Using longitudinal data from the Panel Study of Income Dynamics linked to three decades of census data on immigrant settlement patterns, this study examines how the migration behaviors of native-born whites and blacks are related to local immigrant concentrations, and how this relationship varies across traditional and nontraditional metropolitan gateways. Our results indicate that regardless of gateway type, the likelihood of neighborhood out-migration among natives increases as the local immigrant population grows—an association that is not explained by sociodemographic characteristics of householders or by features of the neighborhoods and metropolitan areas in which they reside. Most importantly, we find that this tendency to move away from immigrants is pronounced for natives living in metropolitan areas that are developing into a major gateway—that is, a community that has experienced rapid recent growth in foreign-born populations. We also demonstrate that among mobile natives, the neighborhoods that they move to have substantially smaller immigrant concentrations than the ones they left, a finding that is especially evident in new gateway areas.
The past several decades have been an extraordinary period of immigration to the United States. Although the growth in the foreign-born population has had profound economic, social, and political consequences, the redistribution of America’s immigrants away from a handful of long-standing ports of entry (e.g., New York City, Los Angeles) and into other communities (e.g., Raleigh, Tulsa) has arguably been the most important demographic trend of the last 20 years. The dispersion of immigrants into a wider mix of communities means that a large and diverse set of places is experiencing the benefits and challenges of immigration. The spatial diffusion of immigrants has also contributed to unmatched levels of racial/ethnic diversity (Fasenfest et al. 2006; Singer 2005) and increased exposure between majority and minority groups (Timberlake and Iceland 2007).
Although mounting contact between racial/ethnic groups portends the possibility of a more racially inclusive society, there are signs that the growth in immigration and their spread throughout the nation has led to the emergence of new forms of stratification. Black–white residential segregation has declined considerably in recent decades, but residential segregation of Latinos and Asians—groups comprising a bulk of new immigrants—from non-Latino whites has remained virtually unchanged (Logan and Stults 2011). One explanation for the stall in Latino and Asian segregation is that these groups tend to be more segregated in nontraditional areas, where they are increasingly settling (Fischer and Tienda 2006; Hall 2013; Lichter et al. 2010).
The heightened segregation of Asians and Hispanics in nontraditional areas partially reflects a tendency of newly arrived immigrants to band together, perhaps to ease the transition into community life (Iceland and Nelson 2008; Iceland and Scopilliti 2008). However, segregation is also created by the migration behaviors of native residents and their decisions to remain or flee in the face of swelling immigrant populations. Thus, variations across gateways in natives’ migration responses may be crucial to understanding variation in residential incorporation across metropolitan gateways.
Our goal in this article is to offer evidence on this issue by evaluating variation across established, new, and developing gateways in the relationship between neighborhood immigrant concentration and native residential mobility. We focus on three main research questions: (1) Does the association between neighborhood immigrant concentration and native out-migration differ in traditional and nontraditional gateway areas?; (2) Can observed variations in mobility reactions to immigrants be explained by variations in characteristics of householders or by features of the tracts and metropolitan areas where they reside?; and (3) Does the association between immigrant concentration in origin and destination neighborhoods differ for native movers living in traditional and nontraditional areas? To answer these questions, we link individual-level longitudinal data from the Panel Study of Income Dynamics (PSID) to three decades of census data describing neighborhood and metropolitan immigrant populations, and use cross-classified multilevel models to predict native out-migration and the concentration of immigrants in mobile natives’ neighborhoods of destination.
New Destinations and Residential Incorporation
Residential segregation of immigrants from the native-born has long been seen as a temporary phenomenon, progressing from the isolation of immigrants in dense central-city enclaves of major gateways to residential integration as immigrants accumulate the social and economic resources necessary to exit urban cores for more advantaged areas containing fewer immigrants and, typically, more white residents (see Massey 1985). A considerable body of research has emerged to test the implications of this and related theoretical arguments for processes of neighborhood attainment and segregation (see Charles 2006; Rosenbaum and Friedman 2007). More recently, interest in immigrant settlement patterns has shifted toward understanding their dispersion out of traditional gateways and into nontraditional communities as well as the corresponding rise of “new destinations” (see Gozdziak and Martin 2005; Massey 2008; Singer 2005; Zúñiga and Hernández-León 2005).
Despite earlier speculation that divergent migration patterns of immigrant and native populations will lead to “regional balkanization” (Frey 1995, 1996; Frey and Liaw 1998), Frey and Liaw (2005:212) argued that the diffusion of the foreign-born is symbolic of their successful spatial incorporation; they predicted that “minorities undergoing spatially assimilated long-distance migration will be residing in more integrated neighborhoods locally.” Empirical support for this hypothesis is mixed. On the one hand, Park and Iceland (2011) found that immigrant dissimilarity from native whites is lower in metropolitan areas having recently emerged as immigrant destinations than in more traditional ones. On the other hand, the same authors found that immigrant-native dissimilarity increased in new immigrant gateways during the 1990s while remaining largely unchanged in traditional gateways. Using an expanded set of both urban and rural communities, Lichter et al. (2010) found that segregation between Latinos and whites is substantially higher in new Hispanic destinations than in established Hispanic areas (i.e., unweighted dissimilarity difference of 14 points), and that the gap cannot be explained by structural or demographic characteristics of communities. Likewise, Fischer and Tienda (2006) found that Latino immigrants are more segregated from other groups in a selected set of new metropolitan Hispanic destinations than in major traditional Hispanic metropolises (dissimilarity difference of 4.9 points). Examining 10 specific immigrant groups, Hall (2013) also found evidence that immigrants—not just Hispanics—tend to be more segregated from native whites in nontraditional destinations, even after accounting for group differences in acculturation and socioeconomic status across areas. That residential integration may be taking root in traditional gateways more rapidly than in nontraditional ones is also implied in Clark and Blue’s (2004) analysis of segregation in five major immigrant-receiving cities.
In sum, mounting evidence suggests that residential integration is heightened in the emergent communities where immigrants are increasingly settling. What drives the uneven segregation across types of areas, however, remains unclear. According to the classic model of spatial assimilation, residential separation between immigrant and native populations materializes as a result of the tendency for immigrants to choose residence near other coethnics. This proclivity for own-group clustering is due not simply to ethnocentric attraction (Clark 1992; Clark and Blue 2004) but to the social and economic benefits of enclave residence (see Edin et al. 2003). Although it may seem logically inconsistent that these tendencies to cluster would operate more strongly in new areas where opportunities for enclave residence are reduced, limited access to ethnic goods and services or a desire to improve well-being in unfamiliar territories may encourage immigrants in new destinations to congregate in a handful of neighborhoods. Indeed, patterns of “heterolocalism,” whereby the existence of strong ethnic communities facilitate the settlement of immigrants in non-enclaves, have been observed in several traditional gateways (see Zelinsky and Lee 1998). Thus, one possibility for heightened segregation in new destinations is an increased propensity among immigrants to cluster in coethnic neighborhoods.
Tendencies to band together may be particularly acute if the reception climate is unwelcoming. In fact, it is in these emerging destinations where some of the more hostile policies toward immigrants have been proposed or enacted (Broder 2007; Hopkins 2010; Steil and Vasi 2014), and a growing body of work details how social and political backlash in these areas stalls integration (Carr et al. 2012; Fennelly 2008; Johnson et al. 1999; Kirk et al. 2012; Marrow 2011; Winders 2008). By comparison, natives in established destinations are well accustomed to the diversity of faces and cultures that immigration brings (Massey 2008).
Importantly, however, variation in local contexts of reception highlights the importance of considering how natives may respond differently to immigrants across areas. Scholarship has long documented how majority group members’ migration behaviors fuel residential separation between groups (see Crowder 2000; Quillian 2002). And given the long history in the United States of neighborhood retreat in the face of “foreign invasion” (e.g., Northern whites fleeing from the influx of Southern blacks (Boustan 2010)), it is plausible not only that native populations are resistant to immigrant neighbors but also that their reactions are shaped by the broader historical context of foreign encroachment into their communities.
Previous research has observed a positive association between neighborhood immigration and native out-migration (Crowder et al. 2011) but has not considered whether and how natives’ migration responses vary across types of settlement areas. One possibility for the link between immigrant concentration and native out-migration is that natives residing in immigrant-rich neighborhoods possess traits—such as being young, childless, or renters—conducive to migration. Crowder et al. (2011) found some support for this argument, which may suggest that any differences in migration across gateway types simply reflect compositional variation in the native populations located in different gateway areas.
Other arguments have suggested that any link between immigrant concentration and out-migration is mediated by characteristics of neighborhoods. The neighborhood-socioeconomic thesis holds that large concentrations of immigrants encourage native out-mobility by lowering the socioeconomic quality of the neighborhood. Because immigrants have, on average, lower levels of education and higher poverty rates than the native-born (Grieco et al. 2012), high immigrant concentrations are likely to reduce neighborhood income levels. If neighborhood income is also linked to physical surroundings, exposure to crime, and the quality of local amenities and services (Logan and Alba 1993), natives may be especially prone to migrate from lower-income areas.
Another possibility is that large immigrant concentrations reshape local housing markets in ways that influence the mobility behavior of native residents (Ley 2007; Ley and Tutchener 2001). Local immigration may, for example, reduce the stock of vacant housing available in the neighborhood and increase local housing costs, which in turn may push some native residents out of the neighborhood. Previous work on this neighborhood-housing hypothesis has found that housing competition generated as a result of growing immigrant populations is an important mediating factor in the relationship between neighborhood immigrant concentrations and native out-migration, especially among African Americans (Crowder et al. 2011; Wilson and Taub 2006).
To the extent that native-born householders prefer racially homogenous neighborhoods, local immigration may spur out-migration by altering the ethnic makeup and social fabric of the neighborhood. Although research on natives’ preferences for foreign-born neighbors is limited, survey research shows that whites rate neighborhoods with relatively few members of other groups as the most desirable (Charles 2006; Krysan 2002; Krysan and Bader 2007) and those neighborhoods with large black or Latino populations as least desirable (Charles 2001; Clark 2009; Emerson et al. 2001). Black survey respondents, by contrast, tend to be more open to racially mixed neighborhoods (Krysan and Bader 2007; Krysan and Farley 2002), although some evidence suggests that blacks harbor negative sentiments toward immigrants settling in their neighborhoods (Johnson et al. 1999; McClain et al. 2011; Marrow 2011; Wilson and Taub 2006). Thus, given that more than 8 in 10 immigrants hail from Asia or Latin America (Grieco et al. 2012), large concentrations of immigrants are likely to affect the concentration of racial/ethnic minorities in the neighborhood, and any differential response to local immigration on the part of natives could reflect differences in how residents respond to this neighborhood racial diversity.
A final possibility for understanding variations in the relationship between immigrant concentration and out-migration is prompted by research documenting how variations in metropolitan structure affect residential mobility (Crowder et al. 2012; Pais et al 2012). In particular, two arguments are relevant to this analysis. First, work by Frey and others suggests that the in-migration of immigrants triggers native out-migration because of the economic dislocations that are created by increased job competition (Frey 1995, 1996; Frey and Liaw 1998). Although this labor market competition hypothesis has generated considerable debate (Card and DiNardo 2000; Kritz and Gurak 2001; Wright et al. 1997), it does highlight the need to consider local labor market conditions. Second, across metropolitan areas, natives have different opportunities to locate into neighborhoods with different immigrant concentrations. This is especially relevant to understanding differences across metropolitan gateways because natives dissatisfied with the presence of immigrants in their neighborhoods may be less likely to out-migrate if the relative number of neighborhoods with few immigrants is minimal. Thus, natives in established gateways seeking to escape immigrants may be constrained in doing so if there are relatively few potential neighborhood destinations lacking immigrants. By contrast, natives in developing gateways who find themselves living in neighborhoods with too many immigrants may be better able to actuate their desires to move because there are many opportunities within the broader constellation of neighborhoods to find an area with small concentrations of foreign-born residents. Thus, this neighborhood opportunity thesis suggests that variation across metropolitan gateways in natives’ responses to local immigration can be explained by metropolitan differences in the stock of nonimmigrant neighborhoods.
Data and Methods
We explore these issues using data from the Panel Study of Income Dynamics (PSID) linked to contextual data drawn from the U.S. Census. The PSID is a longitudinal survey of U.S. residents and their families that originated in 1968 with a sample approximately 5,000 families drawn from a stratified multistage design representative of the United States.1 Members of panel families were interviewed annually between 1968 and 1997 and every two years thereafter, and new families have been added to the panel as children and other members of original panel families form their own households. The PSID is well suited for our analysis because its longitudinal nature makes it possible to track the migration behavior of individual householders over time, and the data contain rich information on characteristics known to influence geographic mobility. Because many residential moves are undertaken by families, a move by one household member often means a move by other family members. Our focus on household heads allows us to avoid counting as unique those moves made by members of the same family. At the same time, moves by family members who were not the household head at one interview but become the head of a household by the subsequent interview are included in our effective sample. Given that rapid immigrant dispersion was not underway until the 1980s (Singer 2005), we limit our analysis to householders in PSID panels between 1980 and 2009.
The availability of restricted-access Geocode Match Files, which link individual respondents to census codes indicating their place of residence at each interview, allows us to identify PSID respondents’ metropolitan and neighborhood location, trace their migration across neighborhoods between successive interviews, and attach detailed census data about their current and previous neighborhoods. We follow much of the prior work in this area (e.g., Crowder et al. 2012; Quillian 2002) by using census tracts to represent neighborhoods, which provide imperfect operationalizations of neighborhoods but nevertheless provide comprehensive coverage of the entire nation during our study period, are summarized for a variety of theoretically relevant measures, and generally approximate the usual conception of a neighborhood (White 1987). Potential problems associated with changes in tract boundaries across decennial censuses are overcome by our use of the Longitudinal Tract Data Base (LTDB; Logan et al. 2014), which normalizes census tract data between 1970 and 2000 to 2010 boundaries. We use the LTDB data on tracts from the 1980, 1990, and 2000 censuses, as well as the 2006–2010 American Community Survey (ACS); we use linear interpolation/extrapolation to estimate values for noncensus years.
Our effective sample for this analysis consists of 9,693 native-born non-Latino white and 6,830 native-born non-Latino black heads of PSID households. Consistent with most research on the topic, our typology of immigrant gateways focuses on the categorization of metropolitan areas, so we include only those PSID observations in which the householder originated in one of the 366 metropolitan areas (defined consistently during the study period using boundaries set by the Office of Management and Budget in 2010).
We take advantage of the longitudinal nature of the PSID by segmenting each respondent’s data record into a series of person-period observations, with each observation referring to the one- or two-year period between PSID interviews. On average, the individual householders in the sample contribute 6.3 person-periods for a total sample of 104,787 person-periods, after observations with missing data are deleted listwise.
Our analysis focuses on two migration behaviors: the likelihood of neighborhood out-migration and the immigrant concentration of movers’ destination neighborhoods. Neighborhood out-migration is a binary indicator of whether a householder moved out of the tract of origin during the interval between interviews, taking a value of 1 for those who moved during the interval and a value of 0 for those who remained in the same tract. The vast majority of migration events among PSID householders are within metropolitan areas (74.0 %), although our measure includes those moving to different metropolitan areas (26.0 %).2 Among those who moved, we assess the concentration of immigrants in their new neighborhoods by examining the percentage of the population that is foreign-born in the tract to which the householder moved.
Our primary independent variables refer to the concentration of immigrants in native householders’ tracts of residence and the history of immigrant settlement in the metropolitan area in which the householders live. Local immigrant concentration is gauged by the percentage of the tract population made up of individuals born outside the United States.3 Our measure of metropolitan gateway type uses data on immigrant settlement between 1970 and 2008 (the midpoint of the 2006–2010 ACS). Our typology differs from previous classification schemes in two important ways. First, in contrast to earlier work on immigrant settlement evaluated at a cross section (Lichter and Johnson 2009; McConnell 2008; Singer 2005), our approach recognizes that areas’ status as a particular type changes over time (e.g., from a new gateway to an established gateway). Second, although other research has often distinguished simply between traditional and nontraditional areas, we take a slightly more refined approach and recognize the variation across nontraditional areas. Doing so is consistent with Singer (2005), who emphasized the importance of distinguishing between newly emerged and “pre-emerging” immigrant gateway areas, and is particular relevant in the context of neighborhood migration if one anticipates that the initial arrival of immigrants should produce the strongest migration response. More specifically, we categorize metropolitan areas separately in 1980, 1990, 2000, and 2008 into one of four types. Established gateways refer to metropolitan areas that, in a given year, met one of three criteria: (1) had a total immigrant population of 200,000 or more; (2) had an immigrant population share that was twice as large as the national share; or (3) had at least 100,000 immigrants and an immigrant share that was larger than the national share. New gateways are areas with at least 50,000 immigrants that either experienced 10-year growth in immigrant populations that was at least twice as fast as the national average or had both above-average growth and above-average immigrant shares. Developing gateways are metros with at least 1,000 immigrants and 10-year growth rates twice as fast as the nation. All remaining metropolitan areas, in a given year, are considered nongateways.4 We recognize that a few of the metros we classify as new or developing gateways are areas with earlier histories of immigration but that have only recently re-emerged as destinations for post–World War II immigrants (e.g., Allentown, Cincinnati). As shown in Table 1, 26 metropolitan areas were defined as established immigrant gateways in 1980, including New York, Chicago, and Los Angeles. During the same time, five areas were considered new gateways (e.g., Dallas, Phoenix), and 28 were developing gateways (e.g., Atlanta, Charlotte). By 2008, nearly twice as many metropolitan areas were considered established gateways, including Atlanta and Dallas; 20 were new gateways (e.g., Cape Coral, Nashville); and 69 were developing (e.g., Boise, Reading). A complete list of metropolitan areas and their gateway classification in each year is shown in Table S1 in Online Resource 1. Gateway types are linked to PSID data, with noncensus years inferred based on the nearest census year: for example, destination type for person-period observations in 1992 is based on the classification of the householder’s metropolitan area in 1990.
We consider a variety of other characteristics of native respondents and their households, neighborhoods, and metropolitan areas in order to test theoretical arguments related to the relationship between local immigrant concentrations and native out-mobility. Descriptive statistics for all variables in the analysis, as well as their correlations, are shown in Table S2 in Online Resource 1. Key demographic predictors of residential mobility include age (in years) and marital status (taking a value of 1 for respondents who were married or permanently cohabiting at the beginning of the interval). The effect of children is tapped with an indicator variable for individuals living in a family with members under 18. We also control for the education of the householder (measured by years of school completed) and the total family taxable income (measured in thousands of constant Year 2000 dollars). Homeownership is coded as 1 for those in an owner-occupied housing unit, household crowding is assessed by the number of persons per room, and prior mobility experience takes a value of 1 for respondents who have moved within the previous three years. In all models, we also include an indicator for the year of observation in order to account for trends in inter-neighborhood migration and the length of the migration interval to control for the switch to a biennial survey.
We present models with controls for several characteristics of origin tracts to test theoretically implicated mechanisms through which local immigrant concentrations may prompt native out-migration, and to assess whether these neighborhood characteristics help to explain variations in mobility responses across destinations. To test the argument that mobility away from immigrant populations reflects a reaction to socioeconomic change, we consider the tract poverty rate (percentage of population at or below 100 % of the federal poverty line). We include controls for tract vacancy rate and median housing value (in $10,000s) to test whether local housing conditions associated with immigrant concentrations influence mobility decisions. To explore the possibility that variations across metropolitan destination types in reactions to immigrants reflect differential exposure to racial diversity, we control for racial/ethnic diversity of the tract using a five-group (Hispanics and non-Hispanic whites, blacks, Asians, and others) entropy measure.5
To explore arguments that variation in mobility to immigrants across gateways reflects differences in labor market conditions, we include job growth over the preceding calendar year and the average wage of salaried jobs (in $1,000s). The possibility that metropolitan areas differ in opportunities for native-born householders to escape immigrants is assessed by a measure of the percentage of all tracts in the metropolitan area with immigrant concentration of less than 5 %. In addition, to account for differences across metropolitan areas in the availability of newly developed neighborhoods, we include the percentage of all tracts in the metropolitan area in which 10 % or more of all housing units were built in the last 10 years.
Given the hierarchical nature of the data and our interest in how metropolitan settlement histories influence migration behaviors, we use a multilevel modeling design in which observations (migration intervals) are nested within individual householders and householders are nested within metropolitan areas. For the first part of our analysis, we estimate a three-level cross-classified random-coefficients logistic model predicting the log-odds of neighborhood out-migration as a function of individual, tract, and metropolitan characteristics.6 In these models, out-migration is allowed to vary across respondents and metropolitan areas, and the effect of immigrant concentration on out-mobility is set to differ across metropolitan areas. Our primary interest in these models is in the differential effect of neighborhood immigrant concentration across metropolitan gateway types (i.e., the interaction between gateway type and tract percentage immigrant). In all tables, we show average marginal effects (AME) rather than logit coefficients to provide a clear understanding of the relative magnitudes of the estimated associations and to facilitate comparisons across models (Mood 2010). Standard errors for the AMEs are calculated using the delta approach (Greene 2011). In the second stage of our analysis, we compare the immigrant concentrations of movers’ destination neighborhoods to the areas they moved away from. Similar to our first set of analyses, our central interest lies in differential destinations across gateway types while controlling for basic individual- and metropolitan-level factors. These variations are assessed via the interaction between metropolitan gateway types and percentage immigrant in the tract of origin in three-level random-intercept models of immigrant destination concentrations.
Native Exposure to Immigrants
We begin our analysis with a descriptive account of native respondents’ exposure to immigrant populations across gateway types and over time. Figure 1 shows average neighborhood immigrant concentrations (dark gray bars) and metropolitan immigrant concentrations (the sum of the light and dark gray bars) for sample members living in established, new, and developing gateways as well as in nongateways in 1980, 1990, 2001, and 2007. The temporal trend in immigration into metropolitan areas and into the neighborhoods where native whites and blacks reside is clear: in 1980, the typical native PSID householder in an established gateway lived in a tract that was 9.0 % immigrant and in a metropolitan area that was 11.3 % immigrant. By 2007, these numbers had grown to 13.7 % and 17.6 %, respectively. Similar upward trends are observed for the other gateway types. Also visible in Fig. 1 are the differences in immigrant population shares across gateway types. As expected, native respondents in established gateways have the highest exposure to immigrants, both at the metropolitan and neighborhood levels, followed by those in newly emerged gateways. Natives in developing gateways and nongateways had similarly low levels of exposure to immigrants during the early part of the observation period, but larger differences in immigrant exposure emerged in the 2000s. That immigrant population shares grew very rapidly in developing areas during the 1990s and rose quickly in new gateways during the 2000s partially reflects the diffusion of immigrants toward nontraditional destinations and has potentially important consequences for native out-migration if natives are sensitive to large foreign-born populations.
Also notable is that in each type of area, metropolitan immigrant concentration is higher and has generally grown at a faster pace than neighborhood immigrant concentration; this serves as a reminder that native householders in our sample are somewhat shielded from the more general residential repercussions of increasing immigrant concentrations, finding themselves in neighborhoods in which immigrants are underrepresented relative to overall metropolitan concentrations. This does not diminish the increases in overall native exposure to immigrants occurring over the past three decades, nor does it lessen the importance of foreign-born deconcentration, but it does highlight the fact that natives have maintained some residential distance from immigrant populations even as their metropolitan areas have diversified.
Immigration and Native Out-Migration
What the descriptive patterns in natives’ exposure to immigrants cannot reveal, however, is the extent to which residential separation in established, new, and developing gateways is maintained through natives’ mobility away from large immigrant concentrations. In this stage of the analysis, we set out to address the issue by exploring the association between neighborhood immigrant concentration and the likelihood that native white and black PSID householders will change neighborhoods, whether this relationship varies by metropolitan gateway type, and whether observed associations can be attenuated by theoretically informed covariates.
The marginal effects shown in Table 2 provide a basic answer to these questions, indicating how the probability of natives’ neighborhood out-migration varies according to tract immigrant concentration and gateway type. The first model includes the measure of immigrant concentration in the tract of residence at the beginning of the observation interval, revealing a tendency for natives to out-migrate as the concentration of foreign-born residents in their neighborhoods grows. Specifically, the marginal effect indicates that the probability of out-migration increases by about 0.8 percentage points for each percentage point increase in neighborhood immigrant concentration.
The second model incorporates product terms between tract immigrant concentration and metropolitan gateway type to assess whether natives’ mobility responses to local immigration vary across gateway types. The results suggest that they do. The marginal effect of tract percentage immigrant on out-migration is smaller in Model 2 than in Model 1, suggesting that the association in established areas is smaller than the average across all types of areas, but the association remains positive and statistically significant.7 The results also show that the association between tract immigrant concentration and native out-migration is similar in new gateways, suggesting that natives’ mobility responses to foreign-born neighbors in established and new gateways are statistically indistinguishable. In contrast, natives in developing gateways and nongateways are more likely than their counterparts in established gateways to out-migrate in the face of substantial immigrant concentrations in the neighborhood. (The superscript a on these estimates indicates that the interaction between gateway type and tract percentage immigrant is significantly different from established gateways.) Specifically, a one-point increase in tract percentage immigrant increases the probability of out-migration for natives in developing gateways by 1.3 points and increases the probability of out-migration for natives in nongateways by about 0.8 points. Thus, although this finding is consistent with previous work showing a tendency for natives to out-migrate as neighborhood immigrant populations increase (Crowder et al. 2011), we find that natives in developing gateways and nongateways are more sensitive to local immigrant concentrations than are their counterparts in established or newly emerged gateways, which is a result that sheds light on the heightened patterns of segregation in nontraditional areas. The results in Model 2 also indicate that, on average, mobility is modestly higher in developing and nongateway areas than in established gateways.
Demographic and socioeconomic characteristics of householders are included in the third model to examine whether the observed differentials in the association between immigrant concentration and native mobility reflect compositional differences in the native-born persons populating immigrant-rich neighborhoods. The results provide some support for this argument. Although micro-level characteristics of householders—such as race, age, family structure, income, and tenure—shape natives’ mobility patterns in expected ways, they only partially attenuate the marginal effects of immigrant concentration. Specifically, the addition of individual-level characteristics reduces the estimate of tract percentage immigrant in established and new gateways by about two-thirds and the total effect of tract immigrant concentration in developing gateways by about 54 %. Nevertheless, the marginal effects of neighborhood immigrant concentration in developing gateways and nongateways remain substantial in size and statistically significant, suggesting that the association between out-mobility and location near immigrant neighbors in these areas cannot be fully explained by variation in householder characteristics.
The fourth and fifth models in Table 2 incorporate features of native householders’ tracts of residence. Neighborhood housing values, vacancy rates, poverty rates, and population density are included in Model 4. Several of these tract characteristics significantly influence migration, but their inclusion does not alter the size or significance of the marginal effects of tract immigrant concentration in developing or nongateway areas. In Model 5, tract racial diversity (entropy score) is added to test for the possibility that native migration responses to immigrants are driven by the association with the racial/ethnic mix of the neighborhood. Results indicate that tract racial diversity is positively related to native out-migration; in comparison with those in more homogeneous neighborhoods, native householders living in racially diverse neighborhoods are more likely to change neighborhoods. Most importantly, the inclusion of the diversity measure modestly attenuates the association between immigrant concentration and out-migration, reducing the marginal effect in developing gateways by about one-sixth and the effect in nongateways by about one-third (from .0035 to .0023). Nevertheless, tract immigrant concentration remains positively and significantly associated with native out-migration in established, developing, and nongateway areas, and the association in developing gateways remains significantly different from that in established gateways.
The final model for the racially pooled sample includes variables assessing labor market conditions and the distribution of nonimmigrant and newly developed neighborhoods within metros. The neighborhood-opportunity variables work in expected directions, with natives more likely to out-migrate when there is an abundance of newly developed and nonimmigrant neighborhoods in the metropolitan area. The economic context variables indicate that mobility is lower in areas with higher wages but is not significantly associated with job growth. The addition of these variables, however, does not alter the effects of immigrant concentration in any meaningful way. With the full set of individual, tract, and metropolitan controls, the results indicate that a one-point increase in tract percentage immigrant is associated with a 0.1 point increase in the probability of out-migration for natives in established gateways. However, for natives living in metropolitan areas without a significant history of immigration, the association between mobility and the presence of foreign-born populations is much stronger: for natives in developing gateways, a one-point increase in neighborhood immigrant concentration corresponds with a 0.5 point increase in the probability of changing neighborhoods, and for those in nongateways, a 0.2 point increase in the probability of out-migration.
Given the historical preponderance of white flight from minority communities, as well as the possibility of tensions between immigrant groups and African Americans (Johnson et al. 1999; Marrow 2011; McClain et al. 2011; McDermott 2011; Vaca 2004), there are compelling reasons to expect the effects of immigrant concentration on native mobility to vary by householder race/ethnicity. To evaluate this possibility, we report results from models separately for native-born white and black householders in the last two columns of Table 2. For the most part, the predictors of out-migration work similarly for native whites and blacks; one of the few important exceptions is tract racial diversity, which is positively related to white out-migration but negatively (and nonsignificantly) related to black out-migration. Most importantly, however, is that the effects of neighborhood immigrant concentration do not vary substantially by race. For both white and black natives in established gateways, the marginal effect of tract percentage immigrant is modestly positive.8 Similarly, both black and white householders in developing gateways exhibit heightened mobility responses to neighborhood immigrant concentrations, with associations in these areas that differ significantly from those in established gateways. Specifically, a one-point increase in tract percentage immigrant increases the probability of out-migration by about 0.5 points for both black natives and white natives in developing gateways.
Supplemental Analysis of Latino Out-Migration
Because the original PSID sample was drawn in the late 1960s, Hispanic respondents in the PSID are not representative of the Latino population today. Nevertheless, we re-analyze our final out-migration model for Hispanic householders—both foreign- and U.S.-born—appearing in the data during the same time horizon in order to offer some tentative evidence on whether the associations we observe in Table 2 are limited to black and white natives. The results for Latinos are shown in Table 3 and indicate that without controls for individual, tract, or metropolitan correlates of mobility, tract immigrant concentration is negatively associated with the probability of out-migration for Hispanic householders. Complete models—with the full set of controls and interaction with destination type—suggest that for Hispanic householders, tract immigrant concentration is not significantly associated with out-migration. Notwithstanding concerns about the representativeness of Hispanics in the PSID, this is suggestive that the results shown in Table 2 do not simply reflect a general tendency for local immigration to be linked to the out-migration of existing populations, but one that is unique to native white and black households.
Destination Immigrant Concentration
The findings up to now point to the heightened probability of out-migration for natives living in neighborhoods with large immigrant concentrations, especially for those in metropolitan areas developing into immigrant gateways. Thus, the results imply that part of the explanation for pronounced segregation of immigrants in nontraditional destinations is that natives leave neighborhoods as the immigrant population swells. The ultimate impact of these mobility behaviors on segregation, however, would be offset if native migrants moved to neighborhoods with immigrant concentrations similar to those of the neighborhoods they left. In the final stage of our analysis, we offer a simple descriptive assessment of the destinations where native movers settle. Our goal in this analysis is to assess the potential effects of native mobility patterns on residential integration of immigrants and native-born populations, not to provide a comprehensive assessment of factors affecting destination choices.
Table 4 shows results from three-level random-intercepts models describing the association between the immigrant concentration of native migrants’ origin and destination tracts and whether this relationship varies across gateway types. The coefficient for tract percentage immigrant in the first model indicates that the concentration of immigrants in destination tracts is positively associated with the immigrant concentration in the tract of origin, indicating that householders originating in immigrant-rich neighborhoods are also more likely to settle in neighborhoods with relatively large immigrant populations. However, that the coefficient is less than 1 indicates that native migrants tend to move to neighborhoods with immigrant concentrations that are smaller than those of the tracts they are leaving. Specifically, native householders move to neighborhoods in which the concentration of immigrants is, on average, about 33 % of the concentration in the tracts they vacated.
The second model in Table 4 introduces metropolitan gateway types and their interactions with origin-tract immigrant concentration. The main effect of origin tract percentage immigrant indicates that in established areas, mobile native-born householders are likely to move into neighborhoods with smaller immigrant shares than the ones they left. The interactions with gateway type indicate that this tendency is observed across in all types of areas but is significantly more pronounced only in new gateway areas. Mobile natives in these new gateways tend to move into tracts with immigrant concentrations that are only about one-fifth (.314 – .107 = .207) as large as the concentrations in the tracts from which they moved. The comparable figures are 31.4 % for mobile natives in established gateways, 33.6 % (.314 + .022 = .336) for those in developing gateways, and 33.2 % (.314 + .018 = .332) for those in nongateway metros.
Also shown in Table 4 are models predicting destination-tract immigrant concentration for white and black householders separately. In general, the racially specific results indicate that the association between immigrant concentrations in origin and destination tracts works similarly for black and white movers. More specifically, both black and white mobile householders tend to move to neighborhoods with fewer immigrants than the ones they originated in, and only among whites in new gateway areas is this tendency amplified.9
To show this relationship in another way, Fig. 2 converts the coefficients in Table 3 to predicted associations between percentage immigrant at origin and destination. Given the similarity of the relationship for native-born blacks and whites, the graph is based on the pooled-group sample (Model 2). The dotted diagonal line represents the hypothetical scenario in which native migrants move to tracts with the same percentage immigrant as in their origin tracts. Given the overall larger immigrant populations in established areas, it is no surprise that both the origin and destination neighborhoods of native movers in established gateways have higher immigrant shares. Most important from our standpoint is that although the lines are upward sloping—indicating a positive association between immigrant concentrations in origins and destinations—native movers’ destination tracts contain substantially smaller immigrant shares than the neighborhoods they exited, at least among those originating in neighborhoods with more than a modest threshold of immigrant presence. In established metropolitan areas, this threshold is about 11 %; that is, natives leaving neighborhoods that are more than about 11 % immigrant move to neighborhoods that, on average, have smaller immigrant concentrations than those they left. This same general pattern exists for natives in other gateway types, but the intersection is even lower: for natives in new gateways, it is about 8 %, and for those in developing gateways and nongateways, it is closer to 6 %. Perhaps more important is that in all gateway types, the gap between immigrant concentrations in origins and immigrant concentrations in destinations is greatest for natives leaving areas with the highest immigrant percentages. Also noteworthy, however, is that the line for natives in new gateways is significantly flatter than the lines for those in other gateways. The upshot is that natives not only show a heightened tendency to move from neighborhoods with large immigrant concentrations but also tend to migrate to neighborhoods with far fewer immigrants when they do move, and this tendency is heightened for natives residing in new gateways (but not for those in developing ones).
Over the past several decades, immigration has brought millions of new faces to the United States and transformed the social fabric of the country. Unlike previous waves, however, today’s immigrants are not strictly concentrated in a handful of coastal cities. Rather, the impacts of immigration are being felt across a diverse mix of U.S. communities. Yet, as immigrants disperse from traditional gateways to other communities, there are worries that they face challenges integrating into local institutions (Dondero and Muller 2012; Fischer 2010) and provoke political backlash (Hopkins 2010; O’Neil 2012; Steil and Vasi 2014), and that their incorporation into the mainstream is stalling.
The findings of this article provide some empirical basis for these anxieties. Our analysis indicates that local immigrant concentrations are associated with native white and black neighborhood out-migration. This propensity to out-migrate, however, is particularly pronounced in developing gateways, where immigrant populations have grown very recently and rapidly. Even accounting for compositional differences in the natives that occupy these different types of areas, differences in socioeconomic and housing features of neighborhoods, economic conditions of labor markets, and opportunities to find nonimmigrant neighborhoods, we find that natives in developing gateways residing in a neighborhood that is 10 % foreign-born have rates of out-migration that are five points higher than their counterparts in a nonimmigrant neighborhood.
Our analysis does not find evidence that the association between local immigration and native out-migration is conditioned by householder race; marginal effects of tract immigrant concentration on mobility are similar in size for both black and white natives. Similarly, our models suggest that the pronounced association between local immigration and out-migration in developing immigrant gateways operates similarly for blacks and whites. Our analyses do indicate, however, that native migrants leaving tracts with modest immigrant concentrations tend to move to neighborhoods with substantially lower immigrant shares. This tendency is somewhat more pronounced in newly emerged immigrant gateways.
Overall, these findings—that local immigration is linked to native out-migration and that mobile natives move to neighborhoods with fewer immigrants than the ones they left—potentially inform the heightened levels of immigrant segregation in nontraditional destinations (Fischer and Tienda 2006; Hall 2013; Lichter et al. 2010) by implicating migration behaviors of native residents as one contributing factor separating immigrants and natives.
Although we were unable to uncover the specific mechanisms that drive natives’ mobility away from immigrant concentrations in developing areas, this finding is consistent with sociological arguments regarding group threat, which hold that dominant groups’ position in society is threatened by the encroachment of new minority group members (Blalock 1967; Bobo 1999). The accelerated contact with immigrants that natives residing in developing destinations experience in their day-to-day lives may make them especially prone to attribute personal and social misfortunes to the recent influx of foreign persons. This may be especially true in the context of neighborhood settings where resources are limited and the tendency to blame immigrants for neighborhood transformations (e.g., on school quality, civic/neighborly relations, or crime/safety concerns) may be magnified. This contrasts with more-established gateways where despite being the sites for expansive ethnic enclaves, natives are likely more accustomed to, or even embrace, the diversity associated with the historical settlement of immigrants (Massey 2008). Native populations in these areas also tend to be more tolerant and view pro-immigrant policies more favorably (De Jong and Tran 2001; Graefe et al. 2008; Haubert and Fussell 2006). Similarly, although many newly emerged metropolitan areas, such as Nashville and Salt Lake City, do not have the long histories of sustained immigration seen in established destinations, they have transitioned into major gateways and have immigrant populations that play vital economic and political roles. Thus, while further research is needed on the precise mechanisms that trigger heightened mobility responses in developing gateways, our results indicate that one of the consequences of the spatial diffusion of the foreign-born population is that natives living in the neighborhoods where immigrants settle may move to neighborhoods with fewer immigrants.
We are grateful to Sam Friedman, Jacob Hibel, John Iceland, Bob Kaestner, Maria Krysan, Barry Lee, Dan Lichter, Giovanni Peri, Emily Rosenbaum, Jeff Timberlake, and Stew Tolnay for comments on earlier versions of this article and to Brian Stults for generous support with the Longitudinal Tract Data Base. This research was supported by infrastructure grants to the Cornell University Cornell Population Center (R24 HD058488) and to the University of Washington Center for Studies in Demography and Ecology (R24 HD042828) by the Eunice Kennedy Shriver National Institute of Child Health and Human Development.
Although the PSID was designed to be nationally representative of the U.S. population in 1968, the sample has maintained close correspondence with characteristics of the U.S. population (see Duffy and Sastry 2012), and supplemental analyses indicate that the geographic distribution of our analytic sample closely matches the distribution observed in census data between 1980 and 2009.
Models restricted to households remaining in the same metropolitan area between interviews yield results that are substantively and statistically similar to those presented here.
We also tested for associations with recent changes in local immigrant populations. We found that in models including both percentage immigrant and change in percentage immigrant over the last five years, immigrant change had a comparatively small association with out-migration, and its inclusion does not alter the results in a meaningful way.
We considered several alternative ways of defining gateways, including approaches that relax or stiffen requirements to be considered an established, new, or developing gateway. The results from these specifications are substantively consistent with those shown here. Our employed typology has the added benefit of approximating the midpoints of the range of coefficients on key variables and produces a classification of metropolitan areas that is consistent with Singer’s (2005) typology.
Racial/ethnic entropy is defined, for each tract, as , where pr refers to the proportion of group r in the tract. High values of E refer to tracts where the distribution of racial groups is relatively uniform. Unfortunately, tract data on the race or birth country of immigrant populations are not available for our entire study period. We did, however, consider (in supplemental analyses) the percentage of the tract that is ethnic Mexican, and found weak associations with native out-migration, suggesting that our observed associations do not simply reflect out-migration from local Mexican populations.
Characteristics of householders’ tracts of residence are treated as a Level 1 characteristic because there is too little clustering of PSID respondents within census tracts to warrant an additional level.
The correlations between neighborhood percentage immigrant and gateway types are all moderate to weak (under |.50|), reflecting the fact that there is substantial heterogeneity in neighborhood immigrant concentrations within gateway types.
Although the marginal effect of tract percentage immigrant in new gateways is small and nonsignifcant, the logit models indicate that for both blacks and whites, the effect is not significantly different from that in established gateways.
Racially pooled models with two- and three-way interactions indicate that the link between immigrant concentrations in origin and destination tracts varies significantly by race (p = .005), and differences in the association across different gateway types also vary significantly by race.