Toulmin Essay

Locational Choices of the Legal and Illegal: The Case of Mexican Agricultural Workers in the U.S.

Author(s): Anita Alves Pena

Source: The International Migration Review , Winter 2009, Vol. 43, No. 4 (Winter 2009), pp. 850-880

Published by: Sage Publications, Inc. on behalf of the Center for Migration Studies of New York, Inc.

Stable URL: https://www.jstor.org/stable/20681733

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

Sage Publications, Inc. and Center for Migration Studies of New York, Inc. are collaborating with JSTOR to digitize, preserve and extend access to The International Migration Review

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

Locational Choices of the Legal and Illegal: The Case of Mexican Agricultural Workers in the U.S.1 Anita Alves Pena

Colorado State University

This paper examines relationships between legal and illegal farm worker migration from Mexico and state-level labor market, agricul tural, demographic, and public policy variables. The study uses a nationally representative farmworker survey providing direct legal status data. Consistent with previous literature, results indicate that personal and community networks are primary determinants of loca tional choices. Conversely, border enforcement is negatively related to migration to certain areas. Results are strongest for California migrants and for those with previous migration experience. Potential welfare and education program values are uncorrelated with locations of recent Mexican agricultural workers.

INTRODUCTION

Immigration policy, especially that relating to illegal immigration, is a heated topic in United States policy debates.2 Proponents of open borders argue that illegal immigrants form a crucial part of the labor force of the

ll thank Michael J. Boskin, John B. Shoven, Aprajit Mahajan, Christina Gathmann, Giacomo De Giorgi, Colleen Manchester, Gopi Shah Goda, Kevin Mumford, seminar participants at University of California, Riverside and at the Pacific Conference for Devel opment Economics, and anonymous referees for helpful comments. I am indebted to Daniel J. Carroll of the Office of Policy Development and Research, Employment and Training Administration at the U.S. Department of Labor for permission to use the NAWS data and to Susan Gabbard and her staff at Aguirre International for resources and hospitality. I acknowledge Christina Gathmann and Gordon H. Hanson for sharing unpublished sector-level border patrol data and Lorin Kusmin of USDA for providing necessary inputs to calculate state-level rural unemployment rates.

"Illegal" and "undocumented" refer to those unauthorized to reside and work in the United States. These terms are used interchangeably in this paper. "Legal" and "documented" refer to naturalized citizens, green card holders, and those with other authorization.

? 2009 by the Center for Migration Studies of New York. All rights reserved. DOI: 10.11111). 1747-7379.2009.00786.X

IMR Volume 43 Number 4 (Winter 2009):850-880

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATioNAL Choices of the Legal and Illegal 851

United States as a whole, or of individual states, and contribute to the

greater economy. Opponents worry about fiscal burdens, decreases in pub lic safety, ethnic segregation, and linguistic and cultural barriers. While some recent policy discussions have focused on potential federal-level actions such as the introduction of a new guest worker program reminis cent of the bracero era, questions regarding if and how individual states can persuade or dissuade migration are also relevant.

Migrants to the United States form a mobile and many times circumspect population. Direct data on these persons and their activities therefore are often unavailable or unreliable. Estimates that do exist of the

magnitude of the illegal population present in the United States and its dispersion (or lack thereof) across receiving states are striking. Passel (2006) estimates that 11.1 million illegal immigrants were present in the

United States in March 2005.3 This estimate is up from 10.3 million in 2003. Of the new total, approximately 6.2 million (56%) were from

Mexico. In terms of spatial distribution within the country, 2.5-2.75 mil lion illegal immigrants resided in California, followed by Texas (1.4-1.6

million), Florida (0.8-0.95 million), New York (0.55-0.65 million), and Arizona (0.4-0.45 million). These five states account for more than 50% of the estimated total.4

Patterns for legal immigrants parallel those for the illegal population. Of the 148,640 persons from Mexico who became legal permanent resi dents in fiscal year 2007, more than 57,000 took residence in California, followed by more than 31,000 in Texas, 8,000 in Arizona, and 4,000 in Florida.5 Naturalizations from Mexico totaled 122,258. Almost 58,000 newly naturalized citizens lived in California, and more than 21,000 lived in Texas. Approximately 6,000 and 2,000 resided in Arizona and Florida, respectively. The Census Bureau reports that the top four states by

3Passel and Cohn (2008) estimate an undocumented population of 11.9 million for March 2008. Comparable state-level breakdowns were not reported for this latter date at the time of this writing.

4The estimation strategy involves subtracting an estimate of legal foreign-born residents from one of the total foreign-born population. Asylum applicants and those with tempo rary protected status comprise as much as 10% of the estimate.

5U.S. Department of Homeland Security, 2007 Yearbook of Immigration Statistics. In fiscal year 2008, 189,989 new legal permanent residents and 231,815 new naturalized citizens were from Mexico. Supplemental data tables indicating state breakdowns, however, were not yet available.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

852 International Migration Review

Hispanic population growth from 2000 to 2006 were California, Texas, Florida, and Arizona.

While previous literature on immigrant geographic distributions has focused on legal immigrants due to data availability, this paper examines state-level locational choices of legal and illegal Mexican migrants by exploiting nationally and regionally representative data from U.S. agricul ture. Focus is on how state-level factors such as changes in labor market conditions, public service provisions, and law enforcement efforts are cor related with decisions of where to locate after successfully crossing the border. Examinations of legal immigrants alone leave out a significant portion of the immigrant population that is present in the United States, and illegal and legal immigrants may or may not have similar determi nants of locational choice. Analysis in this paper focuses on the key immi grant-receiving (and illegal immigrant-receiving) states of California, Texas, Florida, and Arizona, and unlike previous literature, asks what are state-level determinants of immigrant locational distributions while controlling for legal status group. In addition to providing evidence on aggregate location-specific characteristics, this paper updates previous literature by also exploring individual demographic (such as family struc ture differences) and temporal factors associated with migration decisions. This is complementary to recent studies of the destinations of Mexican immigrants using alternate datasets (e.g., McConnell, 2008).

Overall results indicate that personal and community networks are primary determinants of locational choice, and that border enforcement is significantly and negatively correlated with migration by agricultural

workers to certain areas. These results are strongest for California migrants and for experienced migrants relative to new ones. Potential welfare and education program values are found uncorrelated with locational choices of Mexican migrants.

BENEFITS AND COSTS OF MIGRATION AND IMMIGRANT LOCATIONAL CHOICE

If immigrants make migration decisions by weighing expected benefits and costs, then both individual-specific and location-specific variables should influence locational patterns. Expected benefits include expected income, which may include both wage earnings and public aid, and the probabilities of employment and of receiving aid are taken into account

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 853

respectively. Expected costs include monetary and psychological costs. These incorporate financial costs associated with travel such as the value of travel time (foregone income at the origin) and payments to border smugglers or "coyotes" for assistance in the trip (if one is migrating ille gally).6 If a temporary migration is intended, then round-trip costs are considered. Expected psychological costs of migration include values asso ciated with leaving family and home in order to undertake a risky migra tion. Furthermore, expected costs may include monetary and psychological costs associated with apprehension or deportation for cases in which migrants do not successfully cross the border.

State economic, labor market, and demographic conditions, as well as state policy instruments, can be hypothesized and shown to affect the locational distribution of immigrants. Jaeger (2000), for example, uses micro-level admissions data from the Immigration and Naturaliza tion Service (INS, now Department of Homeland Security) and 1980 and 1990 Census data and finds that wage levels and ethnic concentra tions are key determinants of locational choice. Furthermore, he finds that immigrants' responsiveness to labor market and demographic con ditions differs across admission categories. Employment category immi grants, for example, are more likely to locate in areas with low unemployment rates. Since differences are observed among those in various legal admission groups, a hypothesis is that determinants of locational choice also differ for illegal immigrants. This suggests value added from controlling for legal status in a study of immigrant loca tional choice.

A growing literature has examined how information networks affect migration decisions and outcomes conditional on arrival. Carrington, Detragiache, and Vishwanath (1996), for example, present a model where migration costs decrease with the number of previously settled migrants in an area, and Munshi (2003) finds that Mexican immigrants with larger networks are more likely to be employed and to hold a higher paying non-agricultural job.

In the literature on general immigrant locational choice, several studies have debated whether public aid program generosity increases immigrant flows to U.S. destinations. On the affirmative side, Buckley (1996), using INS admissions data from 1985 to 1991, finds a strongly

6Gathmann (2008) shows that illegal migrants may increase their probability of success fully crossing by making a costly investment in a coyote.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

854 International Migration Review

positive, significant relationship between legal immigration and payment levels from the Aid for Families with Dependent Children (AFDC) pro gram, and Dodson (2001) does likewise using INS data from 1991 and revised econometric techniques. In contrast, Zavodny (1999) finds welfare levels only to have a significant positive effect on the location choices of refugees and asylees. Furthermore, Kaushal (2005) adds a state-level policy dummy variable for whether or not new immigrants are eligible for means-tested programs and concludes that means-tested programs have minimal effects of locational decisions.

Evidence on whether border enforcement affects the locational distri

bution of illegal immigrants also is mixed. Some authors conclude that border enforcement causes migrants to make several attempts to cross the border as opposed to deterring migration, or that the composition of illegal migrants may respond to increases in border patrol and the distri bution of destinations may be sensitive to border patrol intensity (e.g., Gathmann, 2008). Others do find a deterrence effect (e.g., Orrenius, 2004).

Studies on agricultural labor specifically have focused primarily on aspects other than locational choice. Farmworkers, as a subpopulation of all immigrants present in the United States, may face incentives different from those in other occupations due to the seasonal and migratory nature of their work and to historical U.S. public policy pertaining to these workers, especially that relating to visa programs.

The history of temporary U.S. farmworker programs dates to Roose velt's Bracero Program of 1942. The Bracero Program increased the supply of Mexican farmworkers in the United States but depressed wages for both international and domestic workers (Martin, 1994). The program was replaced with the H-2 program in 1964 and subsequently with the H-2A program in 1986 as part of the Immigration Reform and Control Act (IRCA). H-2A visas established by the act allow temporary or seasonal entry and employment of foreign workers in U.S. agriculture and therefore legally bring some seasonal immigrant workers to U.S. farms. In fiscal year 2007, non-immigrant temporary worker admissions from

Mexico under H-2A totaled 79,394.

IRCA also allowed for legalization or amnesty of long-term undocumented immigrants. The act included a general program (1-687) granting legal status on the basis of continuous U.S. residence for the 5 years leading up to the program, and the Seasonal Agricultural

Worker (SAW) program (1-700) for farmworkers employed at least

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATioNAL Choices of the Legal and Illegal 855

90 days in the previous year (Martin, 1990). Although temporary farmworker programs have tried to dissuade illegal immigration by promoting legal routes to U.S. work, estimates of the illegal population have continued to grow exponentially. While the effects of IRCA are beyond the scope of this paper, it should be noted that recent discus sions of legalization also have included special provisions for agriculture and that this continuing possibility may affect the probability of work ing illegally in the sector.

DATA

Data on Illegal and Legal Immigrants

Appropriate data for studies of immigrant populations, especially illegal immigrant populations, are scarce. Johnson (2006) writes that "there are no nationally (or state) representative surveys that include questions about legal status." In the macroeconomics literature, a common solution is to proxy for numbers of illegal persons using measures of border apprehen sions or enforcement. In the microeconomics literature, researchers have

used data from immigrant respondents to the Current Population Survey (CPS), the U.S. Census (some of whom are illegal), or household surveys of sending or receiving communities of illegal migrants. U.S. household surveys prove problematic for the study of illegal immigrants as many in this group live in non-standard housing situations and are less likely to be sampled.8 Surveys specifically targeting migrant communities are similarly imperfect. One dataset popular in the literature is the Mexican Migration

applications for amnesty totaled 1.8 million under 1-687 and 1.3 million under 1-700, and a total of 2.7 million received legal permanent residency. Applicants were first granted temporary resident status, followed by permanent residency after passing English language and U.S. civics requirements. 8Gabbard, Mines, and Perloff (1991) explain that while CPS sampling methodology focuses on household location, NAWS focuses on employment and may avoid biases due to undersampling migratory and immigrant farmworkers. They write: "The CPS is based on a random sample of housing units. Though all types of housing are to be included, critics claim that agricultural workers who live in non-standard housing units or who may be illegal tenants or sub-tenants are likely to be missed." The authors compare 1988 NAWS and CPS data and find that NAWS workers are more likely foreign-born and less likely to own or rent houses. By U.S. Census Bureau estimates, its undercount of illegal immigrants is around 15% (Hanson, 2006). Other estimates of undercount range from 2% to 25%.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

856 International Migration Review

Project (MMP), which provides cross-sectional retrospective migration data. The MMP is problematic for studies of migrants present in the Uni ted States at any point in time, however, due to small sample sizes of those with U.S. migration experience. Although a series of cross sections can be reconstructed, the data are not nationally representative of either the sending or receiving country.9 Representative case study data from

U.S. agriculture, however, are available.

The National Agricultural Workers Survey

Primary data used in this paper come from the National Agricultural Workers Survey (NAWS), a nationally representative dataset of employed farmworkers established by IRCA and conducted by the U.S. Department of Labor. Advantages of the NAWS include its sample design, which, unlike traditional micro-level data sources, specifically accounts for migra tory behavior by sampling from work sites as opposed to from houses, and the fact that it contains direct information relating to the legal status of its respondents. NAWS workers are employed by growers and farm labor contractors in crop agriculture. NAWS has sampled from work sites three times per year (fall, winter/spring, summer) since the fall of 1988.10

As of 2008, the NAWS dataset comprised 50,259 observations. Of the 49,494 observations with non-missing legal status information, 20.1% are U.S. born, 4.4% are naturalized citizens, 25.0% are green card holders,

9A newer dataset, the Mexico National Rural Household Survey (Encuesta Nacional a Hogares Rurales de Mexico (ENHRUM)) conducted by El Colegio de Mexico and the University of California, Davis may be more appropriate as those data are reported to be nationally and regionally representative of rural Mexico. 10The sampling procedure of the NAWS is based on four levels: region, crop reporting

district, county, and employer, with probabilities proportional to size at each level. Specifi cally, NAWS uses 12 geographic regions based on USDA Quarterly Agricultural Labor Survey of farm employers. USDA information also is used for cyclical allocation. There are 47 crop reporting districts (county aggregates with similar characteristics) from which

sampling locations are selected. Within crop reporting district, counties are selected ran domly without replacement. The number of interviews per site is determined by a propor tional distribution to total number of workers. Workers are selected randomly when arriving for work, at lunch, or when leaving, and interviews are scheduled for convenient times away from work site at locations chosen by the workers. Due to confidentiality restrictions, the full NAWS dataset can only be accessed on site at the Department of Labor or at the offices of its contractor, the Aguirre division of JBS International. Data were accessed at the Aguirre office in Burlingame, California, for this paper.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 857

and 7.6% have other work authorization. The remaining 43.0% report being illegal.11 Mexican workers total 36,647 (72.9%), and 54.5% of

Mexican workers report being illegal. The NAWS is nationally and regionally representative of agricultural

workers (with sampling weights) within the 12 spatial divisions covering the 48 continental U.S. states as defined in Table 1. These cross-sectional

data are designed to be representative of farmworkers as a population for each year and season of the sample. Despite individual workers being observed only once and although these workers may follow the "migrant stream" depending on time of year, each point of time within the sample is representative of the population (and therefore of particular worker characteristics) present for that year and season.

Because sampling is at the regional level, states that can be cleanly separated and matched to aggregate state-specific data are the focus of this paper. Specifically, California, Arizona, Florida, and Texas are used. These

TABLE 1 NAWS Agricultural Regions

Observations Region (Abbreviation) States included 1989-2008 California (CA) CA 17,172 Southern Plains (SP) TX, OK 2,419 Florida (FL) FL 6,050 Mountain III (MN3) AZ, NM 1,908 Appalachia I, II (AP 12) NC, VA, KY, TN, WV 3,160 Cornbelt Northern Plains IL, IN, OH, IA, MO, KS, NE, ND, SD 3,912

(CBNP) Delta Southeast (DLSE) AR, LA, MS, AL, GA, SC 3,176 Lake (LK) MI, MN, WI 2,543 Mountain I, II (MN12) ID, MT, WY, CO, NV, UT 1,903 Northeast I (NEI) CT, ME, MA, NH, NY, RI, VT 1,264 Northeast II (NE2) DE, MD, NJ, PA 2,192 Pacific (PC) OR, WA 4,560 Total 50,259

nThe Department of Labor in its survey documentation writes: "As farm workers are often reticent, considerable effort is made to maximize both employer and worker response

rates. Growers' associations, extension and social services, training institutions, as well as individual employers and farm workers are informed of the importance of the survey and the need for their voluntary cooperation. Respondents are provided a pledge of confidenti ality and $20 for their participation." Of the 50,259 workers in the 1989-2008 sample, approximately 1.5% (765 workers) declined to answer the legal status questions. Unfortu nately, other approximations of numbers of illegal immigrants in agriculture are not read ily available in the literature and therefore opportunities to cross-check numbers are limited.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

858 International Migration Review

states correspond to the common Mexican immigrant destinations cited above. While Arizona and New Mexico are technically one region in the

NAWS framework, only Arizona respondents were surveyed over the sam ple period used. Likewise, Texas is grouped with the much smaller agri cultural producer of Oklahoma, which will be treated as a Texas sample in what follows. Other states are grouped together regionally in sets of several states as indicated in Table 1. These states cannot be separately

matched to state-level aggregate data. The restriction to four states/regions implies that analysis using this subsample should be interpreted as repre sentative of particular states or regions in relation to each other. Of the

Mexican immigrant farmworker population working in the United States, 62.3% of the total unweighted sample (54.3% weighted) is in the four state subsample. Individual-level determinants of locational choice across all 12 regions, however, are shown for comparison and for additional results in what follows.

Table 2 shows key demographic and employment variables by legal status after pooling the cross-sectional data. Immigrants in all legal status groups working in agriculture are more likely to be male than are U.S. born citizens. Naturalized citizens and green card holders are older on

TABLE 2 Means of Key Demographic and Employment Variables, by Legal Status

Native Nat. citizen Green card Other auth. Illegal Female (%) 34.59 18.54 23.10 13.92 15.94

Age (years) 33.20 39.39 38.92 31.50 27.93 Married, spouse in U.S. (%) 42.31 49.20 58.60 34.33 20.07 Married, spouse anywhere (%) 44.23 62.84 77.29 64.11 48.14 Children in U.S. (#) 0.71 1.11 1.38 0.95 0.40

None(%) 66.12 56.16 46.77 65.26 81.58 One(%) 12.11 10.50 12.20 10.39 6.85

More than one (%) 21.77 33.34 41.02 24.35 11.57 Children anywhere (#) 0.77 1.31 1.75 1.02 0.98 Education (years) 10.88 7.54 5.88 5.54 6.27 U.S. farmwork experience (years) 13.66 17.13 16.24 9.57 4.43 Hourly wage ($1,982-4) 4.08 4.16 4.13 4.06 3.72 Speaks English (%) 95.71 43.12 21.93 16.45 7.01 Reads English (%) 93.98 35.65 17.31 11.89 5.31 Has work network (%) 55.33 60.21 60.10 62.86 78.83 Hispanic (%) 33.82 95.34 96.91 98.24 98.82 In California (%) 6.10 26.59 49.96 34.88 35.98 In Southern Plains (TX, OK) (%) 10.02 7.08 7.66 5.33 2.73 In Florida (%) 3.06 12.36 4.88 8.24 7.71 In Arizona or New Mexico (%) 0.86 2.08 4.55 3.15 1.70

From Mexico (%) 54.16 94.88 94.54 94.10 Observations 6,784 1,791 10,773 2,582 19,104 Source: National Agricultural Workers Survey, pooled cross sections, 1989-2008.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 859

average than are natives, and illegal immigrants and those with other work authorization (e.g., temporary legal workers) are younger. All immigrant groups have fewer years of education and are drastically less likely to report

English language ability than are natives.12 Illegal immigrants report fewer years of U.S. farmwork experience than do legal immigrants.13 Almost 79% of illegal workers report having a work network while around 60% of legal immigrants report the same. The work network variable equals one if the worker was referred to his or her job by a relative, friend, or fellow

worker. In terms of locational distributions across U.S. regions, larger percentages of immigrant than of native agricultural workers are observed in California, Florida, and the Arizona/New Mexico regions. The opposite is true of Southern Plains (TX, OK) farmworkers.14

State Economic and Demographic Conditions

Table 3 presents a description of state economic, labor market, demo graphic, and public policy variables. While 2004 values are presented, data are merged based on specific year of observation. The 2004 values in the table, for example, are matched only to workers observed in 2004. Those observed in earlier years are matched to their survey year values of these aggregate characteristics.15 Since some aggregate variables are not available after 2004, analysis incorporating external sources is restricted to the 1989-2004 fiscal year period.

State labor market and demographic variables include rural unemploy ment rates, farm employment, Hispanic share of the state's population, and average wage rates. Unemployment rates serve as indicators of employment probabilities (and of more general labor market conditions) in various locations at given times. If migrants are attracted to tighter labor markets (those with lower unemployment rates), then higher unemployment rates in a state should dissuade migrants. Rural rates are based on 2003 non-metro

12Workers are asked to rate their English speaking and writing ability on a scale of one to four. The variables for language ability here pool responses one and two as "no" and three and four as a "yes" answer. 13The experience variable is calculated as survey year minus reported first year of U.S. farmwork.

14Unfortunately, the NAWS does not survey workers in agriculture-related occupations such as livestock. This may account for lower than expected percentages of Southern Plains respondents. 15A11 aggregate data are available from the author upon request.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

860 International Migration Review

TABLE 3 State-Level Characteristics, 2004

CA TX FL AZ

Rural unemployment rate (%) 7.12 6.23 4.90 6.36 Farm employment (1,000s) 213,384 45,037 52,915 13,144 State population Hispanic share (%) 34.67 34.60 19.00 28.01 Mean hired farmworker wage (U.S. dollars/hour) 8.41 7.73 7.97 7.08 Minimum wage (U.S. dollars/hour) 6.75 5.15 5.15 5.15 Maximum monthly welfare (U.S. dollars) 697 201 303 347 Annual education value (U.S. dollars) 7,860 7,698 7,181 5,595 Linewatch hours per mile 17,649 3,446 - 7,304 Sources: Rural unemployment rates are calculated using inputs from Economic Research Service of USDA. Farm

employment figures are from Economic Research Service of USDA and the Bureau of Economic Analysis of the Department of Commerce. Hispanic share of state's population is from the U.S. Census. Annual average wage rates of hired field workers are from USDA. Minimum wage data are from the Department of Labor. Maximum monthly AFDC/TANF benefit levels and FSP values are from the U.S. House of Representatives Committee of Ways and Means. Data on current expenditure per pupil in public elementary and secondary schools are from the U.S. Department of Education, Digest of Education Statistics and from the National Education Association, Rankings and Estimates. Border Patrol linewatch hours per mile are from unpub lished INS/Homeland Security data shared by Gordon H. Hanson and Christina Gathmann.

classifications and the author's calculations using data from Economic Research Service (ERS) of the U.S. Department of Agriculture (USDA). State-level unemployment rates have generally moved together cyclically and are highly correlated with statewide unemployment rates.

Data on total farm wage and salary workers employment totals also are from ERS (with the most recent years' data from the Bureau of Economic Analysis, U.S. Department of Commerce). As indicated by the data in the table, California's farm workforce base is magnitudes larger than any other state. Texas and Florida are also relatively large producers, while Arizona has a smaller farm workforce (but is still a common Mexican immigrant destination as previously noted).

Hispanic share of the population in the border states approximates the general size of cultural and linguistic network available in a state and proxies for the potential size of a migrant's greater U.S. social network. These demographic data are from the U.S. Census Bureau. This variable is included in addition to the personal work network variable above.

Annual average wage rates for hired field workers are from USDA farm labor publications. In an expected benefits and cost framework, workers should be attracted to higher wages all else equal.

State Policy Instruments

State policy variables include maximum public aid program generosities, education expenditure, minimum wages, and border patrol intensity.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATioNAL Choices of the Legal and Illegal 861

As United States hired farmwork is a low-wage occupation, minimum wages and welfare programs can be hypothesized to be applicable to this population.

The federal minimum wage was created as part of the 1938 Fair Labor Standards Act. The minimum wage was originally set at 25 cents per hour and has been increased 27 times. Certain states, and in some cases certain cities, have enacted their own minimum wages. In cases in

which a local government sets a minimum wage in addition to the feder ally mandated one, by federal law the greater prevails. Arizona and Texas do not currently have a state minimum wage. Florida raised its minimum wage for the first time (to 6.15) in May 2005. In contrast, California has reset its rate 21 times since February 1943, six times since the start of the NAWS.

U.S. farm employers are not obligated to provide minimum wages to all agricultural workers. Instead, employees of large farms and H-2A

workers are protected by minimum wage legislation, while other migrant workers are exempt. Furthermore, U.S. agriculture is characterized by a large percentage of illegal migrants, and undocumented workers may not receive wages above minimum levels.

Uncovered workers, in addition to covered workers, may receive a wage boost in response to minimum wage increases if the existence of a more generous outside option puts upward pressure on uncovered sector wage rates and workers are free to move between sectors. However, the minimum wage also affects the opportunity cost of working in agriculture by influencing pay and employment probabilities in non-exempt jobs. This suggests that a higher minimum wage might force workers from the non-exempt sector to the lower wage exempt sector. While these factors motivate the inclusion of this variable, the predicted sign of the coefficient is uncertain. Federal minimum wage rates are matched to Arizona, Texas, and Florida migrants in the empirical section since these states did not have state-level minimum wages during the NAWS sample period.16

l6It should be noted that employers may interpret the presence or absence of state mini mum wages differently. Texas, for example, adopts the federal minimum wage by reference

while Florida and Arizona did not have any minimum wage for the sample period studied here (Florida until 2006 and Arizona until 2007). If employers are uncertain about appli cability of state and federal minimum wages to their workers, the absence of a state mini

mum wage may create different incentives than a state law specifying that the minimum wage corresponds to that set at the federal level.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

862 International Migration Review

As stressed in the literature, an often hypothesized factor of loca tional choice is state differences in public aid payments. If migrants expect to receive forms of welfare while in the United States, interstate differ

ences in welfare payments affect the expected income calculation.17 Although many illegal immigrants are excluded from receiving payments from cash welfare programs, the presence of U.S.-born children may allow mixed-status families to receive assistance legally. In addition, workers who have obtained false documents may be able to get benefits.

Maximum monthly cash welfare payments plus food stamp benefit values are from the U.S. House of Representatives Committee on Ways and Means.18 AFDC was replaced with Temporary Aid for Needy Fami lies (TANF) as part of the 1996 Personal Responsibility and Work

Opportunity Reconciliation Act (PRWORA). PRWORA strengthened immigrant eligibility requirements for means-tested benefits and made immigrants ineligible for TANF for the first five years after arrival.19 The Food Stamp Program (FSP) provides vouchers redeemable for food prod ucts.20 Face values of these vouchers are set as a percentage of the federal poverty line. Maximum monthly benefits are constant across states but vary across family structures. Maximum monthly cash welfare plus FSP value for a family of the particular size applicable to each individual sur vey respondent forms the basis of this variable. This variable is designed as an upper-bound proxy for what an immigrant with a particular family size could potentially expect in welfare payments. If other program gener osities (including those available to illegal immigrants, e.g., migrant help

17State-level education expenditures are used here as a proxy for education "value" that may be received by migrants with children. No argument regarding wealth equalization policies is made. 18The Green Book was published annually from 1982 to 1994, at which point it was pub lished biannually until 2000, followed by early in 2004. Linear imputation was used to construct missing-year values.

19There are several state-level policies that may have affected geographic patterns of Mexican agricultural workers during the years of the NAWS data. Examples include Prop osition 187 in California, which denied illegal immigrants access to a large number of public aid and educational programs and services and state policies providing in-state tui tion and driver's licenses to illegal immigrants. While analysis of these programs is beyond the scope of this paper, pro- and anti-immigrant attitudes reflected in these policies should be captured in state fixed effects included in the present analysis. 20FSP changed its name to Supplemental Nutrition Assistance Program (SNAP) in October 2008.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATioNAL Choices of the Legal and Illegal 863

centers) are correlated with these more traditional welfare values, then

these variables should proxy for additional forms of public aid. The offered benefit levels in Arizona, Florida, and Texas are low

and have been relatively constant over the last two decades (supporting exogeneity of this variable). While Arizona has traditionally offered bene fits similar to the median U.S. state, Florida and Texas have had strictly lower benefits than the median state (Texas to the greatest extent).

California has traditionally offered above the median state level of bene fits, suggesting that any welfare migration that does exist should be most evident for that state.

Another public service program dimension on which a state can diverge is that of public education. Like welfare, the value of education may be considered in the expected benefits calculation of choosing one state over another. Current expenditure per pupil based on average daily attendance in public elementary and secondary schools is from the U.S. Department of Education and the National Education Association.21 Cur rent expenditure per pupil based on fall enrollment is highly correlated

with this variable and results using this alternative are qualitatively similar (not shown).

Although unrefunded hospital visits are another state-borne cost of migration, these are not explicitly modeled here. While migrants may plan to utilize welfare and education programs to supplement their expected net benefits from migration, it is unlikely that migrants predict the use of emergency room services (except potentially in the case of pregnancy), and the value of unrefunded emergency room care to a migrant across states is largely equivalent (even if costs to the state are not). Regressions adding Medicaid value per patient (not shown) found Medicaid to be an insignificant factor of locational choice.

Differences in state border patrol intensities are measured via line watch hours (person hours spent patrolling the border). An increase in

21 Current expenditures for elementary/secondary education are "expenditures for operating

local public schools, excluding capital outlay and interest on school debt. These expendi tures include such items as salaries for school personnel, fixed charges, student transporta tion, schoolbooks and materials, and energy costs. ... expenditures for state administration are excluded." Average daily attendance is "the aggregate attendance of a school during a reporting period (normally a school year) divided by the number of days school is in ses sion during this period" (Digest of Education Statistics). The implicit assumption is that education expenditure and quality are positively correlated (or at least expected to be posi tively correlated by migrants).

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

864 International Migration Review

border patrol intensity over certain border patrol sectors should be associ ated with a decrease in the probability of successful crossing (and there fore an increase in costs of reaching particular localities). Border patrol data are from the INS, now Department of Homeland Security (DHS). The DHS divides the U.S.-Mexico border into nine sectors: San Diego and El Centro in California, Yuma and Tucson in Arizona, and El Paso,

Marfa, Del Rio, Laredo, and McAllen in Texas. Linewatch hours are adjusted for border mile coverage and averaged by state, where state is defined as that state housing the sector's central city. Note that the New Mexico portion of the border is implicitly included since the nine border sectors cover the entire U.S.-Mexico border.

EMPIRICAL MODEL

The migration data described above are empirically modeled in a discrete choice framework following Bartel (1989) and Jaeger (2000). Utility for person / migrating to alternative d comprises two components, a system atic observable utility term and a random error term:

Uf=vf + e? (1)

The expected net benefit to person / from making a migration to destination d is Vf. Demographic characteristics of the sampled farm workers and destination-specific attributes are included in Vf. The proba bility that destination option d is chosen by person / (Pf) is an increasing function of Vf by assumption and

Pf = vt(uf>ufyd' eD,d? d') (2) where D denotes the choice set available to agent i. The migrant is assumed to select that location d in his or her individual choice set which

offers the highest utility.22 Utility levels associated with each potential location, although unob

servable, are assumed to be functions of a set of personal attributes (w?) and locational characteristics (xf). Personal attributes include gender, age,

22A potential concern with using the state as the unit of geographic observation is that smaller geographic units may be better approximations to homogeneous labor markets. See discussion in Bartel (1989).

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

Lo cation al Choices of the Legal and Illegal 865

existence of a spouse and/or children, education, U.S. migration or work experience, legal status, and presence of work networks. Locational charac teristics are as specified above and are matched to individuals by year of observation. Namely, a migrant observed in 2000 is matched to year 2000 values of each of the aggregate characteristics for each of the four states and likewise for those observed in other years.

Welfare benefit and education expenditure levels are matched by year of observation and by reported family structure characteristics. For example, a worker with no spouse and two children is matched to welfare values for a family size of three and an education value corresponding to two children (two times per pupil expenditure). Previous studies have used cash welfare values for a family of three (or four) as a regressor, despite differences in family sizes in the actual population. Instead of fac ing a welfare value for a generic family size, each migrant here is matched to the specific maximum welfare values for his or her particular family size. This matching is more appropriate than that used in previous litera ture since migrants may jointly decide where to locate and whether or not to bring family members on a migration. A time trend variable is included to account for time-sensitive changes for each state (e.g., infla tion), and state fixed effects also are included. Independent variables there

fore can be written zf ? [xf, w?\ and

where a and ? (or ) are parameters to be estimated and fi? is the error term.

Maximum likelihood conditional logistic regression allows effects of indi vidual characteristics and state-level attributes to be estimated simulta

neously.23 In this model, data are grouped by unordered receiving states and the likelihood is calculated relative to each group. Specifically, data are reformatted into a panel across individuals and across states. Data con sist of X D observations where TV is the number of individuals in the

sample and D is the number of locations in the choice set. The estimation

Uf = odxf + ?'wt + fi? = ?'zf + fi? (3)

Conditional Logistic Regression

23McFadden (1974) first developed this model, and it is sometimes referred to as McFad den's choice model.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

866 International Migration Review

strategy involves interacting individual attributes with dummy variables for the choices in order to examine how individual attributes apply to choices. As there are D observations corresponding to each individual, the dependent variable is an indicator for the realized location. The variable yf ? 1 if an individual locates in d and alternately yf =1 if he or she locates in df d.

The model estimates the probability of being employed in d as a function of individual and state characteristics. This probability is written:

( ? = l\Zi ) = VZ) Jzd = yrD a'xU?'wi Jxd

? where d ? 1,2,D (4) d=l e 1

Parameters estimated from maximizing the log likelihood show the impact of the vector of variables in a particular state on the individual's underlying utility associated with the particular location.24 Positive coeffi cients indicate that variables increase utility and have a positive effect on the probability that a specific location is chosen over the other possibilities in the choice set. Fixed effects and individual specific characteristics cannot be estimated without modification. In order to allow for individ

ual-specific effects, dummy variables for the choices are interacted with each w?. Because a complete set of interaction terms creates a singularity, defining a reference category is necessary: California is the base category in the analysis. Standard errors are robust and account for multinomial correlation, heteroskedasticity, and clustering at the state level.

REDUCED-FORM RESULTS

Table 4 presents both reduced-form coefficients and odds ratios (exponen tiated coefficients), and their respective standard errors, from the regres sion for the determinants of state choice over the destination set of California, Texas, Florida, and Arizona. The odds ratio increases with the

24Error is assumed to be distributed i.i.d. Weibull. The Weibull distribution is an extreme

value distribution. McFadden (1974) argues for the use of extreme value errors to exploit computational advantages.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATioNAL Choices of the Legal and Illegal 867

t?

OO00i-hOOC\i/-\ r-iOOOOOOO

rH rH <N \D ^ O O O

ooooooooooo

vo \ \ ? m <

m m t\ ^

O O O 00 O O ? -h*

OOOh-.'-'OON.OO ^ooooooo

????????>_^>__ **^***<SJ**n^,-h;LJ OfOo^ocorSr-icri^l-Hoo iT\ O O O O _; i/^ _; _? ? o

I o o o I I

o o I I

^ r-HOOOOO-HOOOO ?od?ooddodo

'O X? ON ? es ^

NO00hOOC\C\NC\NN poooooooooooo dododododooc?

* h S * * ^ * * * ) < roO ? O O ? XT 00 ( d I idd i-n' d d

1 I I III

(\ ^ vjh O C\(N

O O "?i O 00 ^ O (

r-h O O t\ O IT\ O O O O i- -? O

OOOOOOOOOOO

sa;-* m o o d -h"

* * * * V0 *0 * * * * m ers m O O g ? ^ CO O 00 >?1 o o C\ G> ? odd

^ ^ > O ^ CT)

- O O OOOO^Hr-lOCO -

* * / ON ?rs h -? g ( ? O ( O O * o I

I

????r^o^^ * * * 00 2! ? rC t\rOr-H_HOOOO 99^o III O O O I I I 1

ri /il"?

^ (

* <7\ * CT) CO * en \

^

^ v 1* ^ m 00

( ^ 1

? ? ? ? ? ? ? * G> * ( fO O ; ? r-h VO O ?p, O ^ ( ( O O ? 'O O n^h d d d d

Oh (u

? o ^ > * o S a

G S

S ^ Oh S ? S 8,-3 "S a ? S C< (JL, LO

w ? g

?: si fi ^

SA, ? 2 -2

^ 5 3 3

3 o S

d .9

M 3 o

-2 > li < o

rte..

sii " H ?Co

?1 ^

^ ? s ? ? S

c i3 o 8

Ml

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

868 International Migration Review

probability of a positive outcome and decreases with the probability of a negative outcome. An odds ratio of one is interpreted as a zero effect.25

State Attributes

Correlations between various state attributes and migrant flows are esti mated holding individual characteristics constant. State fixed effects account for unobserved attributes affecting locational choice.

A strong positive correlation is found for state population Hispanic share. Assuming a higher share of Hispanic population indicates greater networking opportunities for Mexican immigrants, the significant and positive correlation is representative of network effects. The odds ratio for state population Hispanic share is 1.605. The odds ratio should be inter preted as the ratio of the odds of choosing a state with a 1% increase in the percent Hispanic to the odds of choosing a state at its current percent

Hispanic level. The result indicates therefore that the odds of a Mexican migrant choosing a state with a 1% greater percent Hispanic are 60.5% greater. As shown in Table 3, in 2004, for example, the California per cent Hispanic was 34.7, while the Texas, Florida, and Arizona values were 34.6, 19.0, and 28.0, respectively. Results here therefore are consistent with previous literature suggesting that settlement patterns of earlier immigrants are determinants of current immigrant locational choices. Since not all Hispanic groups may be in one common network, the inclu sion of Hispanic share should be considered an upper-bound approxima tion and the coefficients here are interpretable as upper-bound estimates.

A strong negative correlation is noted for linewatch hours per mile. Migrants have 7% lower odds of choosing a state with 1,000 more line watch hours per mile than a state with less rigid border enforcement all else equal.26 Border enforcement therefore may seem to be an adjustable margin related to illegal immigration. However, there may be decreasing returns to scale. Linewatch hours, for example, were increasing over the same years that there were decreases in apprehensions.

Gathmann (2008) argues that there is an endogenous relationship between migrant flows and border enforcement. She instruments for

25Odds ratios are reported instead of marginal effects for computational reasons. 26Boeri, Hanson, and McCormick (2002) document that non-border patrol apprehensions are low in comparison with border apprehensions. Florida border patrol values are approx imated using Texas values.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 869

border enforcement with Drug Enforcement Administration budgets and finds that enforcement has shifted illegal migrant flows to remote crossing places. The results here are consistent with her story. Unfortunately, 2SLS approaches do not work in the conditional logistic regression framework and therefore testing for endogeneity of this variable here is not possible. It can be argued, however, that border patrol is not conflated with eco nomic effects at the state level. With minor exceptions, border patrol intensities are determined federally.

While identified social network and border patrol effects may be expected, unemployment rate estimates in Table 4 seem counterintui tive. The estimates suggest that all else equal, migrants have 18.0% higher odds of choosing a state with a 1% higher unemployment rate. Previous studies {e.g., Buckley, 1996; Zavodny, 1997) find similar posi tive correlations between unemployment rates and migration choices.

Dodson (2001) hypothesizes that either the time lag on the unemploy ment variable used in these studies is inappropriate, or that if all U.S. state-level unemployment rates are of much lower magnitude than unemployment rates in origin countries, differentials between states may not be relevant in locational decision making. Furthermore, the positive coefficient on the unemployment rate may be due to endogeneity if farmworkers drive up the unemployment rate because they are seasonal workers and spend long periods of time unemployed. Establishing a causal story would necessitate altering the empirical methods here and should be a question for future research in this area. Results using statewide unemployment rates instead of rural ones are qualitatively similar.

The labor market variables in the model (rural unemployment rate, farm employment, and mean hired farmworker wage) are included as one-year lags. Including these variables at their current levels or at two year lags (not shown) leads to similar results, as does using statewide unemployment rates in current or lagged form instead of rural rates. Labor market variables such as unemployment rates represent general equilibrium outcomes. Therefore, the non-intuitive positive coefficient associated with the unemployment rate should not be interpreted only in light of farm labor supply. The supply and demand factors driving the sign and magnitude of this coefficient are not separately identified here. In addition, there may be friction in arriving to the equilibrium outcome.

Another possibility is that because of the presence of personal work networks in agriculture, workers are willing to choose high unemployment

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

870 International Migration Review

areas if those areas are the same locations in which workers have connec

tions. Another is that rural unemployment rates may vary significantly by season and the annual rates are inappropriate.

Other state attribute variables (farm employment, mean wages, mini mum wages, and welfare and education values) do not have significant predictive power for Mexican immigrants in the NAWS. Correlations between migration and welfare benefit levels and education expenditure, for example, are not significantly different from zero in Table 4.27 This IS notable given that welfare values are entered as upper-bound estimates. As far as these variables are valid proxies for program values (as anticipated and understood by migrants), state-level welfare and education program generosities are not important determinants of locational choice for work ers in states reviewed here.

Individual Characteristics

With the exception of the marital status variable, all individual-level vari ables presented in the table are significant for at least one alternative state over the base state of California. Most individual characteristics are highly significant at the 1% level.

Illegal migrants are generally less likely to choose Texas and more likely to choose Florida or Arizona than California. This can be seen by observing the negative coefficient on illegal in the Texas column of the results. Likewise, the coefficients on illegal are positive for Florida and for

Arizona relative to the base case of California. The odds of an illegal worker choosing Texas over California are 0.6 times the odds of a legal worker choosing Texas over California. The odds of an illegal worker choosing Florida over California are 1.2 times the odds of a legal worker doing so. Similarly, the odds of an illegal worker choosing Arizona over California are 1.2 times the odds of a legal worker making that decision.

Female migrants are significantly more likely to choose Florida over California and are less likely to choose Texas or Arizona over California. Older workers are more likely to choose Texas or Arizona and less likely to choose Florida. The presence of children in the United States is of little consequence to locational choice. Mexican immigrants with

more children are less likely to choose any of the alternative states over California than are those with fewer children. More highly educated

This result is also true when restricting to those with family members.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 871

immigrants are less likely to choose Texas or Florida and more likely to choose Arizona than California. More experienced workers, however, are less likely to choose any of these alternative states over California, and summer and fall workers are less likely to choose Florida or Arizona than are springtime workers. The time trend indicates that Mexican migrants are less likely to choose any of the alternative states over California over the period of study, but this relationship is only statistically significant at conventional levels for the Arizona case.28

The work network variable deserves special consideration. Results indicate that those using work networks to obtain employment are less likely to choose any of the alternative states over California. Texas

workers are more than 12% less likely to use a work network than are those going to California. Florida workers have 86% lower odds and

Arizona workers have 41% lower odds of using a work network than do California workers. These results are highly statistically significant, indicating that personal networks are an important determinant of loca tional choice. In recent literature, McConnell (2008) reconfirms the importance of social networks, human capital, and temporal context for

Mexican migrants to particular types of U.S. sites. In contrast to this study, her paper focuses on categories of U.S. destinations (instead of particular ones), stratifying on large and small, urban and rural, among other characteristics.

New Versus Experienced Mexican Farmworkers

Migrants may change destinations as their U.S. experience accumulates, and therefore an extension is the separate characterization of new and experienced migrants. New migrants are defined as those with zero or one year of U.S. farmwork experience when surveyed. Experienced migrants are those with two or more years of experience. Results for new and expe rienced migrant samples are presented in Tables 5 and 6, respectively.

Coefficients on state population Hispanic share and linewatch hours per mile take the predicted directions but are insignificant at conventional levels. The rural unemployment rate shows a positive effect and mean hired farmworker wage a negative effect. Minimum wage is statistically

28While the time trend controls for unobserved time-varying factors, this control may be imperfect. Including multiple linear time trends for subgroups of years during the sample period leads to qualitatively similar results (not shown).

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

International Migration Review

M00M^OONvt\iA00(N--h (NOCS^OOcOOOO-h ???????OOO? C\ Vf * 00

^ o m vo ?? \ ( - ^ ,? en ( ?< t-h

vf co CO vf

CN NrOONrntTMA^fO^^lM CN O - - O O Vf t-H ( - t-h V?)

CO t\ O O O O ;

t-HCO* t-H * * (\ * O M ? C5 ? ? CN ? ? OO^ot?iCOO^f^^Ovf I I ' ?0 I C\ OO

O ? O -^ o I III

V?)^OOCNcOcOt-hcOCNOi/-n Nt th fOC\(N C?\^^\? NC\ ( ( ( OOOOC50C50000

oo o m cr? IN OO m ( t-h CN 00 O C\

00

oo o t-h o ?i - - 1?i ^ ON ri VO O rO vf

?h' o o o

^^^^^^^^^^^^ ^

- ^ ^^ ^ ^^ OO?OOOO?OOOt-i ( VO * * * * * CO t\

CN t-h oo CN CN * \ h-?

OOOOOOOcOOhvCNOCN Il I ^ ^? ?? ^ ^

CN t-h I I I

NfOiANCSNONNinONON

HqinoqcsqoTHors oooo??o??oo

R g i S-o L i f g -H1. Vfr-HN^COVrNOVf^fH o ""'od d

GNOcOvfcNvrNCNOOvfvfVD^O VDcNOCNCNvfOcOcOcOO^O t-HOCOT-H0?ICOCOCNT-Hr-I\0 ??O?OOO??OOO

' ? ? ? ? ? ^ 1 s vtn v^ <N oo ^ Vf

* - *

S ? S ^ ? ^ . ~ ^ , rCiO-^OrCj^^O^rC?O ? ^ ? a ? O CN -

1 I III

t? o -o t?

33 O ? ' > t?

3 "rt -r? ? <u S G < o -a 3 Sfu S3 g ?

OCNcOh-^vfcNCN NtN OOO^cfiONO T-Hr-IVfOOOT-HO

???S59S9 * COCO* * * OflO * ^0 Vf * CN CN O

S - " CN ? 5 S O d rovo o

(f)^^ ONV? cNO^OvfOcomt^ t-?t-hvO(NOOCNO

* * * * M VO ( ?v > * fpO CN CO S^ddcN^Sod o 9 ^ ? I I -

I

? Q D ' D o <=>

-(

Oh ^> e -

"S g Dh Ph

Oh ?

-a ~o

Is

t? ?

c .5 o S t? jo 3 O

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 873

-2 i S

C\OCNtNOOOiAfOCO^ OOOOOOr-jOOOO SSSSSSSS9SS *#,-HNHH*****[\* ? ? oo* * * * I xf* / ^^ ^^^ 0^^ ( ( ) ^ v/ ? \ 00

O rH o o

ooooddooo

-- o / ON

o o o

r-Hvn,-?0.cnvOoOrn1-. . . cstnooHH(N^tnoffio vq o _? co ^ ,_, do do do d I lili

/ 00 V0

(N en (N O 00 O O O O O O O

O O O O O O

^ co oo o o o o o ? ? ? ? ?

* * * * *

O 0 c\ CO / ? ? \ \

N?

CTn o 00 0 . G> V0 ( (N ^ ^ O O d O O

fOMD 00 (N C7\ O 00 CN O O O O O O O O

00 1 \0 \0 ^ ^ fo o m

O O O h O 0 OOOOOOOOOOOfN r?

^O.iA^rH<^N?0NrH. oooo^^^oor-voco ^. o d^^^-^^^lK O O O o O

1 1 I I III

(

* * 5^ * * * rH^H^f

V?V?OOGNinCNONijNCNMM rsOrH(NHO\A^^^inoo

-

* ?I

iAN(NO?N?tfiOrriO^ m odd I I I

d d

>> O

< ? ? fe- < 5 3^ 5 ? 5 < ?

OM(NfONN<0(N oorscocoooo dddddddd

, oo o o ; ^ ( o O

v/

/ 00 \ / \0 / ( / ( (

qq-h^^ooo 3?d???<d<d * co* -sP h- cr\ * * o* cavo o* t\ cOO<NoOOOt\cO

sHH I I O v/ do d

I

g fi ? ? ? e -2 -3

? i a-5 a

co

.a Q D ?

il iJ7^ 13 > <

13 -t? .

e ? e -s ' 'S ?

: S S 4j 3 o

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

874 International Migration Review

significant for this group in the positive direction, and the maximum wel fare value variable is significant in the negative direction.

Many of the significant individual characteristics in the full sample regression are not significant for new migrants alone. Specifically, the strong patterns by gender, age, and education do not appear. Instead, family structure characteristics appear to be of greater importance. The presence of a spouse increases the probability of choosing Texas or Arizona over California, and the presence of more children decreases the

probability of choosing these states. New migrants in the fall are less likely to choose Texas, but over time new migrants have been more likely to choose Texas over California all else equal. As experience increases from zero to one year, migrants are more likely to choose Texas or Florida over California and are less likely to choose Arizona over California. Illegal new workers have 87% lower odds of choosing Texas over California than do legal new workers.

Results for more experienced migrants are presented in Table 6. These results approximate the full-sample results. One difference is that among the more experienced population, workers were more likely to use a work network when migrating to Texas over California. Hispanic share and linewatch hours are strong predictors of locational choice, consistent with a story in which migrants learn about state characteristics through the process of migration and therefore are more responsive to differences 29 on return trips.

Individual-Level Determinants of Locational Choice, 12-Region Sample

Conditional logistic regression imposes an independence of irrelevant alternatives (IIA) assumption. Because this may not be desirable, destina tion set variations were considered. Robustness tests (available upon request) excluding Texas, Florida, and Arizona, respectively, yield similar results (coefficients direction and magnitudes) to those in Table 4.

29When restricting the sample first to illegal return migrants and then to legal return migrants (not shown), rural unemployment rates and farm employment totals are positive predictors of the locational choices of illegal return migrants. Farm employment totals are negative predictors for legal return migrants. State Hispanic share is a positive and highly

significant predictor for both groups.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATioNAL Choices of the Legal and Illegal 875

3:

?si et;

/ * - r-t -? ^ * m * * o ON

-.-, . -.?p:q?q = ?S2-S inOoOooOoOGNOONONOOCNOi/NOooOooONOOoOr-iO.^O ' ( 00 ^ ? ' ' ? ;

* cn * co * m* m { ? 5 \ * ^ ( * m* no * cn * * hJ ^ \ OCAO^Oc\OfoOMOooOoOooHNjtOoO\oOxfOoOhh(N ^. o^o^opd0.o0.?^o^o^o^o^o^o^?^o^o r^'^O^O^^^-H'^r-J^?^O^O^O^r^'^r-l^O^

5 Co * ^ oo * ?n Co ?n ? ( * Co t\ * cn * t\ * cn * ( * co vrNOo^OvoOcNOr-iOoOKOr-irHr^OfOOoOr^Or^Or^OcoO ^oONoONO^?qoqo^O^O^?q?^?^dq000,~"lO

* * CN NO * u"N * o* NO * * * ^ oo" * ?O * ^ * 00 NOO????.???CNOinOcoOir\0|\OcOO?????OcNO ?^r-H^r-H^Cl-^rH^O^O^c^^t-?^?^rH

es hv* r\ * oo* * ^ # i\ cn * \ * m * no

r-H O ^ O O^fN^rH^r-H^

no

o o o on

? * 2> O 00 ^ ? cn ? O <N o

* r\ ?^? o? h| o? inj (d? \?|[ m ?>ONfQ(f1oooN^H(nHo^tn^N^in<Nr, 00Oc7\OOOt\Ot\fr>hv'?'OOOcoOt^OoOO q?0N??o'CN?cN?NoNq?^?!\?c0?^ r-H ^ ? r^'^O^c?^^^O^fN^r-H^c'""

CN : * oo

CN O ? o

? h j cn * cn * no * co] O^^VOONtlCNr^Oh^Or-H^ r-iOoc~5v>oOooOir\<r5'rv

n 6 ??, 6 9 o 9 ^ ? ^ ^

CN v> j * ^ ? r-lOKOr-lOCNjf^-CNOOOOfSl

_ _ _ _ ooOit\Og>Ocn|Oo<NvO'-h'-h OQOOdpdpdON?^?1^?^?? r-^^r-H^O^O^^^O^CN

; its ; ?* < tv \? ?0

? ?o ? ? <n ?" '~ ~ ! ^ o 00 O PH ?

o ^ o

ojo 1 no

O r-H CO ? ^ ?

* * O )0 cN O O CN

? Cn ' ^ o

CN i * 00 ;=; ^ S O m O o ?9 o w 0 ^

* oo * m o o CN O NHH O o on r-h* ^ ?

On O* N? N? ^ ^ ^ ^ ^ ^

.d^dPd^. d^o^.o^tdpdpd^o ?^?^pH^r-'^(-s^d^d^ds-^'-'v-^'-;s?"

O CO * PH IT\ O r O 0N ? O ON

no CN ? (N? OO CN O vo CN

( O O

* no * CN ?4 ^ O \? O 0N O

no

o ? ^ ? ^ 6 ^ o

* } \?? CN * t-h ? iT\ * i/"N r-ir-IOOOOOr-lVncNONr-l^ mON^qtxOvA'NxfONrHVO ^ ^ ^ ^ ? ? ?

vTN cO * >? ? re? ? ^ ? ^NOt^cooor^cooOxj .,^^\flOc\OoOHNiOc >, ?; O

o q o co cO ? no o m m o q

? oo* i- ? no * ** * cN * * m * ? ? cn * oo* m * cn* ? KHoOx(NvoOffiOKO^NiA(N^CAxfin\?NKN\?O?)OOinM coOr^OfoOoON^Ot\OcOOinC^oOr-iOooON^O^OoOcNlfN ^oqo^oqoqoqo^o'q?cr}?T~H. ?oo,o^o9?^ ^, d

* b * CN O o ^ o o

CN * 00 O m h O in O o q o

cn * o cn ' o

iT\* co* O* * On* V0 Ot-hOnNOno^N^i-hCN 0\ ) ^ }<0^ ^ r^^nn^N^^^^h'cNoOOo q?qoqo^o^.?

?tn ^ no ~

NO* no* 00 * cn* t\ * CN ONonOoNv?OoO"^^ "00 \ (

ooo?uN?qo^d^o " CN '

S S ? rH Cl

2 -6 2 D

t?

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

International Migration Review

, <U oi

\3 oo ^ i?l O I ^ 9

oo 00 rd

?0Off|Or-(O^n 9d9dcidrlo

* ^ * ^? * * 2? * * (N'-|o'Nfnfriinr-|fr)rHiAO

( w cn '

? rn ? oo ? ? (? oo OOOooOoOvoOooOxfVO rod?dHdrf>d!S|d^^

co * <n * . t-h * o * ; s h ? s s g s s s ?

* O ? VA ? CO * r-H ? r-H $ |\ , / '? ?( O

~ ~' ~" ~* ~' ^ o <~i 9 ??. d ^ d <~1 :

$ f\ ? * " * in \ ^co^cn^^^^^^^co ^dindridrHo^d0o

* 0 CN ^ V

co en ? * CN ^ ^( tfidodrHd'Hd0d

* ^ * \0 I ^ -, ^ ^

rld9drid9r^1/" - VO m ^ Vf >

O \? ( 9 *0 ' ^

( ][ rO 0~ * ^ 5 / * ( c^po^^eo?OcN'-'cNt^o^o ^ ^--1

m ? t\ * ? cn * t\ ?OOmO^h^OkNhO ^d^d^d^! cd^cdPcd d

t\ O [\ CN ? w, w , O |\ CN r-H O d ^ ? ^ o^d9o

^ITlNfSOfCl'Hv / O o O cN ^ o 1 ? ~

* 00 $ ^0 ^ ,, ^ \ )Q _ m 9 ^ ^ m 9 j ?~ d ?i d o o ;

eu

r-l D O

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 877

As a robustness exercise, relative risk ratios from a standard multino

mial logistic regression of individual-level determinants of locational choice over the full 12-region sample are presented in Table 7. FL, MN3, and SP correspond to the previous Florida, Arizona, and Texas categories. Although some magnitudes differ in the absence of state-level regressors, the general pattern indicated by the estimated relative risk ratios is consis tent with the conditional logistic regression over the restricted four-state sample.

Of interest, the role of personal networks, shown via the work net work variable, is more pronounced for several non-traditional immigrant destinations than in California (the base category). Appalachia I, II (API2); Cornbelt Northern Plains (CBNP); and the Lake (LK) regions, for example, all have statistically significant relative risk ratios greater than one for the work network variable. This is consistent with the importance of networks to new immigrant destinations as stressed in the most recent literature. Indicators of region of origin within Mexico also are highly significant for most cases.30 This again stresses the importance of net

works. Of the 11 equations, 10 display odds ratios insignificant or signifi cant and less than one for the U.S. farmwork experience variable. This suggests selection by more experienced workers to the base case region of California.

DISCUSSION AND CONCLUSIONS

In a recent report, the Federation for American Immigration Reform found that the illegal immigrant population in California in 2004 cost the state $10.5 billion for education, uncompensated medical care, and incarceration of illegal immigrants and their children. After accounting for tax payments made by illegal immigrants to the state, the total net cost for this population was estimated to be $8.8 billion. In similar reports, net outlays were $3.7 billion per year in Texas, $1.3 billion per year in Arizona, and $1 billion per year in Florida. Given these estimated

30BAJA includes Baja California and Baja California Sur. NW includes the northwestern states of Chihuahua, Durango, Sinaloa, and Sonora. NE includes Aguascalientes, Coahuila,

Nuevo Leon, San Luis Potos?, Tamaulipas, and Zacatecas. CH includes Mexico Distrito Federal, Estado do Mexico, Guanajuato, Hidalgo, Morelos, Puebla, Queretaro, and Tlax cala. PAC includes Colima, Guerrero, Jalisco, Michoacan, and Nayarit. GULF includes Tabasco and Veracruz. OAX is Oaxaca. SE includes Campeche, Chiapas, Quintana Roo, and Yucatan.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

International Migration Review

magnitudes of fiscal costs associated with illegal migration, an understand ing of the relationship between specific characteristics of key illegal immi grant - receiving states and migrant flows themselves is important for policy making.

This paper contributes to that effort by examining the determinants of locational choice of illegal and legal Mexican migrants to the United States. Specifically, this paper illustrates regional differences in agricultural worker populations and correlations between locational choice and regional attributes, including economic conditions and public spending.

Understanding these relationships is important for public policy, especially regional planning. Identified locational choice determinants may be useful predictors of future migrant flows to be used in more sophisticated policy analysis. Results suggest that migrants' destinations are correlated with networks (seen via the significance of Hispanic share of a state's popula tion, personal network variables, and Mexican sending-state controls). Linewatch hours are important determinants of state choice in many of the regressions. These variables are shown to be more important to experi enced migrants than to first-time migrants. This suggests that informa tional networks may be imperfect, and information may be conveyed at a lag.

In the extreme, legal status may be thought of as following a life-cycle pattern and to be a function of a migrant's time in the Uni ted States. This is notable in Table 2. Illegal immigrant farmworkers are notably younger and are less likely to be married and to have chil dren than, for example, naturalized citizens and green card holders. Those with other work authorization present an intermediate case. Locational choices are also a function of time in the United States and

therefore migrants at different stages may be attracted to locations dif ferentially.

Some additional caveats should be mentioned. Given seasonality of work and the observation that a majority of this population is male and within the United States without family members, farmworkers as a group may not be the most appropriate set of migrants within the United States to study how welfare and education benefits affect locational decisions.

Therefore, the data in this paper represent a compromise in order to say anything about the often invisible population of illegal immigrants. Although the agricultural industry is a major player in the overall labor market for illegal workers, using the NAWS for a study of migration, spe cifically illegal migration, has its limitations. Because NAWS is a survey

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

LocATiONAL Choices of the Legal and Illegal 879

only of farmworkers, those employed in other sectors of the economy and the unemployed are excluded.31 An additional consideration is that since the survey relies on end-point sampling within the destination country, data are only representative of successful border crossers. Third, workers are observed only once, and it is uncertain to what extent observed loca tions correspond to points of entry.

REFERENCES

Bartel, A. P. 1989 "Where Do the New U.S. Immigrants Live?" Journal of Labor Economics 7(4):371

391 (October).

Boeri, T., G. Hanson, and . McCormick, eds. 2002 Immigration Policy and the Welfare System, Oxford: Oxford University Press.

Buckley, F. H. 1996 "The Political Economy of Immigration Policies." International Review of Law and

Economics 16:81-99.

Carrington, W. J., E. Detragiache, and T. Vishwanath 1996 "Migration with Endogenous Moving Costs." American Economic Review

86(4):909-930 (September). Dodson III, M. E. 2001 "Welfare Generosity and Location Choices among New United States Immigrants."

International Review of Law and Economics 21 7-67.

Gabbard, S. M., R. Mines, and J. M. Perloff 1991 A Comparison of the CPS and NAWS Surveys of Agricultural Workers. Berkeley: Insti

tute of Industrial Relations Working Paper Series, University of California. Gathmann, C. 2008 "The Effects of Enforcement on Illegal Markets: Evidence from Migrant Smuggling

along the Southwestern Border." Journal of Public Economics 92(10-11): 1926 1941.

Hanson, G. H. 2006 "Illegal Migration from Mexico to the United States." Journal of Economic Litera

ture XLIV:869-924.

Jaeger, D. A. 2000. "Local Labor Markets, Admission Categories, and Immigrant Location Choice."

Unpublished paper.

Johnson, H. P. 2006 At Issue: Critical Facts on Critical Issues, Illegal Immigration. San Francisco: Public

Policy Institute of California. <http://www.ppic.org/content/pubs/atissue/AI_406H JAI.pdf>.

31 An additional point is that selection into agriculture may have changed over time as agri culture becomes a smaller share of U.S. production. While time trends are included in the analysis in this paper, the sample may become more selected over time.

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms

880 International Migration Review

Kaushal, . 2005 "New Immigrants' Location Choices: Magnets Without Welfare." Journal of Labor

Economics 23(l):59-80. Martin, P. L. 1990 "Harvest of Confusion: Immigration Reform and California Agriculture." Inter

national Migration Review 24( 1 ) :65-95 (Spring).

1994 "Good Intentions Gone Awry: IRCA and U.S. Agriculture." Annals of the American Academy of Political and Social Science 534:44-57.

McConnell, E. D. 2008 "The U.S. Destinations of Contemporary Mexican Immigrants." International

Migration Review 42(4):767-802 (Winter). McFadden, D. 1974 "Conditional Logit Analysis of Qualitative Choice Behavior." In Frontiers in Econo

metrics. Ed. P. Zarembka. New York: Academic Press. Pp. 105-142. Munshi, K. 2003 "Networks in the Modern Economy: Mexican Migrants in the U.S. Labor Market."

Quarterly Journal of Economics 118(2): 549-599 (May). Orrenius, P. M. 2004 "The Effect of U.S. Border Enforcement on the Crossing Behavior of Mexican

Migrants." In Crossing the Border: Research from the Mexican Migration Project. Ed. J. Durand and D. S. Massey. New York: Russell Sage Foundation. Pp. 281-298.

Passel, J. S. 2006 "The Size and Characteristics of the Unauthorized Migrant Population in the U.S.:

Estimates Based on the March 2005 Current Population Survey." Technical Report, Pew Hispanic Center.

Passel, J. S. and D. Cohn 2008 "Trends in Unauthorized Immigration: Undocumented Inflow Now Trails Legal

Inflow." Technical Report, Pew Hispanic Center.

Zavodny, M. 1999 "Determinants of Recent Immigrants' Locational Choices." International Migration

Review 33(4): 1014-1030 (Winter).

1997 "Welfare and the Locational Choices of New Immigrants." Federal Reserve Bank of DalUs Economic and Financial Policy Review QIL2-10. <http://www.dallasfed.org/ research/er/1997/er9702a.pdf>

This content downloaded from ��������������66.77.17.54 on Sun, 08 Nov 2020 15:11:00 UTC��������������

All use subject to https://about.jstor.org/terms