Undocumented Migration and the Residential Segregation of Mexicans in New Destinations

Abstract

This study uses data from the 2000 Census and 2005–2009 American Community Survey to examine the impact of undocumented Mexican migration to new destinations on residential segregation between Mexican immigrants and native-born whites and native-born blacks. We find that Mexican-white and Mexican-black segregation is higher in new Mexican gateways than in established areas and that, for Mexican-immigrant segregation from whites, this heightened level of residential segregation in new destinations can be explained by the high presence of unauthorized Mexican immigrants living there which tends to bolster segregation between the two groups. By contrast, Mexican-immigrant segregation from native-born blacks tends to be lower in areas with larger undocumented populations, a pattern that is especially true in new destinations. Neither of these opposing effects of legal status on Mexican-immigrant segregation can be explained by compositional differences in assimilation (English ability and earnings) between documented and undocumented immigrants nor by structural variation in metropolitan areas, suggesting a unique association between legal status and segregation.

Over the past four decades, America has been witness to a demographic transformation brought about by enormous growth in the number of immigrants. While the U.S. foreign-born population is much more diverse than popularly perceived, Mexican immigrants accounted for more than a quarter of all arrivals since 1970 and currently compose nearly a third of all immigrants living in the U.S. This momentous and sustained growth in the Mexican population has had profound consequences for American social, economic, and political life, but perhaps none as visible as those occurring in U.S. neighborhoods. The arrival of Mexican and other immigrants has been credited for reductions in multigroup segregation (Iceland 2009) and in the declining isolation of both white and black Americans (Logan et al. 2004; White and Glick 1999). The segregation of Latinos, however, has been stubborn to change, remaining relatively stable or even increasing (Logan and Stults 2011). Timberlake and Iceland (2007) even foresee a future in which Latinos will overtake blacks as the most segregated group in America.

The recent dispersal of Latinos and Mexicans out of gateway cities and into non-traditional destinations has raised questions about whether the durability of Mexican segregation can be linked to their emergence in communities unfamiliar to Mexican faces and immigrant issues. While not entirely conclusive, there is growing evidence that Mexicans are more highly segregated in these new destinations than in more established ports of entry (Fischer and Tienda 2006; Hall 2013; Lichter et al. 2010; but see Alba et al. 2010; Park and Iceland 2011). Existing research, however, has been unable to explain the heightened levels of segregation in emerging destinations. One critically-important omission for understanding residential inequality among immigrants is recognition that Mexican immigrant populations in new destination areas are composed heavily by individuals lacking legal authorization. Current estimates indicate that there are more than 6 million undocumented Mexicans in the U.S. (Passel et al. 2013); and while a majority of them live in traditional immigrant-receiving states, undocumented migrants as a percent of the local Mexican immigrant population are more strongly represented in states and communities outside of traditional gateways (Massey et al. 2010; Passel and Cohn 2009a).

Due in part to the fears associated with being identified as an unlawful resident and to potential challenges in securing housing, undocumented immigrants may be more likely to congregate in segregated, ethnic communities where they are better able to hide away in the shadows of their peers or to tap into ethnic-based networks that ease the acquisition of housing. Similar processes of occupational segregation have been reported in studies of undocumented immigrants’ economic well-being (see Donato and Armenta 2011; Flippen 2012; Hall et al. 2010), but research assessing how legal status influences residential sorting within housing markets is lacking. This potential for stratification by legal status may be especially pronounced in new destinations where the arrival of Mexicans has sometimes been met with nativism, political backlash, and native flight.

Evaluating the connection between undocumented migration and neighborhood inequality is important not only because it sheds light on potentially new forms of residential stratification, but because patterns of segregation often correspond with broader disparities and can be used as a lens through which to view the pace of Mexican incorporation into the mainstream (Iceland 2009; Telles and Ortiz 2008; White and Glick 2009). It is also relevant to ongoing discussions about the changing terrain of race/ethnicity, particularly whether Mexicans and other Latinos will fit within the existing hierarchy or will forge a new one (Lee and Bean 2010). Thus, an understanding of how undocumented migration influences residential patterns can provide insight into the evolving shape of the American color line.

Our aim in this study is to examine the connection between legal status and residential inequality by exploring how undocumented migration is related to patterns of segregation among Mexican immigrants. Guiding our efforts are four key questions: First, are Mexican immigrants more segregated from native whites and native blacks in new destinations than in established ones? Second, can the share of unauthorized migrants among local Mexican immigrant populations explain observed differences in segregation between new and established destinations? Third, does undocumented migration impact segregation differently in new gateways than in established ones? Lastly, does undocumented migration influence Mexican-immigrant segregation from native whites and native blacks in different ways? To answer these questions, we use tract- and metropolitan-level data from Census 2000 and the 2005–2009 American Community Survey to estimate levels of and trends in Mexican-immigrant dissimilarity from native-born whites and native-born blacks, linked to metro-level estimates of the undocumented population derived from the 2000 Census and 2006–2008 ACS Public Use Microdata Samples.²

Background

Over the last several decades, Mexican migration to the U.S. has surged. In California alone, the Mexican (both foreign- and U.S.-born) population grew from 1.1 million to 11.4 million and their share of all Californians rose from 5.5% to 30.7%, between 1970 and 2010. However, the rapid growth in this population was especially pronounced in areas further from the border. For example, the 1960 Census recorded only 1,562 Mexicans in Georgia, but by 2010 more than half a million called the state home. Thus, while the Mexican population in the U.S. remains concentrated in areas with long-standing ties to Mexico, it is considerably more dispersed than in previous times.³ The redistribution of Mexicans, and other immigrant groups, has motioned in a substantial research literature focused on “new destinations” (Go dziak and Martin 2005; Lichter and Johnson 2009; Light 2006; McConnell 2008; Marrow 2011; Massey 2008; Singer et al. 2008; Zúñiga and Hernández-León 2005) and the implications of their emergence for immigrant incorporation and racial/ethnic dynamics.

Why exactly the great dispersion of Mexican immigrants occurred remains uncertain, but a complex array of processes was likely at work. Some scholars have drawn attention to the need for low-skill laborers in non-traditional areas resulting from the restructuring of labor intensive industries (Hernández-León and Zúñiga 2000; Kandel and Parrado 2005; Parrado and Kandel 2008) or from the demands created by native population growth (Kaushal and Kaestner 2010). Others have focused on the saturation of labor pools in established areas (Durand et al. 2005; Hernández-León 2008), elevated fertility rates and limited opportunities for work in Mexican states with little prior history of migration to the U.S. (Hernández-León 2008; Riosmena and Massey 2010; Weeks, Stoler, and Jankowski 2011), the militarization of major border entry points (Massey et al. 2002; Orrenius 2004), and municipal regulations governing employment and housing (Light 2006).

Residential Segregation in New Destinations

Regardless of the underlying explanation, the dispersion of Mexicans has been described as a broad-scale process of regional desegregation (Massey and Capofero 2008; Mouw and Sharma 2009), one that resulted in the influx of Mexicans to communities throughout the American landscape. Yet, it does not appear that the integrative forces operating at the regional level are underway at more local ones. Parisi and Lichter (2007) first reported that emerging Latino destinations throughout the southeastern U.S. were marked by high levels of residential segregation. Using an expanded set of both urban and rural communities, they confirmed that Latino segregation from whites is especially high in areas where the Latino population grew rapidly during the 1990s and that these elevated levels could not be explained by characteristics of local Latino populations nor by structural features of the communities themselves (Lichter et al. 2010). Similarly, Hall (2013) finds that among those in the 100-largest metros, Mexicans immigrants are more highly segregated in new than established destinations, and neither acculturation nor socioeconomic characteristics of local Mexican populations can explain the observed difference (see also Fischer and Tienda 2006). Findings reported by Park and Iceland (2011) are somewhat different: they find that Latino immigrant segregation is slightly lower in new immigrant destinations, but that patterns of Latino segregation increased in these areas during the 1990s, while staying stationary in established areas. The unevenness in some of these findings likely reflects differences in the classification schemes used to distinguish areas as “established” or “new” and differences in the geospatial units used in calculating segregation.⁴ Despite some of these inconsistencies, the growing consensus is that the redistribution of Latinos away from traditional ports of entry has not translated into Latino residential integration.

Two general theoretical models have been developed to explain differences across areas in residential segregation: spatial assimilation and place stratification. The spatial assimilation model argues that new immigrants initially band together in dense ethnic enclaves but, as they accumulate economic resources and become more familiar with American institutions, norms, and values, seek out newer neighborhoods where not just schools, public services and housing are superior, but where majority group members (i.e., whites) are more plentiful (Massey 1985). From an empirical standpoint, the assimilation perspective maintains that groups should be less segregated when they possess characteristics that are associated with residence in majority-group neighborhoods, such as higher incomes, greater English proficiency, and longer durations in the U.S. Evidence of assimilation in the spatial patterning of Mexican immigrants is substantial: they are significantly less segregated in areas with fewer recent arrivals, where their earnings are closer to whites, and where higher shares speak English (Hall 2013; Iceland and Nelson 2008; White and Glick 2010).

In contrast to the assimilation model, the place stratification perspective draws attention to barriers and constraints that minorities face in navigating American housing markets, namely those created by discriminatory practices of banks, lenders, and real estate agents (Ross and Turner 2005; Squires and Kim 1995; Yinger 1995). Perhaps more salient to new immigrants, the stratification perspective also highlights how racial preferences maintain neighborhood separation. Survey research finds that whites remain reluctant to share neighborhoods with even moderately-sized minority populations (Charles 2006; Krysan 2002). No empirical work that we are aware of has specifically examined natives’ residential preferences toward immigrants; however, research indicates that whites typically have less averse attitudes toward Latino neighbors than black ones, though most whites still indicate that they would not move into majority-Latino neighborhoods (Charles 2001; Emerson et al. 2001; but see Lewis et al. 2011). Blacks, on the other hand, tend to express higher preferences for integrated neighborhoods (Krysan and Bader 2007; Krysan and Farlay 2002), but have weaker preferences for Latino neighbors than whites (Charles 2006).

While the place stratification model has conventionally focused on racially-based constraints that limit opportunities for minority groups – racial discrimination in housing and lending practices, racially-charged residential preferences, and animosity or hostility toward particular groups – it is relevant to other (non-racial) forms of discrimination, or what some refer to as “horizontal exclusion” that restrict opportunities or marginalize groups based on citizenship and state-granted rights (see Kim 1999; Marrow 2009a; Telles and Ortiz 2008). This is particularly relevant to Mexicans who compose more than half (52%) of the 11.7 million unauthorized immigrants living in the U.S. (Passel et al. 2013). Indeed, this population is so large that over half (55%) of Mexican immigrants lack documentation to live and work in the U.S. (Passel and Cohn 2009b). Expanding the stratification model to incorporate marginalization based on citizenship does not disregard racial exclusion. Indeed residential stratification of Mexicans potentially operates along both racial and legal dimensions (see Chavez 2008; Massey and Pren 2012a, 2012b). Thus, evidence of undocumented migration leading to heightened Mexican segregation is theoretically consistent with the stratification model.

The inattention to legal status in residential research is a major omission given that the heavy presence of undocumented Mexicans could have potentially profound consequences for residential sorting if unauthorized immigrants are more likely to live in segregated neighborhoods than their legal counterparts. There are several reasons for why this may be true. First, undocumented immigrants may possess certain characteristics that make them more reliant on or confined to ethnically-dense neighborhoods, such as low socioeconomic status or poor language skills. A sizeable body of research has documented the low earnings and other barriers to gainful employment that undocumented Mexicans face (Donato and Massey 1993; Hall et al. 2010; Kossoudji and Cobb-Clark 2002; Rivera-Batiz 1999). Thus, high levels of segregation among undocumented immigrants may simply be a compositional artifact; that is, unauthorized immigrants are only more segregated because they possess traits conducive to enclave residence, not because of their lack of legal status.

Alternatively, a direct effect of legal status may be observed if real or perceived fears of deportation or imprisonment lead undocumented Mexicans to cluster in ethnic neighborhoods where their specific status can be concealed and exposure to native populations and law enforcement avoided. These fears of the Mexicans seem to be real: about one in ten Latinos immigrants report being stopped and asked about their immigration status by police or other authorities, and a majority worry “a lot” about deportation (Lopez and Minushkin 2008). Not only do these fears appear to be real, but they also seem to be well founded. Recent research has pointed to police profiling of Latinos (Gelman et al. 2007) and to their disparate treatment when stopped (Alpert et al. 2007). Moreover, the incidence of profiling appears to be dramatically lower in segregated, minority neighborhoods (Gelman et al. 2007; Tomaskovic-Devey and Warren 2009). Aside from the fear of discovery, undocumented immigrants may well face logistical hurdles in obtaining housing, such as lacking a bank account or valid social security card which is required not only to obtain credit profiles, but also to activate housing utilities (e.g., electricity, gas, water). If these barriers are more easily overcome in co-ethnic neighborhoods, undocumented immigrants may be more likely to settle there.

The limited body of research that has addressed this issue offers some clues that legal status does in fact impede residential attainment. Undocumented Mexicans, for example, have lower levels of homeownership (Hall and Greenman 2013; McConnell and Akresh 2008; Paral 2008) and spend more on housing (McConnell 2012, 2013) than their documented counterparts. Recent research also finds that undocumented Latinos tend to live in lower quality housing units and less desirable neighborhood settings (Cort 2011; Hall and Greenman 2013). Thus, there are empirically-supported, theoretical reasons to suggest that legal status plays a primary role in the residential attainment of Mexican immigrants.

Crucial for this research is that undocumented migrants are not distributed evenly across American communities. While like their legal counterparts, undocumented Mexicans are concentrated in established points along the U.S.-Mexico border; about 4 in 10 unauthorized Mexicans live in California and nearly two-thirds reside in a border state (Massey et al. 2010). Yet, despite their clustering in major gateways, many of the forces that prompted the geographic shift in the Mexican population – strong demand for low-skill labor, especially in industries with little regulatory oversight (Johnson-Webb 2003; Kandel and Parrado 2005; Leach and Bean 2008; Mize and Swords 2010), and changes in border enforcement policies (Massey and Capoferro 2008) – led to migration streams into many non-traditional areas dominated by undocumented migrants. Indeed, growth in the unauthorized population was especially high in emerging destinations between 1990 and 2010, outpacing growth rates in traditional areas by a factor of 2.6 (Passel and Cohn 2011).⁵ This has led some to describe the Mexican dispersal as one driven largely by undocumented migrants (Riosmena and Massey 2010). And, because a disproportionate share of Mexican population growth in new areas was due to growth in the unauthorized population, the portion of the Mexican immigrant population that is undocumented in newly-emerging destinations is especially high. Recent case studies of new destinations have reached similar conclusions about the size of the unauthorized population among local Mexican populations (Flippen 2012; Kasarda and Johnson 2006; Marrow 2011).

Fears and obstacles to residential attainment are potentially intensified in new destination areas where reception contexts may unwelcoming (or perceived to be) because of either anti-immigrant sentiments expressed by locals or to local policies seeking to discourage undocumented migration (see Boushey and Luedtke 2011; Marrow 2009b; Shahani and Greene 2009; Winders 2012). If these policies increase anxiety about lacking documentation either directly (e.g., Alabama’s HR56 or local 287[g] programs allowing police officers to verify residency) or indirectly (e.g., housing occupancy ordinances), undocumented immigrants may be more likely to congregate in co-ethnic neighborhoods where contact with natives can be avoided. While there is heterogeneity in community responses to immigration, there has been a clear rise in anti-immigrant policies in non-traditional locations undergoing rapid growth in their foreign-born populations (Hopkins 2010; O’Neil 2010). If these restrictions create additional logistical barriers for undocumented immigrants’ residential attainment or convey that undocumented immigrants are not welcome in an area, then the salience of legal status is likely to be pronounced in new destinations. Consistent with these arguments, Rugh and Massey (2013) find that Hispanic segregation is magnified in areas with high levels of anti-Latino sentiment. Analytically, these arguments suggest that destination type (whether an area is a new or established destination) moderates the relationship between undocumented migration and segregation.

Our analysis considers these different theoretical arguments, but because of our use of contextual data on the concentration of undocumented immigrants in metropolitan areas, we are unable to make conclusive statements about the extent to which undocumented immigrants are residentially segregated from other groups. Despite this cautionary note, we do not have any theoretical reasons to suspect that authorized Mexican immigrants are highly segregated in metropolitan areas with large shares of unauthorized migrants, while unauthorized migrants are not. Nevertheless, readers should be mindful of the ecological nature of the data being analyzed while interpreting the results.

Data and Methods

To estimate patterns and trends in Mexican segregation, we use data from the summary files of Census 2000 and the 2005–2009 American Community Survey.⁶ Census tracts serve as geographic subunits in this analysis. While they are imperfect operationalizations of neighborhoods, they are assumed to do a better job of approximating the usual conception of neighborhood than any other spatial entity provided by the Census Bureau (White 1987).⁷ Metropolitan areas are defined using the standards set by the Office of Management and Budget in 2008.

Using these files, we extract information from 52,943 tracts in all 366 metropolitan areas, excluding tracts at the start or end of the decade with no population.⁸ We measure segregation using the dissimilarity index, a measure of distributional evenness and the mainstay of segregation analysis:⁹

$$D_{jm}=\frac{1}{2}\sum _{t=1}^T|\frac{p_{t,j}}{P_j}-\frac{p_{t,k}}{P_k}|$$

where t refers to tracts within an MSA, j and k to population groups, p_t,j to the population of group j in tract t, and P_j to the total population of group j in metro m. The index ranges from 0 (no segregation) to 1 (total segregation), and can be interpreted as the proportion of one group that would have to relocate in order for each neighborhood (tract) to have the same racial/ethnic distribution as the broader metropolitan area. We use the dissimilarity index to gauge the residential distribution of Mexican immigrants from two groups: native-born, non-Hispanic whites (hereafter “native whites”) and native-born, non-Hispanic blacks (hereafter “native blacks”).¹⁰ For ease of interpretation, we multiply all D scores by 100. In order to alleviate bias due to geocoding and random error (which will influence small populations more than large populations), we limit the calculation of D scores to metropolitan areas containing at least 1,000 members of both relevant groups (e.g, for Mexican-black D, a metro area must have 1,000 Mexican immigrants and 1,000 native blacks).¹¹

Our two primary explanatory variables are destination type and the share of Mexican immigrants who are unauthorized. Data from the 1980, 1990, and 2000 decennial censuses and the 2005–2009 ACS are used to map the settlement histories of Mexicans into all metros.¹² The categorization that we use here follows related work (e.g., McConnell 2008, Suro and Singer 2002), especially that of Lichter and colleagues (Lichter and Johnson 2009; Lichter et al. 2010). Specifically, we define Established Mexican destinations as metropolitan areas in which the Mexican immigrant percent of the total population surpassed the national average in every census year 1980–2009. New destinations are metropolitan areas with a smaller percent Mexican immigrant than the national average in 1980 and that met one of the following two sets of criteria: 1) contained at least 10,000 Mexican immigrants, whose share of the total population exceeded the national average in 2005–2009, and growth of the Mexican-immigrant population during the 1980s, 1990s, or 2000s that was at least 1.5 times the national average; or 2) had at least 5,000 Mexican immigrants in 2005–2009 and Mexican immigrant growth rates during the 1980s, 1990, or 2000s that were at least 3 times the national average.¹³

Our other key variable attempts to gauge the percentage of a metropolitan area’s Mexican population that is unauthorized. To do so, we use the 2000 Census and 2006–2008 ACS Public Use Microdata Samples. We start with all adult (18+) Mexican immigrants and using the methodologies outlined by Passel et al. (2004) and Hall et al. (2010) as foundations, we identify respondents who meet each of the following criteria: 1) they are noncitizens¹⁴; 2) because at least 75% of currently unauthorized Mexican immigrants arrived since 1990, we require that they came to the U.S. during the 1990s or 2000s; 3) they have no more than a high school education, given that a large majority (∼96%) of undocumented Mexicans are not college educated; 4) to prevent the inclusion of those on student visas, they are not currently enrolled in school; and 5) because undocumented workers are ineligible to work as civil servants, they are not employed by a local or state, or the federal government. To account for the undercoverage of undocumented immigrants, we adjust our estimates assuming an undercount of 15% (Costanzo et al. 2001; Robinson 2001; Van Hook and Bean 1998a, 1998b).¹⁵ At the national level, this approach estimates the undocumented Mexican adult population to be 5.52 million in 2006–2008, who make up 52.9% of all Mexican-born adults in the U.S. This estimate is very close to Passel and Cohn’s (2009a) estimate of the 2008 undocumented population. Their approach – based on a comparison of data from the Current Population Survey with records of legal immigration – estimated 7 million total (children and adult) undocumented Mexican immigrants. Assuming that 12.6% are children – a figure based on another estimate of the authors of the under-18 share of all undocumented immigrants – their estimate of the undocumented, adult Mexican immigrant population is approximately 6.16 million. The remaining discrepancy is likely due to the small number of undocumented Mexicans with a college education (approximated by Passel and Cohn [2009b] to be about 4%) and to those who arrived before 1990 (but after legalization afforded by the Immigration Reform and Control Act of 1986). Using our approach at the state-level also generates undocumented Mexican counts that correlate highly (r = .97) with estimates of the unauthorized population tabulated by Warren and Warren (2013), as well as Kasarda and Johnson’s (2006) estimates of undocumented Hispanics in North Carolina.

Our interest lies in documenting how the relative size of metropolitan-level undocumented Mexican populations relate to residential sorting processes, but since the PUMS-based metropolitan samples from which these estimates are drawn are in some instances small, we restrict all analyses to metropolitan areas that contain at least 100 (unweighted) adult Mexican immigrant respondents. This restriction removed 20 metropolitan areas from our analysis (from 156 to 136).¹⁶

It is widely known that undocumented Mexican immigrants are less likely to speak English proficiently and have lower wages (Hall et al. 2010; Kossoudji and Cobb-Clark 2002; Rivera-Batiz 1999) than their legal counterparts. To account for these compositional differences, we include the percent of Mexican immigrants who speak English “only” or “very well” and the ratio of median Mexican immigrant to median native-white household income in our models. We additionally conduct a supplemental analysis of another important compositional variable – the percent of Mexican immigrants arriving to the U.S. in the last 10 years– which is highly correlated with percent undocumented and leads to model instability when both variables are included in the same model.

Metropolitan characteristics incorporated in our analysis include total population (logged) and percent black. The total percent foreign born (from any country) is included to account for differences in segregation that might emerge from variation in the depth of immigrant-group populations (see Hall 2013). Increased opportunities in housing – a factor strongly related to racial segregation (Farley and Frey 1994; Logan et al. 2004; Timberlake and Iceland 2007) – is assessed by the percentage of the housing supply built during the last ten years.¹⁷ Measures capturing the economic structure of labor markets include the percent of the 25+ population who are college-educated and the share of workers in industries that Mexicans concentrate in: agriculture, construction, low-skill service, and manufacturing.¹⁸

To evaluate differences in Mexican segregation across metropolitan destinations and to assess the impact of undocumented population shares, we develop OLS models that regress Mexican-immigrant dissimilarity from native whites and blacks in 2005–2009 on group and metropolitan characteristics in the same period. These models take the following general form:¹⁹

$$D_{mk}={\beta }_0+{\beta }_1{\mathit{\text{newdest}}}_m+{\beta }_2{\mathit{\text{undoc}}}_m+{\beta }_3{\mathit{\text{assim}}}_m+{\beta }_4{\mathit{\text{metro}}}_m+e$$

where D_mk is Mexican immigrant dissimilarity from native-born group k (i.e., whites or blacks) in metro m; newdest_m is a dummy indicator of whether metro m is a new destination; undoc_m is the percent of metropolitan Mexican immigrants who are undocumented; assim_m is a vector of metro-level assimilation characteristics of Mexican immigrants – percent proficient in English and the Mexican immigrant income ratio; and metro_m is a vector of metropolitan population, housing, and economic characteristics. Continuous covariates are grand-mean centered in all analyses. As is standard in related work on the topic (Hall 2013; Iceland and Nelson 2008; Licther et al. 2010; Park and Iceland 2011), our multivariate models are unweighted.

To offer a stronger test of the relationship between segregation and undocumented Mexican population shares, we estimate fixed effects models using data from 2000 to 2009. These models are well suited to address our research questions because they absorb all time-invariant characteristics of metros that may lead to differences in the settlement patterns of documented and undocumented Mexicans, and bias estimates of the relationship between undocumented migration and segregation. Specifically, the fixed effects design examines variation within metropolitan areas and evaluates how changes in undocumented immigrant shares shape changes in segregation. These models are specified as:

$$D_{mtk}={\beta }_0+{\beta }_1{\mathit{\text{undoc}}}_{mt}+{\beta }_2{\mathit{\text{undoc}}}_{mt}*{\mathit{\text{newdest}}}_m+{\beta }_3{\mathit{\text{assim}}}_{mt}+{\beta }_4{\mathit{\text{metro}}}_{mt}+a_m+{\mathit{\text{year}}}_t+e_{mt}$$

where m refers to metro areas and t to year (i.e., 2000 or 2005–2009); undoc_mt*newdest_m is the interaction between percent undocumented in metro m in year t and the destination type of metro m; year_t is the period fixed effect (2000=0; 2005–2009=1); and a_m is a vector of dummy variables for each metropolitan area.

Results

Mexican Metropolitan Destinations

Established and new Mexican destinations are distributed unevenly across the American landscape. Figure 1 maps the 136 metropolitan areas we analyze. (A list of these metropolitan areas by destination type is shown in Appendix A.) As expected, established Mexican destinations (black shading) are concentrated in the South West, especially in border states, but are also located in southeastern Washington, Illinois, and Florida. New Mexican destinations (grey shading) are checkered across the U.S., but are prominently clustered in Southeastern states, particularly in Florida and the Carolinas, and along the Boston-Washington corridor.

Figure 1.

Established and New Mexican Metropolitan Areas

A corresponding map of the percent of Mexican immigrants who are undocumented is shown in Figure 2, with lighter shading referring to low shares and darker shading to higher shares. As expected, the undocumented population makes up a larger portion of the Mexican population in Southeastern metros and in those throughout the Great Plains than in established Mexican destinations along the U.S-Mexico border. While there is strong regional clustering to the relative size of the unauthorized population, there is some variation within regions, including areas of California and Texas with higher undocumented shares.

Figure 2.

Undocumented Share of Mexican Immigrant Population, 2005–2009

Mexican-Immigrant Segregation, Destination Type, and Legal Status

As shown in Table 1, across the 136 metropolises in our sample, Mexican-immigrant dissimilarity from native whites was 57.7 in 2005–2009, indicating that nearly three-fifths of Mexican immigrants would have to change neighborhoods to reach integration with native whites in their metropolitan area. This represents a small decline from 2000 when Mexican-immigrant dissimilarity from native whites was 58.4. Mexican-immigrant segregation from native blacks is modestly lower (D = 54.9), and stayed mostly stable during the 2000s.²⁰

Table 1.

Mexican segregation from native-born whites and blacks and percent undocumented, by destination type, 2000 and 2005–2009

	Mexican immigrant dissimilarity from:
	Native whites			Native blacks			Percent undocumented
	2000	2005–09	change	2000	2005–09	change	2000	2005–09	change
All metro areas	58.4	57.7	−0.8	54.5	54.9	0.4	50.1	62.0	11.9
Established destinations	55.2	53.9	−1.2	51.6	52.5	0.8	35.5	48.8	13.4
New destinations	60.9	60.4	−0.5	56.8	56.1	−0.7	60.9	71.8	10.9

Regardless of year or reference group, Mexican-immigrant segregation is higher in new destinations than in established ones. In 2005–2009, Mexican-immigrant segregation from whites was 6.5 points higher in new than in established destinations; similarly, Mexican-immigrant dissimilarity from native blacks was 3.6 points higher in new than established areas. It is noteworthy that Mexican segregation from native whites declined more during the 2000s in established areas than in new destinations; continuing a trend observed during the 1990s (Park and Iceland 2011).

Differences across destinations in the percent of Mexican immigrants who are undocumented loosely corresponds to these patterns in segregation. While, on average, about three-fifths of Mexican immigrants in the typical metro area were unauthorized in 2005–2009, undocumented persons made up larger shares of the Mexican-immigrant population in new areas than in established ones, with nearly three-quarters of Mexican immigrants being unauthorized in new destinations while less than half in established gateways. Thus, that undocumented immigrants are so much more strongly represented among local Mexican immigrant populations in new destinations begs the question of whether the elevated levels of segregation are a reflection of undocumented migration.

Multivariate Models of Mexican Immigrant-White Segregation, 2005–09

Coefficients and standard errors (in parentheses) from regression models predicting Mexican immigrant dissimilarity from native whites in 2005–09 are shown in Table 2.²¹ The first column reiterates from Table 1 that Mexican immigrants are more segregated, and significantly so, from native whites in new than established Mexican destinations.

Table 2.

OLS Predictors of Mexican-immigrant dissimilarity from native-born, non-Hispanic whites, 2005–2009

	(1)	(2)	(3)	(4)	(5)
New Mexican destination	6.497 *** (1.778)	1.768 (2.436)	3.843 (2.352)	2.182 (2.100)	1.942 (2.172)
Percent undocumented		.206 *** (.074)	.181 ** (.076)	.138 * (.081)	.169 * (.087)
Mexican-immigrant assimilation
Percent proficient in English			−.012 (.165)	−.244 * (.145)	−.243 * (.145)
Mexican-white income ratio			−.383 *** (.080)	−.242 *** (.074)	−.243 *** (.075)
Population and housing
Total population (ln)				4.856 *** (.843)	4.815 *** (.851)
Percent immigrant				−.219 (.135)	−.212 (.136)
Percent black				.067 (.103)	.077 (.106)
Percent new housing				−.655 *** (.140)	−.658 *** (.140)
Economic structure
Percent college-educated				−.136 (.170)	−.158 (.178)
Percent in low-skill service				.384 (.481)	.346 (.490)
Percent in construction				.012 (.642)	−.030 (.650)
Percent in manfacturing				−.262 (.286)	−.280 (.289)
Percent in farming				.249 (.377)	.214 (.387)
Destination type interaction
New × Percent undocumented					−.056 (.122)
Constant	53.931 *** (1.346)	56.643 *** (1.640)	55.452 *** (1.567)	56.405 (1.358)	56.855 ** (1.684)
R-squared	.091	.140	.270	.579	.579
Adjusted R-squared	.084	.127	.247	.534	.531

Notes:

^*p < .10;

^**p < .05;

^***p < .01;

standard errors in parentheses; N = 136 metropolitan areas with at least 1,000 total Mexican immigrants and 100 or more unweighted adult Mexican immigrants in ACS PUMS; continuous covariates are centered

In the second model, we consider whether the heightened levels of Mexican immigrant segregation in new destinations can be explained by the high share of undocumented persons. The coefficient on percent undocumented indicates that Mexican immigrants are significantly more segregated from whites in metros where the undocumented make up a larger portion of the Mexican population: a one standard deviation increase in percent undocumented is associated with a modest increase in Mexican immigrant dissimilarity from native whites of 3.3 points (.206*16.3). Importantly, the addition of the variable reduces the new destination coefficient by nearly three-quarters (72.8%) and to statistical non-significance, underscoring the importance of compositional features of Mexican immigrant populations in bolstering segregation.

The third model adds assimilation characteristics of metropolitan Mexican-immigrant populations to ascertain whether the effect of percent undocumented is a compositional artifact of undocumented immigrants possessing characteristics that make them prone to enclave residence. Consistent with the spatial assimilation model, where Mexican immigrants’ incomes (relative to whites) are higher, they are less segregated from native whites. This variable partially suppresses Mexican-immigrant/white dissimilarity in new destinations since incomes between the two groups tend to be closer there. The percent of Mexican immigrants who are English proficient also registers a negative sign, but is small and nonsignificant. Combined, the assimilation variables reduce the effect of percent undocumented by 12.1%, but – consistent a model of place stratification – it remains positive and significant suggesting that the relationship undocumented migration and Mexican-immigrant/white segregation is not spurious to Mexican immigrants’ language ability or socioeconomic position.

In Model 4, structural features of metropolitan areas are incorporated to evaluate whether demographic, housing, or economic characteristics of the metros where Mexican immigrants live explain the relationship between undocumented migration and segregation. The coefficients are generally consistent with previous research, finding that larger metropolises are more segregated and that new housing construction is associated with lower segregation. The addition of these metropolitan characteristics does not, however, explain the positive relationship between undocumented migration and segregation, which remains statistically significant.

Lastly, the final model in Table 2 tests whether percent undocumented affects Mexican-white segregation differently in new than established destinations. The coefficient for the interaction term is small and not statistically significant, indicating that the positive association between undocumented Mexican shares and dissimilarity between Mexican immigrants and native whites tends to work similarly in new and established destinations, at least in the cross section.

Multivariate Models of Mexican Immigrant-Black Segregation, 2005–09

Results from parallel models predicting Mexican immigrant dissimilarity from native blacks are shown in Table 3. The first model restates that Mexican immigrants are more segregated from blacks in new destinations than in established areas. The percent of Mexican immigrants who are undocumented is added to the equation in the second column. Unlike with Mexican-white segregation, the coefficient on percent undocumented is significantly negative, implying that Mexican immigrants are less segregated from blacks in metropolitan areas with large undocumented shares. Moreover, this effect actually suppresses even larger differences in segregation between established and new destinations.

Table 3.

Predictors of Mexican immigrant dissimilarity from native-born, non-Hispanic blacks, 2005–2009

	(1)	(2)	(3)	(4)	(5)
New Mexican destination	3.634 * (1.963)	5.621 ** (2.814)	5.379 *** (2.909)	5.545 ** (2.819)	5.151 ** (2.884)
Percent undocumented		−.085 * (.046)	−.043 (.075)	−.076 (.105)	−.016 (.138)
Mexican-immigrant assimilation
Percent proficient in English			.242 (.205)	−.017 (.200)	−.021 (.201)
Mexican-white income ratio			−.086 (.098)	.066 (.098)	.069 (.098)
Population and housing
Total population (ln)				4.623 *** (1.132)	4.737 *** (1.148)
Percent immigrant				−.310 * (.177)	−.327 * (.179)
Percent black				.031 (.133)	.011 (.136)
Percent new housing				−.335 * (.184)	−.324 * (.185)
Economic structure
Percent college-educated				−.354 (.223)	−.310 (.232)
Percent in low-skill service				.331 (.632)	.395 (.640)
Percent in construction				−1.361 (.867)	−1.271 (.879)
Percent in manfacturing				−1.119 *** (.370)	−1.079 *** (.376)
Percent in farming				1.017 * (.527)	1.085 ** (.538)
Destination type interaction
New × Percent undocumented					−.110 * (.062)

The third and fourth models incorporate Mexican assimilation and metropolitan covariates into the model. The income variable works in the expected negative direction but is non-significant; language proficiency, however, exerts a positive (albeit non-significant) effect, which reduces the negative coefficient of percent undocumented by about half and to non-significance. Metropolitan characteristics generally work as expected: Mexican-immigrants and blacks are more segregated in larger metros and in those with large agricultural sectors, but are less segregated in areas with more immigrants, newer housing stocks, and larger manufacturing bases.

The final model in Table 3 explores whether the effect of undocumented Mexican immigrant shares varies in new and established destinations. Here it is shown that percent undocumented has a stronger – and statistically significant – association with Mexican-black segregation in new destinations than in established ones.²² Thus, the negative relationship between the share of Mexican immigrants who are undocumented and Mexican segregation from native blacks appears to be unique to new Mexican destination areas.

Multivariate Models of Mexican-White and Mexican-Black Segregation during the 2000s

To offer a stronger test of these results, we bring in data from 2000 and reanalyze our main models with metropolitan (and year) fixed effects. The advantage of this approach is that time-invariant characteristics of metros (even immeasurable ones) that potentially bias estimates of our key variables are eliminated. Because destination type is time-invariant, its main effect is absorbed within the model but its joint effect with other characteristics (e.g., percent undocumented) can be estimated because their product does vary over time. Under this framework, metro areas serve as their own controls and the coefficients can be interpreted as how changes in group or metro characteristics are associated with changes in segregation (Allison 2009).

Table 4 shows fixed effects coefficients predicting Mexican immigrant segregation from native whites (left half of table) and native blacks (right half of table). The first model includes percent undocumented only (as well as the fixed effects, which include destination type) and suggests that increases in the percent of Mexican immigrants who are undocumented is positively associated with changes in Mexican immigrant-white segregation between 2000 and 2005–09 but not associated with changes in Mexican immigrant-black segregation. When changes in the undocumented population are allowed to vary by destination type – the second model for each group – we see that its relationship with segregation operates differently in new and established destinations. As our cross-sectional models indicated, the fixed effects models find that undocumented shares, as predicted by the stratification model, are positively related to segregation between Mexican immigrants and native whites in new destinations, but negatively to segregation between Mexican immigrants and blacks living in those areas.

Table 4.

Fixed-effects predictors of Mexican immigrant dissimilarity from native whites and native blacks, 2000 and 2005–2009

	Dissimilarity from native whites			Dissimilarity from native blacks
	(1)	(2)	(3)	(1)	(2)	(3)
Percent undocumented	.079 * (.040)	.033 (.058)	.048 (.067)	.000 (.079)	.138 (.092)	.161 (.109)
Destination type interaction
New × Percent undocumented		.074 ** (.035)	.062 * (.037)		−.226 *** (.081)	−.214 ** (.103)
Mexican-immigrant characteristics
Percent proficient in English			−.112 * (.062)			−.044 (.096)
Mexican-white income ratio			−.041 (.036)			−.002 (.025)
Population and housing
Total population (ln)			2.169 (5.100)			−11.993 (8.948)
Percent immigrant			−.506 (.319)			−.817 (.543)
Percent black			−.284 (.486)			.356 (.729)
Percent new housing			−.111 (.081)			−.202 (.156)
Economic structure
Percent college-educated			.281			.099
			.298			.478
Percent in low-skill service			−.429			−.923 (.754)
Percent in construction			.021 (.394)			−.390 (.641)
Percent in manfacturing			−.465 (.287)			−.242 (.453)
Percent in farming			1.115 * (.604)			.591 (.918)
Year	−1.717 *** (.590)	−1.631 *** (.591)	−3.283 2.077	−.088 (.950)	−.341 (.929)	.678 3.692
Constant	58.928 *** (.335)	58.504 *** (.447)	59.400 *** (1.168)	54.451 *** (.522)	55.793 *** (.701)	56.256 * (2.105)

The final models (column 3) in Table 4 include the entire set of Mexican assimilation and metropolitan control variables. Only a few of these terms reach statistical significance – changes in percent of Mexican immigrants who speak English proficiently and changes in farming on Mexican-white segregation – but, most important for our purposes, the coefficients on the interaction between new destination and percent undocumented are not explained by the addition of these other characteristics of Mexican immigrant populations nor by features of metropolitan areas. Indeed, in the case of Mexican-black segregation, the interaction between percent undocumented and new destination is the only coefficient that reaches statistical significance.

Supplemental Analyses

Following the spatial assimilation model, recent arrivals are more likely than their better-established counterparts to congregate in co-ethnic neighborhoods, and an inherent weakness of our analysis is that we are unable to distinguish effects of legal status from arrival recency because in contemporary migration patterns the two are closely related (r = .86 in 2005–2009). Unfortunately, this very high correlation between undocumented shares and the percent of Mexican immigrants who are recent arrivals generates multicollinearity in models including both measures. Our analytic approach, however, seeks to partially address this issue by including acculturation and human capital characteristics – language proficiency and income – that are theoretically-implicated as the mechanisms that prompt better-established immigrants to move out of enclave neighborhoods (see Alba and Logan 1991, Iceland 2009, Massey 1985; Rosenbaum and Friedman 2007; White and Glick 2009).

Nevertheless, we have conducted additional analysis that cautiously suggests that undocumented migration is independently associated with Mexican immigrants’ segregation patterns. Table 5 presents results from a series of fixed-effects models predicting Mexican-immigrant segregation from whites (left panel) and blacks (right panel). The first model shows the basic relationship between percent undocumented and segregation and is equivalent to the first model in Table 4. The second model in Table 5 shows the basic relationship between percent of Mexican immigrants who are new arrivals and segregation. For Mexican-immigrant dissimilarity from whites, the new arrivals coefficient is about half the size of the percent undocumented coefficient and is not statistically significant. By contrast, for Mexican-immigrant segregation from blacks, the new arrivals coefficient is comparatively larger in size, but is not significant. When both variables are included, telltale signs of multicollinearity emerge: extremely large variance inflation scores, increases in standard errors, and – for Mexican segregation from blacks – coefficients that flip signs. However, for Mexican-immigrant segregation from whites, the percent undocumented coefficient remain nearly three times as large as the new arrivals coefficient and is close to reaching statistical significance (p=.13). Most importantly, when all other covariates are added to the equation, the estimated interactions between percent undocumented and new destination are consistent with those shown in Table 4, although only that for Mexican segregation from blacks is statistically significant. Future research may be able to further disentangle the roles of legal status and arrival recency by using microdata containing larger samples and finer grained distinctions in arrival period (e.g., Cort 2011, 2014; Hall and Greenman 2013).

Table 5.

Fixed-effects predictors of Mexican immigrant dissimilarity from native whites and native blacks, 2000 and 2005–2009

	Dissimilarity from native whites				Dissimilarity from native blacks
	(1)	(2)	(3)	(4)	(1)	(2)	(3)	(4)
Percent undocumented	.079 * (.040)	-- --	.068 (.046)	−.014 (.078)	.000 (.079)	-- --	−.052 (.090)	.146 (.129)
Percent new arrivals	-- --	.040 (.039)	.026 (.045)	.055 (.050)	-- --	.068 (.061)	.087 (.071)	.017 (.082)
New × Percent undocumented	-- --	-- --	-- --	.086 (.064)	-- --	-- --	-- --	−.208 ** (.104)
Additional controls:
Metro and year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Mexican immigrant	No	No	No	Yes	No	No	No	Yes
Metropolitan structural	No	No	No	Yes	No	No	No	Yes
R-squared within	.077	.061	.082	.209	.000	.010	.013	.209
Intraclass correlation (rho)	.943	.943	.942	.942	.894	.894	.893	.942
Variance inflation factors
Percent undocumented	--	--	28.14	60.96	--	--	32.15	70.40
Percent new arrivals	--	--	25.49	14.59	--	--	26.75	37.41
N of metro areas	136	136	136	136	130	130	130	130

Notes:

^*p < .10;

^**p < .05;

^***p < .01;

standard errors in parentheses; includes metro areas with 1,000 total Mexican immigrants and 1,000 reference group members (white or blacks), and 100 or more unweighted adult Mexican immigrants in PUMS; continuous covariates are centered; all models include metropolitan and year fixed effects

Discussion

The rapid growth of the Mexican population in the U.S. and its subsequent dispersal out of long-standing communities along the border and into a wider mix of locales has had profound transformative effects on social, economic, and political life in American society. While the movement of Mexicans immigrants into non-traditional places should be heralded as a sign of increased tolerance and of Mexicans’ growing ability to establish homes outside of traditional barrios, there is escalating concern that integration may be stalled in areas with limited experience dealing with immigration and where Mexicans are an unknown (Massey 2008). Residential integration – the extent to which groups share neighborhoods – has long been viewed as providing a snapshot of broader group incorporation processes (Iceland 2009; Massey 1985; Park 1924; White and Glick 2009); thus, it is significant that levels of segregation are heightened in the new destinations where Mexicans are headed (Fischer and Tienda 2008; Hall 2013; Lichter et al. 2010). A critical question is whether undocumented immigration shapes processes of residential attainment and, more generally, to what extent legal status has emerged as a central axis of racial residential stratification.

Using data from the 2000 Census and 2005–2009 American Community Survey, we explored how legal status contributes to the segregation of Mexican immigrants in new and established destination areas. Our results demonstrate the importance of legal status for Mexican incorporation, finding that the high levels of segregation observed between Mexican immigrants and native whites in new destinations can be explained by the share of the Mexican-immigrant population that is unauthorized. While our ecological research design precludes us from stating that undocumented Mexican immigrants are more highly segregated than their documented counterparts, our research suggests – at a minimum – that areas where undocumented migrants make up a larger portion of the Mexican-immigrant population are more highly segregated than other areas.

Our analysis also suggests that while standard spatial assimilation factors (English proficiency and earnings) are negatively tied to Mexican segregation, the effect of legal status – as expected by the place stratification model – operates above and beyond their influences, as well as those of demographic, housing, and occupational features of metropolitan areas. The implication that one of undocumented migration’s consequences is residential segregation is reasonable if lacking legal documentation to live and work in the U.S. makes immigrants more likely to settle in nonwhite neighborhoods where they are less likely to come into contact with government officials, vigilante neighbors, or otherwise unlikely to draw attention to themselves. While some proclivity among undocumented migrants to self-segregate is likely at work, it is also possible that where concentrations of undocumented migrants are known to be especially high, white households may be prone to flee in the face of immigrant encroachment into their neighborhoods. The tendency of whites to leave neighborhoods with growing immigrant populations has been shown to be especially pronounced in emerging gateways (Hall and Crowder 2013), where undocumented concentrations are high. Clearly, the underlying micro-level and behavioral processes deserve more attention from future research, but regardless of the specific mechanisms, this study contributes to a growing body of work on the sweeping impacts of lacking authorization and further implicates legal status as an important source of social inequality (Massey 2007; Menjívar 2006; Menjívar and Abrego 2012).

Given the multi-group racial structure of the new areas that Mexicans are settling in and ongoing discussions of black-brown relations (see Marrow 2011; Telles et al. 2011; Vaca 2004), we also examined the residential proximity of Mexican immigrants from native-born blacks. Like from whites, we found that Mexican immigrants are more segregated from native blacks in new than established destination areas. However, for Mexican-black segregation, undocumented migration seems to be an integrating force; that is, Mexican immigrants in areas with high undocumented shares have lower levels of segregation from native blacks than in other areas. Our analysis revealed, however, that this negative relationship between undocumented immigration and Mexican-black segregation is unique to new destinations.

Our speculation is that housing costs and availability are central to understanding the presumed coresidence between undocumented Mexican immigrants and native blacks. While our multivariate model accounted for socioeconomic deficits among Mexican immigrants and general conditions of metropolitan housing markets, the implied clustering of undocumented Mexicans in black neighborhoods could result from undocumented immigrants’ need for low-cost housing. It is widely known that the cost of housing in mostly-black neighborhoods is substantially lower than it is in other neighborhoods. Moreover, given undocumented migrants’ constraints in purchasing housing, they are especially likely to seek residence in rental units, which are both more widely available and cheaper in black than in non-black neighborhoods.²³ These same housing processes have been discussed as important factors shaping the historical formation of overlapping Latino and black neighborhood clusters (Rodriquez and Mindiola 2011) and the mobility patterns of blacks and immigrants (Crowder et al. 2011). This integration by virtue of housing unaffordability may be especially pronounced in new destinations where co-ethnic barrios have not reached critical mass. Future research should explore these issues at more refined geographic levels, consider how specific housing amenities and other local conditions, including proximity to work, transportation networks, and ecological distance shape group settlement patterns, and explore how neighborhood migration behaviors of undocumented migrants are linked to local racial compositions and opportunities for affordable housing.

Another possible explanation for the settling of undocumented Mexicans in black neighborhoods relates to the perceived likelihood of detection. While law enforcement practices have changed over the last several decades and the policing of the most-disadvantaged black neighborhoods is arguably better characterized by surveillance than abandonment (Goffman 2009), police protection in black neighborhoods is viewed by residents to be severely lacking, either because of racial bias or municipal inequity (Peterson and Krivo 2010). Under this backdrop, in new destinations lacking co-ethnic enclaves, undocumented Mexicans may perceive that oversight of their daily lives (e.g., contact with police, other governmental officials, or landlords) may be reduced in neighborhoods with large black concentrations. Alternatively, with a choice between having white or black neighbors, undocumented Mexicans may prefer the latter, due to perceptions of shared minority status (Telles et al. 2011) or to blacks holding less negative attitudes about undocumented migration (Pew 2011). While there are several reports of Latino-black discord in new destinations (McDermott 2011; Marrow 2011; Rich and Miranda 2005), the social and political backlash, often vocalized most strongly by whites, levied against Mexicans in these areas may drive undocumented Mexicans into black neighborhoods.

Overall our findings stress the importance of taking legal status seriously in evaluating the incorporation of Mexican immigrants. There is extensive evidence demonstrating that undocumented immigrants are more likely to work in dangerous settings, garner lower wages, and less likely to be provided with ancillary benefits than their legal counterparts (Donato et al. 2008; Donato and Massey 1993; Flippen 2012; Hall et al. 2010; Kaushal 2006; Kossoudji and Cobb-Clark 2002; Massey 1987; Massey and Bartley 2005; Phillips and Massey 1999; Rivera-Batiz 1999; Telles and Ortiz 2008). There is also considerable work detailing the detrimental impact of lacking papers on educational achievement and attainment (Abrego 2006; Bean et al. 2011; Gonzales 2011; Greenman and Hall 2012; Kaushal 2008), as well as on health care access and health outcomes (Arbona et al. 2010; Cavazos-Rehg et al. 2007). The results from this analysis add to this list and suggest that legal status is a stratifying force in the residential attainment process. Theoretically, this analysis also suggests a broadening of the place stratification model to incorporate citizenship and legal rights. Given the powerful impacts that neighborhood context has on other dimensions of social life, the residential separation of undocumented persons could further exacerbate inequality between those with and without legal standing. Importantly, the presumed segregation of undocumented migrants in new destinations potentially impedes the incorporation of Mexican migrants, to the extent that it limits contact with majority group members, reinforces social anxieties between groups, and limits material opportunities. Whether these residential patterns are hardened as new destinations transition into major gateways, and whether policymakers are successful in bringing undocumented migrants out of the shadows, will determine whether incorporation is being slowed momentarily or is more-permanently threatened.

The connection between undocumented migration and segregation also has important implications for broader patterns of neighborhood integration – including the potential for reduced contact between whites and blacks if whites are avoiding neighborhoods populated by both African Americans and Mexicans – and raises questions about the relationships between Mexicans and other Latino groups and the extent to which they are coresiding in isolated barrios with large concentrations of undocumented migrants.²⁴ Ultimately, while the precise mechanisms that underlie the relationship between undocumented migration and segregation remain unclear - whether to native avoidance or to self-segregation – the size of the unauthorized population and the mostly-negative response that their arrival provokes make its significance that much more relevant, visible, and potent.

An active – and sometimes contentious – discussion within sociology is whether America’s historically-rigid white-black color line is being transformed. Some foresee a continuation of a white/nonwhite divide in which people of color share a common fate (Portes et al. 2005); others posit one of black exceptionalism, where the boundaries of whiteness will be blurred to include Asians and Latinos while African Americans hold place on the lowest rung (Gans 1999; Yancey 2003; Lee and Bean 2007); and some see a more complex hierarchy cut along multiple lines (Bonilla Silva 2004). Our analysis does not speak directly to this debate, but highlights the importance of citizenship in structuring the emerging contours of the color line. Whether the apparent coresidence of blacks and undocumented Mexicans is indicative of the emergence of a white-nonwhite divide or rather sets the stage for future ethnic conflict, and whether the presumed segmentation of Mexicans with and without legal authorization into different neighborhoods suggests horizontal racialization within the Mexican population, are questions future research would profit from tackling.

Highlights.

• We estimate the share of Mexicans who are unauthorized in metropolitan destinations

• Undocumented migration promotes segregation between Mexicans and whites

• Undocumented migration diminishes segregation between Mexicans and blacks

• Legal status has a strong impact on segregation in new destinations

Appendix A: Metropolitan Areas in Sample, by Mexican Destination Type

Established
Albuquerque, NM	Killeen-Temple-Fort Hood, TX	San Antonio, TX
Amarillo, TX	Laredo, TX*	San Diego-Carlsbad-San Marcos, CA
Austin-Round Rock, TX	Las Cruces, NM	San Francisco-Oakland-Fremont, CA
Bakersfield, CA	Las Vegas-Paradise, NV	San Jose-Sunnyvale-Santa Clara, CA
Boulder, CO	Los Angeles-Long Beach-Santa Ana, CA	San Luis Obispo-Paso Robles, CA
Brownsville-Harlingen, TX	Madera-Chowchilla, CA	Santa Barbara-Santa Maria-Goleta, CA
Chicago-Naperville-Joliet, IL-IN-WI	McAllen-Edinburg-Mission, TX	Santa Cruz-Watsonville, CA
Chico, CA	Merced, CA	Santa Fe, NM
College Station-Bryan, TX	Midland, TX	Santa Rosa-Petaluma, CA
Colorado Springs, CO	Modesto, CA	Stockton, CA
Corpus Christi, TX	Napa, CA	Tucson, AZ
Dallas-Fort Worth-Arlington, TX	Naples-Marco Island, FL	Vallejo-Fairfield, CA
Denver-Aurora-Broomfield, CO	Odessa, TX	Victoria, TX
El Centro, CA	Oxnard-Thousand Oaks-Ventura, CA	Visalia-Porterville, CA
El Paso, TX	Phoenix-Mesa-Scottsdale, AZ	Waco, TX
Fresno, CA	Prescott, AZ	Yakima, WA
Greeley, CO	Riverside-San Bernardino-Ontario, CA	Yuba City, CA
Hanford-Corcoran, CA	Sacramento--Arden-Arcade--Roseville, CA	Yuma, AZ
Houston-Sugar Land-Baytown, TX	Salinas, CA
Kennewick-Pasco-Richland, WA	San Angelo, TX
New
Asheville, NC	Greensboro-High Point, NC	Orlando-Kissimmee, FL
Atlanta-Sandy Springs-Marietta, GA	Greenville-Mauldin-Easley, SC	Philadelphia-Camden, PA-NJ-DE-MD
Baltimore-Towson, MD	Hickory-Lenoir-Morganton, NC	Port St. Lucie, FL
Beaumont-Port Arthur, TX	Idaho Falls, ID*	Portland-Vancouver-Beaverton, OR-WA
Birmingham-Hoover, AL	Indianapolis-Carmel, IN	Poughkeepsie-Newburgh-Middletown, NY
Bradenton-Sarasota-Venice, FL	Jacksonville, FL	Provo-Orem, UT
Bridgeport-Stamford-Norwalk, CT	Kansas City, MO-KS	Raleigh-Cary, NC
Burlington, NC	Lafayette, IN	Reno-Sparks, NV
Cape Coral-Fort Myers, FL	Lake Havasu City-Kingman, AZ	Richmond, VA
Carson City, NV*	Lakeland-Winter Haven, FL	Rockford, IL
Charleston-North Charleston, SC	Lexington-Fayette, KY	Salem, OR
Charlotte-Gastonia-Concord, NC-SC	Little Rock-North Little Rock, AR	Salt Lake City, UT
Cincinnati-Middletown, OH-KY-IN	Longview, TX	Seattle-Tacoma-Bellevue, WA
Columbia, SC	Louisville/Jefferson County, KY-IN	Sherman-Denison, TX
Columbus, OH	Madison, WI	South Bend-Mishawaka, IN-MI
Dalton, GA	Memphis, TN-MS-AR	Spartanburg, SC
Deltona-Daytona Beach, FL	Miami-Fort Lauderdale-Pompano Beach, FL	St. George, UT*
Des Moines-West Des Moines, IA	Minneapolis-St. Paul-Bloomington, MN-WI	Tampa-St. Petersburg-Clearwater, FL
Durham-Chapel Hill, NC	Mount Vernon-Anacortes, WA*	Tulsa, OK
Elkhart-Goshen, IN	Myrtle Beach-North Myrtle Beach, SC	Tyler, TX
Eugene-Springfield, OR	Nashville-Davidson--Murfreesboro, TN	Vineland-Millville-Bridgeton, NJ
Fayetteville-Springdale, AR-MO	New Haven-Milford, CT	Washington-Arlington, DC-VA-MD-WV
Fort Smith, AR-OK	New York-Northern New Jersey, NY-NJ-PA	Wenatchee-East Wenatchee, WA*
Fort Wayne, IN	Ogden-Clearfield, UT	Wichita, KS
Gainesville, GA	Oklahoma City, OK	Wilmington, NC
Grand Rapids-Wyoming, MI	Omaha-Council Bluffs, NE-IA	Winston-Salem, NC

Note: Full names of some MSAs shortened;

^*contains fewer than 1,000 blacks for Mexican-white D calculation.