Ethnoracial Diversity across the Rural-Urban Continuum

Abstract

Both scholarship and popular opinion have long held that cities are more racially and ethnically diverse than rural communities. However, recent trends hint at the potential for less distinctive diversity profiles on either side of the metro-nonmetro divide. To explore this possibility, we compare the magnitude and structure of ethnoracial diversity in 2010 for over 27,000 census-defined places arrayed across 10 different types of county contexts that span the rural-urban continuum. Although the average resident’s exposure to diversity steadily declines as contexts become more rural and remote, place-based (or unweighted) results show an uneven pattern of diversity across most of the continuum. Our multivariate analysis also supports the unevenness scenario: when detailed characteristics of places are taken into account, many of the associations between the context indicators and diversity weaken to the point of non-significance. Taken together, these findings suggest a blurring of rural-urban boundaries with respect to community ethnoracial composition.

Population diversity has long been considered a defining feature of the urban environment. As Louis Wirth (1938) put the matter during the heyday of the Chicago School, cities are not only large and densely settled but heterogeneous in composition. Age, household type, socioeconomic status, and race-ethnicity rank among the major demographic dimensions of diversity. The last dimension is arguably the most consequential, given its correlations with the others. Driven by immigration and higher rates of natural increase among minority groups, ethnoracial diversity has risen impressively in recent decades, reshaping metropolitan America (Frey 2015; Lee et al. 2014). Whatever the effects of the diversification trend—on the economy, politics, education, and social relations (see, e.g., Alesina and La Ferrara 2002; Lichter 2013; Portes and Vickstrom 2011; Waters and Pineau 2015)—they are assumed to be strongest in the nation’s gateway cities, a number of which now exhibit minority-majority compositions.

The multiethnic character of large cities contrasts sharply with accepted wisdom about rural communities. These communities are often perceived as homogeneous in many respects, including their racial-ethnic mix (Lichter and Brown 2011). Certainly the ‘Lake Wobegon’ image of places in the nonmetropolitan Midwest inhabited by descendants of Northern European immigrants has a factual historical basis (Lieberson and Waters 1988). Rural white homogeneity can also be traced to blatant forms of discrimination that excluded people of color from ‘sundown towns’ (Loewen 2005) and, more recently, to municipal underbounding (Lichter et al. 2007). Even when non-whites are present, they may live in equally homogeneous settings; think African Americans in small ‘Black Belt’ places, Latinos (mainly Mexicans) throughout South Texas, or American Indian residents of reservation communities. Such examples, not to mention popular stereotypes, stress the wide gulf between low-diversity nonmetro communities and their high-diversity metropolitan counterparts.

Our research proceeds from the premise that much has changed over the past half-century to challenge this conventional view. Subsequent to passage of the Immigration and Nationality Act of 1965, shifts in immigration policy have increased foreign-born persons’ freedom of movement within the United States (via IRCA-induced legalization of undocumented persons) while deflecting them from established gateways (via border-hardening measures and state and local legislation) (Light 2006; Massey 2008; Tienda and Sanchez 2013). Many members of immigrant-rich groups are leaving saturated urban markets behind for the economic opportunities, affordable housing, and quality of life available in nonmetropolitan places. Their ruralward shift is partly a function of low-skill labor demand in agricultural processing, oil and natural gas production, and other sectors (Kandel and Parrado 2005). Similarly, rural retirement and amenity destinations are drawing immigrants to fill construction and service jobs (Johnson and Lichter 2013; Nelson et al. 2009). Recruitment efforts and ethnic social networks continue to increase awareness of rural employment among Hispanics in particular, who show signs of further geographic dispersion (Kandel and Cromartie 2004; Lichter and Johnson 2006).

As a result, nonmetro places have become more ethnoracially diverse, sometimes in dramatic fashion. A robust literature documents the nonmetro diversification trend and weighs its potential consequences for communities unaccustomed to incorporating minorities (Crowley and Ebert 2014; Lichter 2012; Sharp and Lee, forthcoming). Some of these communities qualify as ‘urban’ in Census Bureau terms, exceeding a population threshold of 2,500 by themselves or as part of a cluster of adjacent places. Others are close to metro areas, a fact that increases their chances of being absorbed through centrifugal expansion or, at minimum, of being subject to metropolitan influences. This urbanization of rural America has been fueled by transportation advances, innovations in information technology, globalization, industrial restructuring, growing corporate dominance, and the pull—for recreation or extraction—of natural resources (Brown 2014; Lichter and Brown 2011). At the same time, many metro areas contain swaths of sparsely settled territory and non-trivial rural populations. The persistence of rurality in metropolitan settings is most evident in exurban or fringe zones, which tend to attract affluent white inhabitants. More generally, evidence suggests that the share of residential segregation due to place-level sorting has increased over time, yielding homogeneous communities in the midst of metro-wide diversity (Farrell 2008; Lichter et al. 2015).

We evaluate a key demographic implication of these overlapping trends: that the diversity profiles of rural and urban communities may no longer be as distinct as commonly thought. Our analysis is designed to address several limitations of existing diversity research. Prominent among these are the inconsistent conceptualization and measurement of diversity across studies and the rarity with which both rural and urban diversity are considered in the same study. Moreover, when rural-urban comparisons on any characteristic (not just diversity) are made, they are often crude, e.g., between places inside and outside metro areas or of different population sizes. Such approaches, which treat ‘rural’ and ‘urban’ in dichotomous or unidimensional fashion, may not adequately reflect the complexity of community types nationwide. Neither do they recognize the potential for variation among communities within different kinds of metropolitan or nonmetropolitan settings. A preferable strategy, adopted here, emphasizes gradations along a rural-urban continuum, with a community’s position on the continuum determined by aspects of the surrounding context in which it is embedded.

Specifically, our analysis compares 2010 patterns of ethnoracial diversity for over 27,000 places found in the largest metro areas, the most remote rural counties, and a variety of locations in between. Most of the places are incorporated as cities, suburbs, towns, or villages while others (termed census-designated places) lack municipal status; both types of places constitute meaningful social, symbolic, and institutional entities. We assign each place to one of 10 rural-urban continuum categories based on the metropolitan status or proximity of its host county and, in certain instances, the size of that county’s urban population. Comparisons across categories allow us to address three central questions. First, how does the magnitude and racial-ethnic structure of diversity for places vary by position on the continuum? Second, are the average residents of places along the continuum exposed to similar or different diversity magnitudes and structures? And third, is rural-urban continuum category associated with ethnoracial diversity when more detailed place characteristics are taken into account?

Background

Dimensions of Diversity

The term ‘diversity’ is regularly used in loose fashion, to refer to the presence of African Americans, Latinos, or some other minority group in a community. Here we stick to a more precise demographic definition that emphasizes two aspects of the ethnoracial composition of the local population. The magnitude or level of diversity is determined by the number of racial-ethnic groups that make up the population and their relative sizes (White 1986). A place comprising many groups of equal size would be judged highly diverse. At the opposite extreme, homogeneity prevails when all residents belong to the same group, i.e., when the population exhibits an absence of diversity. Several statistics are available with which to capture the magnitude dimension of diversity. We favor the entropy index, both for its conceptual congruence and its desirable statistical properties.

Diversity also varies in terms of racial-ethnic structure, or the specific groups constituting a population. The importance of the structural dimension lies in the fact that it may differ among places with identical diversity magnitudes. Consider three communities that each has a 50–50 composition: one is half white and half Asian, one is half black and half Hispanic, and one is half Hispanic and half Native American. Their entropy index values would all be exactly the same, yet they are likely to display divergent socioeconomic mixes, interethnic relations, and the like. In short, to understand ethnoracial diversity, we must know which particular groups are present, not just their number and sizes. A ‘majority rule’ typology, introduced later, allows us to distinguish among places on the structural dimension.

Another valuable distinction can be made between place-based (or unweighted) and person-based (or weighted) diversity. In the former instance, all communities—from the principal cities of major metropolitan areas to the smallest rural villages—are considered equally significant. As the rank-size rule implies, however, more people in total may reside in the largest cities despite the much greater number of villages. This uneven distribution of population across types of communities means that the level or structure of place diversity experienced by the average American could differ from the average level or structure of place diversity if, for example, we assume a positive relationship between community size and diversity in line with conventional Wirthian wisdom. Similar concentrations of people in the largest places within each category of the rural-urban continuum seem likely as well. Thus, our analysis compares person-based diversity exposure to place-based diversity across and within continuum categories. Both approaches are valid depending upon one’s purpose, but all previous studies of which we are aware have chosen one or the other.

Urban vs. Rural?

Perhaps the most glaring lacuna in the diversity literature is the failure to examine communities along the full range of the rural-urban continuum in a single analysis. Due to minority overrepresentation in cities and the perceived homogeneity of rural places, diversity research focuses primarily on metropolitan areas, places, and neighborhoods (Farrell and Lee 2011; Hall et al. 2016; Lee et al. 2014; Logan and Zhang 2010). Across metro units of varying spatial scale, rapid Hispanic and Asian growth rates have combined with absolute or relative white declines to erode the demographic primacy of whites, especially in the South and West. Most large cities are now marked by multigroup racial-ethnic structures, and fewer suburbs conform to the all-white image embedded in popular culture (Berube 2003; Hall and Lee 2010). Nevertheless, many metro places continue to be dominated by a single ethnoracial group, and some have even become more homogeneous over time (Lee and Hughes 2015).

A few recent studies do include both metropolitan and nonmetropolitan geographic units. However, the degree of differentiation among such units remains coarse. Relying on the entropy index and a typology of racial-ethnic structure, Parisi et al. (2015) document a substantial increase between 1990 and 2010 in the number of metro and nonmetro places with four-group structures (white-black-Hispanic-Asian), although the metro-nonmetro gap in mean diversity magnitude widened during the period. A 1980–2010 comparison of diversity in metropolitan and micropolitan areas rather than places yields similar findings (Lee et al. 2014). Hall and his associates (2016), who emphasize the variability in place-level patterns of diversity change, nevertheless show that the steepest upward-sloping trajectories (toward greater diversity) have disproportionately involved metro places instead of micropolitan or rural ones. Population size, an alternative to the metro-nonmetro dichotomy, exhibits a consistent positive correlation with ethnoracial diversity (Allen and Turner 1989; Farrell 2005; Hall and Lee 2010; Lee et al. 2012). But it only taps a single dimension of the rural-urban continuum, and it tends to be used in analyses of subsets of communities (e.g., metro areas, suburbs, rural places).

What is missing—and what we aim to provide—are finer-grained distinctions among the larger settings in which places are located. Metropolitan areas, for example, vary dramatically in population size, as the contrast between New York (18.9 million residents in 2010) and Carson City, NV (55,274) attest. Nonmetro counties vary as well, not only in size but in degree of urbanization and distance to the nearest metropolis. Intuitively, such features of areal contexts would appear to have implications for the diversity of their constituent places. Imagine three towns of a few thousand inhabitants each, one situated in a top-10 metro area, one in a nonmetro county adjacent to a medium-sized metro area, and one far removed from any type of metropolitan settlement. Because urbanism-related influences of the surrounding context are presumably strongest in the first town, Wirthian wisdom would lead us to anticipate high diversity there as immigrants and minority groups—and the forces attracting them—disperse throughout the metropolis. These dynamics may operate to some degree in the second hypothetical town but be weakest in the third.

Fortunately, a handful of classification systems are available that capture basic differences across metropolitan and nonmetropolitan contexts. The USDA’s rural-urban continuum (or RUC) codes strike us as especially attractive for this purpose. Created by Economic Research Service staff in 1975 and updated periodically since then, the scheme consists of nine categories. The first three categories, which we expand to four, differentiate metropolitan counties by the total population size of the metro area to which they belong. The remaining six distinguish among nonmetropolitan counties based on the size of their urban population and proximity (adjacency) to a metro area (U.S. Department of Agriculture 2013). In essence, the RUC scheme facilitates a shift from dichotomous, urban vs. rural thinking to a perspective that identifies gradations along a continuum of county contexts which are multidimensionally defined. The scheme’s conceptual and operational advantages have made it popular in research on other topics. To date, though, it has not been used to study place variation in ethnoracial diversity (but see Winkler and Johnson 2016).

Research Questions

By sorting a large number of places into their appropriate RUC categories, we can shed empirical light on three descriptive research questions. The first question asks how the magnitude and structure of place-based diversity varies along the rural-urban continuum. From a Wirthian point of view, the most likely scenario would be a linear decline in diversity as one moves away from the largest metropolitan settings and toward the least urbanized, most isolated rural contexts. Alternatively, a threshold or stairstep pattern could occur if the metro-nonmetro distinction is meaningful, i.e., if the drivers of ethnoracial diversity are present in all metropolises regardless of size but are absent from nonmetropolitan America. The blurring of rural and urban domains noted earlier, however, suggests more unevenness across continuum categories. As an illustration, places in highly urbanized nonmetro counties that are adjacent to large metropolitan areas might be as diverse as—if not more diverse than—their counterparts located in small metro contexts. Carried to an extreme, blurring logic yields the null hypothesis: that diversity levels and racial-ethnic structures for places do not vary by context, given the limited relevance of traditional urban-rural or metro-nonmetro distinctions in the contemporary U.S.

Our second question reframes the first in person-based terms. Namely, do the average residents of places at various points along the rural-urban continuum experience different or similar forms of ethnoracial diversity? The same hypothetical patterns or relationships just mentioned—linear decline, threshold, uneven, null—also apply to this question, yet the pattern receiving the most support need not be the same as for the first question. Indeed, to the extent that people are disproportionately concentrated in larger places across and within the RUC categories, the exposure (weighted) results will diverge from the place-based (unweighted) results.

The final research question is broader than the previous two, inquiring about correlates of the magnitude and structure of diversity among places. In particular, we are interested in the robustness of any ‘effect’ registered by position on the rural-urban continuum: does a place’s RUC category independently predict diversity after controlling for more detailed characteristics of the place that its RUC category might be proxying? Three general types of place characteristics have been identified in prior diversity investigations (Allen and Turner 1989; Farrell 2005; Lee et al. 2012; Hall and Lee 2010; Sharp and Lee forthcoming). With respect to the context of reception provided by a place, location in the West or South (closer to Hispanic and Asian countries of origin) and a critical mass of foreign-born residents are related to higher diversity and more balanced racial-ethnic structures. A larger retirement-age population, however, tends to undermine ethnoracial diversity, perhaps because it signals a stagnant economy or because its members are uncomfortable living near immigrants and minorities.

A place’s housing and labor market characteristics, broadly reflective of opportunity, must also be incorporated as controls. Access to housing—indexed by features such as affordability, an abundance of rental units, and new construction activity (which creates vacancies in existing units via chains of moves)—is a draw for most ethnoracial groups and hence should promote diversity. Higher incomes and lower unemployment rates have widespread appeal as well. So does a local economy with types of jobs that suit a wide range of educational and skill levels. Lastly, places serving as institutional hubs for government, the military, and higher education are more likely to have diverse, multigroup compositions. Committed to affirmative action, these institutions provide avenues of upward mobility for people of color. The presence of correctional facilities, a less voluntary kind of institution, is also positively associated with diversity in a community.

Methodology

Places as Units

To address our three questions about variation in ethnoracial diversity across the rural-urban continuum, we extract data from the 2010 decennial census and the 2008–2012 American Community Survey (ACS) 5-year summary file for 27,163 places with at least 100 residents. These places comprise cities, suburbs, towns, and villages. Two-thirds of them are incorporated, with legally vested powers and obligations, and the largest—principal cities of major metropolises—often approximate housing and labor markets. Among other duties, incorporated places are responsible for developing fiscal or policy responses to diversity-related issues that occur inside their boundaries. The remaining third of the sample consists of unincorporated or census-designated places. Residents recognize both incorporated and census-designated places by name and may feel some degree of attachment to them. More concretely, the population of a place influences the ethnoracial composition of local neighborhoods, schools, work settings, and voluntary organizations, not to mention the social relationships that form in these venues. On a number of criteria, then, places qualify as ‘real’ communities in addition to being convenient statistical aggregations.

Measuring Diversity

We employ 2010 census tabulations to delineate six panethnic groups that form the building blocks for our diversity measures: Hispanics of any race, non-Hispanic whites, non-Hispanic blacks, non-Hispanic Asians (including Pacific Islanders), non-Hispanic Native Americans (American Indians and Alaska Natives), and all other non-Hispanics (multi-race and other-race individuals). Counts of these panethnic groups have been assembled for each place and, taken together, provide exhaustive, mutually exclusive coverage of the local population.

The panethnic counts are used to capture the two diversity dimensions of interest. The first dimension, magnitude, reflects how evenly residents of a place are divided among the six panethnic groups. We operationalize diversity magnitude with the entropy index, symbolized by E (White 1986). E takes a maximum value equal to the natural log of the total number of groups, or 1.792 in our six-group case. For ease of interpretation, we divide each diversity score by this theoretical maximum and multiply by 100, resulting in a 0–100 range of possible values. A diversity score of zero indicates complete homogeneity, that only one group inhabits a place. Of the 221 such places in our sample, 176 have all-white populations, 43 are entirely Hispanic, and two are entirely Native American. At the opposite extreme, a community that contains identical shares (16.7%) of each of the six groups would receive a score of 100. Although none of our places reaches that level, the village of Hillburn, NY (near New York City) comes closest with an E score of 87.1.

Racial-ethnic structure, the second dimension of diversity, refers to the specific groups that live in a community. Our investigation captures this dimension in a couple of ways. At various points E scores are accompanied by bar charts that visually convey the group proportions underlying the magnitude of diversity. We also utilize a ‘majority rule’ typology (see Farrell and Lee 2011; Holloway et al. 2011) that classifies communities as group-majority (white-majority, Hispanic-majority, etc.) if one ethnoracial group constitutes over 50% of the total population. White-majority places are further subdivided into white-dominant, in which whites are at least 90% of the local population, and white-shared, in which the white percentage is over one-half but less than nine-tenths. We name white-shared subtypes based on any minority groups that constitute 10% or more of all residents (e.g., white-Hispanic, white-black-Hispanic). No-majority places are defined as those where none of the ethnoracial groups exceeds the 50% threshold. Applying this typology, we are able to gauge how places with particular kinds of racial-ethnic structures are distributed along the rural-urban continuum.

Rural-Urban Classification

As noted previously, the nine-category classification scheme developed by USDA’s Economic Research Service enables us to assign places to different types of metropolitan and nonmetropolitan county contexts. We have revised the original scheme slightly, subdividing the most urban category (RUC1) into counties in metro areas of 3 million or more people (RUC1a) and counties in metro areas that fall in the 1–3 million range (RUC1b). We have also attached a brief verbal descriptor to the numerical code for each category. Operational definitions of the 10 county contexts are as follows:

• Super metro (RUC1a)—County in metro area of 3+ million population; place N=3,585

• Large metro (RUC1b)—County in metro area of 1 million to 2,999,999 population; place N=3,393

• Medium metro (RUC2)—County in metro area of 250,000 to 999,999 population; place N=4,639

• Small metro (RUC3)—County in metro area of less than 250,000 population; place N=3,372

• High-urban proximate (RUC4)—Nonmetro county with urban population of 20,000+, adjacent to metro area; place N=2,472

• High-urban distant (RUC5)—Nonmetro county with urban population of 20,000+, not adjacent to metro area; place N=808

• Low-urban proximate (RUC6)—Nonmetro county with urban population of 2,500 to 19,999, adjacent to metro area; place N=4,008

• Low-urban distant (RUC7)—Nonmetro county with urban population of 2,500 to 19,999, not adjacent to metro area; place N=2,474

• Rural proximate (RUC8)—Nonmetro county with urban population less than 2,500, adjacent to metro area; place N=939

• Rural distant (RUC8)—Nonmetro county with urban population less than 2,500, not adjacent to metro area; place N=1,473

Each continuum category contains a substantial number of places (from a low of 808 in RUC5 to a high of 4,639 in RUC2) and thus permits more nuanced comparisons of diversity magnitude and structure across a wider range of community contexts than usual.

Other Variables

In the multivariate portion of the analysis, we utilize three broad sets of place characteristics to assess the robustness of any zero-order associations detected between diversity and position on the rural-urban continuum. The local context of reception is captured with indicators of region, immigrant presence, and the retirement-age population. Four conventional census-defined regions are recognized: Northeast, Midwest, South, and West. We operationalize immigrant presence as the percentage of foreign-born residents in a place, and the percentage of residents 65 years of age or older constitutes our measure of the retirement-age population.

The second set of place variables covers selected features of the housing and labor market that may make a place more or less attractive to multiple ethnoracial groups. Housing characteristics include the stock of new homes (measured as the percentage of units built since 2000), the percentage of renter-occupied units, and rent burden (median rent as a percentage of household income). We represent labor market opportunities with median household income in the past 12 months, the percentage of civilians 16+ years old who report being unemployed, and an occupational diversity variable that reflects the range of job types available in a community. The diversity variable, constructed using the entropy index, quantifies how evenly workers are distributed across five general occupational categories extending from professional (management, business, science and arts occupations) to blue collar (production, transportation, and material moving occupations).

Finally, we develop four measures of institutional hub status that denote whether a place specializes in government, military, higher education, or correctional functions. Places qualify as government hubs if the percentage of their employed residents holding federal, state, or local government jobs is at least double the percentage in the total (summed) place population nationally (coded 1 if yes, 0 otherwise). The same threshold has been incorporated in the remaining measures. For military specialization, the share of a place’s labor force participants employed in the armed forces must be two times (or more) greater than the national percentage. In the case of educational and correctional specialization, we apply the doubling rule to the percentage of local residents who are enrolled in college or who are incarcerated in adult or juvenile facilities, respectively.

Results

Place-Based Diversity

Our initial research question concerns how ethnoracial diversity varies across places in different contexts along the rural-urban continuum. Average E scores, which tap diversity magnitude, are reported for the 10 RUC categories at the right edge of Figure 1. At first glance, the diversity of places located in super-metro counties (RUC1a) stands out: their mean E is twice that for the nonmetro places located in rural distant counties (RUC9). But changes in diversity are hardly monotonic as one moves from the top to the bottom of the figure. Nonmetro places in the RUC5 and 6 categories, for example, reach the same diversity levels as their metro RUC2 and 3 counterparts. In fact, the mean E for the RUC5 places—found in highly urban but distant nonmetro settings—equals the mean for the large metro places (RUC1b). Taken together, these mean patterns best conform to the unevenness scenario.

Figure 1.

Mean Place Diversity by RUC Category

The overlap in diversity across the RUC categories becomes clear when we examine the distribution of E scores within each category. A comparison of the boxplots in Figure 2 reveals the elevated diversity levels of super-metro places (RUC1a). However, the value of E at the 75^th percentile for these places falls between the 25^th and 75^th percentiles of the distributions for the nine other continuum categories. Places at all points on the continuum have long top ‘whiskers’ as well, with the most diverse places in five of the six nonmetro contexts exhibiting Es between 76 and 85. Thus, there seems to be little evidence of linear or stair-step diversity declines with decreasing urban character, aside from the pronounced downward shift between the super metro (RUC1a) and large metro (1b) contexts.

Figure 2.

Distribution of Place Diversity (E) by RUC Category

The racial-ethnic structures of communities in different continuum categories also appear broadly similar. Returning to Figure 1, the segments that make up each compositional bar reflect the representation of panethnic groups in the average place within that category. Whites constitute seven-tenths or more of place residents across the board, and Native Americans tend to be over-represented in nonmetro places, especially the low-urban distant (RUC7) and rural distant (RUC9) types. In the case of other groups, generalizations are difficult. Blacks constitute a larger share of place populations in rural proximate settings (RUC8) than anywhere else, Hispanics reach double-digit percentages in selected metro (RUC1a and 2) and nonmetro (RUC5) places, and Asian percentages—though relatively small—are highest in RUC1a and RUC5 places, similar to the Hispanic pattern.

Application of our majority-rule typology unpacks the means summarized by the compositional bars. In Table 1 we show the distribution of places across types within each continuum category. The first column highlights the distinctive ethnoracial structures of places in super-metro counties (RUC1a). Almost three-fifths of these places are white-shared (whites in the numerical majority but less than 90% of the total population), with white-multigroup and white-Hispanic communities the most common subtypes. Super metro places are also the least likely to be white-dominant and the most likely to qualify as no-majority (lacking a panethnic group to which 50% or more of all residents belong). In contrast, the frequency of white dominance exceeds that of a white-shared composition in every continuum category other than the super-metro context, sometimes by wide margins. While white-multigroup places are the modal white-shared subtype in RUC1b through RUC5 settings, either white-black or white-Hispanic places tend to be modal toward the rural end of the continuum (RUC6–9). With respect to minority-majority structures, rural proximate counties (RUC8) boast the greatest share of black-majority places, medium-sized metro areas (RUC2) the greatest share of Hispanic-majority places, and rural distant counties (RUC9) the greatest share of Native American-majority places. Asian-majority communities are scarce in every type of context.

Table 1.

Distribution of Places by Majority-Rule Structure and RUC Category

Type of Structure	RUC Category
	1a	1b	2	3	4	5	6	7	8	9
White dominant	19.8	47.5	43.1	54.5	56.8	43.1	55.3	58.0	58.6	61.4
White shared	56.2	41.4	41.0	35.4	30.2	37.0	29.9	27.5	29.4	25.6
White-Black	4.1	10.8	9.4	8.2	5.7	4.3	8.5	4.7	11.6	4.4
White-Hispanic	16.3	8.3	11.2	8.6	9.9	13.1	8.5	10.1	5.5	9.5
White-Native American	0.0	0.2	1.4	0.4	0.9	2.4	2.3	2.3	1.2	3.1
White-Black-Hispanic	4.2	2.7	2.5	2.2	1.1	1.5	1.5	0.7	1.2	0.1
White-Multigroup	21.4	17.6	15.2	15.0	11.9	13.7	8.2	8.8	9.7	7.5
Black majority	5.0	5.0	2.7	2.9	2.8	3.8	5.4	4.5	8.1	2.0
Hispanic majority	6.6	2.0	8.5	3.3	5.7	5.3	3.4	3.4	0.9	1.8
Asian majority	0.5	0.1	0.4	0.1	0.0	1.4	0.0	0.0	0.0	0.1
Native American majority	0.5	0.0	0.7	1.0	1.9	2.0	2.1	4.9	1.8	6.9
No majority	11.3	4.1	3.6	2.7	2.6	7.4	3.8	1.8	1.3	2.2

Note: N = 27,163 places.

Two lessons emerge from the place-focused portion of the analysis. First, places in super-metro areas have higher diversity levels and more complex racial-ethnic structures than places elsewhere, on average. Oakland and Jersey City exemplify this point: their E scores fall in the mid-80s and they contain roughly equal percentages of white, black, Hispanic, and Asian residents. The second lesson is that sharp deviations in diversity patterns are not apparent along the rest of the rural-urban continuum despite some degree of unevenness. Put another way, nonmetro places can be quite diverse, rivaling their metro siblings. As an illustration, Unalaska, AK—in the rural distant or RUC9 category—displays a high diversity magnitude (E = 84) that is driven by nontrivial proportions of Asians (34.1%), whites (33.7%), and Hispanics (15.2%). Other no-majority places in nonmetropolitan America include Winslow, AZ (RUC4), Nanawale Estates, HI (RUC5), Andarko, OK (RUC6), and Crescent City, CA (RUC7). All have entropy scores above 75 but varying combinations of panethnic groups.

Diversity Exposure

Unlike the first question, which treats every place as equal, the second question guiding our research recognizes the uneven distribution of population among communities. Specifically, it asks how person-based (or weighted) estimates of ethnoracial diversity compare with place-based (unweighted) estimates not only for the total sample but across and within the 10 contextual categories that comprise the rural-urban continuum. For the sample as a whole, the weighted E reaches 50.5. This reflects the diversity level to which the average inhabitant of our 27,163 places was exposed in 2010. The racial-ethnic structure experienced by that hypothetical inhabitant remains primarily white (58.0%) but with non-trivial shares of Hispanic (19.5%), black (13.7%), and Asian (5.8%) dwellers. By contrast, unweighted or place-based means for the total sample indicate a much lower magnitude of diversity (E = 28.1), greater white representation (77.6%), and mean minority-group shares below 10%.

Such differences are consistent with the notion that exposure to diversity falls in a roughly linear manner as county contexts become less urban. The weighted E scores and compositional bars in Figure 3 support this inference. With the exception of the RUC5 or high-urban distant category, diversity magnitude declines rather steadily from the super-metro to the rural-distant end of the continuum. While the typical denizens of rural nonmetro places (RUC8 and 9) encounter white-dominated homogeneity, ethnoracial heterogeneity is the norm for people living in super-, large-, and medium-sized metro areas (RUC1a, 1b, and 2). These people experience high levels of diversity—Es ranging from the mid-40s to nearly 60—and racial-ethnic mixes in which 40% to 50% of their fellow residents are people of color. Hispanics constitute the largest minority in the RUC1a, 1b, and 1c categories (roughly one-fifth of the population), followed by blacks (10–15%) and Asians (4–8%).

Figure 3.

Exposure to Place Diversity by RUC Category

A category-by-category comparison of the Figure 3 results with those in Figure 2 shows person-based diversity levels to be higher than place-based ones across the board. That is, within each type of RUC context more individuals live in bigger places that tend to be more diverse. We illustrate the principle in Table 2 by examining the diversity magnitude and structure of the 10 largest places in the U.S. Eight of the 10 are located in super-metro areas (RUC1a); the two that are not, San Antonio and San Jose, anchor large metro contexts. Diversity levels (column 2) substantially exceed that of the average RUC1a resident (E = 59.0), with New York leading the way (79.7). Only San Antonio slips below the person-based average, owing to its Hispanic-majority composition. The nine other cities are all no-majority in nature: the fact that three of them have white pluralities, two have black pluralities, and four have Hispanic pluralities attests to their varied multigroup racial-ethnic structures. Asian representation also reaches double-digit percentages in four cities, including the three in California.

Table 2.

Diversity of 10 Largest Places

	Total Population	E Score	% White	% Black	% Hispanic	% Asian	% Native American	% Other
New York	8,175,133	79.7	33.3	22.8	28.6	12.6	0.2	2.5
Los Angeles	3,792,621	71.0	28.7	9.2	48.5	11.2	0.2	2.3
Chicago	2,695,598	73.6	31.7	32.4	28.9	5.4	0.2	1.5
Houston	2,099,451	71.7	25.6	23.1	43.8	6.0	0.2	1.3
Philadelphia	1,526,006	70.2	36.9	42.2	12.3	6.3	0.2	2.1
Phoenix	1,445,632	63.7	46.5	6.0	40.8	3.2	1.6	1.9
San Antonio	1,327,407	54.4	26.6	6.3	63.2	2.4	0.2	1.3
San Diego	1,307,402	73.6	45.1	6.3	28.8	16.0	0.3	3.5
Dallas	1,197,816	69.0	28.8	24.6	42.4	2.8	0.3	1.2
San Jose	945,942	73.1	28.7	2.9	33.2	32.1	0.2	2.9

Accounting for Diversity

Thus far the relationship between a place’s ethnoracial diversity and its position on the rural-urban continuum has been examined in bivariate fashion. Our final research question requires a multivariate approach: are differences in diversity across continuum categories accounted for by other community characteristics, or do the differences persist? To address this question, we first regress diversity level (E) on nine dummy variables tapping the 10 RUC categories (with the rural distant category, RUC9, the omitted reference), then add the context of reception, housing and labor market, and institutional hub measures identified earlier. Table 3 summarizes these OLS regression results. In the partial model, all RUC variables have significant positive associations with E relative to RUC9, and the largest effect is location in a super-metro setting (RUC1a). After the other place attributes are entered, however, the size of the RUC1a coefficient shrinks by roughly 50% and the RUC1b through 8 coefficients are diminished as well, some becoming non-significant or negative in sign.

Table 3.

OLS Regression of Diversity Magnitude on RUC Categories and Other Place Characteristics

Place Characteristic	Partial Model		Full Model
	b	(SE)	b	(SE)
RUC Category [ref: Rural distant (9)]
Super metro (1a)	20.741	(.538) ***	10.303	(.460) ***
Large metro (1b)	9.504	(.536) ***	2.726	(.425) ***
Medium metro (2)	8.503	(.517) ***	0.920	(.406) *
Small metro (3)	4.397	(.540) ***	0.026	(.414)
High-urban proximate (4)	1.366	(.569) *	−1.104	(.435) *
High-urban distant (5)	9.038	(.757) ***	1.869	(.572) **
Low-urban proximate (6)	3.535	(.527) ***	0.886	(.398) *
Low-urban distant (7)	1.315	(.569) *	−1.263	(.428) **
Rural proximate (8)	1.734	(.722) *	−0.086	(.541)
Region (ref: Midwest)
Northeast			−0.449	(.481)
South			13.680	(.481) ***
West			11.252	(.481) ***
% Foreign-born (In)			3.094	(.066) ***
% 65+ years			−0.309	(.002) ***
% New housing (In)			−0.015	(.021)
% Rental housing			0.288	(.010) ***
Rent burden (In)			0.093	(.018)
Median income (In)			1.479	(.002) ***
% Unemployed			0.066	(.002) ***
Occupational diversity			0.023	(.002) ***
Government hub			2.173	(.002) ***
Military hub			3.718	(1.133) ***
Education hub			0.825	(.486)
Corrections hub			4.482	(.787) ***
Intercept	21.123	(.450) ***	−4.545	(2.924)
Adjusted R2	0.111		0.504

Notes: p < .001; p < .01; p < 05. N = 27,163. See text for operationalization of place characteristics.

The remaining predictors in Table 3 operate as anticipated for the most part. Greater ethnoracial diversity appears more likely in places with receptive contexts (i.e., located in the South or West, having more immigrants and few older residents). Abundant rental housing, higher incomes, and higher occupational diversity are also conducive to ethnoracial diversity, but the positive coefficient for the local unemployment rate contradicts our reasoning about the appeal of economic opportunity. Finally, most types of institutional hubs are significantly and positively correlated with a place’s E score.

A similar pattern of results can be seen in the first pair of logistic regression models in Table 4, in which no-majority status serves as the dependent variable. This similarity is not surprising since, by definition, E values will rise as community racial-ethnic structures incorporate more groups. The key finding for our purposes concerns the RUC indicators. Although positive and significant RUC coefficients are common in the partial model, most of them—including the RUC1a coefficient—fail to attain statistical significance in the full model or they reverse direction. Thus, position along the rural-urban continuum hardly seems decisive in determining which places have no-majority racial-ethnic structures. On the other hand, place measures that capture the local context of reception, housing and labor market characteristics, and the presence of institutional hubs play a central role, much as they do with respect to diversity magnitude.

Table 4.

Logistic Regression of Selected Majority-Rule Structure on RUC Categories and Other Place Characteristics

Place Characteristic	No Majority				White Dominant				White Shared
	Partial		Full		Partial		Full		Partial		Full
	b	(SE)	b	(SE)	b	(SE)	b	(SE)	b	(SE)	b	(SE)
RUC Category [ref: Rural distant (9)]
Super metro (1a)	1.714	(.184) ***	0.314	(.208)	−1.869	(.069) ***	−1.638	(.102) ***	1.325	(.069) ***	0.351	(.083) ***
Large metro (1b)	0.783	(.193) ***	−0.158	(.209)	−0.666	(.063) ***	−0.219	(.090) *	0.767	(.069) ***	0.139	(.078)
Medium metro (2)	0.495	(.193) *	−0.568	(.206) **	−0.742	(.061) ***	−0.135	(.086)	0.702	(.067) ***	0.043	(.075)
Small metro (3)	0.202	(.205)	−0.416	(.217)	−0.285	(.064) ***	0.119	(.088)	0.467	(.070) ***	0.058	(.077)
High-urban proximate (4)	0.132	(.218)	−0.337	(.229)	−0.191	(.067) **	0.060	(.093)	0.230	(.074) **	0.025	(.081)
High-urban distant (5)	1.253	(.221) ***	0.422	(.236)	−0.745	(.089) ***	0.000	(.122)	0.535	(.094) ***	0.001	(.103)
Low-urban proximate (6)	0.556	(.194) **	0.186	(.205)	−0.255	(.062) ***	−0.011	(.085)	0.215	(.069) **	0.038	(.075)
Low-urban distant (7)	−0.236	(.233)	−0.568	(.241) *	−0.143	(.067) *	0.241	(.091) **	0.095	(.075)	−0.101	(.081)
Rural proximate (8)	−0.571	(.340)	−0.721	(.350) *	−0.120	(.085)	0.122	(.114)	0.190	(.093) *	0.038	(.101)
Region (ref: Midwest)
Northeast			0.815	(.161) ***			0.186	(.055) ***			0.193	(.048) ***
South			1.775	(.136) ***			−2.316	(.044) ***			1.331	(.038) ***
West			1.385	(.145) ***			−2.221	(.054) ***			1.098	(.044) ***
% Foreign-born (In)			0.547	(.029) ***			−0.548	(.012) ***			0.216	(.009) ***
% 65+ years			−0.051	(.007) ***			0.084	(.003) ***			−0.003	(.002)
% New housing (In)			−0.032	(.025)			0.011	(.011)			0.051	(.010) ***
% Rental housing			0.023	(.002) ***			−0.037	(.002) ***			0.012	(.001) ***
Rent burden (In)			0.025	(.036)			0.024	(.012) *			0.062	(.010) ***
Median income (In)			0.145	(.115)			0.549	(.055) ***			0.596	(.045) ***
% Unemployed			0.022	(.005) ***			−0.028	(.002) ***			−0.011	(.002) ***
Occupational diversity			0.029	(.004) ***			0.014	(.001) ***			−0.005	(.001) ***
Government hub			0.172	(.093)			−0.538	(.050) ***			−0.021	(.041)
Military hub			−0.070	(.133)			−0.154	(.088)			0.348	(.066) ***
Education hub			−0.638	(.216) **			−0.005	(.115)			0.728	(.090) ***
Corrections hub			0.311	(.111) **			−0.467	(.082) ***			0.462	(.061) ***
Intercept	−3.776	(.176) ***	−9.448	(1.431) ***	0.466	(.054) ***	−5.660	(.643) ***	−1.067	(.060) ***	−7.920	(.531) ***
−2 Log Likelihood	9276		7482		35879		22562		34907		30761

Notes: p < .001; p < .01; p < 05. N = 27,163. See text for operationalization of place characteristics.

In the rest of Table 4, we consider whether the RUC contexts are better at predicting the most prevalent group-majority types of racial-ethnic structures. The second and third pairs of models regress white-dominant (90% or more white) and white-shared (less than 90% white) indicator variables on the RUC dummies and other place characteristics. Contrasting signs are evident in the partial models: while location in nearly every RUC category significantly decreases the likelihood of white dominance relative to rural distant places, the RUC categories have positive associations with white-shared status. However, adding the other characteristics reduces most of the RUC coefficients to non-significance and flips a few of their signs. Indeed, the super-metro coefficient (RUC1a) is the only one to remain both positive and significant in the full model for the white-shared type of structure. Among the non-RUC characteristics, many influence the odds of white-dominant and white-shared structures in the same way, but some discrepancies emerge. For example, location in the South or West, substantial foreign-born representation, an abundant supply of rental housing, and functional specialization in corrections all increase the chance that a place will possess a white-shared composition yet make white dominance less likely.

We have also estimated logistic regression models for black-majority and Hispanic-majority types of structures and for selected white-shared structures (i.e., white-black, white-Hispanic, and white-multigroup) that exist in an adequate number of places. These models (not shown) document more consistently positive, significant relationships between RUC context and type of ethnoracial structure than do those summarized in Tables 3 and 4. Once again, though, the individual RUC categories are of secondary importance to other community characteristics when predicting diversity structure or magnitude.

Conclusion

To the best of our knowledge, this is the only study to examine racial and ethnic diversity across the rural-urban continuum for a full range of communities, encompassing small hamlets in the countryside as well as teeming cities with millions of inhabitants. A key lesson learned from the study is that the distinction between person- and place-based diversity matters. The former approach yields results consistent with the linear decline scenario and conventional wisdom about differences in diversity experienced by average metropolitan and nonmetropolitan residents. But when the focus shifts to places as equivalent (unweighted) units, diversity variation among RUC contexts is modest and uneven, with places in super-metro settings the lone outlier. Thus, despite their apparent redundancy, our first two research questions turn out to be worth asking. The strength of the evidence favoring similarity or divergence in rural and urban diversity patterns depends upon how ‘place’ and ‘diversity’ are conceived and operationalized.

The place-based results supportive of unevenness are reinforced by the multivariate analysis undertaken to address our third question. When we include relevant characteristics of places as controls, many of the associations between the RUC context measures and diversity become non-significant or take signs contrary to the linear decline hypothesis. Intriguingly, the RUC categories also do not fare well in comparison with place population size, a simple unidimensional measure of rurality-urbanism criticized earlier in the paper. A supplemental analysis (available upon request) reveals that the population of a place—expressed either as a continuous logged variable or a series of size categories—is strongly related to diversity as Wirth would have expected: the larger the place, the higher its diversity level and more complex its racial-ethnic structure. Moreover, the positive and significant effect of size persists when separate multivariate models are estimated for places within each of the RUC contexts.

This last finding suggests that something about the demographic scale of a place influences ethnoracial diversity irrespective of the context in which the place is embedded. Perhaps the ‘critical mass’ principle proposed by Fischer (1976) operates well below typically urban thresholds. That is, for smaller communities an incremental increase in population size may be sufficient to boost the likelihood that groups will coalesce around a shared attribute or interest like race-ethnicity, form supportive networks and organizations, and ultimately attract additional members from elsewhere. However, such a possibility should not preclude further consideration of the larger settings that surround places. The USDA’s RUC codes, for example, could be modified in fruitful ways (Winkler and Johnson 2016) or other classification schemes tried (Champion and Hugo 2004; Brown and Cromartie 2004). Ideally, future investigators will move beyond dummy-variable representations of context, determining whether the detailed social and economic aspects of counties or areas shape the diversity of places located within them.

Pending these steps, the research reported here undermines long-held assumptions about urban heterogeneity and rural homogeneity. We regard the broadly similar diversity patterns among places in most types of contexts as more evidence for the blurring of traditional metro-nonmetro boundaries. An important caveat to this interpretation is that similar diversity levels and structures may still have quite different consequences. Because of their extensive histories as destinations for immigrants and minorities, many metropolitan communities are accustomed to dealing with the educational, healthcare, housing, and other needs of a diverse population. Yet the recency and pace of diversification experienced by some nonmetro places poses challenges in virtually every institutional domain (Carr et al. 2014; Grey and Woodrick 2005; Kandel and Parrado 2006). What remains to be seen is whether the potential benefits of diversity—especially the demographic, economic, and cultural revitalization that ethnic newcomers can bring—outweigh the perceived costs to cohesion and quality of life.