Our study investigates the diversification and fragmentation theses, fueled by claims that greater diversity is reshaping the social fabric of American life and that the United States is an increasingly fragmented nation. We take a multidimensional view of heterogeneity that considers whether growing ethnoracial diversity within U.S. communities (i.e., incorporated and unincorporated places) has resulted in the consolidation or differentiation of demographic, sociocultural, and economic distinctions between 1980 and 2010. As communities have become more ethnoracially diverse, they have become more heterogeneous in language and nativity—two characteristics tied closely to Latino and Asian population growth. However, ethnoracial diversity within communities is only weakly associated with household, age, educational, occupational, or income heterogeneity despite large racial/ethnic differences in these characteristics nationally. This trend does not apply to all forms of ethnoracial diversity equally: Hispanic and especially Asian population growth is more likely to generate community sociodemographic and economic heterogeneity than is black population growth. Consistent with the fragmentation hypothesis, we also find that broader geographic context matters, with more ethnoracially diverse metropolitan and micropolitan areas experiencing reduced social and economic heterogeneity inside their constituent places. We conclude by discussing the social implications of these patterns for intergroup relations, spatial exclusion, and ethnoracial inequality.
Demographic changes have propelled the United States to become a more racially and ethnically diverse nation over the past half-century (Colby and Ortman 2015). Some politicians, public intellectuals, and academics have offered a positive assessment of this trend, arguing that the rise in ethnoracial diversity has benefited the nation by fueling economic growth, promoting tolerance, and blurring racial boundaries (Alba and Nee 2003; Card 2005; Lee and Bean 2004; Portes and Vickstrom 2011; Rieder and Steinlight 2003). Others have been more negative, concluding that greater ethnoracial diversity has increased political polarization, diminished cultural cohesion, and weakened the fabric of American society (Alesina et al. 1999; Bishop 2008; Huntington 2005; Luttmer 2001; Putnam 2007).
Those raising concerns about diversity tend to do so for two reasons, contingent on the type of population sorting they find most troubling. The first camp argues that national trends have increased the average American’s exposure to diverse others, thereby undermining traditional American values, decreasing cohesion and trust, and fostering interpersonal and institutional conflict (Alesina and La Ferrara 2002; Buchanan 2011; Huntington 2005; Putnam 2007). The second camp contends that national trends have reduced the average American’s exposure to diverse others, a result of the fragmentation of American life in which similar individuals increasingly cluster within distinct geographies, institutions, and subcultures (Bishop 2008; Fischer et al. 2004; Jargowsky 2015; Massey 1996; Murray 2012).
Growing ethnoracial diversity has the potential to reshape American society, but the consequences of greater diversity depend in part on how it restructures the social and economic composition of communities, the settings where people live and are exposed to residents who exhibit a range of characteristics. In particular, does growing ethnoracial diversity result in greater differentiation in other aspects of social and economic life at the local level? Our study tackles the structural foundations of this question by asking whether the shifting ethnoracial composition of American communities (defined here as incorporated and census-designated places) resulted in the consolidation or differentiation of the local population along social, demographic, and economic lines between 1980 and 2010. We consider indicators of economic heterogeneity as measured by education, occupation, income, and housing tenure; and indicators of sociodemographic heterogeneity in terms of age, language, nativity, and family structure. Prior research has typically examined diversification or fragmentation along a single parameter, such as race, income, or education, but this univariate approach fails to consider how multiple population attributes come together within communities. Our approach follows that of Blau (1977, 1994; Blau and Schwartz 1984), who conceptualized heterogeneity as the distribution of a population across multiple categories of group membership.
Most prior research has also examined diversity at either the micro level (such as neighborhoods) or the macro level (such as metropolitan areas), overlooking places as important meso-level geographic units that align with governmental jurisdictions, housing markets, school districts, and community reputations. Throughout this article, we define places as the Census Bureau does: as cities, suburbs, towns, boroughs, and villages. Most places are incorporated, which means that they have legal standing and provide local government functions to residents. Unincorporated, or census-designated, places are defined to capture local population concentrations that are recognizable by name but lack a formal legal basis. Our examination of the correlations among multiple dimensions of population composition within local places provides a foundation for understanding the broad consequences of greater ethnoracial diversity. We also look at how places are influenced by the diversity of the metropolitan or micropolitan contexts in which they are embedded.
For our purposes, diversity refers to the distribution of the population across racial/ethnic categories, with minimum levels of diversity when only one group is present in a place and maximum levels achieved when each group is represented equally. Although we do not consider it here, some recent scholarship has considered the diversity within specific racial/ethnic groups as well. For example, black immigrants are viewed differently in the United States than U.S.-born African Americans and thus are treated differently, contributing to distinct economic and social outcomes for the two groups (Freeman 2002; Habecker 2012; Massey et al. 2007; Waters 1994, 2000). Immigrant populations themselves can also be quite diverse in terms of national origin, socioeconomic status, and legal status—a phenomenon sometimes known as super-diversity (Meissner and Vertovec 2015; Vertovec 2007). To date, empirical examinations of super-diversity have been more common in European nations than in the United States.
Trends in Diversification and Fragmentation
The United States has always been a heterogeneous and geographically dispersed nation, and Americans have oscillated between viewing these characteristics as a source of national strength or weakness (Fischer and Mattson 2009; Rieder and Steinlight 2003). Two distinct streams of scholarly literature address the demographic underpinnings of American diversity. One stream has explored the growing ethnoracial diversification of American society at various geographic scales, and the other has highlighted increasing geographic fragmentation of the population along demographic, economic, and cultural dimensions.
The U.S. white population share has declined since the 1980s, and it is expected to continue its decline over the next several decades. By 2060, the Census Bureau projects that whites will lose their majority status, Hispanics will constitute almost one-third of the population, and the combined representation of Asian and multiracial individuals (9 % and 5 %, respectively) will rival or surpass that of blacks (Colby and Ortman 2015). These changes have the potential to affect economic and social heterogeneity because large differences exist in the economic and social characteristics of ethnoracial groups nationally. If growth in ethnoracial diversity at the national level were translated to the local level, residents of majority-white towns and cities undergoing ethnoracial diversification might observe their neighbors become poorer, find their schools filling with nonwhite children, and see and hear languages other than English during their daily activities.1
Some scholars have argued that racial and ethnic diversification is a trend that has been shared broadly across different regions of the country and in geographic units ranging from the neighborhood to the metropolis (Farrell and Lee 2011; Frey 2015; Lee et al. 2014; Wright et al. 2014). Large metropolitan areas have had populations diversify and neighborhood segregation levels fall in response to minority and immigrant population growth (Logan and Stults 2011; Logan and Zhang 2010; Rugh and Massey 2013) and, in some instances, to white population loss (Frey 2011b; Johnson and Lichter 2010). Although most growth in ethnoracial diversity has been concentrated in urban cores, inner-ring suburbs have experienced growing shares of blacks, Latinos, and Asians as well (Frey 2011a; Hall and Lee 2010; Lichter 2013; Logan and Zhang 2010; Singer et al. 2008). Diversity has also spread to smaller metro areas and rural locations, with selected places becoming new immigrant destinations due to influxes in nonwhite populations and high nonwhite fertility rates alongside stagnating white populations (Hall 2013; Lee and Sharp 2017; Lichter 2012; Marrow 2011, 2013; Massey 2008).
Although national ethnoracial diversity is clearly on the rise, scholars have debated the uneven spread of diversity across the different geographic scales of metropolitan areas, places (e.g., cities, suburbs, towns), and neighborhoods (Frey 1995; Lichter et al. 2015). Segregation at the neighborhood level has declined for certain racial groups since the 1980s, but this has been offset in part by growing segregation among larger geographic units, including places (Farrell 2008; Fischer et al. 2004; Holloway et al. 2012; Iceland 2004; Lichter et al. 2015; Logan and Stults 2011). In particular, prior research has shown that growing ethnoracial diversity at the metropolitan level leads to growing segregation of whites from other groups at the place level (Iceland 2004; Lichter et al. 2015; cf. Parisi et al. 2015). Similarly, researchers have found that adjacent places tend to move together toward greater or lesser diversity and that this process plays out at both the metropolitan and submetropolitan level (Martin and Fowler 2018). Together, this body of work underscores the growing importance of places as a distinct geographic context relevant for understanding patterns of racial/ethnic sorting.
Many empirical examinations of fragmentation also have considered the particular economic and sociodemographic dimensions along which Americans segregate (see Fischer and Mattson 2009 for a review). On this front, there is clear evidence that the United States has become more segregated along economic lines, fueled primarily by growing spatial separation of high-income and highly educated households from others across large geographic scales (Domina 2006; Fischer et al. 2004; Massey 1996; Owens 2016; Quillian 2012; Reardon and Bischoff 2011; Vesselinov et al. 2007). Only a handful of studies have considered the fragmentation of population characteristics other than race and economic status, but this work suggests that segregation is lower by age or family type in the United States than by race and class and that fewer changes in segregation have occurred along the former dimensions since the 1960s (Fischer and Hout 2006; Fischer et al. 2004; Tittle and Rotolo 2010).
Pundits and scholars concerned with fragmentation have highlighted the geographic and demographic underpinnings of American cultural and political divides in their discussions of “red states versus blue states,” the “culture wars,” and what Bishop termed “the big sort” (Bishop 2008; Gelman 2009; Wuthnow 1989), yet the empirical evidence on geographic, cultural, and political polarization shows much less change over time than media accounts or public beliefs would suggest (DiMaggio et al. 1996; Fiorina and Abrams 2008; Fischer and Mattson 2009; Klinkner 2004). Most of this work has examined national-level trends, however, so the sociopolitical consequences of racial fragmentation at the local level remain unclear. Researchers have argued that such sorting could undermine the quality of public institutions and amenities by eroding the local tax base and exacerbating disparities in education, health, employment, and housing market outcomes (Ananat 2011; Hall et al. 2015; Lichter al. 2015; Mayer 2001; Quillian 2014).
Prior research has typically examined diversification or fragmentation of a single parameter, such as race/ethnicity, income, or education. This approach lacks a multidimensional view of heterogeneity that considers how multiple population characteristics covary within communities. In his seminal work on heterogeneity and inequality, Blau (1977, 1994; Blau and Schwartz 1984) argued that heterogeneity—the distribution of a population across multiple categories of group membership—is positively associated with the likelihood of intergroup relations. When heterogeneity along different dimensions is consolidated, or highly correlated, members of one particular in-group are more likely to share group membership along other dimensions as well, such as when members of a distinct ethnoracial group also share a distinct religion, occupation, or family type. When parameters are only weakly correlated, however—what Blau called multiform heterogeneity—fewer subgroups are perfectly homogeneous in every way. People who belong to different groups in one respect are likely to share group membership along other dimensions.
Blau’s distinction between consolidated and multiform heterogeneity suggests two alternative empirical consequences of growing ethnoracial diversity for other forms of sociodemographic and economic heterogeneity. If individuals of different races also differ along social and economic lines, as national-level statistics indicate, growing ethnoracial diversity should lead to greater social and economic heterogeneity within a place. However, the sorting of racial/ethnic groups may not reproduce national-level patterns of social and economic differentiation if groups select into places that match their social and economic characteristics or if selective outmigration occurs in response to growing racial/ethnic diversity. This would lead to either a null or a negative association between growing racial/ethnic diversity and other forms of population heterogeneity.
Certain types of social and economic heterogeneity may be particularly salient for social relations between groups. Residents of racially diverse communities are less likely to interact and offer social support when they differ along social and demographic characteristics, such as occupation, parental status, or life cycle stage (Maly 2008; Nyden et al. 1997; Wilson and Taub 2006). Family type and life cycle appear to be particularly salient for white households, given that flight from areas undergoing racial and socioeconomic diversification appears to be strongest among whites with school-aged children (Goyette et al. 2014; Hall and Hibel 2017; Owens 2016; Van Hook and Snyder 2007) and support for investments in local institutions like public education decline among elderly whites in areas undergoing rapid growth among youthful Latino populations (Figlio and Fletcher 2012; Poterba 1997, 1998). Linguistic and citizenship differences between Hispanic and white populations can further trigger interpersonal and political tensions in local communities (Hopkins 2009, 2010; Marrow 2011, 2013; Winders 2013).
Variation by Racial/Ethnic Structure
The processes of diversification may not unfold in the same fashion for all racial/ethnic groups given their distinct modes of incorporation and historical experiences of exclusion in American society, the latter ranging from enslavement (for African Americans) to banishment and internment (for the Chinese and Japanese, respectively) (Fong 2008; Iceland 2017; Lieberson 1980; Portes and Rumbaut 2006).
Spatial assimilation theory posits that nonwhites convert socioeconomic gains and acculturation into geographic gains in the form of higher-income and whiter neighborhoods (Alba and Logan 1991, 1993; Massey 1985; Rosenbaum and Friedman 2007; South et al. 2008). This process of locational attainment may provide such groups with access to more advantaged neighbors and local institutions. In examinations of individual-level residential mobility patterns, empirical support for the spatial assimilation hypothesis has been found for Latinos and Asians, for whom upward mobility and greater acculturation correspond with moves to neighborhoods and towns with higher incomes and more integration with whites (Alba et al. 2000; Crowder and South 2005; South et al. 2008). Despite the emphasis in spatial assimilation research on individual mobility patterns and integration at the neighborhood scale, the original theory was developed and tested on individual mobility patterns in suburban communities comparable to the places examined in this article (Alba and Logan 1993; Logan and Alba 1993).
Although spatial assimilation theory has been used to explain the divergent residential attainment of racial/ethnic groups, the theory does not consider how communities themselves might change in response to racial/ethnic population shifts. In other words, this literature helps explain the neighborhood attainment of individuals but not how those neighborhoods change as a result of individual residential decisions. At the community level, spatial assimilation theory might imply that diversification fueled by Latino and Asian population growth will result in greater social or economic heterogeneity as members of those groups move to whiter areas of higher socioeconomic standing. However, it might also imply that Latino and Asian population growth will do little to alter community social or economic characteristics as members of these groups move to areas that reflect their newly attained status.
Place stratification theory provides an alternative model of locational sorting, in which groups with more power actively seek to maintain their physical and social separation from other lower-status groups (Charles 2003; Logan and Molotch 1987; Massey and Denton 1993). Exclusionary land use regulations, housing market practices, and racial discrimination create strong barriers to entry into higher-income or higher-opportunity communities for some minority groups, particularly African Americans (Pager and Shepherd 2008; Rothwell and Massey 2009; Sharkey 2013; South and Crowder 1997; Turner and Ross 2005; Yinger 1995). Furthermore, when African Americans do move to racially integrated communities, white flight and white avoidance can erode community economic resources (Crowder and South 2008; Farley et al. 1994; Frey 1995; Krysan et al. 2009; Quillian 1999). Applied to broader patterns of community racial/ethnic diversification, place stratification theory implies that racial/ethnic diversification fueled by African American population growth may not yield greater social or economic heterogeneity because residents are excluded from communities with greater relative economic and institutional resources. It may even imply a negative association between African American population growth and other types of heterogeneity if it prompts the out-migration of more socioeconomically advantaged residents or reductions in other community resources (Quillian 1999; Wilson 1987).
Geographic Scale and the Importance of Place
We regard census places—individual cities, suburbs, towns, boroughs, and villages—as an especially relevant type of community for an analysis of diversity. Incorporated places are legally responsible for developing fiscal or policy responses to diversity-related issues that arise within their borders. They can also take steps to discourage or promote diversification. For example, some places have used annexation, zoning, or other measures to deter minority growth and preserve economic homogeneity (Lichter et al. 2007; Pendall 2000; Rothwell and Massey 2009). At the opposite extreme, immigrant-fueled diversification is occasionally pursued as a revitalization strategy by places experiencing demographic and economic decline (Carr et al. 2012). Consistent with sociological definitions of place (Gieryn 2000), census places constitute symbolic as well as political entities: residents recognize them by name and feel attached to them (Bader and Krysan 2015). The sociopolitical power accorded to place influences the ethnoracial composition of schools, work settings, and voluntary organizations, and ultimately, the social relationships that form in these locales. Of course, residents will sort into residential neighborhoods that are more homogeneous than the entire place, particularly in large cities. However, as shown in activity space research, people frequently traverse across many census tracts during the course of a day, and those spaces tend to be more diverse than one’s residential census tract (Browning and Soller 2014; Hall et al. 2019; Jones and Pebley 2014; Matthews 2011).
In addition to serving as meaningful units of analysis for studying the consequences of diversity, places are also nested within larger regional housing and labor markets and respond to conditions within those surrounding areas, which the Census Bureau defines as core-based statistical areas (CBSAs). CBSAs can be metropolitan or micropolitan areas, containing at least one county with an urban center plus any adjacent counties connected by employment and commuting patterns. CBSAs approximate regional housing and labor markets, and they are the macro-level geographic units used by researchers and policymakers to measure dynamics of population sorting, such as residential segregation (Charles 2003; Massey and Denton 1988). A growing body of research has identified that people are increasingly sorting among places within metropolitan areas rather than neighborhoods, making places important “segregation-generating” actors (Lichter et al. 2015:851). For example, growing ethnoracial diversity at the metropolitan level leads to growing segregation of whites from other groups at the place level (Iceland 2004; Lichter et al. 2015; cf. Parisi et al. 2015). In a recent analysis, Lichter et al. (2015) decomposed changes in metropolitan racial segregation since 1990 across different geographic scales and found that segregation among places has increased, offsetting declines in segregation at the tract level within places. They determined that segregation among places now accounts for at least one-half of metropolitan segregation in the most segregated areas of the country. This meso-level spatial sorting is particularly pronounced in metropolitan areas marked by political fractionalization, which have a large number of independent municipal governments that can use local policies, such as exclusionary zoning, to restrict racial and socioeconomic diversification (Massey et al. 2009; Parisi et al. 2015; Pendall 2000; Rothwell and Massey 2009). If growing ethnoracial diversity at the metropolitan level leads to ethnoracial fragmentation among the constituent places, metro-level ethnoracial diversification may have hastened place-level social and economic fragmentation as well.
The Present Study
In this study, we extend prior research by examining the ethnoracial diversification and fragmentation of the U.S. population within local communities in a multidimensional context. Prior research has typically taken a univariate approach to this topic, which overlooks how distinct population characteristics come together within a community. Specifically, we ask whether growing ethnoracial diversity at the metropolitan or place level leads to growing heterogeneity within places along other sociodemographic and economic dimensions. Our analysis attempts to adjudicate among three competing hypotheses derived from the preceding discussion:
Hypothesis 1 (H1): Consolidated diversity. Given significant differences in the social and economic profiles of ethnoracial groups at the national level, growing ethnoracial diversity within places will be positively correlated with growth in other forms of economic and social heterogeneity.
Hypothesis 2 (H2): Multiform diversity. If racial/ethnic groups sort into places that reflect their social or economic characteristics, a null association between ethnoracial diversity and other forms of heterogeneity would occur.
Hypothesis 3 (H3): Fragmentation. Growing racial/ethnic diversity may hasten economic and social fragmentation at the local level, resulting in a negative association between ethnoracial diversification and other forms of sociodemographic and economic heterogeneity.
We expect that the processes of interest will not be uniform across racial/ethnic groups in light of their distinct historical legacies of incorporation and exclusion in the United States. Spatial assimilation and place stratification theories, discussed earlier, suggest that Asians and Hispanics should be able to access communities that are either equally or more advantaged than their own status, resulting in consolidated diversity. African Americans may be less able to do so, and may be more likely to spur outmigration among more advantaged groups, resulting in either null or negative associations with other dimensions of population heterogeneity. These theories suggest the following specific hypothesis:
Hypothesis 4 (H4): Group variation. The ethnoracial diversification of places due to growth in Asian and Hispanic populations will yield greater social and economic heterogeneity than will diversification due to African American population growth.
Finally, researchers have found that growing ethnoracial diversity at the metropolitan level is correlated with greater ethnoracial fragmentation among towns and cities. Put differently, metro diversity may hasten the sorting of racial/ethnic groups into more homogeneous localities. This leads us to our final hypothesis:
Hypothesis 5 (H5): Multilevel fragmentation. Growing ethnoracial diversity at the metropolitan or micropolitan geographic scale will result in less economic and social heterogeneity within constituent places, in line with the fragmentation thesis.
Data and Sample
Following most prior studies of community-level racial/ethnic diversity, our analyses employ summary file data from the decennial census and the American Community Survey (ACS). These data offer unparalleled coverage of racial/ethnic composition by asking racial/ethnic identity questions of the entire U.S. population at each decennial census. This full population coverage allows us to generate reliable measures of racial/ethnic diversity for small geographic units. We focus on the period 1980–2010 for both substantive and data-driven reasons. Substantively, this period covers a time of remarkable racial/ethnic diversification of the U.S. population (Colby and Ortman 2015). Although racial/ethnic diversity was increasing prior to 1980, the Hispanic origin question in the decennial census has been asked of the full population only since 1980. For 1980–2000, we use both long- and short-form decennial census data for racial and nonracial measures. For 2010, we use the short-form decennial census for demographic and racial measures. Because most economic and social measures are not available in the decennial census after 2000, we rely on the 2008–2012 five-year ACS estimates for the 2010 measures.2
Places serve as our geographic unit of analysis, approximating local communities. As mentioned earlier, the Census Bureau defines places as specific, identifiable communities that serve as county subunits. Incorporated and unincorporated places accounted for about 74 % of the U.S. population in 2010; the remaining 26 % lived in open country outside areas of population concentration. To obtain a stable set of places across the time frame for our study, we omit places that (1) had total populations less than 1,000 in 1980 because the Census Bureau suppressed detailed data for these small places in that year (N = 9,349); (2) first appeared in census tabulations in 1990 or later (N = 8,444); (3) changed designation after 2010 and are not included in the 2008–2012 ACS place data (N = 16); or (4) were combined into another statistical unit for that decade but then were reestablished at a later time (N = 9). Taken together, the selection criteria yield a final analytic sample of 11,439 places with 1,000 or more inhabitants in 1980 and complete data for all four time points.3
Although the number of places remains constant in our analysis, the same cannot be said for the areal size of individual places. Most places are economically, politically, and socially relevant entities partly because of their jurisdictional status. Unlike census tracts, their boundaries are not derived simply for statistical aggregation purposes; rather, place boundaries affect how residents experience diversity and how local governments and other institutions respond to its effects. Changes in diversity caused by the annexation or ceding of land would matter throughout the community as reconstituted, not just inside its old or new territory. Thus, we allow the boundaries of our sample places to vary across census years. We tested the sensitivity of our analyses to this decision and found that major boundary shifts are relatively infrequent and leave basic diversity patterns intact.4 We nevertheless include land area size as a control in the multivariate models. We also include a control for population size that accounts for the differing geographic scales of places, ranging from the large population centers of principal cities to smaller suburban and exurban jurisdictions.5
Part of our analysis examines places embedded within larger CBSAs, which are metropolitan and micropolitan areas that consist of counties with urban centers of at least 10,000 (micropolitan) or 50,000 (metropolitan) people and the adjacent counties that are tied economically to the urban centers through commuting patterns. We omit a few CBSAs because all constituent places within that CBSA had populations less than 1,000 in 1980 or were omitted for one of the aforementioned reasons. This results in a final sample of 932 CBSAs, 366 metropolitan and 576 micropolitan. In 2010, 84 % of the U.S. population resided in metropolitan areas, and another 10 % of the population lived in a micropolitan area (Mackun and Wilson 2011).
Table A1 in the online appendix details the categories we use to construct each diversity measure in our analysis. Our five-group ethnoracial diversity measure comprises Hispanics; non-Hispanic whites; non-Hispanic blacks; non-Hispanic Asians or Pacific Islanders; and non-Hispanic others (including American Indians/Alaska Natives, multirace persons, and individuals reporting another race). We also create nonracial measures to capture the demographic, economic, and social heterogeneity of the population. These measures include four sociodemographic characteristics (household type, age, language, and nativity) and four economic characteristics (educational attainment, occupation, household income, and housing tenure).
In this fixed-effects model, the interpretation of λ reflects the association between changes in ethnoracial diversity and changes in nonracial heterogeneity within a place. The place fixed effects also absorb all time-constant differences across places. Positive and significant values on the racial-diversity coefficient would support the consolidated diversity hypothesis that growing ethnoracial diversity is associated with social and economic (i.e., nonracial) heterogeneity. A null coefficient would support the multiform diversity hypothesis that changes in racial/ethnic diversity do not alter nonracial heterogeneity. Negative values would support the fragmentation hypothesis that growth in racial/ethnic diversity yields populations that are more sociodemographically or economically homogeneous.
Note that the temporal structure of the data captures the contemporaneous associations between changes in racial diversity and nonracial dimensions of population heterogeneity over the course of a decade. Some of the change may result because those moving into (or being born into) a given place have characteristics that differ from the existing community, but some change may be due to selective outmigration of residents in response to growing ethnoracial diversity. Census data do not allow us to distinguish between these two different mobility processes; we are able to observe only the end result.8 The analyses presented here are therefore descriptive rather than causal because the changes in ethnoracial diversity occur during the same period as the changes in sociodemographic or economic heterogeneity.
Finally, the third part of our analysis examines whether the associations between ethnoracial diversity and nonracial heterogeneity depend on geographic scale. To test the multilevel fragmentation hypothesis that more ethnoracially diverse CBSAs (metropolitan and micropolitan areas) contain less sociodemographically and economically heterogeneous places, we estimate a multilevel model where census decades are nested within places, which are in turn nested within CBSAs. We add CBSA-level measures for ethnoracial diversity and nonracial heterogeneity to Eq. (1). We also include a control for political fractionalization, measured as the number of distinct census-defined places within the CBSA, to account for variation in the potential for place-based sorting among CBSAs. Negative and significant coefficients on the CBSA-level diversity variables would support the multilevel fragmentation hypothesis that community-level heterogeneity is diminished in CBSAs with more ethnoracially diverse populations.
National Patterns of Ethnoracial Diversity and Nonracial Heterogeneity
Table 1 highlights differences in the distributions of the social and economic characteristics among ethnoracial groups at the national level in 2000. Despite a great deal of heterogeneity in household type within each group, whites are considerably more likely to live in married-couple households without children than other groups, and Hispanics and Asians are more likely than other groups to live in married-couple households with children. By contrast, a much larger share of the black population lives in nonmarital family types with children. Age distributions also vary across racial/ethnic groups, with Hispanics exhibiting the most youthful age structure and whites having the oldest. Even more extreme divergences are apparent in linguistic and nativity characteristics across groups: whites and blacks have small shares of immigrants and non-English speakers, but Hispanics and Asians have great heterogeneity on these features. However, these last divergences are quite predictable—and relatively uninteresting—given the bundled nature of race, nativity, and language. The Other Race category also has high levels of heterogeneity, which is unsurprising because this residual category contains multiple distinct groups, such as American Indians and multiracial individuals. We present results for this category to achieve mutually exclusive and exhaustive coverage of the population, but we do not devote much space to the interpretation of the Other Race results because of the multifaceted composition of this residual category.
Ethnoracial groups also differ considerably in their economic characteristics. This is reflected in the educational profiles of each group, with almost half of Hispanics having no high school credential but more than 40 % of Asians having at least a four-year college degree. It is also reflected in the occupational distributions of each group, with blacks and Hispanics overrepresented in the service industries, and whites and Asians overrepresented in the managerial and professional sectors. These differences among groups translate into racial/ethnic income inequality: almost one in three blacks and one in five Hispanics have household incomes less than 40 % of the national median, whereas almost one in three whites and two in five Asians have household incomes 1.6 times the national median. Economic differences are further manifested in the types of housing in which different groups reside: the modal white household is a homeowner, whereas the modal black, Asian, and Latino household is a renter.
Consistent with prior literature, Table 2 shows that the average ethnoracial diversity of U.S. places, measured by a five-group entropy score, almost doubled in the past four decades, from 19.95 in 1980 to 36.58 in 2010. The linguistic and nativity heterogeneity of communities also grew, although the pace of change was more modest. Less change is evident in the average heterogeneity of communities along other demographic and economic dimensions. Household type, age, educational, occupational, and income heterogeneity within communities were persistently high over this period. However, the average diversity score across all places can obscure substantial year-to-year change on these characteristics within communities over time. The within-place correlations between 1980 and 2010 on many indicators of heterogeneity are high (r > .7), but correlations over time are much lower for measures of household type, and educational and income heterogeneity, indicating greater temporal variability.
Fragmentation or Diversification?
Because of differences in the social and economic characteristics of ethnoracial groups at the national level, we might expect greater ethnoracial diversity at the place level to result in greater heterogeneity on social and economic indicators as well, consistent with the consolidated diversification hypothesis (H1). However, this assumes that there is little sorting along nonracial lines, and the multiform diversification and fragmentation hypotheses (H2 and H3) predict that such sorting has indeed occurred: the former in a way that reproduces existing levels of heterogeneity (null association) and the latter in a way that actually reduces heterogeneity (negative association).
Table 3 shows the bivariate correlations between ethnoracial diversity and each nonracial heterogeneity measure in each decade. In this table and in subsequent analyses with standardized regression coefficients, we follow Cohen’s (1988) widely used conventions for interpreting effect sizes: r < .1 = none; .1 < r < .3 = weak; .3 < r < .5 = moderate; and r > .5 = strong. Racial/ethnic diversity is strongly correlated with linguistic and nativity heterogeneity (r > .5), and this association became even stronger between 1980 and 2010. These strong correlations are to be expected given the pronounced linguistic and nativity differences that exist among racial/ethnic groups at the national level. But this is where the strong positive correlations end. Racial/ethnic diversity is weakly correlated with household type heterogeneity and tenure type heterogeneity and—aside from a weak association with education heterogeneity—is uncorrelated with the economic heterogeneity of the community in terms of income and occupation. Finally, racial/ethnic diversity has a weak negative correlation with age heterogeneity in the cross section, indicating that places where racial/ethnic diversity is higher have less diverse age distributions.
We see a similar pattern for the correlations between changes in ethnoracial diversity and changes on the other diversity indicators in Table 4. Changes in ethnoracial diversity are moderately to strongly correlated with changes in linguistic and nativity heterogeneity, indicating that place populations became more diverse in terms of language and citizenship when they became more racially or ethnically diverse. However, changes in community-level ethnoracial diversity are either weakly correlated (household type and income; .1 < r < .3) or uncorrelated (age, education, occupation, and tenure; r < .1) with other diversity measures.
The multivariate fixed-effects regressions in Table 5 expand on the bivariate correlations discussed thus far by regressing each nonracial heterogeneity measure on ethnoracial diversity, dummy variables for census years, interactions between ethnoracial diversity and census year, controls for land area and population size, and place fixed effects. These racial/ethnic diversity coefficients can be interpreted as the association between changes in racial/ethnic diversity and changes in nonracial heterogeneity within a place over time. We convert the racial/ethnic and nonracial entropy measures into standardized scores, which allows the regression coefficients to be interpreted in the same metric as the correlations reported earlier and to be compared across models.
The fixed-effects results reinforce what we observe in the bivariate correlations in Tables 3 and 4. Ethnoracial diversity is strongly associated with linguistic heterogeneity and moderately associated with nativity heterogeneity, and these associations become significantly larger over time with the growth in the national share of foreign-born residents. By contrast, changes in ethnoracial diversity are weakly and positively associated with changes in household type heterogeneity, tenure type heterogeneity, and household income heterogeneity (.1 < r < .3); unassociated with changes in age heterogeneity and occupational heterogeneity (r < .1); and weakly but negatively associated with changes in educational heterogeneity (–.1 < r < –.3).
Although many of the ethnoracial diversity coefficients in Table 5 reach conventional levels of statistical significance because of our large sample size, the substantive significance of the coefficients is modest. We illustrate this by plotting the slopes for each nonracial heterogeneity variable from Table 5 in Fig. 1 for the year 2010. Characteristics strongly associated with more racial/ethnic diversity (r > .5) are in black, those with weak associations are in gray (.1 < r < .3), and those with no association (r < .1) are in dashed lines. Except for nativity and linguistic heterogeneity, the nonracial measures of heterogeneity vary little across the distribution of racial/ethnic diversity. In sum, the results from Tables 3, 4, and 5 offer support for the consolidated diversification hypothesis (H1) for linguistic and nativity heterogeneity, and the weak associations between changes in ethnoracial diversity and other dimensions of sociodemographic and economic heterogeneity are more consistent with the multiform diversity hypothesis (H2).
Variation by Racial/Ethnic Composition
The next phase of our analysis examines whether racial-nonracial associations vary by the underlying racial/ethnic composition of places (H4). Table 6 presents fixed-effects regressions of each nonracial heterogeneity measure on race-specific population shares. Each coefficient comes from a separate regression, and the coefficients report how a 1 percentage point increase in the share of non-Hispanic blacks (for example) changes each measure of nonracial heterogeneity.9 We also summarize these results in Fig. 2. Consistent with place stratification theory, we see little association between growing non-Hispanic black population shares and growing sociodemographic heterogeneity in line with multiform diversification. A 10 percentage point increase in the share of black residents in a community is associated with about a 0.1 standard deviation increase in household type heterogeneity but less than a 0.1 standard deviation change in any other form of social, economic, or demographic heterogeneity.
Increasing Hispanic and Asian population shares do more to alter community demographic and economic heterogeneity. A 10 percentage point increase in the share of Hispanic residents in a place is associated with about a 0.5 standard deviation increase in linguistic and nativity heterogeneity, about a 0.2 standard deviation increase in family and educational heterogeneity, and about a 0.1 standard deviation increase in occupational heterogeneity. Growing Asian population shares exert an even stronger influence on other dimensions of place heterogeneity. A 10 percentage point increase in the share of Asian residents is associated with a 0.8 standard deviation increase in linguistic heterogeneity, about a 0.6 standard deviation increase in nativity heterogeneity, about a 0.4 standard deviation increase in education and income heterogeneity, and a 0.3 standard deviation increase in age and occupational heterogeneity.
Taken together, these findings support the group variation hypothesis, indicating that the social and economic consequences of ethnoracial diversification depend on the type of racial/ethnic change occurring within the community. The strong associations between racial/ethnic diversity and linguistic and nativity heterogeneity are confined to diversification from Latino and especially Asian growth, and are unsurprising given the strong linguistic and nativity differences between these groups and non-Hispanic whites at the national level. The modest-to-weak associations between racial/ethnic diversity and other forms of demographic and economic heterogeneity that we identify for the population as a whole are driven by growing Hispanic and particularly Asian population shares, with growing black population shares contributing little to change in social or economic heterogeneity. Beyond the scope of H4 but noteworthy, increases in white representation reduce every type of sociodemographic and economic heterogeneity at the place level (first column of Table 6).
Metropolitan and Micropolitan Fragmentation
Although our preceding analysis focuses on the consequences of place-level racial/ethnic diversification, we also consider the consequences of diversity within the broader metropolitan or micropolitan area. Our final research question asks whether greater ethnoracial diversity at the metropolitan and micropolitan level promotes more localized sorting that results in fragmentation of the constituent places. We hypothesize that CBSA-level ethnoracial diversity will be associated with greater fragmentation among communities, leading to less sociodemographic and economic heterogeneity in places located within more ethnoracially diverse metro and micro areas (H5). We find support for this multilevel fragmentation hypothesis in Table 7, which presents the results of multilevel random effects regression models. Model 1 for each measure regresses levels of place nonracial heterogeneity on contemporaneous levels of CBSA racial/ethnic diversity. Model 2 for each measure regresses changes in place-level nonracial heterogeneity on changes in CBSA racial/ethnic diversity between decades.
Across all models in the table, place sociodemographic and economic heterogeneity decline as CBSA ethnoracial diversity increases. In other words, communities located in more ethnoracially diverse CBSAs are less socially and economically heterogeneous (Model 1), and they become even less heterogeneous as CBSA-level ethnoracial diversity grows over time (Model 2). This occurs net of the ethnoracial diversity and areal and population sizes of the individual places, and controlling for CBSA-level nonracial heterogeneity and political fractionalization (number of places in the CBSA). The magnitudes of these coefficients are moderate to strong for linguistic and nativity heterogeneity, and weak to moderate for all other measures. In a supplemental analysis (not shown), we found that this result held for both metropolitan and micropolitan areas when we analyzed them separately. Further supplemental analyses revealed that this result for the average place in a CBSA also translates into greater fragmentation of the metro or micro area as a whole, as measured by a population-weighted information theory index (online appendix, Table A6).
Our study investigates the structural underpinnings of the diversification and fragmentation of the U.S. population within local communities. We contend that how ethnoracial diversity reshapes social relations depends in part on how it affects the social and economic composition of communities. Prior research has typically examined diversification or fragmentation on a single characteristic, such as race, income, or education, but this univariate approach fails to consider how multiple population characteristics covary within places. We find strong support for the consolidated diversity hypothesis that growing ethnoracial diversity produces greater heterogeneity in language and nativity. This result is unsurprising given that language and nativity are in many ways part and parcel of racial/ethnic identity. However, the hypothesis receives less support for the other measures of sociodemographic and economic heterogeneity. Growing racial/ethnic diversity is only weakly correlated with household type, tenure type, and educational heterogeneity, and is virtually uncorrelated with the income, occupation, or age heterogeneity of a community, consistent with a process of multiform diversification. When the average American town or city became more racially or ethnically diverse between 1980 and 2010, it did not alter sociodemographic or economic heterogeneity in substantively meaningful ways, other than for language and nativity. Although we are unable to observe the sociodemographic and economic differences between or within racial/ethnic groups directly, the results found here are consistent with the predictions of the multiform diversification hypothesis.
Our examination of the strength of correlations among distinct population parameters provides a foundation for understanding the social consequences of trends toward greater ethnoracial fragmentation or diversification. When different groups persistently reside in separate locations, the resulting homogeneity offers fewer chances for interpersonal contact or conflict (Alesina and La Ferrara 2002; Costa and Kahn 2003; Fossett and Kiecolt 1989; Putnam 2007; Stein et al. 2000; Taylor 1998). Such spatial fragmentation, though, may still generate severe inequities: disadvantaged groups in segregated areas face an unequal geography of opportunity that shapes prospects for economic mobility, health, and life expectancy (Chetty et al. 2014; Mayer 2001; Quillian 2014; Williams and Collins 1995). Thus, our finding that the ethnoracial diversification of metropolitan and micropolitan areas reduces sociodemographic and economic heterogeneity within constituent places raises concerns about the long-term consequences of macro-level ethnoracial diversification for socioeconomic segregation and racial equality.
When different ethnoracial groups do integrate geographically at the local level, the social consequences may depend in part on whether those groups remain far apart in sociodemographic or economic status. Diversity that results in consolidated lines of differentiation could heighten social conflict and make social structures more rigid and resistant to change, reinforcing and sustaining the status quo (Blau 1977, 1994). For example, communities with greater racial income inequality tend to have higher victimization rates and lower rates of intermarriage, and their residents report less social trust and residential satisfaction than communities with less racial income inequality (Abascal and Baldassarri 2015; Hipp 2007; Sampson 1984; South and Messner 1986; Wilson and Taub 2006). Case studies of “new destination” immigrant communities also show how the consolidation of linguistic and citizenship differences between Hispanic and white populations can serve as a lightning rod for interpersonal and institutionalized race relations (Marrow 2011, 2013; Winders 2013).
By contrast, overlapping demographic and economic characteristics within a community may reduce the salience of group boundaries over time given that residents who differ along one dimension often have other characteristics in common. This is particularly likely because many racial/ethnic groups are themselves quite heterogeneous (Freeman 2002; Habecker 2012; Lee et al. 2017; Massey et al. 2007; Portes and Truelove 1987; Waters 1994, 2000). Intersecting social categories increase the likelihood of interracial friendships and marriage (Blau et al. 1984; Blum 1985; Lauman 1973; Moody 2001; Smith et al. 2014). Neighborhood residents of different races are more likely to interact and offer social support when they have other characteristics in common, such as occupation, parental status, or religious or organizational membership (Maly 2008; Nyden et al. 1997; Wilson and Taub 2006). Of course, physical proximity alone is not sufficient to produce contact or interaction; residents of diverse communities could continue to occupy distinct geographic, institutional, and cultural spaces (Tach 2014). But the likelihood of overcoming such boundaries may be greater when individuals share common interests and equality of status (Allport 1954; Gaertner et al. 1996; Pettigrew and Tropp 2000).
The findings presented here raise the question of why the large national differences in the sociodemographic and economic composition of different ethnoracial groups do not translate to the community scale. One potential explanation is that migration streams are self-selective in ways that sort racial/ethnic groups so that they more closely resemble one another at the local level, perhaps because of the profiles of jobs and structures of opportunity available in the local labor, housing, and marriage markets. These selective sorting processes may be the product of differences in community knowledge and preferences across racial/ethnic groups (Bader and Krysan 2015) or the influence of social networks (Garip 2008), but they could also reflect exclusionary practices at the local level that limit the viability of residing in certain communities. Although some places have pursued diversification as a revitalization strategy (Carr et al. 2012), others have used local zoning laws to restrict the availability of affordable housing options (Rothwell and Massey 2009) or limited the goods and services available in languages other than English (Hopkins et al. 2014).
A second possible explanation is that racial or ethnic newcomers and their descendants come to resemble their native counterparts via assimilation, economic incorporation, and subsequent residential moves reflecting these changes. Immigrants’ integration in the United States generally increases over time across most social and economic measures, including educational and occupational status, income, and language ability. They become more like native-born residents the longer they are in the country, with second and third generations continuing that process (Waters and Gerstein Pineau 2015). Thus, although the consolidation along language and citizenship lines identified in this study may be politically and culturally significant in places undergoing rapid diversification in the present day (Hopkins 2009, 2010), it might not endure over time or across generations as socioeconomic gains translate into locational outcomes. Future research will require more detailed data on group-specific residential mobility to disentangle which of these two distinct migration patterns—selective sorting or attainment-driven moves—are producing the aggregate results found in our study.10
The diversification trend captures the experience of some ethnoracial groups better than others, in line with our fourth hypothesis. Consistent with place stratification theory, there is little association between growing non-Hispanic black population shares and growing sociodemographic or economic heterogeneity. Consistent with spatial assimilation theory, increasing Hispanic and Asian population shares alter community demographic and economic heterogeneity in moderate to strong ways. Despite our inability to observe the individual-level mobility decisions that produce these divergent results, prior research on race-specific mobility patterns offers some interpretations. As Latinos and Asians experience upward mobility and greater acculturation, they tend to move to neighborhoods with higher incomes and more whites (Alba et al. 2000; Crowder and South 2005; South et al. 2008). For African Americans, residential mobility is less likely to result in proximity to higher-income or whiter neighbors (Pager and Shepherd 2008; Rothwell and Massey 2009; Turner and Ross 2005; Yinger 1995). Furthermore, whites are more likely to move out of—and avoid moving into—areas with significant shares of African American residents (Crowder and South 2008; Krysan et al. 2009; Quillian 1999). Future research should aim to uncover how the underlying racial/ethnic structures that generate diversity shape particular nonracial outcomes.
Finally, we demonstrate that the geographic scale of diversification matters. Whereas growing ethnoracial diversity at the place level resulted in weak positive correlations with other forms of sociodemographic and economic heterogeneity, growing ethnoracial diversity at large geographic scales (in our case, the level of metropolitan and micropolitan areas) was weakly to moderately negatively associated with other forms of heterogeneity among the constituent places within the CBSA. These findings support the fragmentation hypothesis in a multilevel context: as metropolitan and micropolitan areas have become more racially or ethnically diverse, the cities and towns within them have become less sociodemographically and economically diverse. Our work thus extends prior research finding a positive association between ethnoracial diversity at the metro level and ethnoracial segregation at the community and neighborhood level (Iceland 2004; Lichter et al. 2015), showing that this pattern of sorting applies to sociodemographic and economic fragmentation as well. This result also has implications for geographic sorting at other scales, which should be examined in future research. For example, as a place becomes more racially diverse, do smaller geographic units such as census tracts or blocks become more segregated? Although tract-level racial segregation has declined nationally, it is possible that micro-level segregation has persisted, or even grown, in places undergoing rapid population diversification.
Because our study examines only one set of diversity indicators, more work is needed to understand the multidimensional types of community change that might occur in the wake of growing ethnoracial diversity. Census data provide information about a basic set of demographic and economic indicators, but they do not measure many other potentially salient forms of social differentiation, such as religion, political affiliation, or wealth. Our article focuses on diversity among racial/ethnic groups, but future scholarship should consider diversity within specific racial/ethnic groups as well, much as empirical examinations of super-diversity have done in the European context (Meissner and Vertovec 2015; Vertovec 2007). The evidence presented here does lead to some testable hypotheses, however. For example, ethnoracial diversity should be more likely to blur racial boundaries in places with multiform heterogeneity, where members of different racial groups share demographic or economic commonalities. By contrast, we might expect ethnoracial diversity to generate interpersonal conflict, less tolerance, and less inclusive policies when heterogeneity is consolidated—that is, when members of different racial/ethnic groups also differ along demographic and economic lines. As we have shown in this article, the consequences of diversity vary across spatial units, which implies variation across political structures as well. Additional research on this topic would be both timely and important because it could inform the decisions of local planners and community leaders and assuage local reactions that are often driven by fear and stereotypes.
Support for this research has been provided by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01HD074605). Additional support comes from the Population Research Institute of Penn State University, which receives infrastructure funding from NICHHD (P2CHD041025). The content of this article is solely the responsibility of the authors and does not reflect the official views of the National Institutes of Health. The authors thank Chad Farrell, Chris Fowler, Matthew Hall, Stephen Matthews, and Gregory Sharp for feedback on this article.
Nonwhite population growth does not always lead to greater ethnoracial diversity, particularly if a place is already majority nonwhite. However, about three-quarters of incorporated and unincorporated places were majority non-Hispanic white in 1980 (Hall et al. 2016), so white population loss and nonwhite population growth at the place level since 1980 have usually led to greater ethnoracial diversity.
The ACS estimates for some communities have large margins of error. Following Spielman et al. (2014), we computed coefficients of variation and assessed the sensitivity of our results to the inclusion of places with low-reliability estimates (coefficient of variation > 40). Our results are robust to excluding these places.
Places that were excluded in 1980 because they did not meet the population threshold were less racially diverse on average than our analytic sample. Newly created places (1990 or later) tended to be less ethnoracially diverse than existing places, but they also became more diverse over time. Sensitivity analyses show that the results are robust to these two sample restrictions.
We computed a ratio of the 2010 to 1980 area size of each place (in square miles) to determine which places had lost land, remained stable, or annexed land during the observation period. Mean 1980 and 2010 diversity levels were quite similar across the shrinking, stable, and expanding categories of places. In Table A2 of the online appendix, we also reran our main analysis on the sample of stable places and found results that are substantively and statistically similar to those using the full sample.
Supplemental analyses indicated that the associations between nonracial diversity and other forms of population heterogeneity were similar in magnitude for large principal cities and other types of places with smaller populations (results available in the online appendix, Table A3).
The results of this supplemental analysis indicate that changes in racial diversity and nonracial population heterogeneity are strongest in places that start with low levels of racial diversity in 1980 and are weaker in places with high levels of starting racial diversity.
Although theoretically distinct, the nominal and ordinal entropy scores are highly correlated for our ordinal measures. The results we present are robust to our use of either nominal or ordinal entropy scores to measure diversity for the ordered measures of diversity.
In supplemental analyses presented in Table A4 of the online appendix, we also examined a set of models that used lagged measures of ethnoracial diversity to predict subsequent changes in nonracial heterogeneity. These results were generally statistically significant although of somewhat smaller magnitude than the results presented here.
Changes in non-Hispanic white population shares are negatively correlated with changes in the shares of other racial/ethnic groups, but population changes among nonwhite racial/ethnic groups are very weakly correlated (r < .1).
Unfortunately, the Census Bureau did not publish detailed tabulations by race or ethnicity (STF2 and STF4) prior to 2000, which are needed to construct measures of within racial/ethnic group heterogeneity at the place level.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.