns are systematically related to public aid program participation. This is important to empirically demonstrate when state legislative initiatives may be misdirected and anti-immigrant public opinion may relate to notions of welfare-seeking behavior.

A final question is why undocumented agricultural immigrants appear to exhibit less welfare clustering than do other immigrants. One reason is that many immigrants in this occupation are transitory and therefore less likely to settle permanently in an area and to integrate into local communities.16 While lack of permanence may suggest lower mobility costs (and therefore welfare migration), decreased probabilities of receiving benefits before moving to another location may push expected benefits below expected costs. Because of application-processing times, transitory workers likely have low probabilities of physically receiving benefits before moving if they apply. Second, established agricultural work and housing networks may decrease the need for public aid as may certain family structure differences across undocumented and documented populations. Finally, immigrant agricultural workers often face language barriers (and therefore linguistic challenges applying for benefits) at greater rates than other immigrants, thus contributing to why aid program usage among immigrants is sensitive to the “permanence” of the specific immigrant population of interest. Notably, the negative estimates of the welfare-clustering gap between undocumented immigrants and natives to California are strengthened (become more negative) in the presence of controls for types of crop worked, specific agricultural tasks, and migrant behavior. This leaves eligibility differences and the possibility of social networks serving as substitutes for welfare as explanations. It also should be noted, however, that it is possible that the welfare magnet effect is underestimated for the undocumented population despite the broad definition inclusive of several types of programs that is used here. Particularly, it remains possible that undocumented workers are attracted to less formal assistance, such as soup kitchens, clothing banks, and church aid, which may not be picked up in the survey questions pertaining to public aid as given. Given the magnitude of negative estimates for undocumented workers, however, it is unlikely that the inclusion of unobserved less formal public aid mechanisms would change the direction of the conclusions.

Finally, it is relevant to note that state and local governments may decide, after the fact, that it is in their best interest to support undocumented immigrants if migrant workers are critical for production. In this case, instead of a supply-driven story for welfare clustering, a demand-driven story could be told where public aid availability and generosity is a function of the population inflows necessary to support local industry. Still, the findings in this paper are important in either case for quantifying fiscal costs of immigration and for dispelling attitudes toward undocumented workers when the economy slows.

  1. 1

    The Tiebout (1956) conjecture is that people locate in the jurisdiction that best satisfies their tastes for local public goods. This leads to efficient scale and allocation in equilibrium.

  2. 2

    Committee on Ways and Means 2011 Green Book.

  3. 3

    An extended description is presented in Martin (1996). Proposition 187 barred illegal immigrants from California's public education system (kindergarten through university) and required schools to verify legal status of students and their parents. Second, it required all publicly paid, nonemergency health care providers to verify legal status before treatment in order to be reimbursed by the state. Third, it required welfare benefit offices to verify legal status before benefit transfers. Fourth, it required a broad classification of service providers to report suspected undocumented immigrants to the state's attorney general and to the Immigration and Naturalization Service (now U.S. Citizenship and Immigration Services with the Department of Homeland Security). Finally, Proposition 187 declared production, distribution, and use of false documents to be a state felony.

  4. 4

    National Conference of State Legislatures.

  5. 5

    FSP changed its name to Supplemental Nutrition Assistance Program in October 2008. Despite the general patterns of legal immigrants having higher participation rates than natives and likewise natives displaying higher rates than undocumented workers, there is variation in participation by legal status across particular welfare programs. Highest participation rates among undocumented immigrants, for example, are in Medicaid, WIC, and FSP.

  6. 6

    Borjas (1999) introduces welfare programs as a minimum-guaranteed income and considers the case where welfare recipients and workers are mutually exclusive. Lower skilled persons opt into welfare when the minimum guarantee exceeds expected wage earnings. Because many low-income immigrants receive both wage earnings and supplemental public aid and because aid is not guaranteed for immigrants, a more general case is considered here.

  7. 7

    While restricting to agricultural workers is nonoptimal and a function of data availability by legal status, welfare clustering disaggregated by occupation tells a qualitatively similar story (see discussion under “Self-Selection and Locational Choice” in Section 'THE NAWS AND PUBLIC AID PARTICIPATION') and therefore the restriction to agricultural workers here is not a serious compromise. Thus, these data allow for the study of illegal immigrants in a more systematic way than before.

  8. 8

    The NAWS sampling procedure is based on four levels: region, crop reporting district, county, and employer with probabilities proportional to size at each level. Specifically, NAWS uses 12 geographic regions based on U.S. Department of Agriculture Quarterly Agricultural Labor Survey of farm employers. There are 47 crop reporting districts, which are county aggregates from which sampling locations are selected. Within crop reporting district, counties are selected randomly without replacement. The number of interviews per site is determined by a proportional distribution to total workforce. Workers are chosen randomly and interviews scheduled at times and locations chosen by respondents. Respondents receive a small payment for participation.

  9. 9

    Income, which may be associated with eligibility for aid programs, is not explicitly accounted for in this framework. Given that the sample comprises farm laborers who are low income irrespective of legal status, this is less of a concern than it would be for higher income persons who are excluded from participating. Furthermore, it is noted in the literature that higher relative benefits cover a larger range of reservation incomes than do lower benefits, and therefore the estimation method may integrate over different distributions for some individuals who are observationally equivalent. Given data on employed workers, public aid is assumed to be a supplement to income, as opposed to a replacement of income, thus minimizing this concern.

  10. 10

    This also is consistent with literature on risk aversion and information in the context of migration (e.g., Allen and Eaton, 2005).

  11. 11

    The inclusion of these extra controls for this case study is appropriate given the sizable concentration of agriculture in California and therefore the high demand for migrant workers during harvest seasons.

  12. 12

    The regional patterns can be hypothesized to relate to traditional south to north migrant streams as the agricultural season progresses. Those participating in the streams (who therefore may change locations more frequently than those who do not migrate in this way) may have decreased propensities to participate in welfare programs which often require a permanent address to receive benefits. If respondents to the survey in the more northern regions therefore are drawn from this more mobile group of agricultural workers, then this story could be consistent with the patterns across Tables 3 and 4.

  13. 13

    An implication is that any welfare clustering that does exist for the undocumented immigrant population is not occurring in the traditional high-benefit area of California and therefore arguably is clustering on other unobservable dimensions that may be correlated with participation.

  14. 14

    In contrast to the treatment here, Christiadi and Cushing (2008) model a joint decision over location and occupation by migrants by simultaneously estimating a multinomial logit model of occupational choice and a conditional logit model of destination choice. This type of joint estimation, while relevant, is not possible given the data limitation of only observing employed agricultural laborers in this case study.

  15. 15

    Chiquiar and Hanson (2005) and Orrenius and Zavodny (2005) find that Mexican immigrants are selected from the intermediate or high-end of the Mexican education distribution. More educated agricultural migrants (the majority of whom are from Mexico) are found to self-select to the Southwest region as indicated by the results in Table 5. Another interpretation is that those who select into the northward migrant stream (as opposed to engaging in return migration) have lower skills.

  16. 16

    One also could argue that the U.S. harvest season is long, given diversity of crops with different harvest schedules. Permanence within the U.S., therefore, may be substantial, more so than in any particular region.

REFERENCES

  1. Top of page
  2. ABSTRACT
  3. 1. INTRODUCTION
  4. 2. IMMIGRANT LOCATIONAL CLUSTERING
  5. 3. DIFFERENCES BETWEEN UNDOCUMENTED AND DOCUMENTED IMMIGRANTS: THEORETICAL CONSIDERATIONS
  6. 4. THE NAWS AND PUBLIC AID PARTICIPATION
  7. 5. CONCLUSIONS
  8. REFERENCES