Abstract
This article relies on local area variation in immigration policies, specifically the local implementation of the 287(g) program, and economic conditions to estimate their impact on changes in the size of local Mexican immigrant populations between 2007 and 2009. The author also investigates the impact of the 287(g) program on the employment prospects of low-skilled native black and white workers. The study finds that outside of four influential outliers (Dallas, Los Angeles, Riverside, and Phoenix), there is no evidence that the 287(g) program impacted the size of the Mexican immigrant population. In addition, there is no evidence that immigration enforcement policies mitigated the negative impact of the economic recession on the native population, even in the four outliers where the program was strongly enforced. The author highlights the limited efficacy of immigration enforcement as a way to resolve the issue of the undocumented immigrant population and for altering the employment opportunities of native workers.
The economic boom of the 1990s and early 2000s coincided with both the rapid growth and dispersion of the foreign-born Mexican population of the United States. Between 2000 and 2007, the foreign-born Mexican population in the United States enjoyed steady and continuous growth from 9.2 to 11.7 million, an increase of more than 27 percent. Between 2007 and 2009, however, this trend reversed; estimates from the American Community Survey indicate that the Mexican foreign-born population in 2009 declined by more than 200,000 to 11.5 million. Two salient changes coincide with this reversal. First, starting in December 2007, the collapse of the housing market triggered a severe recession that significantly altered the economic conditions attracting Mexican workers to the United States. Second, stronger immigration enforcement policies associated with growing anti-immigrant sentiment both heightened the costs and danger associated with crossing the U.S.-Mexico border and produced an ever-growing number of deportations. The efficacy of immigration enforcement policies for reducing the size of the foreign-born Mexican population, though, is highly contested, especially since the relative impact of enforcement and changing economic conditions on migration trends is unclear.
This article takes advantage of local area differences in immigration enforcement policies and economic conditions to estimate their unique impact on changes in the size of local Mexican foreign-born populations between 2007 and 2009. In particular, I evaluate the effect of establishing a 287(g) program, which involves local enforcement agencies in immigration control, on the subsequent size of the Mexican immigrant population of local areas. In addition, I also investigate how changing employment opportunities resulting from the economic recession affected the size of the Mexican immigrant population. A central focus of the analysis is on understanding variation in the effectiveness of the 287(g) program and the role of the recession on the size of local immigrant populations. The final part of the article investigates whether participation in the 287(g) program had a positive effect on the native population, specifically whether whites and blacks in areas that enacted the program were less adversely affected by the recession than their counterparts in other areas.
Background: Economic and Policy Changes since 2007
The impact of the immigrant population has always been highly concentrated geographically, though after 1990 the number of local areas experiencing large immigrant inflows grew dramatically. Previous studies have documented the importance of industrial restructuring and larger economic considerations in shaping the distribution of the immigrant population across the country (Parrado and Kandel 2008, 2011), but it remains unclear whether and to what extent policy actions can influence population movements over and above economic considerations.
The economic prosperity of the 1990s and early 2000s ended abruptly in December 2007 with the advent of a global economic recession that began with a housing crisis in the United States. Prior to the recession, economic growth and expanding employment opportunities contributed to declining U.S. unemployment. In the fall of 2000, unemployment levels reached record lows of 3.9 % they increased to 6.3 percent in 2003 during a short-term recession and then declined again to 4.4 percent for several months between October 2006 and May 2007. However, the collapse of the U.S. housing market in the fall of 2007 rippled through the world economy and produced dramatically higher unemployment, which reached 10.1 percent by October 2009, more than twice the rate prior to the recession (Fronstin 2010; Sum, Khatiwada, and McLaughlin 2009).
Not surprisingly, the economic boom prior to the recession coincided with large inflows of Mexican immigrant workers. Between 2000 and 2007, the share of the U.S. labor force that was foreign-born increased from 12.5 to 15.6 percent (Newburger and Gryn 2009). While the immigrant contribution to the U.S. labor force is evident across all industries and skill levels, Latin American and Mexican immigration was disproportionately low-skilled. While only 9.5 percent of the native civilian labor force had less than a high school diploma in 2007, 28.6 percent among the foreign-born did. Mexican immigrants represented 62 percent of the workers with less than high school, while an additional 23 percent came from other Latin American countries (Newburger and Gryn 2009).
The overrepresentation of Hispanic immigrants in the low-skilled labor force translates into a particular industrial distribution with direct implications for understanding the impact of the recession on migration flows and its connection with the employment prospects of low-skilled native and immigrant workers. In 2007, the industrial sectors with the largest representation of foreign-born relative to native workers were agriculture, forestry, fishing, and hunting (25.7 percent); followed by accommodation and food services (24.1 percent); and construction (23.4 percent). In general, foreign workers were less likely to be represented in high-skilled industries, representing only 6.9 percent of public administration; 10.1 percent of educational services; 11.7 percent of finance and insurance; and 13.7 percent of professional, scientific, and technical services (Newburger and Gryn 2009).
Thus, the industries in which Hispanic immigrants were concentrated were precisely those that were hit particularly hard by the recession. According to data from the Bureau of Labor Statistics, the construction industry gained 865,000 jobs between December 2001 and 2006 but lost close to 2 million jobs between 2007 and 2009. Accommodation and food services likewise gained close to 1 million jobs between December 2001 and 2006 but lost close to 300,000 jobs between 2007 and 2009. Even employment in the professional and service industry, which tends to employ fewer immigrants, declined by 1.4 million between 2007 and 2009, after increasing by 1.1 million between 2001 and 2006. 1
It is important to note, though, that there is considerable geographic variation in recession-related job losses. Areas where the pre-recession housing boom was more pronounced likely attracted more foreign-born Mexican migrants due to expanding labor demand in the construction industry. We could expect, therefore, that regional differences in employment conditions will have differential effects on changes in the size of the foreign-born Mexican population, with areas suffering steeper declines in construction and retail services exhibiting greater reductions in their immigrant Mexican population.
While labor market changes may have motivated return migration to Mexico or redistribution within the United States, it is also possible that increased enforcement of immigration laws contributed to the reduction in the foreign-born, especially Mexican, population. Labeled by proponents as “attrition through enforcement” (Vaughan 2006) and opponents as “the misery strategy” (New York Times 2007), the initial immigration enforcement policies as laid out by the Center for Immigration Studies included proposals such as “mandatory workplace verification of immigration status; measures to curb misuse of Social Security and IRS identification numbers; partnerships with state and local law enforcement officials; expanded entry-exit recording under US-VISIT; increased non-criminal removals; and state and local laws to discourage illegal settlement” (Vaughan 2006, 1).
The overall strategy has been variably enforced, but one of the most salient outcomes has been a dramatic increase in the number of deportations/removals 2 and a decline in the number of border apprehensions/returns. 3 Depending on the perspective, the latter change can be differentially attributed to economic conditions or border enforcement. Figure 1 documents these trends. The figure reports immigration statistics from the Department of Homeland Security. The number of returns (border apprehensions), which is usually interpreted as an indication of migration flows, fluctuates over time. Specifically, the number of returns declined during the 2001 recession, increased immediately after in 2004 to 2006, and then declined again to record low levels. More important for our purposes, the number of removals (deportations) shows a dramatic and continuous increase beginning with the passing of the Illegal Immigration Reform and Immigration Responsibility Act (IIRIRA) in 1996. The increase is particularly pronounced after 2002; the number of immigrants removed more than doubled from 165,000 in that year to 393,000 just 7 years later. As many as 67.3 percent of the removals in 2009 had no prior criminal conviction. The vast majority of persons removed (72 percent) were Mexicans, among whom only 34.3 percent had a criminal conviction in 2009. Together, the decline in in-migration flows and increased deportation should logically reduce the overall size of the foreign-born Mexican population in the United States, especially after 2007, in a manner consistent with the attrition through enforcement strategy.
Trends in Removals and Returns: 1990–2009
These overall trends, however, mask considerable variation in enforcement policies across local areas. One particular program, the 287(g), has been particularly instrumental in the implementation of the attrition through enforcement strategy (Vaughan and Edwards 2009; Capps et al. 2011). In 1996, IIRIRA amended the Immigration and Nationality Act by the addition of Section 287(g). The amendment authorized the federal government to enter into a written memorandum of agreement (MOA) with state and local law enforcement to participate in immigration control, which until then had fallen under the sole purview of federal immigration agents. At the state level, Florida (2002) and Alabama (2003) were the first to sign on, followed by Arizona (2005). At the local county level, Los Angeles and San Bernardino, California, entered the program in 2005, followed by Orange and Riverside, California, and Mecklenburg, North Carolina, in 2006. Enrollment significantly increased in 2007 with twenty-six new jurisdictions, and again in 2008 with the addition of thirty-four jurisdictions. The program has been tightly linked to the number of deportations; by 2011, 186,000 immigrants had been identified for removal through the program, and 126,000 voluntarily departed. 4
The program remains highly controversial. Critics argue that it violates human rights, subjects Hispanics to racial profiling, and suffers from a lack of clear goals and oversight (Government Accountability Office [GAO] 2009; Organization of American States [OAS] 2010). Proponents argue it is successful at removing those with standing deportation notices (such as visa over-stayers) who are otherwise not pursued aggressively by immigration authorities (Baker McNeill 2009). Both critics and proponents agree, however, that the program is likely to create an inhospitable environment that will discourage the entry and encourage the exit of immigrants from participating jurisdictions, potentially shifting the distribution of Hispanic immigrants, documented and undocumented alike, within the United States and possibly abroad.
While these policies have garnered significant media attention and controversy, their effects remain unclear. This analysis takes advantage of the variation across U.S. localities in both the impact of the recession and the implementation of the 287(g) program to assess their contribution to changes in the size of local immigrant Mexican population. I first investigate the unique effect of the 287(g) as a policy intervention affecting immigrants. I then elaborate on how differences in employment conditions across local areas also affected the size of the Mexican immigrant population. I pay particular attention to the particular localities that might be important for understanding overall changes at the national level.
Finally, a common rationale motivating the implementation of stronger immigration enforcement policies is the expectation that restrictions on the supply of immigrants will enhance the employment prospects of low-skilled native workers. This might be particularly salient in the context of recessions since low-skilled workers appear to be more strongly affected by economic downturns than their highly skilled counterparts, and are also expected to more directly compete with low-skilled immigrants. Thus, an anticipated outcome of immigration enforcement policies is that the reduction in the supply of foreign workers resulting from the implementation of the 287(g) program should improve the employment position of low-skilled natives. I therefore investigate the extent to which anti-immigrant policies relate to changes in the employment opportunities for low-skilled native workers.
Data, Analytic Strategy, and Model Specification
The data for the analysis come from the 2005–2009 American Community Survey (ACS). I restrict the sample to the male population between the ages of 20 and 45 to capture the prime working and mobility years. The primary geographic unit of analysis is the metropolitan area. In cases where individuals are not residing in a metro area, the geographic unit becomes the consistent Public Use Microdata Area (PUMA). I limit the analysis to places with at least an estimated 2,500 foreign-born Mexican residents in 2005 to reduce estimation variability. The end product is panel data of 161 metropolitan and consistent PUMAs, spanning 2005–2009, with estimates of population size and employment conditions aggregated for the local area. Thus, there are observations before and after the implementation of the 287(g) program and the initiation of the economic recession.
The analysis focuses on changes in the size of the Mexican foreign-born population at two time points: 2007 and 2009. In addition, I investigate changes in the unemployment rate of low-skilled white and black native workers at those two time points. The temporal and spatial variation in the panel data design is particularly well suited for difference-in-difference (DID) methods to evaluate the effect of the 287(g) program and the recession on outcomes. DID methods have become widespread in the area of policy and program evaluation. They are a type of fixed effects estimation that relies on aggregate data. The basic approach is to compare outcomes before and after a policy intervention in a treatment group and a control group. The method has been applied in a wide variety of areas, including studies on the effect of minimum wage policies on employment (Card and Krueger 1994), the impact of competition in the retail market and gas prices (Hastings 1994), and how immigrant inflows shape the employment and wages of natives (Card 1992).
In this case, signing an MOA to participate in the 287(g) program is treated as a policy intervention that affects the conditions of immigrants and natives in the localities where it is introduced. The DID approach compares the difference in outcomes before and after the introduction of the 287(g) program in the localities affected by the policy, that is, the treatment group, to the same difference for unaffected areas, that is, the control group. Average changes over time in the localities without the 287(g) program are then subtracted from average changes over time in localities with the program. This double difference or difference-in-difference removes the effect that could result from permanent differences between the two groups as well as the effect of changes over time in the intervention group unrelated to the treatment, thus substantially reducing the omitted variable problems in cross-sectional analyses. 5
The DID approach is particularly appropriate when the assignment into the treatment group is close to random, enhancing the comparability between treatment and control groups (Meyer 1995). Recent developments, however, have highlighted that the simple two-period two-group comparison can be improved by expanding the design to multiple groups as well as investigating the role of preintervention conditions and changes other than the policy intervention in affecting the comparability of groups. Such extensions can further check and refine hypotheses and allow for alternative sources of variation to be ruled out. I explore such extensions in this analysis by extending the two-group comparison and modeling change in a regression framework.
I use ordinary least squared (OLS) regression to estimate DID. The regression formulation provides a convenient way to obtain estimates and standard errors. In addition, it facilitates investigating additional comparison groups, pretreatment conditions, and changes other than the policy interventions on outcomes. 6 Since I focus on average changes over two periods, the initial approach follows a simple first difference specification. An alternative approach to policy evaluation introduces lagged dependent variables as covariates into the model. This strategy assumes that the causal effect of the program intervention is conditional not only on the permanent characteristics of the local areas but also on the lagged outcome under consideration, which makes the treatment and control groups comparable in terms of initial conditions. 7 There is considerable debate in the literature about the choice between first difference relative to lagged dependent variable specification. I report results from both approaches to assess their differences and add robustness to the results.
Variable specification
The focus of the analysis is on changes to two main outcomes: the size of the Mexican foreign-born population and the unemployment rate among the low-skilled native population between 2007 and 2009. The intervention group (P i ) includes twenty metropolitan and consolidated PUMA areas that implemented the 287(g) program before 2008. It is important to note that P i need not be constrained to two groups. As will be seen in the analysis, I investigate the particular role of specific localities in affecting the overall changes in the foreign-born Mexican population. Table 1 lists the areas included in the intervention group together with the year in which they signed their agreements with the federal government. Interestingly, no area outside the South or Southwest signed a 287(g) agreement, aside from Monmouth County, New Jersey (part of the New York metropolitan area), and the City of Danbury in Fairfield County, Connecticut, which signed on in 2009. 8
Metropolitan Areas and Counties with 287(g) Agreements Prior to 2009
One of the advantages of the regression specification is that it allows for the inclusion of controls for the role of other time-varying trends different from the policy groups in affecting the size of the Mexican immigrant population or native unemployment. Specifically, it allows me to investigate the role of other employment changes associated with the recession as well as prior population changes that might have resulted in the application of the 287(g) program on changes in outcomes. These simultaneous and prior changes confound the evaluation of the 287(g) program.
The models include the following set of controls. Since the introduction of the 287(g) program might have been influenced by population changes, especially rapid immigrant in-flows, I include as a covariate the rate of change of the Mexican immigrant population in each particular locale between 2005 and 2007, prior to the recession and policy changes.
The effect of the recession is measured by including changes in employment conditions affecting both low- and high-skilled native workers. Specifically, I include as predictors changes in the size of the native population employed in agriculture, construction, and retail, which are the main industries concentrating Mexican immigrants. In addition, I include measures of changes in the employment opportunities for highly skilled native workers, since they reflect the effect of the recession on the wider, non-immigrant-niche economy. Specifically, I include changes in the size of the native college-educated labor force employed in finance, professional, and public administration industries, as well as a measure of the unemployment rate among college-educated natives.
The final part of the article evaluates the impact of the recession and immigration policy on the change in the unemployment rate of native whites and blacks with a high school education or less. The models include the same industrial and employment conditions as predictors, but since the dependent variable is a change in rate, independent predictors are defined as changes in the percentage of the labor force employed in particular industries. Finally, to reduce variability in the estimators, I further restrict the analysis of unemployment change to areas with at least 2,500 low-skilled black and white individuals in 2005. The overlap in the size of the Mexican foreign-born and low-skilled native population restriction results in 89 and 156 geographic areas for the final analysis of blacks and whites, respectively.
Variation in the Decline of the Mexican Foreign-Born Population across Local Areas
The overall pattern of decline in the size of the Mexican immigrant population aged 20 to 45 from 2007 to 2009 masks considerable variation across local areas. Descriptive results plotting the distribution of the dependent variable show that the majority of local areas witnessed a modest decline in Mexican immigrant populations, with the average being a drop of 1,615 men between 2007 and 2009. Positive change is evidenced only in 32 percent of cases. The highest gains occurred in Miami-Hialeah, Florida, where the number of foreign-born Mexicans increased by 4,446 during the period. However, the description also shows four clear outliers that experienced dramatic declines in their Mexican foreign-born population, specifically Los Angeles–Long Beach, California (–40,701); Riverside–San Bernardino, California (–34,776); Phoenix, Arizona (–26,991); and Dallas–Fort Worth, Texas (–14,964). All these areas implemented 287(g) programs during the period. Without these outliers, the distribution is very close to normal. As will be seen below, these outliers are critical to evaluating the effect of the 287(g) on the size of the Mexican immigrant population.
To further investigate the role of the 287(g) intervention, Figure 2 traces the average size of the foreign-born Mexican population, separating the local areas by whether they implemented the 287(g) program. The two-group comparison shows that for both groups, the immigrant Mexican population grew in the years leading up to 2007 and fell afterward. The decline, however, was particularly dramatic in the group with the 287(g). Between 2007 and 2009, the Mexican immigrant population declined by more than 7,000 in areas with the 287(g) program, on average, compared to declines of less than 1,000 in areas without it.
Size of the Mexican Male Immigrant Population According to 287(g) Status of Local Areas: 2005–2009
Since the treatment and control groups differ in the size of their initial immigrant populations, and since larger starting populations imply greater potential for declines, I next expand the two-group typology to account for initial population size in two ways. First, I distinguish within policy groups between areas with more or fewer than 50,000 Mexican foreign-born residents in 2005, which results in a four-group typology. Second, I expand the four-group classification by separating out outliers identified in the description of the dependent variable, all of which were 287(g) implementers with large initial Mexican immigrant populations. This results in five distinct groups: small initial population / no 287(g) (n = 134); small initial population / with 287(g) (n = 13); large initial population / no 287(g) (n = 7); large initial population / with 287(g) but not outlier (n = 3); and large initial population / with 287(g) but outlier (n = 4). The expansion directly investigates the comparability between groups as highlighted in recent developments in the literature on program evaluation.
Table 2 reports the DID estimates of the average decline in the Mexican foreign-born population between 2007 and 2009 according to the two-, four-, and five-group typology. Confirming the pattern shown in Figure 2, the most basic two-group comparison shows that the Mexican immigrant population in areas with the 287(g) program declined by an additional 6,488 relative to areas without the program. However, estimates vary considerably when I separate the groups by initial size of the immigrant population. The four-group typology shows that among areas with small initial immigrant populations, those with the 287(g) program (group 4) exhibit a slightly greater (–710) average reduction in their Mexican immigrant populations than areas without the program (group 3).
Difference-in-Difference Estimates of the Average Change in Mexican Male Immigrant Population by 287(g) Status: 2007–2009
Comparing larger immigrant areas with small areas without the program (group 3) shows that between 2007 and 2009, the Mexican foreign-born population in large areas without the 287(g) (group 5) declined by 2,982. There are seven areas in this category: Chicago, Illinois; Denver-Boulder, Colorado; Fresno, California; New York–Northeastern New Jersey; San Diego, California; San Francisco, California; and San Jose, California. These areas tend to be referred to among immigration opponents as “sanctuary” cities since they do not openly enforce immigration controls. Results from the four-group typology also show that the decline was considerably more pronounced in large areas with the 287(g) program (group 6). As compared to small areas without the program, the DID estimator shows a decline of 17,642 in large areas with the policy intervention.
However, interesting results are obtained when I separate the outliers from other large immigrant areas with the 287(g) program in the bottom panel of Table 2. Three areas fall under the large initial population with 287(g) but not outlier (group 7): Atlanta, Georgia; Houston, Texas; and Las Vegas, Nevada. Among these areas, the DID estimator shows their Mexican foreign-born population declining by 2,902 relative to small areas without the program (group 3), which is almost identical to the –2,982 DID estimator obtained for large areas without the 287g program (group 5). In fact, if one takes areas with large Mexican populations and no 287g program (group 5) as the reference group, the DID estimate for nonoutliers (group 7) is in fact positive, 80.
Most of the decline in the Mexican foreign-born population related to the 287(g) program is accounted for by the four outliers (group 8). The DID calculations show that for this group, the Mexican immigrant population declined by 28,697 and declined by 25,714 relative to small and large areas without the 287(g) program. Even compared to other large areas with the program, the decline in these four outliers was 25,795 higher. The salience of these four outliers (Los Angeles, Riverside, Phoenix, and Dallas) in explaining trends in the size of the Mexican foreign-born population is a recurrent finding in the study.
One possibility is that the impact of the collapse of the housing market was more pronounced in these four outliers than in the rest of the metropolitan areas with large Mexican immigrant populations. To investigate the issue, Table 3 lists changes in the size of the foreign-born Mexican population in the fourteen areas with more than 50,000 foreign-born Mexican men in 2005 together with an indicator of participation in the 287(g) program and absolute changes in the number of native workers employed in the construction industry. The highlighted rows indicate the outlying areas in terms of foreign-born change.
Change in Mexican Male Immigrant Population and Construction Employment in Large Metropolitan Areas: 2007–2009
Outliers are indicated in bold.
Results document considerable lack of correspondence between changes in the size of the foreign-born Mexican population and construction employment. The metropolitan area with the largest losses in construction is Chicago (–29,972); however, it lost only 4,433 Mexican immigrant men between 2007 and 2009. At the other extreme, Los Angeles–Long Beach lost considerably more Mexican immigrants (–40,701) than construction jobs for natives (–12,678). A similar pattern is evident for the neighboring Riverside–San Bernardino metropolitan area. The difference is particularly dramatic as compared to New York, which lost a similar number of construction jobs but only 357 Mexican immigrants.
Dramatic disparities are also evident for the comparison between Phoenix and Denver-Boulder. Both metropolitan areas lost close to 8,000 jobs in the construction industry. However, Phoenix saw its Mexican immigrant population decline by almost 27,000 compared to only 3,715 in Denver. San Francisco and Dallas–Fort Worth also present contrasting images. Both metropolitan areas lost close to 5,000 jobs in construction, but the decline in the Mexican foreign-born population was 14,964 in Dallas, compared to a 597 increase in San Francisco. Overall, results document that the outliers are not unusual in terms of the impact of the recession on construction employment and that the change in the size of the Mexican immigrant population does not perfectly correlate with economic conditions, issues that will be explored more systematically in the multivariate analyses.
Regression Results: First Difference and Lagged Dependent Variable Models
The next set of analyses model group differences in changes in the size of the foreign-born Mexican population, controlling for prior population conditions and changes in employment opportunities associated with the recession. Table 4 reports results from OLS first difference and lagged dependent variable models predicting changes in the size of the local foreign-born Mexican population. Following the descriptive findings, the top panel reports results from the two-group specification, and the bottom panel reports results obtained from the expansion to five groups.
OLS Estimates from Fixed Effects and Lagged Dependent Variable Models Predicting Change in Mexican Male Immigrant Population
p <.05
The model in column 1, which does not include covariates, reproduces the DID estimates reported in Table 2. Adding the changes in employment conditions associated with the recession as predictors in column 2 reduces the overall effect of the 287(g) program by 29 percent, from –6,488 to –4,630. The lagged dependent variable results in column 3 show that even after conditioning the estimates on the prior size of the immigrant population, areas that implemented the 287(g) program reduced their Mexican foreign-born populations by 1,870. In the lagged dependent variable specification, which relates changes in size of the Mexican immigrant population to the prior size of the group in addition to policy changes, the effect completely disappears after controlling for changes in employment conditions associated with the recession (column 4).
Results for the five-group specification reported in the bottom panel of Table 4 adds further precision to the role of policy and the economy in affecting changes in the foreign-born Mexican population. As with the two-group case, column 1 reproduces the DID results reported in Table 2. Small areas with the 287(g) program are no different from small areas without it, but all other area types show significantly larger losses, with the largest deficit by far registered among the large 287(g) outliers, as would be expected. After accounting for socioeconomic conditions, large areas without the policy intervention reduce their immigrant population by 3,196 (column 2). The effect is actually higher among large areas with the 287(g) that are not outliers in the change distribution. This group saw its immigrant Mexican population decline by 2,290. Again, the four outliers experienced a ten-times-larger decline in their Mexican immigrant populations (–27,030) than either of the other large area groups.
Results from the lagged dependent variable specification show that controlling for prior size of the Mexican population eliminates the effect of the 287(g) policies for all except the four-outlier group. Accounting for the negative effect associated with prior population size, the Mexican foreign-born population declined by 20,919 among the group that had large initial immigrant populations and 287(g) but were not outliers (column 3). Accounting for other socioeconomic changes reduces the effect of the group by 25 percent to 18,962 (column 4). Thus, even relating changes in the size of the foreign-born Mexican population to prior conditions in addition to policy does not eliminate the effect of the outliers.
The role of socioeconomic changes in affecting the size of the immigrant Mexican population is consistent across specifications. The main factor reducing the decline is changes in construction employment. Each additional native worker employed in construction reduces the decline in the Mexican foreign-born population by 0.15 to 0.28 persons (columns 2 and 4), depending on specification. Interestingly, results show that the size of the Mexican immigrant population is also affected by changes in employment conditions among natives in highly skilled industries. Specifically, places with larger expansions of the college-educated native population employed in professional industries lost fewer Mexican immigrants than areas that did not gain highly skilled natives. At the same time, more pronounced growth in the unemployment of college-educated natives was associated with larger drops in the Mexican immigrant population (–385 in the bottom panel of column 2). These effects support perspectives that stress the role of labor demand processes, even stemming from changes in employment conditions among the high-skilled native population, in affecting the size of immigrant groups.
Changes in the Unemployment Rate of the Low-Skilled Native Black and White Populations
I next explore the possibility that immigrant enforcement measures may enhance native employment prospects by examining changes in the unemployment rate of low-skilled black and white workers between 2007 and 2009. Estimates from the ACS show the unemployment rate for native black and white men with a high school education or less increased from 17.6 to 26.8 percent and from 8.0 to 14.4 percent between 2007 and 2009, respectively. As with the changes in the size of the immigrant Mexican population, we can expect considerable variation across geographic areas in changes in the unemployment rate.
Table 5 reports results from OLS models predicting change in the unemployment rate of black and white low-skilled native workers from 2007 to 2009 across local areas. The change models follow the specification applied in models predicting change in the foreign-born Mexican population. Since the dependent variable is change in rates, the explanatory covariates are measured as percentage of the labor force, not absolute values. As before, I report both first difference and lagged dependent variable estimates.
OLS Estimates from Fixed Effects and Lagged Dependent Variable Models Predicting Change in the Low-Skilled Black and White Men Unemployment Rate
p <.05
In all cases, results show no effect of the five-area typology on the changes in the unemployment rate of low-skilled native black and white men. One might have expected that given the strong effect of the four area outliers in reducing the local size of the foreign-born Mexican population, the strong enforcement of 287(g) policies would have facilitated the expansion or at least buffered the deterioration of employment opportunities for low-skilled native workers. However, there is no evidence of such an effect. Similar lack of effects is obtained when estimating the models using the simpler two- or four-group typology.
Similarly, the rate of growth of the Mexican immigrant population between 2005 and 2007 also shows no effect on the change in the native unemployment rate resulting from the recession. The main predictor of the impact of the recession on natives is the trend in construction employment. Areas that enjoyed smaller declines (or gains) in construction employment saw smaller increases in native unemployment than areas in which construction contracted more sharply. Since conditions in the local construction industry also affected the size of the Mexican immigrant population, this suggests that the employment prospects of both native and foreign workers are connected. Rather than evidence of competition, this supports a view of foreign and native workers as complements directly affected by the decline in construction employment resulting from the recession.
Conclusion
Estimates from the ACS document a considerable decline in the size of the U.S. Mexican immigrant population since 2007. This is perhaps not surprising given the dramatic increase in immigration control at the federal level, resulting in far more deportations and lower in-migration flows. However, there is considerable variation across areas within the United States in changes in the size of the immigrant Mexican population. Two processes are potentially salient in encouraging immigrants to relocate either within the United States or abroad: the enactment of local immigration control provisions and the variable impact of the economic recession across local areas. In this analysis, I evaluated the relative impact of more stringent immigration enforcement policies—namely, the 287(g) program—and employment change associated with the 2007 recession on the size of local Mexican immigrant populations between 2007 and 2009. In addition, I investigated the extent to which immigration policies and economic conditions affected the unemployment rate of non-Hispanic black and white low-skilled natives.
In investigating the effect of the 287(g) program on changes in the immigrant Mexican population, several conclusions are evident. Overall, the 287(g) is not particularly effective at reducing the local Mexican population. The lack of effect is particularly evident once the initial size of the immigrant population is taken into account. Among areas with small Mexican immigrant populations, 287(g) implementation was not associated with larger reductions in immigrant populations over time. And among areas with large immigrant concentrations, 287(g) programs were associated with significant declines in Mexican immigrant populations in only four localities, namely, Los Angeles, Riverside, Phoenix, and Dallas. These areas experienced a dramatically larger reduction of their Mexican immigrant population than other communities of similar size that also implemented the 287(g) program, such as Atlanta, Houston, and Las Vegas. Interestingly, among this last set of areas, the reduction in the Mexican immigrant population was actually smaller than that experienced by other large communities that as of 2008 had not implemented the program, such as Chicago, Denver, Fresno, New York, San Diego, San Francisco, and San Jose. Thus, aside from four outliers, there is no evidence that 287(g) implementation reduced Mexican immigrant populations at the local level.
The economic recession, on the other hand, was highly effective in reducing the size and growth of the Mexican immigrant population. Deteriorating employment opportunities, particularly in the construction industry but also in high-skilled sectors, had a pronounced effect on the Mexican immigrant population of local areas. However, variation in employment opportunities across local areas does not account for the outlier status of Phoenix, Los Angeles, Riverside, and Dallas. Accounting for the employment trajectories of these locales reduces the effect of the 287(g) program on Mexican immigrant populations by only 6 percent. This reinforces the interpretation that what distinguishes these outliers from the rest of the United States is the particular form of 287(g) implementation and not the economic impact of the recession.
The final part of the article evaluated the proposition that removing immigrants would result in better employment prospects for native low-skilled workers. The analysis found no effect of the 287(g) program on changes in the unemployment rate of low-skilled black and white workers. Even among the four outliers, I found no evidence that accelerated reduction of the foreign-born improved economic opportunities for natives. The key determinant of native unemployment was, not surprisingly, local economic conditions. Changes in employment in the construction industry, not the presence or absence of immigrants, explained the lion’s share of local variation in native unemployment for both blacks and whites. The fact that both the size of the immigrant population and native employment are sensitive to conditions in the construction industry undermines the argument that immigrant competition fuels low-skill native unemployment.
These findings have numerous policy implications. Overall, they indicate that merely implementing a program like the 287(g) alone is not sufficient to secure attrition in the Mexican foreign-born population. Rather, the experience of the outliers suggests that the policies are successful only if they are enforced with extreme vigor. Indeed, the outlier communities have often received extensive media coverage at the local, national, and even international level for the severity of their program implementation. In Phoenix, Maricopa County Sheriff Joe Arpaio, who promotes himself as America’s toughest sheriff, is well known for his outspoken stance against immigration and frequent high-profile raids used to round up suspected undocumented immigrants. His publicity stunts include parading Mexican men awaiting deportation in front of TV cameras dressed in pink underwear and pink handcuffs that are available online for purchase (New York Times 2009). His practices have been the focus of several Justice Department investigations for alleged civil rights violations and have been sharply criticized for diverting resources away from basic law enforcement, even by respected conservative groups such as the Goldwater Institute (Bolick 2008). In spite of the many controversies, they still became a model for Arizona’s controversial anti-immigrant legislation known as SB1070 (which was also closely shaped by the private prison industry 9 ).
Likewise, the sheriff of Los Angeles County, Lee Baca, is another vocal advocate of a get-tough approach to immigration. Baca has opposed many other sheriffs in the state (such as those of Yolo, Sacramento, San Francisco, and Santa Clara counties) and even the local Los Angeles City Council who advocate limiting local participation in deportation programs (Los Angeles Times 2011) and instead has worked to expand the use of his police for immigration enforcement. He is currently facing an FBI investigation for inmate abuse and has publicly argued that immigrants are not entitled to the basic civil rights protections guaranteed to American citizens (CBS Los Angeles 2011).
Another implication is that even when applied with draconian measures, the attrition through enforcement strategy is not very effective at eliminating the undocumented worker problem. For instance, if all fourteen metropolitan areas with large Mexican immigrant populations in 2005 were to become as effective at reducing their immigrant populations as the four 287(g) outliers in this analysis, we could expect an average of 190,000 (27,000 × 14/2) fewer Mexican male immigrants in these areas annually. If one made the highly unlikely assumptions that all immigrants who left these areas returned to Mexico, voluntarily or through deportations, and that undocumented migration into the United States would cease, it would still take almost 30 years to remove all of the estimated 5 million undocumented Mexican men currently estimated to reside in the United States. Thus, even under this highly unlikely, “best-case” scenario, it would still take an inordinate amount of time (not to mention resources) to eliminate the undocumented population in the United States.
Given the lack of obvious economic benefits to natives from such policies, even in their most severe applications, it is particularly sobering that the current drift of our immigration policy continues to be toward “attrition through enforcement.” Examples at the state level include the Arizona SB1070 legislation and the recently enacted Alabama anti-immigration law that went so far as to require public schools to verify the legal status of their pupils, leading to a large-scale withdrawal of Hispanic children from local schools. At the federal level, the Obama administration initiated the Secure Communities program in 2008, a program that builds on the 287(g) experience and relies on partnership among federal, state, and local law enforcement agencies to identify deportable aliens. A main objective is to eliminate the “sanctuary” areas that have resisted pressures to participate in immigration enforcement. However, the Northeast and Midwest, which have struggled since the 1970s with anemic population growth, have received an important boost in recent years from immigrants, who have contributed to demographic vitality, neighborhood revitalization, and small business formation throughout the region. This analysis shows that while slow, the extreme application of these programs does in fact reduce the local immigrant population. It is not surprising, then, that states such as New York, Illinois, and Massachusetts have tried (so far unsuccessfully) to withdraw from the Secure Communities program. For these areas, enforcing immigration laws might go against their demographic interests.