Beyond the City: Exploring the Suburban and Rural Landscapes of Racial Residential Integration across the United States

Abstract

In recent decades, racial and ethnic diversity has expanded from the city into the suburbs, the rural-urban interface, and remote rural places across all regions in the United States. This study examines how these population trends shape the possibility of racial residential integration across the American rural-urban continuum and regions. Using the information theory index (H) and racial and ethnic composition thresholds, we identify integrated cities, suburbs, and rural towns and villages that are stably integrated between the 2000 and 2010 censuses. This study shows a substantial number of diverse places where people of different races and ethnicities live near each other. Further, the largest clusters of integration locate in suburbs, followed by rural places, while central cities show the lowest rates of integration. In addition, the West typically hosts larger numbers of integrated communities compared to other regions. Findings suggest that to better understand shifting patterns of American racial inequality, research must look outside the city and toward the West to investigate residential integration as a new form of 21^st-century race relations.

The history of legal segregation, from Jim Crow to discriminatory lending practices, fundamentally shapes American racial settlement patterns (Rothstein 2017). Despite moderate declines in segregation after the passage of the Fair Housing Act, metropolitan areas across the nation display durable patterns of racial segregation (Logan and Stults 2011). However, attention to metropolitan-level segregation neglects the structure of residential patterns at different spatial scales and the possibility of racial integration in more local contexts. Moreover, metropolitan analyses omit how these local communities vary across regions with different histories of racial settlement. Analyzing central cities, suburbs, and nonmetropolitan places separately may offer new insights regarding the geography of race relations and may help inform policy for the maintenance if not promotion of integration.

Settlement resulting from the expansion of urban spaces may offer new opportunities for residential integration in suburbs and rural areas (Lee and Sharp 2017; Lichter and Ziliak 2017). In the 21^st century, the “chocolate city/vanilla suburbs” dichotomy no longer exists—as of 2010, most metropolitan people of color (Frey 2018) and immigrants (Wilson and Singer 2011) live in suburbs. Black “return migrants”, from the North to the South, often settle directly into suburbs like those in greater DC or Atlanta, contributing to the roughly 80% of Blacks living in suburbs in these areas (Lacy 2016). Beyond metropolitan spaces, rural areas have been home to longstanding, often overlooked communities of color arising from reservations, historical slavery, and agricultural migrations (e.g., the Bracero program), and military installations (Snipp 1996). Taken together, recent trends in settlement and historical population distributions might reasonably promote contemporary racial integration at a spatial scale not exclusive to or even necessarily centered on the metropolis.

Racial and ethnic diversity varies between regions in ways that are specific to migration patterns by racial and ethnic groups. The Southwest and California are traditional immigrant points of entry that have been home to large Latinx and Asian American populations; however, recent immigration drives the local expansion of diversity in “new destinations” (Riosmena and Massey 2012). While one-third of immigrants still live in traditional coastal gateways like New York and Los Angeles, many have dispersed throughout the Midwest, South, and New England (Frey 2018; Zhang and Logan 2016). The West shows little Black representation, yet immigration to the South and East have contributed to the rise of the most multiethnic metropolitan areas that include Blacks, for example, Houston, Atlanta, and the Boston-Washington corridor (Zhang and Logan 2016). With some exceptions like Chicago, racial diversity in the Midwest and inland Northeast (e.g., Pennsylvania, upstate New York) primarily comprises bi-racial Black-white compositions (Zhang and Logan 2016) arising from 20^th century northern migrations. While little research examines how Native Americans contribute to local diversity, places with substantial shares of Native Americans are largely located in the South (e.g., Oklahoma) and around reservations and tribal lands throughout the Midwest and West including Alaska. These settlement patterns likely fuel diversity in predominantly white landscapes with small Black, Latinx, and Asian communities.

The current study investigates whether diversity across the urban gradient and regions translates into racial integration for racial and ethnic groups across the American landscape. Our central question is whether racial integration varies across local, regional, and racial/ethnic contexts. Relatedly, we assess whether the local racial landscape modifies the relationship between the larger spatial context and integration. Where prior research investigates trajectories of change (Bader and Warkentien 2016; Logan and Zhang 2010), we seek to identify places that maintain stable racial and ethnic compositions in the contemporary United States, and to examine how these compositions vary spatially. We use census places to adopt a spatially inclusive approach that explicitly incorporates rural communities, suburbs, and central cities. Census places are cities, towns, villages, boroughs, or other places that are either incorporated with legal boundaries or are unincorporated and defined by local officials. Places especially benefit this analysis because their spatial scale and population densities vary widely, while they concurrently constitute socially recognized spaces across the urban gradient and region.

We pay special attention only to diverse places and extricate racially integrated communities from those that are more segregated. We limit our sample to diverse places that we define as 80% one racial or ethnic group at maximum and calculate a spatially informed evenness index, the information theory index, to measure integration within these places. With this method, we argue that residents on average will be exposed to higher levels of diversity in integrated places compared to other diverse environments. Using place stratification and spatial assimilation perspectives, we assess whether common place-level correlates of diversity associate with residential integration. In doing so, we highlight new trends in residential patterns across myriad local contexts that make up the full American experience and advance a spatially informed perspective of residential sorting and race relations in the United States.

Our study is the first to investigate actively the occurrence of integration across the rural-urban continuum and across regions. We find that local communities covering millions of people display durable integration primarily in suburban and rural areas, largely in the West. While conventional studies focus on the segregation of the industrial metropolis, this study suggests that the time is right to look outside of the city to understand cross-racial exposure and contact. Where racial segregation contributes to racial disparities in all domains of life (Massey and Denton 1993), understanding the local conditions and regional contexts that facilitate residential integration may support movement towards a more equitable society.

Defining Diversity and Integration

We begin with a discussion of the concepts of residential diversity and integration. Currently, the literature lacks a consensus on how to define and measure diversity and integration (Sin and Krysan 2015). In this paper, when we refer to diversity, we mean the aspatial population composition of a place. Whether a place is diverse or not reflects the number of different groups and the size of these groups within a place. A place may be considered highly diverse if it contains several populations, each represented equally; while, alternatively, the opposite of diversity is complete homogeneity.

We use the term integration to represent the spatial distribution of groups within a place. A place displays integration when people of different races and ethnicities live near each other. Conversely, a place is segregated when residents are primarily clustered around other residents of the same group. By these definitions, diverse places can exhibit varying levels of integration. Further, diversity is required for integration—places may only be integrated if they demonstrate some level of diversity. Homogenous places are racially isolated (i.e., segregated) and therefore cannot be integrated across different racialized groups.

While other studies track racial change in communities (Fowler, Lee, and Matthews 2016; Hall, Tach, and Lee 2016), we seek to identify the characteristics that support stable integration across time relative to more segregated forms of diversity. As the US grows increasingly diverse, stably integrated communities likely support greater cross-racial exposure than other contexts, which in turn may mitigate the problems of segregation.

Historical and Contemporary Forces of Residential Integration

Several social forces motivate this analysis. The 1960s saw the passage of legislation that crucially altered the racial geography of the United States, namely The Fair Housing Act (FHA) of 1968 and the Immigration and Nationality Act (INA) of 1965. The FHA struck down legal forms of housing discrimination. Prior to its passage, the government segregated public housing, the Homeowners Loan Corporation redlined cities, and real estate agents blockbusted neighborhoods in the midst of desegregation (Rothstein 2017). Simultaneously, the INA removed discriminatory national origin restrictions in immigration. In the years since the INA’s passage, the United States has grown increasingly multiethnic, fundamentally reworking American race relations. From 2000 to 2010, Latinx population growth accounted for more than half of overall American population growth and, by proportion, Asians were the fastest growing group over the decade (U.S. Census Bureau 2011). As of Census 2010, 36% of the US population identify as people of color.

Despite the passage of antidiscrimination legislation, study after study details the processes that keep the metropolitan US segregated (Charles 2003; Krysan and Crowder 2017). However, relatively little work looks beyond the city to understand racial distributions in other contexts, perhaps most especially rural areas. Historically, studies implicitly assume that urban areas are the sole domain of racial and ethnic diversity. Yet, in the 21^st century, immigration has changed the rural demography of the country. The “new destinations” literature documents Latinx migration to rural areas in the South, Midwest, and Mountain West (Riosmena and Massey 2012). Lichter and colleagues refer to rural Latinx populations as a “demographic lifeline for dying small towns and white natural decrease” (2016:514). These migration patterns blur the symbolic boundaries between rural and urban spaces—rural spaces are beginning to display levels of racial and ethnic diversity on par with those of urban areas (Lee and Sharp 2017).

Unprecedented in-migration to new destinations is not the only reason to investigate these spaces. Since before the founding of the nation-state, some rural areas have long been home to non-European-descended populations. While contemporary Native American populations reside throughout the entirety of the United States, tribal lands are largely located in places like Oklahoma and across the rural Midwest, Southwest, and West. Prior to Juneteenth 1865, enslaved Blacks sustained the agricultural economy throughout the South, from Maryland to Texas, and most notably the “Black Belt.” As of Census 2010, the South is still home to the majority of Blacks (Rastogi et al. 2011). While much smaller, there are long-established rural Asian American populations as well: Punjabi Sikhs in California’s central valley (Gibson 1988); Chinese in Mississippi (Loewen 1988); and Southeast Asian refugee populations (Zhou 2001). The large Mexican population throughout the Southwest arises in part from the fact that the Southwest was Mexico before US settler colonial expansion. In addition, beginning in World War II, the Bracero program recruited Mexican workers to perform agricultural labor. Furthermore, these Braceros were not legally excluded from white areas. Lastly, military installations often are located in rural areas, creating pockets of diversity in largely white landscapes, for example, Minot Air Force Base in Ward County, North Dakota, or Fairchild Air Force Base in Spokane County, Washington.

The racial demography of the broader regions in which rural and urban communities are nested has also shifted. While Blacks still show their largest shares in the South and older industrial cities in the Northeast and Midwest, Latinx and Asian migration has moved east. Zhang and Logan (2016) track the trajectories of racial change in metropolitan areas from 1980 to 2010. The authors find that many coastal eastern and southern metropolitan areas that were primarily bi-racial or bi-ethnic in 1980 became completely multiethnic across Blacks, Latinxs, Asians, and whites by 2010 including Boston, Philadelphia, Raleigh-Durham, and Orlando. Metro areas in the Midwest and New England that were predominantly white, have shown a substantial influx of Latinxs and/or Asian Americans for instance, Minneapolis-St. Paul, Ann Arbor, and western Massachusetts. Notably, in Los Angeles, Latinx and Asian American growth has outpaced Blacks such that Blacks compose a smaller and decreasing share of southern California. Metro areas in the Southwest and West, like Albuquerque, Seattle, and Portland, still primarily show diversity arising from Latinx or Asian American communities.

Given historical communities of color as well as changes brought about by post-1965 immigration and migration, the current study seeks to investigate stable, racially integrated communities across the rural-urban continuum and how these communities vary across region. In the 21^st century, ample time has passed for antidiscrimination legislation to affect the material realities of housing and race. In metropolitan settings, recent declines in segregation are largely owed to declines in within-place segregation (Lichter, Parisi, and Taquino 2015b), suggesting the emergence of pockets of residential integration. Furthermore, scholars document the rise of multiethnic census tracts within urban areas that remain stable over decades (Bader and Warkentien 2016; Crowder, Pais, and South 2012; Farrell and Lee 2011; Logan and Zhang 2010; Lumley-Sapanski and Fowler 2017). Yet scholarship has not engaged a systematic study of integrated places in rural settings, leaving unknown the prevalence and the drivers of residential integration outside of cities and suburbs. Further, studies have yet to investigate how the race relations of integration within the American racial order vary over regions with group-specific histories of settlement. However aberrant integrated environments may be, stably integrated places may inform theories of cross-racial interaction in a multiethnic context as well as policies that mitigate the problems of segregation.

Models of segregation: Place stratification and spatial assimilation

While extralegal barriers to integration continue to exist, we assert, that in the 21^st century, there are openings for new possibilities of residential integration across urbanicity and regions. These barriers and openings are location-type specific and apply to different racial and ethnic groups with distinct histories of settlement. Broadly, the place stratification model argues that legacies of structural discrimination still pervade housing access for populations of color (Massey and Denton 1993). This model is highlighted by the fact that whites specifically targeted Blacks through legal segregation (Rothstein 2017) and, subsequently, Blacks remain the most segregated group in the contemporary US (Logan and Stults 2011). Recent reports indicate that renters of color and buyers still experience real-estate steering (Turner et al. 2013) and, racial disparities in loan denial and mortgage cost has changed minimally since the 1970s (Quillian, Lee, and Honoré 2020). Furthermore, Black segregation is independent of income, suggesting that upward economic mobility does not act as a panacea to segregation (Charles 2003). While whites segregate from Asians and Latinxs to a lesser degree, declines in segregation have stalled in tandem with high growth through immigration (Logan and Stults 2011). These findings point toward the “group threat” hypothesis where whites feel threatened by people of color and express greater discrimination as populations of color expand locally (Blalock 1967; Fussell 2014).

While used to characterize the United States more broadly, we suggest that the place stratification model differentially applies in rural and urban spaces. To illustrate, Blacks are disproportionately situated in the South and, with substantial rural populations, segregation may persist in the South from the direct histories of slavery, Jim Crow, and associated physical intimidation and violence (Reece and O’Connell 2016). At the same time, large Black populations are located in highly segregated northern and midwestern cities with industrial spatial structures and legacy segregation from redlining, white flight, and disinvestment. Asians and Latinxs may experience less segregation with larger populations in the West with newer urban development built after the Fair Housing Act. Furthermore, western cities tend to have sprawling spatial structures that do not follow patterns of concentric circle segregation as observed in older Northern/Eastern cities (Strom 2017).

While discrimination largely governs racial residential patterns, spatial assimilation may influence Black communities in the Sunbelt. This model states that segregation results from socioeconomic differences across groups and that integration will occur with upward mobility and acquisition of normative white, middle-class behaviors (Charles 2003). Black “return migrants” to the South display high educational attainment and have suburbanized (Frey 2018)—the suburbs of Washington, DC and Atlanta boast stable middle-class populations (Lacy 2016). Interviews indicate that Black return migrants experience the “New South” as tolerant and accepting in comparison to the northern sending communities (Tavernise and Gebeloff 2011). Lastly, Sunbelt cities have an abundance of new housing stock built after the passage of antidiscrimination legislation. These factors may promote integration in the South for return migrants and the Black population as a whole.

Place stratification and spatial assimilation models offer complementary explanations for Latinx populations’ heterogeneous distributions. Spatial assimilation reflects the experience of phenotypically white Latinxs that may assimilate with whites based on pigmentation (Charles 2003). Higher Latinx income associates with decreased segregation from whites (Lichter, Parisi, and Taquino 2015a). However, contrary to spatial assimilation, income disparities do not explain overall Latinx-white segregation and some Latinx subgroups display residential integration with Blacks (Lichter et al. 2015a). New destinations throughout the South and Midwest exhibit exceptionally high levels of Latinx-white segregation and show higher segregation scores than established gateways (Lichter et al. 2016). These distributions may reflect place stratification arising from blue-collar labor migrations for example, agricultural manual labor (e.g., the mid-20^th century Bracero program), construction, and, more recently, manufacturing (e.g., meatpacking) and mining industries (Lichter et al. 2016).

The literature on segregation of Native Americans is less robust. Scholars likely neglect Native people and communities because of their small shares of the American population and their distribution throughout the rural US. Histories of discrimination point to greater spatial integration, and neither place stratification nor spatial assimilation models likely fit the experience of indigenous populations. While the “color line” defined Black-white race relations, one dimension of the white solution to the “Indian problem” was assimilation into white society. For example, federal policies placed many native youth in residential boarding schools, separating them from their families and communities, to erase indigenous identity and cultural practices (Wilkinson 2006). The Dawes Act dissolved tribal ownership of land and individualized property rights (Wilkinson 2006). Plots were often sold to whites creating a “checkerboard” pattern throughout tribal lands, thus creating spatially integrated places. Furthermore, the Indian Relocation Act of 1952 attempted to depopulate reservations and encouraged assimilation by offering monetary incentives to leave tribal lands for the city. While integration for other communities may reflect the erosion of traditional race relations, spatial integration among Natives Americans may reflect past histories of discrimination and attempted termination (Snipp 1992).

While Asians are the least segregated group from whites, little work has examined Asian populations in rural areas. Qualitative work suggests that longstanding Asian communities like Punjabi Sikhs in California or Chinese in the Mississippi Delta form tight-knit networks in response to discrimination (Gibson 1988; Loewen 1988). However, these communities display intergenerational upward mobility where later generations move out of rural areas in a way that is consistent with the spatial assimilation perspective. Lastly, some rural Asian communities emerge from refugee relocation programs, and experience racial hostility, or conversely, assimilation by way of sponsorship from voluntary organizations (Zhou 2001).

Spatially Informed Hypotheses of Integration

Taken as a whole, previous research and theory suggests that much of the variation in the spatial structure of race and ethnicity is attributable to discrimination, and points to specific hypotheses regarding integration across the rural-urban continuum and between the nation’s regions. Figure 1 provides a conceptual map for the hypothesized relationships between integration and our main independent variables: location type, racial composition, and region. First, we anticipate integration will structure around the metropolis. Central cities will be least integrated, and the most rural areas will not be integrated, both due to entrenched histories of discrimination. However, we hypothesize that the multiethnic landscape of the suburbs will promote more integration relative to other types of places.

Figure 1.

Hypothesized relationships between integration and location type, race and ethnicity, and region.

Second, we anticipate that place-level race and ethnic composition will influence integration. The place stratification perspective suggests that Black presence is associated with less integration as legal segregation specifically targeted Blacks. Spatial assimilation and past trends indicate that places with Asian presence will be the most integrated. Latinx, and Native presence will fall in between.

Third, we expect that the prevalence of integration will vary regionally. We anticipate that the West will host the largest number of integrated places across racial and ethnic composition due to historical and contemporary multiethnic immigration and new development after the passage of antidiscrimination legislation. The Midwest will show the lowest rates of integration across racial and ethnic composition because of little racial and ethnic diversity, and rigid metropolitan segregation within the region. The South and the Northeast will occupy an intermediate position because of abundant racial and ethnic diversity in the broader historical context of southern segregation and northern industrial cities, respectively.

Fourth, we hypothesize that region will moderate the relationship between racial and ethnic composition and integration.¹ For Blacks, we expect the South to fall in the middle because of high integration in pockets of the “New South” around large metropolitan areas like Texas, South Florida, and DC, contrasting with rural segregation left in the wake of slavery, Jim Crow, and persistent racial hostility. We anticipate the Northeast will host more integrated communities due to sprawling urbanization, and the positive effect of the West will be amplified due to smaller Black populations and the associated lessened perceived group threat. Given large Latinx and Asian populations as well as Native reservation lands, we hypothesize that the interaction will attenuate the effect of the West. For Asians, both the Northeast and South will host more integrated communities due to recent multiethnic immigration to large cities. For Latinxs, the Northeast and South will show lower rates of integration due to high segregation in new destinations. The Northeast will show few integrated communities for Native Americans due to smaller indigenous presence from the history of colonization and displacement, while the South may display higher rates because of large clustering of Native communities and tribal lands in Oklahoma.

Data and methods

Data

All Census and American Community Survey data, including shapefiles, come from the National Historical Geographic Information System (Manson et al. 2017). We begin with all census places in the nation that contain at least 100 residents, normalized to 2010 census boundaries. The Census Bureau defines places as cities, town, villages, suburbs, or other places recognizable by name (U.S. Census Bureau 2005). Incorporated places have legal boundaries with governmental powers and functions, while Census-designated places are identifiable by name but are not legally incorporated and are defined in cooperation with local officials. We specifically use places because they constitute socially meaningful units across the urban gradient. For example, New Llano is a highly diverse rural town in Louisiana that local residents recognize by name.

Racial and ethnic composition data come from the 2000 and 2010 Decennial Censuses and we use the following categories: Latinx and non-Latinx Black, Asian (combined with Pacific Islanders), Native American/Alaskan Native, and white and other.² Other variables come from the American Community Survey’s 2008–12, 5-year estimates: percent renter-occupied housing units, housing stock built after 1990, military and public-sector employment, university enrollment, and percent foreign-born.

The US Department of Agriculture provides data on rural-urban continuum classification (Economic Research Service 2015). We categorize counties into large metropolitan areas, small-medium metropolitan areas, non-metropolitan counties adjacent to a metropolitan area (shares a border with a metropolitan county) and non-metropolitan counties not adjacent to a metropolitan area (does not share a border with a metropolitan county). We geocode places to counties and for those that span more than one county, we assign them to the county based on the places’ central longitude-latitude coordinates. For places within metropolitan areas, following Census definitions, we consider a place to be the central city if it has the largest population within the metro area. All non-central city places within metropolitan areas are suburbs. All places within non-metropolitan areas are rural.

Defining the sample

While this study is motivated by a narrative of historical change, we focus on identifying stably integrated places and not newly emergent settings. We adapt previously established measurement strategies to identify integrated places. We first narrow the sample to places with substantive racial and ethnic heterogeneity using Galster’s (1998) stock/flow model for defining racially integrated spaces. Galster argues that, to be integrated, a space should have a stock of residents that is racially mixed (i.e., comprise no more than a certain percentage of a single group) and the flow of residents is such that the space will continue to be mixed by the same standard 10 years into the future. Because race and ethnic neighborhood-level change from white flight, gentrification, or rapid population growth that may occur over the course of a few years, stability over 10 years arguably provides a time horizon to exclude homogenizing forces of re-segregation (Maly 2005, Freeman 2005). In addition, since we use a comparative measure to understand integration (discussed later in this section), an absolute cutoff guarantees that places dominated by one group will not be classified as integrated.

We choose as a cutoff that a place should be no greater than 80% one group.^3,4 This cutoff excludes many predominantly white places with low segregation scores like Billings, MT and Fort Collins, CO. It also excludes places that were predominantly white in 2000 but showed modest increases in diversity over the decade like the central cities Madison, WI or Portland, OR, and suburbs like Naperville, IL and Newton, MA. At the other extreme, this cutoff excludes predominantly Black places such as Detroit, MI, Gary, IN, and East Orange, NJ. Places excluded due to Latinx dominance include the border city Laredo, TX, as well as East Los Angeles, CA, and Hialeah, FL. Lastly, many places on tribal lands are excluded including Tuba City, AZ, Zuni, NM, and Rosebud, SD. No places are excluded as a result of Asian dominance.

The 2010 Census identifies 29,261 census places across the nation, which cover approximately 74% of the American population.⁵ Of this sample, 2,344 places (8% of total places) do not meet the 100-resident criterion in either 2000 or 2010, leaving 27,510 places. A further 20,536 places (70%) have greater than 80% one racial or ethnic group, of which 94% are predominantly white, 2% are predominantly Black, 2% are predominantly Latinx, and 2% are predominantly Native American. Of the remaining, 6,381 places, 14 to do not have 2008–2012 ACS estimates due to dissolution or annexation between the 2010 Census and 2012 ACS.

The final sample totals 6,367 stably diverse places across the 2000 and 2010 censuses. We calculate the spatial information theory index (H~) to analyze the distribution of race and ethnic groups within these places. Reardon and Firebaugh (2002) popularized the use of the information theory index (H) to investigate multigroup segregation.⁶ H measures how closely the diversity of subareas mirror the diversity of the larger geographic unit of interest. The index ranges on a scale from 0 to 1, where 0 indicates that each subarea matches perfectly the overall diversity (maximum integration) and 1 indicates that each subarea contains only one group (maximum segregation or, given our focus, minimum integration). However, as an aspatial measure, H fails to assess the spatial ordering of the subareas within the larger unit of analysis (Lee et al. 2008). To address this limitation, we calculate H~ by constructing egocentric, local environments using block-level data. We assume the population of each block is concentrated at the centroid and use an exponential distance-decay function to weight closer observations more heavily than farther ones (Hong, O’Sullivan, and Sadahiro 2014). Rather than using administrative units, this method calculates H~ by comparing the diversity of each local environment with the overall place-level diversity.

We categorize places as integrated if they fall within the bottom quartile of H~ (H~ ≤ 0.15) in both 2000 and 2010.⁷ We refer to the rest of the distribution as diverse-low integration places. Combining the stock/flow model with H~ provides an understanding of spatial integration, in addition to evenness and segregation, within substantively diverse places. Again, by excluding predominantly one group places, we exclude the most segregated places in the nation from this analysis. Ultimately, 1,292 (20.5%) places maintain stable integration, housing roughly 17% of the population-weighted sample, and 8% of the total American population.

Analytic strategy

Beyond racial and ethnic composition and population counts, few variables are standardized across censuses to 2010 place boundaries. Thus, while we define integration between 2000 and 2010 to ensure stability, we cannot conduct a longitudinal analysis due to these data limitations. Using 2010 census data and 2008–12 ACS data, we conduct a cross-sectional logistic regression analysis where the dependent variable is the indicator of place integration.

To assess our first hypothesis, we investigate the spatial structure of integration based on location in or proximity to metropolitan areas. We construct a variable based on urban-adjacency: central cities (i.e., the place with the largest population within a metro area), suburbs (non-central city, metropolitan places) within large metropolitan areas (greater than 1 million residents), suburbs within small to medium metropolitan areas (between 50,000 and 1 million residents), places within nonmetropolitan areas in a county adjacent to a metropolitan area; and finally, places within nonmetropolitan areas in a county not adjacent to a metropolitan area.

To test our second hypothesis, we use a comparative approach and classify the racial and ethnic composition of each place using the 25% criterion first proposed by Logan and Zhang (2010). We consider a place to have a substantial presence of a specific race and ethnic group if the share of residents in that place is at minimum 25% of the share of the study population overall. For example, in 2010, Latinx composed 24.6% of the study population; we consider a place to have a substantial Latinx presence if it displays a Latinx proportion of at least 6.2%. However, we exempt Native Americans from this rule since this group composes very small shares of the study population yet compose substantial populations of many small, rural places. For example, in 2010, Native Americans composed approximately 0.6% of the study population, but 580 (9%) places contained shares greater than 5%. Thus, we impose a threshold of 1.9% for Native Americans, the same used for Asians.⁸

To capture variation among regions, the focus of our third hypothesis, we use the census region categorization of Northeast, Midwest, South, and West. To test our fourth hypothesis, we include an interaction term between region and local (place) presence for each racial and ethnic group to examine how region moderates the relationship between race/ethnic composition and integration.

We also account for several key independent variables identified in previous research examining racial integration or, its counterpoint, segregation. The military, public-sector, and university settings facilitate place diversity (Lee and Sharp 2017), are inversely related to segregation (Farley and Frey 1994), and may promote residential integration through employment integration (Diprete and Soule 1986; Moskos and Butler 1996; Moulton 1990). Following Lee and Sharp (2017), we classify a place as a military hub if the share of residents in the labor force employed in the Armed Forces is twice the national share. We use the same rule for public-sector employment and university enrollment. Because Black renters are less segregated from whites than their homeowning counterparts (Friedman, Tsao, and Chen 2013), we use percent renter-occupied units to incorporate the rental market of places. Newer housing stock built after the passage of antidiscrimination legislation may promote coresidence between populations of color and whites (Crowder et al. 2012). We measure new housing stock as the percent of units built after 1990. Since geographically larger places necessarily allow for greater spatial differentiation, we control for geographic size of place by quartiling land area. Lastly, for ease of interpretation, we transform percentage variables (i.e., foreign-born, renter-occupied units, and new housing stock) to units of 10 percentage-points.

Lastly, residential integration likely clusters across places, because nearby places are typically more alike compared to places farther away (Tobler 1970). This clustering potentially biases inferences (i.e., standard errors, confidence intervals, and p-values) because regression analysis assumes observations are independent even when they are correlated in space. We use two methods to address such spatial autocorrelation. First, we use empirical standard errors that are robust to autocorrelation (heteroskedasticity). Second, we create a spatial lag variable that captures the rate of integration among neighboring integrated places to parameterize the spatial autocorrelation (possibly dependence) remaining in the data after accounting for heteroskedasticity (see Anselin 1990; Anselin 1988; and Voss, Curtis White, and Hammer 2006 for a discussion of spatial dependence and spatial heterogeneity). We identify at least 5 nearest places in the sample places using straight-line distances between the central latitude-longitude coordinates of the places. We coerce symmetry (i.e., if place i is neighbors with j, then j must also be neighbors with i), so many places have greater than five nearest neighbors (65%, range: 5 to 13). For the corresponding spatial cluster, we calculate the rate of integration.

Results

Integration across the rural-urban continuum

We begin our analysis of residential integration in local communities by reporting the frequency of the most integrated quartile across the urban adjacency categories in Table 1. Results support our first hypothesis, that integrated places cluster in suburbs more so than in central cities and rural areas. The clear majority of integrated places situate within large metropolitan suburbia (58%). For example, 18% of all integrated places lie within the major Californian metropolitan areas alone (e.g., Los Angeles, San Francisco, Sacramento). For comparison, large metro suburbs compose only 26% of the diverse but not integrated sample. Suburbs in small-medium metropolitan areas also show large representation of the most integrated places (28%), but their share among diverse-low integrated places is more even (25%).

Table 1:

Rural-urban classification of US places by integration status between 2000 and 2010, n (column %) (Census 2000 and 2010)

	Integrated		Diverse-Low Integration
Central city	9	(1)	236	(5)
Suburb in large metro area	743	(58)	1319	(26)
Suburb in small/medium metro area	357	(28)	1286	(25)
Nonmetro adjacent to metro area	91	(7)	1433	(28)
Nonmetro not adjacent to metro area	92	(7)	801	(16)
N places	1,292		5,075

Note: The chi-square test resulted in a p-value < 0.001.

Among integrated places, central cities show disproportionately low representation (1%, n = 9), and they also tend to be institutional hubs like Ithaca, NY (university hub), and Jacksonville, NC (military hub). Moreover, these central cities display small populations [2010 population range: 30,000 (Ithaca, NY), 130,000 (Killeen, TX)]. Clearly, the largest central cities, despite their high racial diversity, do not meet our criteria for integration.

Rural places show similarly low representation in the integrated sample compared to the diverse-low integrated sample. Approximately, 14% of the most-integrated places lie within nonmetropolitan areas with a level split between those in adjacent and non-adjacent categories. However, underrepresentation is less extreme for the most rural places not adjacent to metropolitan areas. Places in counties adjacent to metropolitan areas make up 7% of integrated places, but 28% of diverse-low integration places. Places in counties not adjacent to a metropolitan area comprise 7% of integrated places while they make up 16% of diverse-low integrated places. Results suggest metropolitan segregation may be expanding outwards to adjacent rural places, while many diverse non-adjacent places operate as military hubs (15%) such as Whidbey Island Station, WA, Fallon Station, NV, and Fort Rucker, AL.

Integration across racial/ethnic contexts

Consistent with our second hypothesis, the racial and ethnic composition of integrated places influences the potential for residential integration. We hypothesized that communities with substantial Black presence will show lower rates of integration and Asian presence will promote integration. Table 2 reports the overall racial and ethnic composition of integrated places by the 25% criterion versus diverse-low integration places.

Table 2:

Column percents for 2010 Racial and Ethnic Composition (25% Criterion) by Integration Status (n = 6,367, Census 2010)

	Integrated	Not Integrated
Asian	72.4	23.3
Black	49.8	66.5
Latinx	79.6	57.0
Native American	9.0	14.0
White	95.9	98.3
Total (n)	1,292	5,075

Note: Cells represent the share of places in the column that demonstrated group presence by the 25% criterion. For example, 49.8% of integrated places had Black shares above 5%, the threshold determined by the 25% criteria. The chi-square test displayed a p-value < 0.001.

Results support our proposed hypothesis and are consistent with segregation research that documents the lowest segregation scores for Asians and the high segregation scores for Blacks. Blacks and Native Americans show dramatic underrepresentation among integrated places (49.8 vs. 66.5%; 9.0% vs. 14.0%, respectively), and the converse is true for Asians. Asians are present in more than triple the share of integrated places compared to diverse-low integration places (72.4% vs. 23.3%). Latinxs also show substantially higher representation in integrated places (79.6% vs. 57.0%), while whites show roughly even distributions across integration status (95.9% vs. 98.3%, respectively).

Regional variation in integration

To assess the extent to which integration varies systematically across regions, we map the distribution of integrated places across the United States in Figure 2. This map is generally consistent with our expectations as laid out in our third hypothesis. The West displays large clusters of integrated places, specifically in metropolitan California and the Seattle area. The Census-designated South also shows a preponderance of integrated places: large metropolitan areas in south Florida and Texas, and Washington, DC. However, northeastern Oklahoma proves to be a notable exception with a large cluster of integrated places in smaller metropolitan and nonmetropolitan areas which all contain large Native American populations. For example, Flute Springs, OK in Sequoyah County, was 68% Native American in 2010.

Figure 2.

Map of US sample comparing racially integrated places to diverse-low integration places (total n=6,367, integrated n=1,292).

Table 3 reports the regional distribution of integrated places. Again, confirming our third hypothesis, integrated places are disproportionately located in the West, with underrepresentation in southern and midwestern states. Exactly half of integrated places are in the West compared to 21% of diverse-low integration places. Southern places make up 36% of integrated places while nearly two-thirds of diverse-low integration places are in the South. The Midwest shows low representation in the overall diverse sample—midwestern places compose 4% of integrated places and 8% of diverse-low integration places. Finally, the Northeast shows modest overrepresentation in integrated places compared to diverse-low integrated places (10% vs. 7%, respectively). Results make clear that while each region houses integrated places, racially integrated places predominantly locate in the West.

Table 3:

Region of places by integration status between 2000 and 2010, n (column %) (Census 2000 and 2010)

	Integrated		Not Integrated
North	125	(10)	342	(7)
Midwest	58	(4)	390	(8)
South	463	(36)	3280	(65)
West	646	(50)	1063	(21)
N places	1,292		5,075

Note: The chi-square test resulted in a p-value < 0.001.

Multivariable context of integration

Figure 3 reports the predicted probabilities of residential integration generated by multivariable logistic regression analyses predicting integration among racially diverse US places (for all parameter estimates, see Appendix Table A1). This analysis uses robust standard errors and further adjusts for spatial autocorrelation using a measure of spatially lagged integration (i.e., the rate of integration among the nearest integrated neighbors), and accounts for potential confounding variables identified in previous research that associate with place-level diversity or segregation. After controlling for geographic clustering and potential confounders, the main effects of location type buttress our earlier findings: suburbs are the most integrated location type, while central cities are the least integrated location type, and notably, rural places show higher levels of integration than central cities. Both suburban categories display the highest predicted probabilities (large metro suburb: 23%; small-medium metro suburb: 21%). Central cities display the lowest probability (4%) and are substantially lower than even rural diverse places (adjacent to a metro area: 17%; not adjacent to metro area: 16%). These findings provide further evidence that rural and urban places are becoming more alike in terms of the geography of integration. When limited to diverse contexts, consistent with our first hypothesis, suburban areas maintain higher levels of integration than central cities, and central cities show lower levels of integration than rural areas.

Figure 3.

Bar Graph of Predicted Probabilities of Residential Integration by Location Type (A), Race (B), Region (C), and Race by Region (D), Generated by Logistic Regression

Appendix Table A1.

Odds Ratios, Logistic Regression Predicting Integration status of US places

	Model 1		Model 2
	Odds Ratio	95% CI	Odds Ratio	95% CI
Location type (ref: Nonmetro not adjacent to metro)
Central City	0.12	(0.06, 0.28)	0.13	(0.06, 0.29)
Suburb in large metro	2.04	(1.50, 2.78)	2.09	(1.52, 2.87)
Suburb in small-med metro	1.66	(1.24, 2.23)	1.76	(1.30, 2.38)
Nonmetro adjacent to metro	1.06	(0.76, 1.48)	1.08	(0.76, 1.52)
Place race and ethnic
composition (25% criterion)
Black (5%)	0.66	(0.53, 0.81)	0.53	(0.39, 0.73)
Asian (2%)	2.64	(2.18, 3.20)	3.84	(2.90, 5.08)
Latinx (6%)	1.22	(1.00, 1.50)	2.13	(1.59, 2.85)
Native American (2%)	1.24	(0.94, 1.63)	3.20	(2.14, 4.80)
Region (ref: South)
North	0.98	(0.73, 1.32)	1.39	(0.46, 4.19)
Midwest	1.15	(0.81, 1.64)	1.27	(0.36, 4.53)
West	1.21	(0.96, 1.53)	4.74	(2.61, 8.61)
Race by Region (ref: South)
Black
North			2.09	(1.09, 4.00)
Midwest			4.70	(1.68, 13.12)
West			1.57	(0.99, 2.47)
Latinx
North			0.20	(0.10, 0.40)
Midwest			0.52	(0.24, 1.14)
West			0.32	(0.19, 0.52)
Asian
North			1.45	(0.58, 3.64)
Midwest			0.31	(0.15, 0.61)
West			0.46	(0.31, 0.68)
Native
North			0.00	(0.00, >100)
Midwest			0.09	(0.01, 0.95)
West			0.21	(0.12, 0.36)
Employment hub
Military	2.65	(2.04, 3.43)	2.72	(2.06, 3.58)
University	2.01	(1.38, 2.91)	1.96	(1.34, 2.88)
Public sector	1.68	(1.32, 2.15)	1.78	(1.39, 2.28)
Percent variablesa
Foreign-bom	1.16	(1.08, 1.26)	1.18	(1.09, 1.27)
Renters' share	0.99	(0.94, 1.05)	1.00	(0.95, 1.05)
New housing stock	1.10	(1.06, 1.15)	1.10	(1.06, 1.15)
Spatially lagged rate of integration	27.02	(19.50, 37.45)	21.33	(15.33, 29.68)

^a.Model 1 has the main effects of each variable and Model 2 adds the interaction effects of race and region. Percentage variables were scaled to units of 10 percentage-points for easier interpretation of effect sizes. For example, a 10 percentage-point increase in foreign-born population associates with an odds ratio of 1.18.

The main effects estimates for racial and ethnic composition support our second hypothesis (Figure 3b). Places typed that met the 25% criterion for Asians show the highest representation among integrated places, Blacks the lowest, and Native Americans and Latinx in the middle. The main effects of racial and ethnic composition represent places in the South because it is the reference category for region. Places in the South with Asian presence display the highest probability of integration (26%), Native American and Latinx presence are roughly even with sample representation (21%). Notably, Blacks are the only group with a negative association (below 25% threshold: 22%%; above 25% threshold: 19%), suggesting that places with substantial Black presence are the least likely to display integration.

As for region, the West displays the highest probabilities of integration (25%, Figure 3c) while the Northeast displays the lowest probability (16%) and, notably, the Midwest falls in the middle (18%). This result indicates that once a community becomes diverse in the Midwest, it is as likely to sustain integration as places in the Northeast and South (21%). Further, it appears that integration around large, diverse southern metropolises compensates for low rates of integration across the Deep South. While we hypothesized low levels of integration in the South, areas like Washington, D.C., Florida, northeastern Oklahoma, and Texas show a substantial number of integrated places.

The moderating role of region

Given regional differences in historical patterns of racial settlement, we assess our final hypothesis about the moderating role of regions on the relationship between racial and ethnic composition and integration (Figure 3d). Consistent with descriptive evidence, the West shows the largest probabilities of integration for Blacks, Latinxs, and Asians, and a moderate probability for Native Americans. The West shows striking probabilities for Asians (29%) and Blacks (24%), and still substantially large probabilities for Latinxs (23%), and Native Americans (20%). Across racial and ethnic groups, the new development of the West provides the basis for the highest levels of residential integration. The effect size is particularly notable for Blacks, which suggest that non-Black residents may experience lower group threat among smaller, western Black populations.

The rank order of the associations for other regions varies across racial and ethnic groups. By Black presence, the Northeast has the lowest probability of integration (17%). The Northeast also shows the lowest probability for Latinx presence (17%), which may arise from markedly high segregation in new destinations. For Asian presence, the Northeast falls in the middle (24%), where Asians likely maintain proximity to the most cosmopolitan metropolises. The Northeast does not show any Native presence in integrated communities (there are few diverse communities in the Northeast with substantial Native American presence (n=3, 25% criterion).

Notably, contrary to descriptive analyses, the Midwest displays the second highest probability of integration by Black presence compared to the South (22%). Despite the Midwest being the locus of the most Black-white segregated metropolises in the nation (e.g., Chicago, Detroit, Milwaukee) with few diverse places more generally, residents appear less resistant to integration once a place becomes diverse. The Midwest shows middling probabilities for both Latinxs and Asians (19%). Latinxs in the Midwest may be disproportionately tied to established gateways like Chicago, which show lower segregation scores than new destinations throughout the South. The effect size for Asians is substantially smaller than the West or Northeast, which suggests that the racial climate for Asians in the Midwest is still resistant to integration. For Native Americans, the Midwest displays a strikingly low probability of integration (8%). Spatial sequestration on reservation lands in rural areas may drive the dramatically low odds of integration for Native Americans in the Midwest (8%).

Despite high diversity and large Black populations in the region, the South is still more resistant to integration for Blacks than other regions of the United States (18%). Group threat and communities arising from the history of slavery, legal segregation, and Jim Crow still fundamentally define the racial landscape of the South. However, the South shows high probabilities for both Asians and Latinxs on par with the West (29%, 23% respectively). The South hosts Asian and Latinx populations that may be disproportionately anchored to cosmopolitan regions that are welcoming of integration (e.g., South Florida, Texas, DC) Notably, large clusters of integration in northeastern Oklahoma likely define southern integration for Native Americans, where the South shows the largest probabilities of integration by Native presence (32%). However, this pattern of spatial integration stems from removal programs and discrimination via assimilation, and thus may not reflect a breakdown of racial boundaries.

Conclusion and discussion

While prior research on residential integration largely focuses on the metropolitan area as the unit of analysis, we actively search for integration in local contexts both within and beyond the metropolis and across spatial and racial landscapes. We take places as the unit of analysis to investigate patterns of integration at different spatial scales within central cities, suburbs, and nonmetropolitan places that vary widely in population size and spatial structure. Whereas metropolitan-level analyses highlight the durability of segregation, our place-based approach provides novel insights regarding the possibility of integration between metropolitan and nonmetropolitan contexts, place-level racial and ethnic composition, as well as geographic region. This analytic approach grows in importance with the continual diversification of the nation across these place-types. Our findings show that highly integrated places exist across spatial dimensions, yet still support a place stratification perspective where structural forces of discrimination dictate the emergence of integration within places.

This study generally supports our proposed hypothesis regarding integration across the rural-urban continuum: suburbs generally host integrated places at the highest rates. Contrary to early constructions of ethnically homogenous white suburbs, suburbs now offer a variety of multiethnic contexts that provide terrain for people of different groups to live in proximity to one another. Notably, suburbs inside many of the most-segregated, large metropolitan areas (e.g., Los Angeles, Washington, D.C.) display the highest probabilities of integration after controlling for many place-based characteristics. While qualitative analyses have only investigated integration within the central city (e.g., Lumley-Sapanski and Fowler 2017; Nyden et al. 1998), our national study suggests that scholars may wish to look beyond the metropolis to better understand multiracial spaces and cross-racial interaction. Central cities were perhaps the most diverse places in the 20^th century, however, entrenched dynamics of segregation still pervade them. Overall, our results demonstrate the advantage of adopting a spatially inclusive analytic approach that incorporates areas spanning the rural-urban continuum and the nation’s regional and racial landscapes.

Evidence from this study continues to underscore hierarchy in the American racial order where places with substantial Black presence show the lowest prevalence of integration. Places with Asian presence show the highest prevalence of integration, and Latinx presence occupies an intermediate position. Our study is the first to adopt an approach that considers Native American populations across spatial landscapes, and we find that places with Native American presence show considerable levels of integration that are similar to Asians in the South.

Our results suggest regional patterning of integrated places consistent with specific histories of racial discrimination. Blacks show the lowest probabilities of integration in the South, with higher rates in the Midwest and the West. For the Midwest, despite rigid metropolitan segregation between places and low representation among diverse places, stable integration is more likely to arise after a place becomes diverse. Asians show a slightly different pattern, where the Midwest shows the lowest probabilities of integration, however, it is followed by the Northeast, with the West and South showing a dramatically higher probability of integration. Large Asian communities in Hawaii, California, and Seattle likely drive this association along with cosmopolitan, high diversity suburbs in the South (e.g., Houston, Dallas, DC). This result may also indicate that Asians are less subject to group threat, where broader regions with large Asian populations show higher levels of integration. Similarly, for Latinx presence, the South and West show the highest probabilities of integration, and lower probabilities in the Northeast. This result may reflect Black racialization of Dominican and Puerto Rican populations around New York, Boston, and western Massachusetts while reflecting white racialization in South Florida and non-Black racialization throughout the Southwest and West. For Native communities, the Midwest shows the lowest probabilities of integration, which is suggestive of segregation around reservation lands. While the West may also house large rural communities, low integration in these places may be offset by large coastal Native communities from southern California up to Alaska.

While rural integrated places are relatively diffuse through the United States, Oklahoma provides an interesting case for rural integration given the large cluster of integrated places in the northeastern part of the state. These places compose approximately 17% of nonmetropolitan, integrated places. Due to 19^th century removal programs, Oklahoma is home to many tribal lands disconnected from ancestral locations and all integrated places show Native American presence. Here, contrary to established models of segregation, spatial integration may reflect the unique history of discrimination against Native American communities in Oklahoma through coercive assimilation. Perhaps due to a metropolitan focus, prior research on segregation overlooks the Native American experience and what it may suggest about rural race relations. Future research may seek to understand the local context of Oklahoma to provide a broader picture of the meaning of place stratification or spatial assimilation in rural spaces for Native Americans.

This study is not without limitations. We do not study the dynamic quality of integration. While prior research investigates racial change in metropolitan areas starting in 1970 (Bader and Warkentien 2016; Zhang and Logan 2016), future research may wish to examine these trajectories in rural contexts as more data in the contemporary multiethnic period emerge. We also do not investigate the most segregated places, which we have conceptualized as places dominated by one racial or ethnic group. This choice likely leaves out places with explicit exclusionary policies including, for example, sundown towns (Loewen 2006). Future research may compare predominantly white spaces against diverse-segregated places and places that become integrated to understand factors that influence these divergent trajectories.

Further, because we argue that integration is only possible under conditions of diversity (e.g., a predominantly white place by definition cannot be integrated across different racial and ethnic groups), our approach focuses on the roughly 20% of US places with substantial racial and ethnic diversity and excludes the other 80% of places where less diversity exists. A strength of this approach is that we study places that are sufficiently diverse as to make appreciable integration possible. A caveat is that it necessarily excludes most places, which are not sufficiently diverse.

Our study is the first to investigate actively the occurrence of integration across the American spatial and racial landscape. Metropolitan-level analyses and folk concepts present a picture of ubiquitous segregation and an absence of residential integration in the nation. In contrast, this study shows a substantial number of diverse places where people of different races and ethnicities live near each other. While this study does not examine social integration downstream of residential integration, we hope to lay the foundation for further investigation. By virtue of physical proximity, racially integrated communities may facilitate greater cross-racial contact through religious institutions, public education, or bi-/multiracial romantic relationships and families. As a result, residential integration may transform current notions of racial boundary formation. Further, this study may help inform policy to understand the types of contexts that may be conducive to integration. W.E.B. Du Bois referred to the “color line” as the major defining factor in American racial injustice (Du Bois 1903). Residential integration blurs this color line by disrupting the spatial processes of segregation. Where segregation harms the life chances of people of color in all domains of life, understanding the factors that contribute to spatial integration may help mitigate the vast inequalities between people of color and whites across the rural-urban continuum and the nation’s storied regional contexts.

Appendix

Appendix Table A2.

10 Largest places by population excluded by single race dominance (2010 % share of dominant group)^a

Predominantly White	Predominantly Black	Predominantly Latinx	Predominantly Native
Lincoln, NE (83)	Detroit, MI (82)	Laredo, TX (94)	Tuba City, AZ (90)
Madison, WI (76)	Gary, IN (84)	Hialeah, FL (93)	Shiprock, NM (95)
Scottsdale, AZ (84)	East Orange, NJ (87)	Brownsville, TX (92)	Zuni Pueblo, NM (95)
Spokane, WA (84)	Clinton, MD (80)	McAllen, TX (83)	Kayenta, AZ (91)
Boise, ID (85)	Redan, GA (93)	East Los Angeles, CA (94)	Chinle, AZ (90)
Overland Park, KS (81)	Milford Mill, MD (84)	South Gate, CA (91)	Whiteriver, AZ (96)
Port St. Lucie, FL (62)	East St. Louis, IL (98)	Cicero, IL (83)	San Carlos, AZ (95)
Vancouver, WA (76)	Suitland, MD (91)	Edinburg, TX (87)	Fort Defiance, AZ (92)
Springfield, MO (87)	Lochearn, MD (80)	Mission, TX (84)	Pearl River, MS (82)
Eugene, OR (82)	Maywood, IL (74)	Pharr, TX (91)	Pine Ridge, SD (96)

^a.Places less than 80% one group were greater than 80% in 2000 and underwent racial change during the decade.

Appendix Table A3.

Logistic Regression Sensitivity Anlysis Using a 70% Threshold to Define Diversity (n = 3,927)

	Odds Ratio	95% CI
Location type (ref: Nonmetro not adjacent to metro)
Central City	0.13	(0.06, 0.29)
Suburb in large metro	2.09	(1.52, 2.87)
Suburb in small-med metro	1.76	(1.30, 2.38)
Nonmetro adjacent to metro	1.08	(0.76, 1.52)
Place race and ethnic
composition (25% criterion)
Black (5%)	0.53	(0.39, 0.73)
Asian (2%)	3.84	(2.90, 5.08)
Latinx (6%)	2.13	(1.59, 2.85)
Native American (2%)	3.20	(2.14, 4.80)
Region (ref: South)
North	1.39	(0.46, 4.19)
Midwest	1.27	(0.36, 4.53)
West	4.74	(2.61, 8.61)
Race by Region (ref: South)
Black
North	2.09	(1.09, 4.00)
Midwest	4.70	(1.68, 13.12)
West	1.57	(0.99, 2.47)
Latinx
North	0.20	(0.10, 0.40)
Midwest	0.52	(0.24, 1.14)
West	0.32	(0.19, 0.52)
Asian
North	1.45	(0.58, 3.64)
Midwest	0.31	(0.15, 0.61)
West	0.46	(0.31, 0.68)
Native
North	0.00	(0.00, >100)
Midwest	0.09	(0.01, 0.95)
West	0.21	(0.12, 0.36)
Employment hub
Military	2.72	(2.06, 3.58)
University	1.96	(1.34, 2.88)
Public sector	1.78	(1.39, 2.28)
Percent variablesa
Foreign-born	1.18	(1.09, 1.27)
Renters' share	1.00	(0.95, 1.05)
New housing stock	1.10	(1.06, 1.15)
Spatially lagged rate of integration	21.33	(15.33, 29.68)

^a.Percentage variables were scaled to units of 10 percentage-points for easier interpretation of effect sizes. For example, a 10 percentage-point increase in foreign-born population associates with an odds ratio of 1.20.

Appendix Table A4.

Logistic Regression Sensitivity Anlysis Using a H~ ≤ 0.20 Threshold to Define Integration

	Odds Ratio	95% CI
Location type (ref: Nonmetro not adjacent to metro)
Central City	0.18	(0.10, 0.32)
Suburb in large metro	1.99	(1.50, 2.64)
Suburb in small-med metro	1.83	(1.40, 2.38)
Nonmetro adjacent to metro	1.10	(0.83, 1.46)
Place race and ethnic composition (25% criterion)
Black (5%)	0.52	(0.39, 0.69)
Asian (2%)	4.05	(3.16, 5.18)
Latinx (6%)	1.47	(1.15, 1.88)
Native American (2%)	3.04	(2.15, 4.31)
Region (ref: South)
North	1.74	(0.72, 4.23)
Midwest	2.16	(0.87, 5.37)
West	3.23	(1.84, 5.70)
Race by Region (ref: South)
Black
North	1.74	(0.94, 3.21)
Midwest	2.32	(1.12, 4.81)
West	1.66	(1.07, 2.60)
Latinx
North	0.40	(0.21, 0.74)
Midwest	0.53	(0.27, 1.03)
West	0.54	(0.33, 0.88)
Asian
North	1.01	(0.54, 1.91)
Midwest	0.40	(0.22, 0.71)
West	0.53	(0.37, 0.76)
Native
North	0.34	(0.02, >100)
Midwest	0.09	(0.02, 0.39)
West	0.29	(0.18, 0.47)
Employment hub
Military	2.83	(2.15, 3.74)
University	2.00	(1.39, 2.86)
Public sector	1.50	(1.19, 1.88)
Percent variablesa
Foreign-born	1.20	(1.11, 1.29)
Renters' share	0.97	(0.92, 1.01)
New housing stock	1.11	(1.06, 1.16)
Spatially lagged rate of integration	21.66	(15.70, 29.90)

^a.Percentage variables were scaled to units of 10 percentage-points for easier interpretation of effect sizes. For example, a 10 percentage-point increase in foreign-born population associates with an odds ratio of 1.20.

Appendix Table A5.

Logistic Regression Sensitivity Anlysis Using a H~ ≤ 0.20 Threshold to Define Integration

	Odds Ratio	95% CI
Location type (ref: Nonmetro not adjacent to metro)
Central City	0.18	(0.10, 0.32)
Suburb in large metro	1.99	(1.50, 2.64)
Suburb in small-med metro	1.83	(1.40, 2.38)
Nonmetro adjacent to metro	1.10	(0.83, 1.46)
Place race and ethnic
composition (25% criterion) Black (5%)	0.52	(0.39, 0.69)
Asian (2%)	4.05	(3.16, 5.18)
Latinx (6%)	1.47	(1.15, 1.88)
Native American (2%)	3.04	(2.15, 4.31)
Region (ref: South)
North	1.74	(0.72, 4.23)
Midwest	2.16	(0.87, 5.37)
West	3.23	(1.84, 5.70)
Race by Region (ref: South)
Black
North	1.74	(0.94, 3.21)
Midwest	2.32	(1.12, 4.81)
West	1.66	(1.07, 2.60)
Latinx
North	0.40	(0.21, 0.74)
Midwest	0.53	(0.27, 1.03)
West	0.54	(0.33, 0.88)
Asian
North	1.01	(0.54, 1.91)
Midwest	0.40	(0.22, 0.71)
West	0.53	(0.37, 0.76)
Native
North	0.34	(0.02, >100)
Midwest	0.09	(0.02, 0.39)
West	0.29	(0.18, 0.47)
Employment hub
Military	2.83	(2.15, 3.74)
University	2.00	(1.39, 2.86)
Public sector	1.50	(1.19, 1.88)
Percent variablesa
Foreign-born	1.20	(1.11, 1.29)
Renters' share	0.97	(0.92, 1.01)
New housing stock	1.11	(1.06, 1.16)
Spatially lagged rate of integration	21.66	(15.70, 29.90)

^a.Percentage variables were scaled to units of 10 percentage-points for easier interpretation of effect sizes. For example, a 10 percentage-point increase in foreign-born population associates with an odds ratio of 1.20.