Marriage and cohabitation between members of different racial and ethnic groups has increased in the United States over recent decades. Despite this demographic shift, we know relatively little about how the growing numbers of mixed-race couples are faring in systems of residential stratification. Previous research indicates that mixed-race couples tend to be located in diverse neighborhoods, but because this past research has used cross-sectional data and has not focused on actual residential mobility, it is not clear whether mixed-race couples choose diverse neighborhoods or are just more likely to develop in diverse neighborhoods. To provide a more complete picture of this topic, I conduct a prospective analysis of the residential location and mobility patterns of mixed-race couples, focusing on the extent to which these couples are more likely than monoracial couples to move into, and/or remain in, diverse neighborhoods. The use of longitudinal data between 1985 and 2009 from the Panel Study of Income Dynamics (PSID) linked to neighborhood- and metropolitan-level data from multiple population censuses reveals that in comparison with monoracial couples, mixed-race couples tend to be located in neighborhoods with higher levels of racial and ethnic diversity and tend to enter more diverse residential destinations when they move. However, these outcomes vary substantially across types of mixed-race couples. Moreover, the outcomes associated with individual- and metropolitan-level conditions provide limited support for the common contention that the residential patterns of mixed-race couples reflect differences in residential preferences, and point to the role of broader patterns of racial stratification in shaping their residential outcomes.
Marriage and cohabitation between members of different racial and ethnic groups has increased dramatically over recent decades. By 2010, the U.S. Census Bureau (2010) recorded more than 4.5 million married couples with members of different races or Hispanic origins, representing almost 8 % of all married-couple households. Just 40 years earlier, mixed-race couples represented less than 1 % of husband-wife households, numbering just 321,000 (U.S. Census Bureau 1998). Moreover, in 2010, more than 15 % of the growing number of opposite-sex married couples included partners of different races or Hispanic origins (Wang 2012).
The rapid increase in the number of mixed-race couples has been viewed as an indication of a softening of racial boundaries that were once considered largely impermeable and has coincided with a loosening of social proscriptions of mixed-race coupling (Gullickson 2006; Wright et al. 2003). Growing racial heterogamy has also helped usher in a period of dramatic diversification, with the population of mixed-race children growing by 37 % from 2000 to 2009, an increase from 1.96 million to 2.69 million (U.S. Census Bureau 2012).
Despite these dramatic demographic and corresponding social shifts, we still know relatively little about how the growing numbers of mixed-race couples are faring in existing systems of racial stratification. This is particularly true in the study of residential stratification. Work on the residential location of mixed-race couples has relied on cross-sectional, often aggregate-level data (cf. Holloway et al. 2005; Wright et al. 2011), making it impossible to assess the residential processes shaping the neighborhood location of mixed-race couples or to examine micro-level residential outcomes. Especially important is the inability to test the relative roles of the racial characteristics and economic standing of these couples, or the impact of metropolitan opportunity structures, in shaping their residential mobility patterns and neighborhood location.
The purpose of this study is to assess the residential mobility and attainment of mixed-race couples using multilevel data drawn from almost a quarter-century of data from the Panel Study of Income Dynamics (PSID) linked to neighborhood and metropolitan data from multiple population censuses. I use multilevel models to compare several groupings of mixed-race couples with monoracial couples and examine these couples’ access to, and movement between, neighborhoods characterized by their levels of racial and ethnic compositional diversity.
Background and Theory
The scant evidence on the topic suggests that residential outcomes for mixed-race couples are distinct from those of monoracial couples. For example, in their recent study, Wright et al. (2011) found that in 2000, black-white couples were overrepresented in the most diverse neighborhoods of the 12 metropolitan areas included in their analysis. Holloway et al. (2005) found a similar dynamic in their study of residential patterns reflected in 1990 census data. These general findings suggest that the growth of mixed-race populations might alter patterns of residential segregation in potentially important ways (Ellis et al. 2012). At the very least, past research suggests that the experiences of mixed-race couples represent important challenges to patterns of racial stratification in residential attainment.
However, this past research presents an important issue with the temporal ambiguity between mixed-race coupling and their exposure to neighborhood diversity. Wright and his colleagues argued that the residential patterns of mixed-race couples “reflect a combination of choices and constraints” (2011:6) that lead them into diverse neighborhoods, thereby implying that mixed-race couples exhibit distinct patterns of inter-neighborhood migration that result in their heightened exposure to neighborhood diversity. Yet, given the association between residential diversity and the likelihood of forming a mixed-race relationship (Briggs 2007; Britton 2011; Houston et al. 2005), at least part of the association between mixed-race coupling and neighborhood diversity likely reflects the fact that mixed-race couples are more likely to be formed in diverse neighborhoods than in more homogeneous areas. Investigating this temporal ambiguity can be addressed through the use of longitudinal data that track the residential mobility behaviors of mixed-race couples that are already formed. In other words, looking at the mobility behaviors of mixed-race couples after they are formed removes the possibility that their exposure to higher levels of neighborhood diversity at time t is being partially driven by their formation in already diverse neighborhoods. Thus, investigating common claims in the literature requires a prospective analysis of residential location and mobility patterns, focusing on the extent to which mixed-race couples are more likely to move into, and/or remain in, diverse neighborhoods than are monoracial couples.
Moreover, theoretical arguments of residential attainment point to several factors that might provide insight into mixed-race couples’ greater exposure to neighborhood diversity and the residential mobility processes that drive this exposure. First, the spatial assimilation perspective highlights the potential importance of economic resources. This perspective suggests that as members of racial and ethnic minority groups increase their economic resources, they can convert this capital into upward residential mobility, often moving into neighborhoods containing higher shares of whites (Charles 2003; Moran 2001). Accordingly, group differences in neighborhood attainment can be thought of as attributable to group differences in economic resources. Combined with evidence that mixed-race marriage is somewhat selective of the highest-status members of at least some groups (Gullickson 2006; Kalmijn 1998), this theoretical argument suggests that controlling for income and education will explain part of the heightened exposure to diversity among mixed-race couples compared with monoracial minority couples, as well as group differences in underlying mobility behaviors. Some support for the assimilation argument is found in recent studies of the topic. In their analysis of data from the 1990 U.S. Census, Holloway et al. (2005) found that high-income mixed-race couples were more likely to live in neighborhoods with whites than with nonwhites.
A second theoretical perspective—the place stratification model—suggests that discrimination pervades the housing market, taking a variety of subtle and more blatant forms, including unequal treatment by rental agents (Massey and Lundy 2001), exclusionary zoning (Rothwell and Massey 2009), and persistent racial biases in mortgage lending and foreclosure practices (Clark 2013; Rugh and Massey 2010). Most observers have agreed that the common forms of contemporary housing discrimination are fairly subtle (Roscigno et al. 2009; Squires 2007; Yinger 1995), such as racial steering by real estate agents who guide white clients away from minority areas, and minority home-seekers toward minority-dominated areas (Ross and Turner 2005). Despite the subtlety of modern forms of housing discrimination, the place stratification perspective holds that discrimination continues to constrain the residential options of some minority groups so that racial groups are arranged into neighborhoods hierarchically, with whites on top, followed by Asians and Latinos, and then by blacks at the bottom (Alba and Logan 1993; Logan and Molotch 1987; Massey and Denton 1993).
The central argument of the place stratification perspective is that because of the effects of housing market discrimination, group disparities in residential attainment and mobility outcomes will persist even after economic predictors of these outcomes are controlled for. Furthermore, researchers have extended the place stratification perspective in more detailed terms by distinguishing two forms: the strong and weak versions. The strong version asserts that because of discrimination, minorities are unable to convert their financial resources into higher-status neighborhoods, leading to more pronounced effects of income on residential attainment for whites than for minorities. Alternatively, the weak version of the place stratification theory contends that the effects of income will be less pronounced for whites than for minorities; only highly resourced minorities will gain access to more advantaged areas, whereas even relatively low-income whites will have opportunities to access such places (Logan and Alba 1993).
The implications of the place stratification perspective for the residential attainment of mixed-race couples are not completely clear. On one hand, discrimination may be less intense for some mixed-race couples than for monoracial minority couples. Specifically, the social advantages enjoyed by whites might provide for minority group members some level of protection against the discrimination normally affecting the residential outcomes of members of their race. Under such a dynamic, we would expect residential outcomes, and the effects of economic resources on these outcomes, of mixed-race couples that include a white member to be more similar to those of monoracial white couples than to those of monoracial minority couples.
On the other hand, persistent resistance to racial intermarriage (Bobo et al. 2012; Dalmage 2000) raises the possibility that mixed-race couples may face discrimination that is as strong, or stronger, than that faced by monoracial minority couples. In in-depth interviews of black-white couples, Dalmage found that these couples perceive a form of racism she calls “borderism.” Dalmage (2000:40) defined borderism as “discrimination faced by those who cross the color line . . . or attempt to claim membership (or are placed by others) in more than one racial group.” Based on these observations, we might expect that in comparison with monoracial white couples, mixed-race couples—and especially those that include a black member—may face obstructions in translating their economic resources into high-status neighborhoods.
A third proposition is that mixed-race couples simply have stronger preferences for neighborhood diversity than do monoracial couples (cf. Dalmage 2000; Datzman and Gardner 2000; Holloway et al. 2005; Wright et al. 2011). Certainly, the tendency to view residential outcomes as a reflection of differences in residential preferences is common in the literature on residential stratification (cf. Clark 2009; Fossett 2006). Although there is little data suitable for linking preferences to actual locational outcomes, the idea that preferences play an important role is consistent with observations of sharp racial differences in neighborhood preferences at the individual-level (Charles 2006; Krysan et al. 2009). Additionally, several authors have pointed to social dynamics that may create a strong incentive for mixed-race couples to choose racially diverse neighborhoods. For example, Moran (2001:156) argued that black-white couples choose diverse areas so that their “children develop an appreciation of their complex heritage.” Similarly, Dalmage’s (2000) in-depth interviews indicate that black-white couples may simply feel more comfortable in diverse neighborhoods where their unique status is less conspicuous.
However, we actually know little about the racial-residential preferences of black-white couples, and even less about the preferences of other types of mixed-race couples. One possibility is that residential preferences of mixed-race couples reflect the combination of preferences held by the members of the couple. Available research indicates that whites’ tolerance for living near minority groups has increased over time but remains limited, and whites tend to rank diverse neighborhoods as substantially less desirable than mostly white neighborhoods (Charles 2006; Krysan and Bader 2007). In comparison with whites, blacks express stronger preferences for balanced residential racial compositions (Charles 2006; Krysan and Bader 2007; Krysan and Farley 2002). Available research also suggests that Latinos and, to a lesser extent, Asians are more tolerant of neighborhood diversity than whites but are somewhat less tolerant than blacks (Charles 2006). Thus, to the extent that the residential preferences of mixed-race couples reflect the combined preference of their members and these preferences affect residential decision-making, we might expect exposure to neighborhood diversity to be greatest among couples with a black member and lowest among those with a white member.
Finally, the general housing availability model (South and Crowder 1997) suggests that processes of residential mobility and attainment are shaped by the structure of opportunities afforded by the local housing market. In the context of studying access to diverse neighborhoods, the racial and ethnic composition of the metropolitan population is potentially most important. Diverse metropolitan areas are likely to contain a larger relative number of diverse neighborhoods, increasing the likelihood of individual households selecting this type of neighborhood. Given that mixed-race marriage is also significantly more common in diverse metropolitan areas (Crowder and Tolnay 2000), it is possible that the higher level of neighborhood diversity experienced by mixed-race couples is explained by the racial and ethnic composition of the metropolitan areas in which they are located, not by a heightened preference for diverse neighborhoods as has been argued in past research (Dalmage 2000; Wright et al. 2011).
Together, these theoretical arguments raise a number of important research questions about the residential outcomes of mixed-race couples:
Question 1: How does the level of neighborhood diversity vary for different types of mixed-race couples compared with monoracial couples?
Question 2: Is the association between neighborhood diversity and residential out-mobility different for mixed-race couples than for monoracial couples?
Question 3: Among couples that move, do mixed-race couples move to neighborhoods that are more diverse than the neighborhood destinations of monoracial couples?
Question 4: To what extent are differences in residential outcomes between different types of couples driven by variations in economic resources, other sociodemographic characteristics, and opportunity structures presented by the broader metropolitan area?
Pursuing answers to these questions provides a unique prospect to test, from a new angle, the relative efficacy of core theoretical arguments used to explain segregation by race, and to understand the ways in which growing numbers of mixed-race couples are a part of reshaping the residential landscape of metropolitan America.
Data and Methods
I address these questions using data from the PSID linked to neighborhood-level data drawn from the U.S. Census. The PSID is a longitudinal survey of U.S. residents and their families that began in 1968 with approximately 5,000 families. Members of panel families were interviewed annually between 1968 and 1997 and every two years thereafter; new families are added to the panel as children and other members of original panel families form their own households. The PSID is well suited for my analysis for two reasons. First, it identifies the location of couples at each interview, which allows me to track their mobility between neighborhoods across time. Second, the data contain information on individual- and household-level characteristics known to influence residential mobility and attainment. I focus on observation years between 1985 and 2009, interview years in which the PSID data include information on the race and ethnicity of both the household head and her/his spouse or long-term cohabitor. Because I am interested in comparing mixed-race couples with monoracial couples, I focus only on those householders with a spouse or a long-term cohabitating partner present at both the beginning and the end of an observation period1 (the time between sequential interviews). I include in my analysis only those couples containing a black or white member (but married to members of other groups as well) because of the questionable representation of Asian and Latino groups in the PSID. Finally, given the theoretical debates about the effects of broader metropolitan context on residential outcomes, I focus on couples living in a census-defined metropolitan area at both the beginning and the end of an observation period.
The PSID’s restricted-access Geospatial Match Files allow me to identify the residential location of individual PSID respondents at each interview and attach information about the racial and ethnic composition of their neighborhoods as well as characteristics of the broader metropolitan area. I follow much of the prior work in this area (e.g., Crowder et al. 2012; Massey et al. 1994; Quillian 2002) by using census tracts to represent neighborhoods. Although census tracts are imperfect operationalizations of neighborhoods (Lee et al. 2008; Tienda 1991), they provide near-comprehensive coverage of the nation during my study period, are summarized for a variety of theoretically relevant measures, and approximate the usual conception of a neighborhood (Jargowsky 1997; White 1987). Potential problems associated with changes in tract boundaries across decennial censuses are overcome by my use of the Neighborhood Change Database (NCDB),2 which normalizes 1980–2010 census tract data to 2010 census boundaries. I use the NCDB’s data on tracts from the 1980, 1990, 2000, and 2010 censuses, and use linear interpolation/extrapolation3 (cf. Quillian 2003) to estimate values for all tract and metropolitan characteristics in noncensus years.
I take advantage of the longitudinal nature of the PSID by segmenting each couple’s data record into a series of couple-period observations, with each observation referring to the one- or two-year period between PSID interviews. My effective sample for the analysis consists of 49,814 observations for monoracial couples and 3,183 observations for mixed-race/ethnicity couples.4 I contrast couples containing various combinations of several (potentially diverse) racial/ethnic groups: non-Hispanic black (hereafter, black), non-Hispanic white (white), Latino (of any race), and all other non-Latino racial groups (other).5 For example, the black-white couples (N = 597) in my sample have one non-Latino black and one non-Latino white partner. I construct similar categories for black-Latino (N = 279), black-other (N = 168), white-Latino (N = 1,405), and white-other (N = 734) mixed-race couples. I compare these mixed-race couples with two types of monoracial couples: (1) those in which both partners are black (hereafter, black-black, N = 12,715), and (2) those in which both members are white (hereafter, white-white; N = 37,099). I focus only on couples containing at least one black or one white member6 because members of other groups are underrepresented7 in the PSID data; and given that they were added to the PSID panel using nonrandom sampling, they do not represent the national populations of these groups. The emphasis on couples that contain either a black or white member also enhances the comparability of my research to past aggregate-level studies.
Next, I turn to outcomes related to residential mobility, treating inter-neighborhood residential mobility as a two-stage process involving the decision to move and then the choice of destination (Massey et al. 1994). Accordingly, the second dependent variable in my analysis is a dichotomous variable indicating whether the couple moved out of the census tract of origin between PSID interviews (a value of 1 for those who moved during the mobility interval, and 0 for those who remained in the same tract). The third dependent variable measures the level of racial/ethnic diversity, as measured by the entropy index, in the destination tracts of mobile PSID householders.
I also consider the effects of a number of theoretically relevant micro-level and contextual characteristics that may account for group differences in residential location and mobility. Key demographic predictors include age (in years) of the household head and the presence of children in the family (1 = yes). Financial resources are indicated by the total family taxable income, measured in thousands of constant 2000 dollars. Other measures of socioeconomic conditions include (1) the education (in completed years) of the household head; (2) employment status of the household head, coded as 1 for those employed at least part-time; and (3) homeownership, coded as 1 for those in an owner-occupied housing unit. I also control for household crowding, measured by the number of persons per room, as well as length of residence, indicated with a dummy variable taking a value of 1 for those respondents who had lived in their home for at least three years at the beginning of the observation period. Finally, to explore arguments that group variation in exposure to neighborhood diversity and mobility between neighborhoods characterized by different levels of diversity reflect differences in metropolitan opportunity structures, I include a metropolitan-level entropy index. Paralleling my measure of neighborhood diversity, the metropolitan-level entropy index summarizes the relative concentrations of white, black, Latino, and other populations in the metropolitan area as a whole. All characteristics in the analysis are measured at the beginning of the observation period and are considered time-varying (i.e., Level 1 in my multilevel framework). Additionally, I include an indicator for the year of observation in order to account for trends in neighborhood diversity and mobility, and the length of the migration interval (1 or 2 years) to control for the switch to a biennial survey in 1997.
Given the hierarchical nature of the data, I use a multilevel modeling design9 in which couple-period observations are nested within individual householders and householders are nested within metropolitan areas.10 For the first part of my analysis, I estimate three-level random-intercepts linear regression models predicting the entropy score in the neighborhood of residence at time t as a function of individual and metropolitan characteristics. My central focus in these models is the pattern of neighborhood diversity experienced by various types of mixed-race couples relative to that experienced by monoracial white and black couples.
In the second set of models, I estimate three-level random-intercepts logistic regression models predicting the log-odds of neighborhood out-migration as a function of individual, tract, and metropolitan characteristics. In these models, out-migration is allowed to vary across respondents and metropolitan areas. My primary interest in these models is in the differential effect of neighborhood diversity (entropy) across various types of mixed- and monoracial couples.
In the final stage of my analysis, I predict the racial diversity (entropy) of movers’ destination neighborhoods as a function of individual, tract, and metropolitan characteristics, using three-level random-intercepts linear regression models. Similar to my first set of analyses, my central interest lies in differential destinations across couple-groups. Because models in the final stage of the analysis are based only on inter-tract movers, I include a Heckman correction (inverse Mills ratio) for the selection of householders into the mover category. In my application of the Heckman procedure, the “selection” equation includes all the regressors in the out-mobility models (Heckman 1979), whereas the “substantive” equation (predicting tract diversity) omits those variables assumed to affect the decision to move but not the characteristics of destinations.
Table 1 shows that couple-groups differ dramatically in terms of a number of factors that might affect exposure to neighborhood diversity. For example, mixed-race couples that include a white partner have higher average income than do couples with a black partner. Black-white couples, however, have the highest average income out of couples with black partners. Black-black and mixed-race couples without a white partner, on average, have less education than do white-white couples and their mixed-race counterparts with a white member. Among couples with at least one white member, household heads in black-white couples have the lowest average education. Moreover, all mixed-race couples have lower average levels of homeownership than white-white couples; mixed-race couples with a black partner typically have lower homeownership rates compared with those mixed couples with a nonblack member. When it comes to exposure to metropolitan diversity, all types of mixed-race couples and black-black couples tend to live in metropolitan areas that have higher diversity than those occupied by white-white couples—a factor that likely increases their opportunities to choose diverse neighborhoods.
To convey group differences in exposure to, and types of, neighborhood diversity, Fig. 1 provides means and standard deviations for neighborhood entropy and average neighborhood racial composition for mixed-race and monoracial couples in my sample across all observation periods. Consistent with past research using aggregate data, the table in Fig. 1 reveals a high level of neighborhood diversity at time t for mixed-race couples compared with monoracial couples. For instance, the mean entropy score is higher for each of the mixed-race couple-groups—except for white-other couples (0.601)—than for the average black-black couple (0.603) and especially the average white-white couple (0.407). Among mixed-race couples, however, experiences of neighborhood diversity vary considerably. Most notably, mixed-race couples with black partners have the highest mean entropy scores, with black-Latino couples living in the most diverse neighborhoods (0.754). In contrast, with the exception of black-white couples, couples containing one white partner tend to live in places that are less diverse than are those occupied by other mixed-race couples, although still more diverse than the average neighborhood occupied by white-white couples.
Figure 1 also highlights substantial differences in the specific neighborhood racial composition for couple-groups with similar entropy scores. For instance, white-Latino couples have an average entropy score of 0.645, and black-other couples have a slightly higher value of 0.682. Yet, when looking at the racial composition of their relative neighborhoods, white-Latino couples, on average, have drastically higher concentrations of white neighbors and far fewer black neighbors than do black-other couples. Generally, however, Fig. 1 confirms the central finding from aggregate-level studies, showing that individual mixed-race couples tend to be exposed to higher neighborhood diversity than are monoracial couples.
Table 2 presents the results of a three-level random-intercepts linear regression analysis of neighborhood entropy at time t for the couples in my PSID sample. Model 1 includes dummy variables for black-black couples and five mixed-race couple-groups, with white-white couples set as the reference group. Consistent with the statistics in Fig. 1, all coefficients for the mixed-race couple-groups are positive, indicating that these couple-groups tend to live in more diverse neighborhoods than do white couples. All these group differences in average neighborhood entropy are statistically significant (p < .001). The results also show that about 14 % ((0.0349) + (0.0437) + (0.0127) = 0.091, 0.0127 / 0.091 = 14 %) of the overall variation in neighborhood entropy is related to variation in entropy across the observations associated with the same couple, 48 % (0.0437 / 0.091 = 48 %) is related to variation between couples in the same metropolitan area, and 38 % (0.0349 / 0.091 = 38 %) is attributed to differences in outcomes across metropolitan areas.
In Model 2 of Table 2, I add the theoretically key variable of metropolitan entropy to my regression analysis to test whether mixed-race couples are exposed to higher levels of neighborhood diversity as a result of residing in more diverse metropolitan areas. The coefficient for metropolitan-level entropy is positive (b = 0.7761) and statistically significant (p < .001), indicating that residents of more diverse metropolitan areas tend to experience substantially higher levels of diversity in their neighborhoods of residence. Controlling for this significant effect also explains a sizable portion of the cross-metropolitan variation in the average neighborhood diversity. This is demonstrated by a decrease in the estimate for cross-metropolitan variation from 0.0349 in Model 1 to 0.0072 in Model 2. More important is the fact that metropolitan diversity explains a negligible share of the group differences in neighborhood entropy. The coefficients for black-black, black-white, black-Latino, black-other, and white-other groups decline slightly when metropolitan diversity is controlled, indicating that the relatively high level of neighborhood diversity experienced by these groups partially reflects their location in more diverse metropolitan areas compared with white-white couples. Overall, though, group differences in neighborhood diversity remain pronounced even when metropolitan diversity is controlled.
In subsequent models, I add controls for sociodemographic characteristics that have been known to influence residential attainment (cf. Wright et al. 2012). Model 3 shows a significant negative association between neighborhood entropy and age, indicating that younger couples tend to be located in neighborhoods that are more diverse than those occupied by older couples. Families with children are exposed to significantly lower levels of neighborhood diversity than are those without children, and the average level of neighborhood diversity increased over the period of my study as indicated by the positive and statistically significant coefficient for year of observation. Nevertheless, the addition of these measures causes basically no change in the coefficients for the couple-group categories.
In Model 4, I assess the extent to which neighborhood diversity is shaped by housing and socioeconomic characteristics: homeownership; whether a couple-group was in the same home for three or more years; employment status of the household head; educational attainment (in completed years) of the household head; and total family income (measured in thousands of constant 2000 dollars). All else being equal, homeowners tend to live in less diverse places than do non-owners, and longer-term residence is associated with exposure to more neighborhood diversity. Net of couple-group racial status and other measures, employment status, education, and income have negligible effects on neighborhood diversity. With the addition of these socioeconomic measures, the coefficients for couple-groups do not change substantively, indicating that group differences in socioeconomic resources play little role in explaining differences in mixed-race couples’ higher exposure to neighborhood diversity at any point in time.
Model 5 of Table 2 adds an interactive measure to assess the extent to which the effect of economic resources, as measured by total family income from all sources, vary across couple-group categories, providing a test of the weak and strong versions of the place stratification perspective. The coefficients for these product terms indicate significant differences across couple-groups in the effects of income. Consistent with the weak version of the place stratification perspective, the effect of income on neighborhood diversity is stronger for black-black couples than for white-white couples. Thus, although income has no net effect on the neighborhood diversity of white-white couples, it tends to be only higher-income black-black couples that buy their way into more diverse neighborhoods. In contrast, the effect of income is significantly more negative for black-white couples than for white-white couples. The fact that higher-income black-white couples—those presumably best able to afford a location that matches their preference—tend to be located in areas with significantly less diversity than those occupied by lower-income black-white couples would appear to contradict claims (Wright et al. 2011) that black-white couples tend to prefer living in diverse places. Moreover, supplemental models with black-black couples as the reference category reveal a negative (b = –0.0006) and significant interaction (p < .001) between black-white couples and income, further evincing that black-white couples are not necessarily unique in their preferences for neighborhood diversity.
Mobility Behaviors and Neighborhood Diversity
Like previous cross-sectional analyses, the results from the preceding analyses in Table 2 are temporally ambiguous: it is not clear whether mixed-race couples occupy more diverse locations because they move into such neighborhoods after they form as a couple or simply form as a couple in those neighborhoods. Thus, in the remaining analyses, I investigate couples that are already formed by time t and then examine their subsequent mobility behaviors in relation to neighborhood diversity. If, as past research has indicated, mixed-race couples have a heightened affinity for diverse neighborhoods, we would expect these couples to exhibit a relatively low likelihood of leaving diverse neighborhoods. Hence, Table 3 presents the results of three-level random-intercepts logistic regression models predicting the log-odds of leaving the neighborhood of origin between sequential PSID interviews. Model 1 includes the couple-group categories and a control for neighborhood entropy. The results show fairly modest group differences in the likelihood of inter-neighborhood mobility: in comparison with white-white couples, black-black and black-Latino couples are more likely to move. Controlling for these group differences, the odds of out-mobility are also higher for those originating in diverse neighborhoods. Specifically, on average, a difference in neighborhood entropy of 0.301 (1 standard deviation from the pooled sample) is associated with a 55 % (e(1.463 × 0.301) = 1.553) increase in the odds of out-mobility.
Model 2 includes interaction terms for each couple-group and neighborhood entropy. This approach allows me to assess whether the association between neighborhood entropy and out-mobility varies across couple-groups. For white-white couples, the effect of entropy is strong and positive (b = 2.040), indicating that a 1 standard deviation difference in neighborhood entropy is associated with an 84 % (e(2.040 × 0.301) = 1.847) increase in the odds of a white-white couple moving out of their neighborhood. In comparison, the effect of neighborhood diversity on out-mobility is significantly weaker for black-black couples. For these couples, a comparable difference in neighborhood entropy is associated with just a 28 % (e((2.040 – 1.199) × 0.301) = 1.288) increase in the odds of out-mobility, and supplemental models (not shown) indicate that this effect is not statistically significant (b = 0.5331, p = .083). This basic contrast in the mobility reaction to neighborhood diversity is consistent with evidence from past research suggesting that blacks have greater tolerance for diverse neighborhoods than do whites.
Also, among the significant interactions in Model 2, the reactions to neighborhood diversity for white-other and white-Latino couples more closely resemble those of white-white couples, whereas black-other couples demonstrate less sensitivity to increasingly diverse neighborhood settings. Conversely, only black-other couples differ significantly (b = –2.751, p < .05) from black-black couples in terms of the association between neighborhood entropy and out-mobility. Moreover, supplemental models (not shown) indicate that the association between neighborhood entropy and out-mobility is statistically significantly different from 0 only for white-white couples. These findings are roughly consistent with the argument that, at least among some mixed-race couples, the presence of a nonwhite partner produces a higher level of tolerance for diverse neighborhoods than is typically exhibited by white householders. Overall, these results provide mixed support for the idea that mixed-race couples make mobility decisions that are conducive to exposure to neighborhood diversity. Additionally, it is not clear whether the differences that do emerge reflect stronger preferences for diverse neighborhoods or the effects of sociodemographic or contextual factors that affect mobility decisions.
Model 3 shows that metropolitan entropy has little effect on the likelihood of mobility or, more importantly, group differences in reactions to neighborhood diversity. Model 4 includes controls for a broad array of individual- and household-level sociodemographic variables that have been shown in past research to affect neighborhood out-mobility. As in past research, I find that the likelihood of out-mobility declines with age; is lower for families with children, homeowners, long-term residents, and the employed; and increases with more people per room. The length of time between sequential PSID interviews also influences the likelihood of observing an inter-neighborhood move during this period.
The results in Model 4 show that the addition of these controls for sociodemographic factors reduces almost all the coefficients for the couple-group indicators to statistical non-significance.11 Only black-Latino couples are less likely to move than white-white couples. Thus, among couples with similar characteristics, there is essentially no difference in the likelihood of inter-neighborhood migration. Particularly important is that the coefficients for the interactions between couple-group categories and neighborhood entropy become statistically non-significant.12 Therefore, a substantial portion of the group differences in the association between neighborhood entropy and residential out-mobility is attributed to the fact that members of some groups have characteristics that increase their mobility in general.13 Based on the results of these out-mobility models, there is no evidence that preferences for neighborhood diversity are driving a heightened tendency among mixed-race couples to remain in more diverse neighborhoods.14
Racial Diversity of Destinations for Mobile Couples
Although the preceding analysis raises serious doubts about the argument that mixed-race couples’ exposure to relatively high levels of neighborhood diversity reflects stronger preferences for diverse neighborhoods, it is possible that these differences are reflected in the choice of destinations rather than in out-mobility decisions. Hence, Table 4 presents the results from a three-level random-intercepts linear regression analysis examining the level of diversity in neighborhoods entered by mobile members of each couple-group. In these models, I account for the non-random selection of individuals into the mover category through a Heckman correction strategy in which all predictors included in Model 4 of Table 2 are used in the selection equation.
In Model 1, the coefficient for the sample selection variable (λ) indicates a negative association between the level of neighborhood diversity in the tract of destination and the latent probability of out-mobility. Controlling for this selection process, I find a strong positive association between the level of diversity in the tract of destination and neighborhood entropy at time t. At least in part, this reflects the spatial clustering of neighborhoods with similar racial and ethnic compositions, which, in combination with the distance-dependence of migration (Long 1988), increases the likelihood that those originating in high-entropy neighborhoods will move to other relatively high-entropy neighborhoods.
Particularly salient is the fact that controlling for neighborhood entropy at the origin and the latent probability of moving, the coefficients for almost all the couple-groups are positive and statistically significant, indicating that mixed-race couples tend to move to destinations that are significantly more diverse than those entered by mobile whites. The contrast between black-Latino (b = 0.1719) and white-white couples is most pronounced, but even those mixed-race couples containing a white member tend to move into neighborhoods that are substantially more diverse than those selected by mobile white-white couples. Model 2 shows that controlling for metropolitan diversity—a key determinant of opportunities for the selection of diverse neighborhood destinations (b = 0.3645)—slightly attenuates most of the group differences in destination characteristics: a majority of the coefficients for the couple-groups are smaller in Model 2 than in Model 1. The exceptions to this pattern are the coefficients for black-white (b = 0.1211) and white-other (b = 0.0978) couples, which increase slightly with the control for metropolitan diversity.
In Model 3, I add controls for the sociodemographic variables of age, presence of children, homeownership, employment, education, family income, and year. Net of their influence on the latent probability of out-migration (captured in the coefficient for λ), and controlling for metropolitan diversity and the diversity of origin neighborhoods, the effects of these individual-level characteristics are modest. Only the coefficient for year of observation (b = 0.0051) is statistically significant (p < .001), indicating that mobile couples have tended to enter increasingly diverse neighborhoods over time. Given these modest effects, it is not surprising that controlling for individual characteristics does nothing to explain the relatively higher level of diversity in the destinations selected by mobile mixed-race couples.
Although the pattern of destination differences revealed in Table 4 are roughly consistent with arguments related to group differences in preferences for neighborhood diversity, it is not clear that preferences for neighborhood diversity are uniquely high among mixed-race couples. Furthermore, supplemental analyses indicate that most mobile mixed-race couples that include a black member do not differ significantly from black-black couples in terms of the diversity of their destination neighborhoods, with one exception: black-Latino couples tend to enter slightly more diverse neighborhoods than do mobile black-black couples (b = 0.1072, p < .05). Thus, the positive coefficients for couple-groups in the models in Table 4 may be seen more as a reflection of the strong preferences for low-diversity neighborhoods among couples containing white (but not black) members than the unique destination choices made by mixed-race couples.
In this article, I have endeavored to describe the association between mixed-race coupling and neighborhood location, as well as the role of underlying micro-level residential processes. Following theoretical arguments on neighborhood attainment, I have tested the relative roles of the racial characteristics, in addition to the sociodemographic and economic standing of these couples. Moreover, I have assessed the significance of metropolitan opportunity structures that shape residential patterns.
The results of my analysis of multilevel data from the PSID linked to neighborhood-level data drawn from the U.S. Census confirm past aggregate-level research showing that mixed-race couples face somewhat distinctive residential experiences. Especially in comparison with monoracial white couples, couples containing members of two different racial/ethnic groups tend to reside in neighborhoods with significantly higher levels of racial diversity—and, importantly, these differences remain significant even after controls are added for pronounced group differences in individual- and family-level characteristics and the conditions of the metropolitan areas in which different groups are located.
Following past research, such unique residential experiences might be seen as a reflection of unmeasured racial group differences in residential preferences, with mixed-race couples presumed to choose diverse neighborhoods that match the combined preferences of their individual members or offer some protection against the unique discrimination faced by mixed-race couples. However, my prospective analysis of residential mobility behaviors provides only limited support for this argument. Specifically, I find no support for the argument that mixed-race couples are more likely than monoracial couples to remain in highly diverse neighborhoods, net of the effects of other individual- and contextual-level factors that affect the likelihood of out-mobility. And, although I do find evidence that the relatively high level of diversity faced by mixed-race couples is likely shaped by their movement into neighborhoods that are more diverse than those selected by mobile white couples, the destinations of most types of mixed-race couples do not differ substantially from those of monoracial black couples. Based on these results, it is difficult to claim that mixed-race couples are unique in terms of their affinity for diverse neighborhoods. In fact, the results of my analysis—especially those related to variations in the effects of economic resources on neighborhood location—suggest that forces affecting the residential experiences of mixed-race couples extend beyond the realm of preferences.
These findings are important because they further illuminate the emerging residential patterns of a growing segment of the U.S. population who are part of diversifying American neighborhoods. However, despite the potential for the growing population of mixed-race couples to redefine prevailing patterns of residential stratification, these couples still appear to be subject to the same types of dynamics that maintain sharp distinctions between black and white residential outcomes. In other words, the results presented here, in part, highlight the continued salience of race in residential attainment.
Limitations and Future Directions
This initial investigation provides insight about the relevance of existing theoretical arguments to the residential experiences of mixed-race couples, but there are some limitations in its scope. For instance, past research found that individuals who enter into couples might have different types of residential outcomes than single individuals (Woldoff 2008). Thus, because of the non-random selection into couple status, the results of this analysis cannot be generalized to represent the residential outcomes of single individuals. Additionally, the data used in the present study do not contain information on individual preferences for neighborhood diversity and, therefore, are unable to distinguish the negotiation of personal preferences within couples.
Future research would benefit greatly from data with specific measures of the residential preferences of mixed-race couples: how they differ from those of monoracial couples as well as how they vary by individual characteristics, household composition, and the broader residential context. Similarly, direct measures of the potentially unique forms of discrimination faced by mixed-race couples would prove extremely useful in assessing the relative support for prevailing theoretical arguments on the topic. At this point, no data provide the information necessary to disentangle the relative effects of neighborhood preferences and racial discrimination experienced based on race.
Fruitful research might also be found in data that provide detailed race and ethnicity measures that capture broader dimensions of skin color and country of origin. Moreover, future research would profit from data that provide contextual information on other combinations of mixed-race couples. Furthermore, as noted earlier, the racial and ethnic neighborhood composition of mixed-race couples can be vastly different despite having relatively similar entropy scores. Hence, future work should test various measures of residential composition to illuminate the disparate neighborhood compositions of varying types of mixed-race couples. Additionally, research focusing on neighborhood disadvantage is likely to provide important insights into variations in more general processes of residential attainment. Scholars should also investigate how mobility decisions are negotiated within mixed-race couples and the potentially unique strategies they employ to navigate the housing search and acquisition process. What is apparent is that the study of residential patterns of mixed-race couples provides a unique opportunity to understand the complex and ever-changing dynamics of racial-residential stratification.
Partial support for this research came from a Shanahan Endowment Fellowship and a Eunice Kennedy Shriver National Institute of Child Health and Human Development training grant, T32 HD007543, to the Center for Studies in Demography & Ecology at the University of Washington. I wish to thank Kyle Crowder, Stewart Tolnay, Charles Hirschman, Mark Ellis, the editors of Demography, and anonymous reviewers for their very helpful comments on earlier versions of this article.
Individuals are included in the analysis only during observation periods in which they are a member of a couple. Members of couples that dissolve during the study period are dropped from the analysis for remaining years, returning only if they form another union. Supplemental analysis using a Heckman correction (inverse Mills ratio) for the selection of individuals into couple status produces results that are quite similar to those presented here. Nevertheless, given that this analysis focuses on couples, caution should be exercised in extrapolating the results to residential outcomes for individuals not in couples.
Data are available online (www.geolytics.com/USCensus,Neighborhood-Change-Database-1970-2000,Products.asp).
Values for intercensal years are assumed to fall on a straight line between values for endpoints defined by the most recent past and future decennial censuses.
Individuals are included in the analysis only for years in which they completed a PSID interview. In rare cases in which a respondent left the sample and then returned at a later date, observations for both periods are included in the analysis.
The “other” category includes individuals self-identifying as Asian, Native American, and other.
According to the 2010 U.S. Census, 97.3 % of all mixed-race couples contain a black or a white partner (U.S. Census Bureau 2010).
Despite efforts to enhance the representativeness of the PSID by adding Latino and immigrant subsamples, these groups are still underrepresented numerically and do not represent the broader populations of these groups.
The racial groups used to compose the entropy score are white, black, Latino, and other.
Models are estimated using the gllamm and xt suite commands in Stata 13 (StataCorp 2013).
There is too little clustering of PSID respondents within census tracts to justify the analysis of an additional level of hierarchy.
Supplemental analyses controlling for a wide range of other neighborhood characteristics—including the local poverty rate, labor force participation rate, percentage of adults with a college education, median housing values, and housing vacancy rates—had no impact on the observed group differences in the reaction to neighborhood diversity.
When the reference category is black-black couples, only the interaction for black-Latino couples and neighborhood entropy is positive (b = 1.798) and significant (p < .05), meaning that as neighborhood entropy increases, black-Latino couples have a higher likelihood of out-mobility than black-black couples.
Supplementary analyses find that group differences in homeownership are the main source of group differences in residential mobility and the interaction between couple-groups and neighborhood entropy.
Supplemental models estimating polynomials of both neighborhood and metropolitan entropy provide little indication of nonlinear effects of either neighborhood or metropolitan diversity, and the inclusion of these squared terms does not alter the substantive findings from the main models.