Suburbanizing Segregation? Changes in Racial/Ethnic Diversity and the Geographic Distribution of Metropolitan School Segregation, 2002–2012
by Kori J. Stroub & Meredith P. Richards - 2017
Background: While postwar suburban migration established suburbs as relatively affluent, homogeneous white enclaves distinct from the urban core, recent waves of suburbanization and exurbanization have been spurred largely by rapid growth in the nonwhite population. While these increases in suburban racial/ethnic diversity represent a significant evolution of the traditional “chocolate city, vanilla suburbs” dichotomy, scholars have expressed concern that they are worsening racial/ethnic segregation among suburban public school students.
Objective: In this study, we document shifts in the racial imbalance of suburban schools in terms of several racial/ethnic and geographic dimensions (i.e., multiracial, black–white; between and within suburban districts, among localities). In addition, we extend the urban/suburban dichotomy to provide initial evidence on changes in racial balance in metropolitan exurbs. Finally, we use inferential models to directly examine the impact of changes in racial/ethnic diversity on shifts in racial imbalance.
Research Design: Using demographic data from the National Center of Education Statistics Common Core of Data (NCES CCD) on 209 U.S. metropolitan areas, we provide a descriptive analysis of changes in segregation within and between urban, suburban, and exurban localities from 2002 to 2012. We measure segregation using Theil’s entropy index, which quantifies racial balance across geographic units. We assess the relationship between demographic change and change in segregation via a series of longitudinal fixed-effects models.
Results: Longitudinal analyses indicate that increases in racial/ethnic diversity are positively related to change in racial imbalance. However, observed increases in diversity were generally insufficient to produce meaningful increases in segregation. As a result, suburbs and exurbs, like urban areas, experienced little change in segregation, although trends were generally in a negative direction and more localities experienced meaningful declines in segregation than meaningful increases. Findings are less encouraging for suburbs and exurbs than for urban areas and underscore the intractability of black-white racial imbalance and the emerging spatial imbalance of Asians and whites. We also document an important shift in the geographic distribution of segregation, with suburbs now accounting for a plurality of metropolitan segregation.
Conclusions: Contrary to previous researchers, we do not find evidence that suburban and exurban schools are resegregating, although we fail to document meaningful progress towards racial equity. Moreover, while suburbs are not necessarily resegregating, we find that segregation is suburbanizing, and now accounts for the largest share of segregation of any locality. We conclude with a discussion of recommendations for policy and research.
Over the past decades, the American metropolis has experienced an era of rapid suburbanization. In 1960, less than a third of Americans lived in suburban neighborhoods; by 2010, a slight majority of Americans were suburbanites (Littman, 1998; Mather, Pollard, & Jacobsen, 2011). This population growth in suburbs has been accompanied by geographic expansion and differentiation: In addition to the more established inner-ring suburbs adjacent to urban cores, the continued outward development of metropolitan areas has incorporated outlying towns and rural areas into fast-growing exurbs or outer-ring suburbs on the metropolitan periphery (Berube, Singer, Wilson, & Frey, 2006). While the early wave of postwar suburban migration established suburbs as relatively affluent, homogeneous white enclaves distinct from the urban core, recent waves of suburbanization and exurbanization have been spurred largely by rapid growth in the nonwhite population (Mather et al., 2011). As a result, the racial/ethnic composition of suburban and exurban metropolitan neighborhoods has transformed dramatically. Indeed, the proportion of suburban residents who are nonwhite doubled between 1990 and 2010, from 19% to 39% (Author calculations; U.S. Census, 2010).
Not surprisingly, these increases in suburban and exurban diversity have had profound implications for public schools, which have experienced increases in nonwhite student enrollment mirroring population trends more generally. Indeed, Fry (2009) finds that suburban public school diversity has increased by over 30% since the early 1990s. As Fry notes, the increasing share of nonwhites and corresponding growth in racial/ethnic diversity in suburban schools may be attributed largely to accelerated growth in the enrollment of suburban Hispanic students, which increased by 82% over the past decade alone. The proportion of black and Asian students in suburbs has also increased, albeit at a slower rate than that of Hispanics (20% and 25%, respectively).
While these increases in suburban racial/ethnic diversity represent a significant evolution of the traditional chocolate city, vanilla suburbs dichotomy (Farley, Schuman, Bianchi, Colasanto, & Hatchett, 1978), they have also prompted concern regarding the adverse consequences of this rapid racial/ethnic transformation of suburban schools. Some have argued that these dramatic demographic shifts are precipitating racial/ethnic segregation among suburban public school students (Frankenberg, 2012; Frankenberg & Orfield, 2012). However, most extant evidence on suburban segregation has focused primarily on the typical level of student exposure to students of similar or different racial/ethnic groups and may not capture other patterns of racial segregation in schools, such as racial imbalance across schools. Moreover, while the prior literature has provided some insight into overall trends in the level of segregation in suburban schools, few studies have directly examined how changes in racial/ethnic diversity are related to changes in school segregation (Reardon & Yun, 2001). In addition, while scholars have distinguished among different types of suburbs (Frankenberg & Orfield, 2012), prior empirical research has failed to distinguish between more traditional inner-ring suburbs and newer, fast-growing exurbs.
In this study, we contribute to the emerging literature on the changing metropolitan dynamics of public school segregation, focusing on shifts in suburban and exurban areas. We assess change in suburban, urban, and exurban segregation across a dimension of segregation that has received less empirical attention in the extant literature: unevenness, which quantifies the degree of racial imbalance across schools in districts and across districts in a metropolitan area. As we discuss in greater detail below, it is important to examine change in segregation across multiple dimensions because: 1) not only can the different dimensions yield seemingly contradictory results regarding the severity of segregation in a given context, but 2) they also emphasize different causal mechanisms through which segregation impacts student outcomes and have implications for the types of strategies that might effectively reduce segregation. In addition to documenting changes in segregation over time, we extend descriptive work on segregation trends by employing inferential models to estimate the nature of the relationship between changes in metropolitan demographics and changes in segregation.
Below, we provide a brief discussion of different conceptualizations of segregation and how they are measured, emphasizing the unique contributions of each. We then provide a review of the extant literature on recent trends in suburban school segregation as they relate to these dimensions, and then identify areas where the present study contributes to our existing knowledge.
DIMENSIONS OF SEGREGATION
In their seminal discussion of the measurement of segregation, Massey and Denton (1988) identify five primary dimensions of segregation: 1) exposure/isolation, 2) unevenness, 3) concentration, 4) centralization, and 5) clustering. While an exhaustive review of these indices is beyond the purview of this paper, it is important to note that school racial/ethnic segregationdefined broadly as the degree to which two or more racial/ethnic groups attend school separately from one anothermay be manifest and measured in a number of specific ways. Below we discuss the two dimensions of segregation most commonly studied by researchers: exposure/isolation and unevenness.
The most commonly used measures of segregation, indices of exposure/isolation capture the average racial composition of schools which students of different racial/ethnic groups attend. Specifically, exposure refers to the degree of potential contact between students of different racial/ethnic groups, while isolation refers to the extent to which students of the same racial/ethnic group are exposed only to each other. Because exposure and isolation indices capture the average racial/ethnic composition of the schools attended by students of different races/ethnicities, they are also typically interpreted as measures of the subjective experience of segregation for the average student of a given racial/ethnic group (Massey & Denton, 1988).
Whereas indices of exposure/isolation capture the potential for interaction between students of different racial/ethnic groups, indices of unevenness measure the degree of racial imbalance in a district (or other geographic area) by quantifying the extent to which different racial/ethnic groups are evenly distributed across schools. (We use the terms unevenness and racial imbalance interchangeably throughout this paper.) In terms of unevenness, a district is said to be segregated if students from a given racial/ethnic group are highly asymmetrically distributed or imbalanced across schools (Massey & Denton, 1988). As with exposure and isolation, unevenness can be measured in a number of ways, including indices of dissimilarity and so-called entropy-based measures.
Measures of exposure/isolation and unevenness are generally positively correlated (Massey & Denton, 1988); however, they are conceptually and mathematically distinct, each making unique contributions to our understanding of segregation in a given context. This stems largely from the fact that indices of exposure and isolation are more sensitive to the relative size of racial/ethnic subgroups than measures of unevenness. In contrast, measures of unevenness depend far less on the relative sizes of the groups being studied and focus on the distribution across schools. Indeed, it is possible for both a predominantly black and a predominantly white district to be highly racially balanced, as long as students are evenly distributed by race/ethnicity across schools.
The distinctions between measures of exposure/isolation and unevenness have important implications for research and practice. First, when assessing trends in segregation, changes in the racial/ethnic composition of a district over time may lead to decreases in exposure (or increases in isolation) even if students of different racial/ethnic groups do not become more unevenly distributed across schools (See also Reardon & Owens, 2014; Richards & Stroub, 2014; Stroub & Richards, 2013). Thus, in a longitudinal context, changes in measures of exposure/isolation may reflect changes in the racial/ethnic composition or diversity of the context, whereas measures of unevenness tend to reflect changes in how students of different racial/ethnic groups are distributed across schools or districts, independent of changes in racial composition. To the extent that researchers are interested in isolating changes in the distribution of students over time, net of changes in diversity, researchers may find measures of unevenness relatively useful.
Second, the two dimensions emphasize different causal pathways by which segregation may affect student outcomes. As Reardon and Owens (2014) have recently argued, to the extent that the effects of segregation on student outcomes are compositional in nature (e.g., peer effects) or impact student through mechanisms correlated with school composition, measures of exposure and isolation may be most appropriate. For example, Frankenberg and Orfield (2012) identify several suburban districts facing significant challenges in meeting the needs of their rapidly diversifying student populations. By contrast, to the extent that segregation impacts students through contextual mechanisms (e.g., unequal exposure to high quality school environments), measures of unevenness might be more appropriate. Of course, these two broad causal mechanisms are not mutually exclusive, and segregation may influence student outcomes in a number of ways simultaneously. Given the potential salience of both causal pathways in affecting student outcomes, as Reardon and Owens have observed, it is important to understand and document changes in both dimensions.
Finally, the different ideals against which measures of exposure/isolation and unevenness assess segregation have significant implications for the practice of desegregating schools. Because districts have little control over the racial composition of the students they serve, the courts have generally employed the relative criterion of unevenness in assessing whether districts have desegregated to the maximum extent practicable given the racial composition of their district. Indeed, in the wake of the Green and Swann decisions (Green v. County School Board of New Kent, 391 U.S. 430, 1968; Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1, 1971), the courts generally employed measures of racial imbalance or unevenness to determine whether a district had achieved integration objectives (Armor, 1995; Raffel, 1998; Rossell, 1990). It is important to note that the courts emphasis on unevenness has generally been conceptual, rather than mathematicalinstead of specifying an acceptable value of unevenness, the courts generally establish a maximum deviation by which each school was permitted to deviate from the district average composition, often 10 to 15 percentage points, depending on the context (Armor, 1995).
Taken together, these significant differences underscore the importance of understanding multiple dimensions to elucidate the often complex patterns of school segregation in suburban schools and the differential utility of these measures across contexts. Moreover, they underscore the importance of ensuring that researchers are clear and consistent in reporting which dimension(s) of segregation they employ in their work. Below we review the extant literature related to trends in suburban segregation in terms of both of these major types of measure, outlining our current knowledge of trends in exposure/isolation and highlighting the gaps in our understanding of changes in unevenness or racial balance.
EVIDENCE ON SUBURBAN SCHOOL SEGREGATION
Researchers have fairly comprehensively documented changes in exposure and isolation in suburban districts over the past decades. However, we know relatively little about how suburban schools have changed in terms of racial balance and unevenness. Moreover, while extant research has provided a descriptive account of changes in segregation, only one study has directly examined the relationship between changes in diversity and changes in segregation in schools (Reardon & Yun, 2001). Below, we outline existing knowledge related to changes in both exposure/isolation and unevenness and then turn to the link between changes in diversity and segregation.
A handful of studies have examined changes in the exposure and isolation of suburban public school students. Taken together, such studies generally find troubling evidence that nonwhite students, particularly black students, have lower levels of exposure to whites than in previous decades. Conversely, however, white students have higher levels of exposure to nonwhite students than in previous decades, as schools shift from a more traditional blackwhite dichotomy to a more diverse milieu encompassing larger numbers of racial minorities. Moreover, isolation is stable or declining for racial subgroups with the exception of with the exception of Hispanic students.
Using categorical variants of exposure/isolation measures, Frankenberg (2012) analyzes data on the suburbs of the 25 largest U.S. metropolitan areas, finding that the proportion of students enrolled in suburban schools that were 90% to 100% nonwhite increased from 8% in 2000 to 13% in 2007, suggesting a decline in nonwhite exposure to whites. Frankenberg also finds that the percentage of students enrolled in suburban schools that were 90% to 100% white decreased from 27% to 16% over the same period, suggesting a decline in white isolation and an increase in white exposure to nonwhites.
Using more traditional indices similar to those outlined above, Fry (2009) detects a similar pattern of decreasing nonwhite exposure to whites and increasing white exposure to nonwhites from 1994 to 2007, examining the entire population of U.S. public suburban and exurban districts. Specifically, Fry finds that the average suburban Hispanic or black student attended a school that was 9 percentage points less white in 2007 than 1994, while the average Asian student attended a school that was 7 percentage points less white. This is perhaps not surprising given the rapid increases in suburban diversity over the past decades. Indeed, in interpreting this result, it is helpful to note that nationally, the proportion of suburban public school students who are white declined by 13 percentage points, from 72% to 59%, between 2007 and 1994 (Fry, 2009).
Concurrent with these declines in nonwhite exposure to whites, Fry finds that racial/ethnic isolation decreased for most racial groups over the same period. The average suburban white student in 2007 attended a school that was 8 percentage points less white than in 1994. Fry found little change in the isolation of suburban black or Asian students. The average black student in 2007 attended a school that was 1 percentage point less black than in 1994, while the average Asian student attended a school that was 1 percentage point less Asian. Conversely, the typical Hispanic student attended a school that was 7 percentage points more Hispanic in 2007 than in 1994. Again, in interpreting this finding, it is helpful to note that nationally, the proportion of suburban public school students who are Hispanic increased by 9 percentage points, from 11% to 20%, between 1994 and 2007 (Fry, 2009). Taken together, therefore, Frys findings suggest that racial/ethnic isolation is stable or declining for all racial groups with the exception of the rapidly-growing Hispanic student population.
While recent evidence has examined changes in exposure/isolation in suburban and exurban districts, we know less about changes in the racial balance or unevenness of students across suburbs. Nationally, available data indicates that students have become slightly more evenly distributed across metropolitan areas over the past decades (Fiel, 2013; Stroub & Richards, 2013). However, it is possible that these aggregate declines in unevenness mask important divergences in the segregation trajectories of urban, suburban, and exurban schools within American metropolises. Towards that end, a relatively small body of research has examined trends in the unevenness with which suburban students are distributed by race and ethnicity, although the majority of evidence documents trends from over two decades ago.
Using several measures of unevenness, Reardon and Yun (2001) document trends in overall suburban racial balance between 1987 and 1995. They find that Hispanicwhite and Asianwhite unevenness increased slightly, suggesting that these groups became more asymmetrically distributed across suburbs over the eight-year period. Blackwhite unevenness, however, declined slightly over this period, indicating that blacks and whites became more evenly distributed across suburbs. Importantly, however, Reardon and Yun only report levels of overall suburban segregation, which combine the segregation between suburban districts and within suburban districts (i.e., between schools), and do not afford insight into trends in segregation between and within suburban districts.
In another study on a similar time span, Reardon, Yun, and Eitle (2000) examine how total metropolitan segregation (as measured by Theils  index of unevenness, discussed at length below) is distributed across and within suburban and urban areas. They find that suburban unevenness accounted for an increasing share of total metropolitan segregation, while the proportion of metropolitan segregation attributable to urban segregation declined. In addition, they find that the majority of suburban segregation is attributable to the uneven distribution of students across district boundaries rather than between schools. Reardon et al. only report changes in the components of metropolitan segregation; thus, it is possible that the absolute level of suburban racial imbalance was declining even as the suburban share of total racial imbalance was increasing.
In his study of racial/ethnic segregation trends from 1994 to 2007, Fry (2009) also assesses changes in unevenness across suburban (as well as urban and rural) districts. Employing a dissimilarity index, Fry finds that suburban Hispanics became slightly more segregated from non-Hispanics between 1994 and 2007 (from 0.28 to 0.30). However, blacks became slightly less segregated from nonblacks over the study period (from 0.37 to 0.35). Likewise, Asians became slightly less segregated from non-Asians (from 0.26 to 0.26). As such, Frys findings regarding unevenness generally mirror his findings regarding the isolation of each racial/ethnic group over the same period.
Although Frys analysis provides a more recent examination of trends in suburban unevenness, his analysis, like that of Reardon and Yun (2001), is geographically limited, examining segregation within suburban localities, and fails to examine issues of segregation among suburban localities; segregation among suburbs, cities, and exurbs; or the contribution of suburban segregation to overall metropolitan segregation. In addition, Fry only captures segregation between each focal racial group and other racial groups (e.g., between blacks and nonblacks), and fail to capture other important racial/ethnic dimensions of historical and policy importance (e.g., blackwhite, Hispanicwhite, multiracial, etc.). Moreover, he excludes districts with fewer than 1,000 students in each focal racial/ethnic group, thus excluding less populous and less racially diverse districts. Thus, Frys study provides an initial but incomplete picture of change in the racial and geographic dynamics of suburban racial imbalance within American metropolises.
CHANGES IN DIVERSITY AND SEGREGATION
Although rapid increases in racial/ethnic diversity in the suburbs and, to a lesser extent, the exurbs, have been well documented (e.g., Frankenberg & Orfield, 2012; Frey, 2011; Fry, 2009; Mather et al., 2011), we know surprisingly little about the direct link between demographic shifts and the severity of racial/ethnic segregation of these metropolitan localities. This is a particularly important omission, because it is possible that racial/ethnic diversity may be increasing over time while segregation is declining, even if racial/ethnic diversity is positively related to change in segregation, owing to the magnitude and share of the population experiencing such changes.
A notable exception is the analysis of Reardon and Yun (2001), who examine the impact of changes in racial composition on changes in suburban unevenness between 1987 and 1995. Reardon and Yun find that changes in nonwhite enrollment are associated with changes in the level of unevenness in suburbs. Moreover, they find that increases in all three major nonwhite groups are associated with worsening racial balance. Specifically, they find that increases in the proportion of blacks in a suburb are generally associated with increases in blackwhite unevenness, increases in the proportion of Hispanics are associated with increases in Hispanicwhite unevenness, and increases in the proportion of Asians are associated with increases in Asianwhite unevenness.
Importantly, however, Reardon and Yun (2001) find that the positive relationship between diversity and unevenness is weaker in areas with higher shares of nonwhite enrollment. For example, growth in the black population results in smaller increments in blackwhite unevenness in areas with lower concentrations of blacks than those with higher concentrations of blacks. This finding suggests that the positive relationship between growth in the share of nonwhites and the level of segregation in suburban areas tapers off as the proportion of suburban nonwhite enrollment continues to increase. Given that the share of suburban students who are nonwhite has continued to increase in the intervening decades since Reardon and Yuns study, their findings underscore the importance of reexamining the relationship between diversity and segregation in the emerging suburban context.
As the foregoing demonstrates, despite recent interest in the issue of suburban segregation, our extant knowledge of the dynamics of racial/ethnic balance across suburbs and other metropolitan localities is limited in a number of ways. First, nearly all of our current knowledge on suburban segregation is drawn from measures of exposure and isolation, thus neglecting the potentially important dimension of unevenness (Fry, 2009; Reardon & Yun, 2001). Second, although the unevenness of suburban segregation can be measured in different wayswithin suburban districts, between suburban districts, between suburbs and other localitiesthe existing literature has failed to capture these geographic distinctions (Fry, 2009; Reardon & Yun, 2001). This omission is particularly problematic given that most suburban segregation tends to lie across district boundaries (Reardon et al., 2000).
Third, the prior literature on suburban segregation is limited by its reliance on an increasingly obsolete urban/suburban dichotomy. As discussed above, however, the past decades have witnessed the rapid emergence of exurbs on metropolitan fringes (Berube et al., 2006). Although these exurbs have also experienced significant increases in diversity over the past decade, they remain overwhelmingly white. Moreover, while whites accounted for only 8% of population growth nationally over the past decade, they accounted for 73% of growth in exurbs (Frey, 2011). Despite the increasing importance of exurbs, most prior research focuses exclusively on suburbs and urban areas (e.g., Frankenberg, 2012; Reardon & Yun, 2001; Reardon et al., 2000).
Fourth, prior research has focused on a limited set of dual-group racial/ethnic comparisons. For example, Fry (2009) documents changes in unevenness between blacks and nonblacks, Hispanics and non-Hispanics, and Asians and non-Asians. These dual-group measures neglect several other potentially salient racial/ethnic comparisons, particularly multigroup measures. This is particularly important given that prior research has found that segregation among nonwhite students accounts for an increasing proportion of total school segregation (Stroub & Richards, 2013).
Finally, we know little about the direct relationship between changes in diversity and changes in racial balance. Reardon and Yun (2001) find that growth in the concentration of nonwhitesup to a pointcontributes to increased racial/ethnic imbalance in suburban schools. However, given the significant growth in nonwhite enrollment in suburban localities in the past decadesindeed, many suburbs are now majority-minorityit is unclear whether the positive link between changes in diversity and changes in segregation holds decades later. Indeed, given the curvilinear relationship between diversity and segregation documented by Reardon and Yun, it is plausible that continued increases in diversity may be associated with declines in segregation. In addition, because Reardon and Yuns analysis was limited to suburban schools, it is unclear whether their findings generalize to more diverse urban schools and newer, more homogeneous exurbs.
In this study, we contribute to the emerging body of scholarship on suburban segregation, addressing each of the above limitations of the prior literature. Extending work documenting trends in the racial exposure and isolation of students, we use data from the National Center for Education Statistics Common Core of Data (NCES CCD) to document recent trends in the racial imbalance of suburban schools. In documenting these trends, we seek to provide nuance on the geographic and racial dynamics of changes in the distribution of students. Because racial imbalance may occur at various geographic levels, we examine change along multiple geographic dimensions (i.e., between and within suburban districts, among localities). In addition, we extend the urban/suburban dichotomy to provide initial evidence on changes in racial balance in metropolitan exurbs. Moreover, we attempt to capture the racial/ethnic complexity of racial imbalance by employing several multigroup and dual-group measures of racial/ethnic unevenness. Finally, we use multilevel, longitudinal inferential models to directly explore the relationship between changes in racial/ethnic diversity and racial imbalance in suburban, exurban, and urban localities.
DATA AND SAMPLE
To assess change in the racial balance across and within suburban and exurban areas over time, we computed annual values of unevenness using public school demographic data from the NCES CCD Public Elementary/Secondary School Universe Survey. We used the most recent decade of data available from the NCES CCD, spanning the 200102 to 201112 school years. Previous research found that national trends in metropolitan unevenness increased until the 200102 school year and declined linearly thereafter (Stroub & Richards, 2013). Focusing on this period, during which the racial imbalance of public schools declined overall, allowed us to examine whether overall declines masked disparate patterns for suburban and exurban areas.
Because this study was concerned with changes in segregation in suburban and exurban districts vis-à-vis those in cities, we classified each school district by its NCES CCD locality. These locality definitions, developed by the U.S. Census Bureau at the behest of the NCES (NCES, 2006), assign public schools and districts one of four broad geographic localities: city, suburb, town, and rural. Schools located inside the principal city of a Census-defined urbanized area (i.e., area with 50,000 or more people) are classified as cities, while those outside the principal city of an urbanized area are classified as suburbs. Schools are classified as towns if they are located in a Census-defined urban cluster (i.e., area with 25,000 to 50,000 people), while those in rural territories (i.e., areas with fewer than 25,000 people) are classified as rural. School districts are then assigned locale codes based on the most common locality code of its schools. To align the NCES CCD categories with the terminology used in the segregation literature, in this study we used the term urban to identify districts located in the principal cities of urbanized areas. In addition, we collapsed all districts within metropolitan areas classified as towns and rural into a single exurban locale code.
Although district locality codes may change over time as metropolitan populations shift in accordance with the system above, we fixed the locality of each school district based on its status in 2012. This practice is relatively standard in longitudinal studies of segregation (e.g., Logan, Oakley, & Stowell, 2008; Reardon et al., 2000; Stroub & Richards, 2013), as it is necessary to ensure that observed changes in segregation in urban, suburban, and exurban districts are not conflated with changes in how districts are classified as urban, suburban, and exurban. Thus, if a school districts locality designation changes from suburban to urban, it may have a large impact on segregation independent of any actual change in the distribution of students across localities. By fixing localities, we ensured that our findings reflected changes in the distribution of students and not Census designations.
Several filters were applied to the population of schools and districts in the NCES CCD to arrive at the final analytic sample. First, we restricted our sample to schools and districts located in Census-defined metropolitan statistical areas in the 50 U.S. states and the District of Columbia, excluding those in Puerto Rico and other U.S. territories. Following the procedure employed by Reardon et al. (2000), we retained for analysis only those metropolitan areas for which over 90% of schools reported racial/ethnic data for both the first and last year of the study. Because Tennessee failed to report student racial/ethnic data for the six years between 1999 and 2004, we excluded Tennessees nine metropolitan areas for the entirety of the study. Finally, because we were concerned with the dynamics of segregation across urban, suburban, and exurban portions of metropolitan areas, we restricted our analyses to metropolitan areas containing school districts classified as urban, suburban, and exurban, thereby excluding a number of less populous metropolitan areas. As with locality designations, we fixed metropolitan boundaries at their 2012 values to ensure that we isolated change in school segregation from changes in metropolitan boundaries over time (e.g., from the incorporation of outlying areas or the fragmentation into multiple metropolitan areas).
At the school level, we followed the procedure of Logan (2004) and Stroub and Richards (2013), restricting the sample of schools in each metropolitan area to elementary schools. By calculating segregation values for schools of a single level, we could distinguish actual segregation from cohort effects. (For example, if a school district has only one elementary and one secondary school, it may have a high segregation value; however, the difference between these two schools is better conceptualized as a cohort effect reflecting the different demographic characteristics of older and younger students.) We also excluded nonoperational schools and schools with zero enrollments. Public charter schools were included in the analysis: school districts comprised of only charter schools were analyzed as separate districts within a metropolitan area; charter schools operated by traditional public school districts were counted among their public schools. Magnet and alternative schools were retained for analysis, while Department of Defense and Bureau of Indian Affairs/Tribal schools were excluded. After application of all exclusion criteria, the final analytic sample for 2012 contained 5,362 school districts in 209 metropolitan areas, accounting for more than 14 million elementary school students, or three-fifths of all elementary students enrolled in U.S. public schools.
Table 1 reports characteristics of the sample of urban, suburban, and exurban districts in the first and last year of the study period. The table reveals three important trends. First, although urban districts remained substantially more racially diverse than suburban and exurban districts, suburbs and exurbs experienced more rapid increases in racial/ethnic diversity than urban districts over the past decade. Second, nearly all of the growth in suburbs and exurbs might be attributed to increases in the proportion of Hispanics and, to a lesser extent, Asians, in the context of stable proportions of blacks and declining proportions of whites. Finally, although exurban areas experienced substantial increases in enrollment over the study period, the enrollment of schools in traditional suburbs was relatively stable, and urban areas experienced substantial declines in student enrollment.
We measure the unevenness with which students are distributed across schools, districts and localities via Theils (1972) entropy index. Mathematically, the Theil index quantifies how racially/ethnically diverse schools are, on average, relative to the overall racial/ethnic diversity of their metropolitan area (Reardon et al., 2000). The Theil index varies from 0 to 1, where 0 means that all schools have the same composition as the metro area (i.e., perfect integration) and 1 means that all schools only contain students of a single, unique racial/ethnic group (i.e., perfect segregation) (Iceland, 2004). Readers may refer to Reardon and Firebaugh (2002) for more detailed information regarding the computation of H.
One advantage of the Theil index is that it may be used to quantify the segregation among any number of racial/ethnic groups, as well as between specific pairs of groups (e.g., C. S. Fischer, Stockmayer, Stiles, & Hout, 2004; M. J. Fischer, 2003; Reardon et al., 2008; Reardon et al., 2009; Reardon et al., 2000). We computed segregation along five different racial/ethnic dimensions. First, we computed measures of multiracial segregation encompassing all five racial/ethnic categories tracked by the NCES CCD (i.e., American Indian/Alaska Native, Asian, black, Hispanic, and white). We also computed measures of segregation among nonwhite students. In addition, we computed three dual-group indices capturing the segregation between whites and blacks, whites and Hispanics, and whites and Asians.
Another advantage of the Theil index is its property of geographic decomposability. In the context of this study, this allowed us to examine the contributions of segregation within and between school districts and localities as well as the contributions of urban, suburban, and exurban segregation to total metropolitan segregation. For each of the racial/ethnic dimensions of segregation identified above, we computed segregation at several geographic levels. First, we computed segregation within each urban, suburban, and exurban district. Second, for each metropolitan area, we computed a measure of segregation between urban districts, between suburban districts, and between exurban districts. We then computed measures of total urban, suburban, and exurban segregation as the sum of each localitys between-district component and a population-weighted composite of the within-district segregation for each of its districts. Third, we computed a measure of total metropolitan segregation as a population-weighted composite of its urban, suburban, and exurban segregation, as well as segregation among these localities. Readers may refer to Reardon et al. (2000) for mathematical proofs and a detailed discussion of the geographic decomposition of Theil.
We report population-weighted averages for each measure of segregation. Thus, all measures of segregation were weighted by the proportion of the total population residing in each district or metropolitan area (for a similar weighting procedure, see Reardon & Yun, 2001; Stroub & Richards, 2013). Because of variability in the student population of school districts across metropolitan areas, weighted measures provide better estimates of the level of school segregation to which a typical student is exposed.
ESTIMATING THE RELATIONSHIP BETWEEN CHANGE IN RACIAL/ETHNIC COMPOSITION AND SEGREGATION
To address the relationship between changes in the nonwhite enrollment share of metropolitan localities and changes in segregation, we estimated five sets of longitudinal fixed-effects modelsone each for multiracial, Asian/white, black/white, Hispanic/white, and among nonwhite segregationfor urban, suburban, and exurban metropolitan localities separately. The models were estimated using all 11 years of data from 200102 to 201112.
As has been previously noted by Reardon and Yun (2001), the primary advantage of this type of model is that it controls for any between-metropolitan area differences in segregation, permitting the unbiased estimation of the relationship between within-metropolitan area predictors and trends in segregation for urban, suburban, and exurban localities. The general form of each model is given by the following equation:
where is the estimated level of total segregation at time t, of locality i, in metropolitan area j, and is the metropolitan area specific intercept for metropolitan area j. To account for any non-time varying differences between the localities of metropolitan areas, we include a metropolitan area specific fixed effect, denoted by . We also include a series of covariates that control for school year (), and the natural logarithms of the total enrollment and number of districts at time t, in locality i, in metropolitan area j, respectively.
As the primary predictors of interests, we include , which represents the enrollment share of the focal racial/ethnic group of interest at time t, in locality i, in metropolitan area j, and the mean-centered square of this term. Since was centered prior to squaring it, can be interpreted as the effect of changes in focal racial/ethnic groups enrollment shares on changes in total segregation in localities with average levels of focal group enrollment. Moreover, if is not equal to zero, the magnitude of the effect of changes in focal group enrollment on segregation depends on the proportion of the focal groups enrollment within a locality.
For each metropolitan locality type we estimated five different models, predicting change in segregation along each of the five racial/ethnic dimension of segregation examined in this study. The focal nonwhite racial/ethnic group for each model depended on the dimension of segregation being estimated. For models predicting multiracial and among nonwhite segregation, the focal racial/ethnic group was the proportion of nonwhite students. For models predicting Asianwhite, blackwhite, and Hispanicwhite segregation, the focal racial/ethnic groups were the proportion of Asian, black, and Hispanic students, respectively.
In the sections that follow, we present results of descriptive analyses independently examining the change in total suburban, exurban, and urban racial imbalance/unevenness between 2002 and 2012 for the five racial/ethnic dimensions of interest (i.e., multiracial, white/nonwhite, blackwhite, Hispanicwhite, Asianwhite). We geographically decompose the changes in total racial imbalance in each locality, examining the independent changes in unevenness within and between suburban, exurban, and urban districts. We then turn to the larger metropolitan context of segregation, evaluating the extent to which suburban, exurban, and urban areas are segregated from each other as well as their unique contributions to the overall level of racial imbalance of metropolitan public schools. Because descriptive analyses of aggregate trends at the national level may mask important relationships between changes in racial/ethnic diversity and racial imbalance, we then present results of our inferential models exploring these relationships in depth.
CHANGE IN RACIAL IMBALANCE BY LOCALITY
Table 2 presents changes in the overall racial imbalance of suburban areas between 2002 and 2012. Overall, Table 2 demonstrates that a typical student living in a suburban area experienced a lower or equivalent level of racial imbalance in 2012 than in 2002 on four of the five racial/ethnic dimensions of unevenness. Indeed, suburban areas experienced slight declines in multiracial and nonwhite racial imbalance (6.1% and 4.9%, respectively). Levels of overall suburban blackwhite and Hispanicwhite racial imbalance experienced negligible change over the study period (0.6% and -0.3%, respectively). Only Asianwhite segregation increased in suburban areas (13.8%), although Asians remain the least segregated racial/ethnic group from suburban whites.
Reardon and Yun (2001) argue that an increase in Theil of greater than 0.05 is substantively meaningful, noting that a change in Theil of 0.05 is roughly equivalent to an increase in dissimilarity of 0.10. The latter may be more readily interpreted in terms of the proportion of students that would need to transfer schools to achieve perfect integration (i.e., 10%). Applying this heuristic to the changes documented above, it is important to note that none of the observed changes in suburban racial imbalance, on average, met the 0.05 threshold to be substantively meaningful.
Across four of the five racial/ethnic dimensions of segregation, fewer than one-tenth of suburban localities experienced increases in segregation large enough to be deemed meaningful according to the 0.05 criterion. Moreover, on four of the five racial/ethnic dimensions, suburbs were more likely to experience meaningful declines in racial imbalance than meaningful increases. In terms of multiracial and nonwhite racial imbalance, roughly 5 times as many suburban localities experienced meaningful decreases in segregation than experienced meaningful increases (Multiracial: 12.0% vs. 2.4%; Nonwhite: 23.6% vs. 4.3%). Nearly twice as many suburbs experienced meaningful decreases in Hispanicwhite and Asianwhite racial imbalance as experienced meaningful increases (Hispanicwhite: 15.4% vs. 7.2%; Asianwhite: 8.7% vs. 4.8%). Only in terms of blackwhite racial imbalance did slightly more suburbs experience increases in segregation than decreases (13% vs. 12%).
Table 2 reveals that the small decline in total level of multiracial suburban racial imbalance has been driven largely by declining segregation among suburban districts (-9.8%), while segregation within suburban districts remained relatively stable (-0.9%). Conversely, the decline in total nonwhite suburban racial imbalance was driven largely by declining segregation within suburban districts (-11.5%), although nonwhites also became more evenly distributed across suburban districts (-9.8%). Interestingly, the relatively stable rate of blackwhite imbalance in suburban schools masks divergent trends for unevenness between and within suburban districts. While blacks and whites became less unevenly distributed across suburban districts (-1.9%), they became more unevenly distributed within districts (7.4%). A similar pattern was discerned for Hispanicwhite racial imbalance; while Hispanics and whites became less unevenly distributed across suburban districts (-1.7%), they became more unevenly distributed within districts (0.6%). Increases in total Asianwhite suburban segregation were facilitated by relatively large increases in unevenness both within and between districts (9.4% and 13.5%, respectively).
Exurban areas exhibited a similar pattern of change in racial imbalance as suburban areas (Table 3). Overall, a typical student living in an exurban area experienced a lower level of overall racial imbalance in 2012 than in 2002 on three of the five racial/ethnic dimensions of unevenness. Indeed, exurban areas experienced declines in multiracial, nonwhite, and Hispanicwhite unevenness of 6.8%, 17.3%, and 6.3%, respectively. As a result, Hispanic students, who were already the least segregated group from whites in exurban areas in terms of unevenness, became even less segregated from whites over the study period. Similar to suburban areas, exurban areas experienced a net increase of 15.4% in Asianwhite segregation. Unlike suburban areas, which experienced little change in blackwhite unevenness, exurban areas experienced an increase in blackwhite racial imbalance of 5.9%. Again, however, it is important to note that none of these average changes, positive or negative, met the 0.05 threshold to be classified as substantively meaningful according to Reardon and Yuns (2001) criterion.
As with suburban racial imbalance, few exurban localities experienced increases in segregation large enough to be deemed meaningful according to the 0.05 criterion. Moreover, on four of the five racial/ethnic dimensions, exurbs were more likely to experience meaningful declines in racial imbalance than worsening segregation. In terms of multiracial, nonwhite, and Hispanicwhite racial imbalance, more than 3 times as many exurban localities experienced meaningful decreases in segregation than meaningful increases (Multiracial: 13.0% vs. 4.3%; Nonwhite: 41.3% vs. 9.6%; Hispanicwhite: 21.6% vs. 5.3%). Slightly more suburbs experienced meaningful decreases in Asianwhite racial imbalance as experienced meaningful increases (15.4% vs. 12.0%). As with suburbs, only in terms of blackwhite racial imbalance did slightly more suburban localities experience meaningful increases in segregation than decreases (16.3% vs. 14.4%).
Table 3 reveals that the decline in total level of multiracial exurban racial imbalance has been driven by declines in segregation within exurban districts (-6.2%) as well as among exurban districts (-8.0%). Likewise, declines in total nonwhite exurban racial imbalance were driven both by declines in segregation within exurban districts (-28.8%) and between urban districts (-18.3%). Similarly, declines in total Hispanicwhite exurban racial imbalance were driven by declines in the unevenness of Hispanics and whites both within exurban districts (-5.6%) and between exurban districts (-6.8%). Increases in total blackwhite exurban segregation were facilitated by increases in unevenness both within and between exurban districts (+11.3% and +5.4%, respectively). However, the large observed increases in the total racial imbalance of Asians in exurbs were largely a function of increasing unevenness of the distribution of Asians and whites across districts (19.2%), while Asians and whites became less unevenly distributed within exurban districts (-3.0%).
Trends in racial imbalance for suburban and exurban areas were relatively comparable to those for urban areas (Table 4). A typical student living in an urban area experienced lower overall levels of racial imbalance in 2012 than in 2002 on four of the five racial/ethnic dimensions of unevenness. Overall, only urban racial imbalance between Asians and whites increased over the study period. Trends in the racial imbalance of urban and exurban areas thus diverged from urban areas in one key respect: While suburban and exurban areas experienced stable and increasing levels of blackwhite unevenness, respectively, urban areas experienced declines in blackwhite racial imbalance. Again, however, none of the changes in urban racial imbalance, positive or negative, were sufficiently large to be considered substantively meaningful according to Reardon and Yuns (2001) 0.05 criterion.
Across all five racial/ethnic dimensions of segregation, fewer than one-seventh of urban localities experienced increases in segregation large enough to be deemed meaningful according to the 0.05 criterion. Moreover, on all five racial/ethnic dimensions, urban areas were more likely to experience meaningful declines in racial imbalance than worsening segregation. In terms of multiracial and nonwhite racial imbalance, 3 times as many urban localities experienced meaningful decreases in segregation than experienced meaningful increases in segregation (Multiracial: 17.3% vs. 5.8%; Nonwhite: 23% vs. 6.7%). Nearly twice as many urban localities experienced meaningful decreases in Hispanicwhite racial imbalance as experienced meaningful increases (21.2% vs. 11.1%). Although the disparity was less pronounced for Asianwhite and blackwhite segregation, more urban localities experienced meaningful declines in these dimensions of segregation than experienced increases (Asianwhite: 20.7% vs. 13.5%; blackwhite: 16.8% vs. 13.5%).
Table 4 reveals that the slight decline in the total level of multiracial urban racial imbalance has been driven by declines in segregation within urban districts (-11.1%) as well as among urban districts (-7.6%). Likewise, declines in total blackwhite urban segregation were facilitated by decreases in unevenness both within and between urban districts (-11.3% and -5.4%, respectively). Similarly, declines in total Hispanicwhite urban racial imbalance were driven by declines in the unevenness of Hispanics and whites both within urban districts (-4.1%) and between urban districts (-12.5%). Interestingly, the overall decline in urban nonwhite racial imbalance masks disparate trends for within- and between-district measures unevenness: While nonwhites became substantially less unevenly distributed within urban districts (-15.8%), they became more unevenly distributed across urban districts (18.4). Similarly, while Asians became less unevenly distributed from whites within urban districts (-6.8%), they became more unevenly distributed across urban districts (16.3%).
Taken together, our findings reveal that suburban, exurban, and urban segregation, as measured via unevenness, declined over the past decade on the majority of racial/ethnic indicators measured in this study. None of the observed changes in the Theil index were sufficiently large to be deemed substantively meaningful, however, highlighting the stability in segregation over the past decade. To the extent that suburban and exurban localities did experience meaningful changes in segregation, they were far more likely to experience meaningful decreases than meaningful increases. However, findings may suggest areas of concern. Although the change was not substantively meaningful, the racial imbalance between Asians and whites increased in both suburban and exurban areas. Perhaps even more importantly, blacks and whites became slightly more unevenly distributed in exurban areas and within (albeit not between) suburban districts. Moreover, approximately equivalent numbers of suburbs and exurbs experienced meaningful increases in blackwhite racial imbalance as experienced meaningful decreases.
CHANGES IN THE SUBURBAN, EXURBAN, AND URBAN COMPONENTS OF METROPOLITAN RACIAL IMBALANCE
Because shifts in suburban imbalance might have occurred not only within and between suburban, exurban, and urban districts, but also between suburbs and exurbs and between suburbs and urban areas, we also assessed the changing relationship among suburban, exurban, and urban imbalance at the metropolitan level. Table 6 reports the component values and proportion of total metropolitan unevenness attributable to each geographic dimension of segregation. For the sake of parsimony, only results for the multiracial dimension of unevenness are presented here; decompositions of the other racial/ethnic dimensions yield similar results and are available upon request.
Table 5 reveals three central findings regarding the changes in the distribution of racial imbalance by locality over time. First, the past decade has witnessed convergence among localities in terms of the geographic distribution of students by race/ethnicity across and within districts. Urban segregation is generally a within-district phenomenon, with roughly four-fifths of unevenness occurring within district boundaries. Conversely, suburban and exurban segregation generally lie between districts, with roughly seven-tenths of unevenness across district boundaries. However, over the study period, the proportion of urban segregation that lay between districts increased, while the proportion of suburban and exurban segregation that lay between districts declined slightly.
Second, the table reveals an important shift in the geographic distribution of segregation across metropolitan areas. In 2002, 36% of all metropolitan racial imbalance was attributable to urban areas (between or within urban districts), while 31% was suburban. By 2012, however, just 34% of metropolitan racial imbalance was urban, while 37% was suburban. Thus, while urban areas have historically accounted for the plurality of the racial imbalance in metropolitan areas, the suburbs now account for a larger share than urban areas. Although exurbs still account for a relatively small proportion of metropolitan unevenness, owing to their small share of population and relatively even distribution of students by race/ethnicity, the proportion of metropolitan segregation attributable to exurbs increased by two percentage points over the study period.
Finally, levels of segregation across the three localities are converging. Tables 2 through 4 demonstrate that urban segregation was significantly higher than suburban segregation, which, in turn, was higher than exurban segregation. However, the past decade has witnessed a narrowing of these segregation gaps. As a result, the proportion of metropolitan segregation attributable to segregation among localities declined substantially, from 28% in 2002 to 22% in 2012.
PREDICTING THE EFFECT OF LOCALITY CHANGES ON RACIAL IMBALANCE
The results above provide a descriptive account of the changing nature of the distribution of students by race and ethnicity across metropolitan localities in the broad context of increases in the diversity of U.S. students. However, these ecological associations tell us little about the specific relationship between changes in diversity and changes in segregation at the local level. Towards that end, Table 6 presents coefficients from longitudinal models estimating change in segregation of urban, suburban, and exurban areas over time as a function of changes in the demographics of each locality. For the sake of parsimony, we present results of models for total urban, suburban, and exurban segregation, comprising the segregation between and within the urban, suburban, and exurban districts in each metropolitan area, respectively. Models for each geographic dimension of segregation are available upon request. To aid in the interpretation of the coefficients in the table, Figure 1 graphically depicts the relationship between 10-year changes in the proportion of nonwhites and 10-year changes in school segregation for each locality and racial/ethnic dimension of unevenness.
Our inferential analyses yielded five key findings. First, consistent with the findings of Reardon and Yun (2001) from the 1990s, across all three localities, growth in the proportion of nonwhites was consistently positively associated with racial imbalance on all racial/ethnic dimensions of unevenness (Figure 1). Overall, our models estimated that a locality with an average proportion of nonwhites in 2002 that experienced a 10% increase in its nonwhite population between 2002 and 2012 had a 0.02 (SD = .04) increase in racial imbalance. As we discuss below, the strength of this relationship varied considerably by locality and dimension of racial/ethnic unevenness. The only exception to this generally positive trend was the relationship between change in nonwhite enrollment and changes in nonwhite racial imbalance in exurbs and suburbs. For localities experiencing large declines in their proportion of nonwhites, the relationship between racial change and change in segregation was negative; however, the relationship was positive for localities experiencing smaller declines or increases in nonwhite enrollment (Figure 1E).
Second, although racial/ethnic diversity was positively associated with racial imbalance, a small fraction of localities experienced changes in nonwhite enrollment sufficiently large to produce meaningful increases in segregation. Indeed, across all three locality types and five racial/ethnic dimensions of segregation, fewer than 2 percent of all localities experienced increases in nonwhite enrollment of the magnitude large enough to produce increases in racial imbalance of 0.05 or more. For example, no suburban, exurban, or urban areas experienced increases in the proportion of nonwhites sufficient to produce increases in multiracial imbalance of greater than 0.05, while 0.5% of exurban areas had growth in their Asian populations large enough to produce meaningful increases in Asianwhite racial imbalance. The lone exception to this trend was blackwhite imbalance in suburbs: Nearly 13% of suburban localities experienced increases in the proportion of blacks necessary to precipitate meaningful increases in blackwhite segregation, ceteris paribus. As such, after controlling for demographic changes, our models revealed statistically significant declines on most dimensions of segregation, with the exception of blackwhite segregation, which was declining slightly in urban areas and flat in suburban and exurban areas.
Third, the relationship between demographic change and racial imbalance was generally concave, curvilinear in form. For example, an urban area that was 10% black in 2002 and experienced a 5-percentage-point increase in its share of black enrollment between 2002 and 2012 had an expected increase in racial imbalance of 0.044. However, an urban area that was 20% black in 2002 and 25% black in 2012 had a slightly smaller expected increase in racial imbalance of 0.036. This suggests that as localities become more diverse, the incremental impact of continued increases in diversity on segregation will likely continue to decline. Interestingly, Figures 1B and 1E reveal the opposite relationship for blackwhite racial imbalance in suburban areas and for nonwhite racial imbalance. For example, a suburban area that was 10% black in 2002 and experienced a 5-percentage-point increase in its share of black enrollment had an expected increase in racial imbalance of 0.013. However, a suburban area that was 20% black in 2002 and 25% black in 2012 had a larger expected increase in racial imbalance of 0.017. This suggests that, while black suburbanization over the past decade has not been large enough to produce meaningfully significant increases in blackwhite unevenness, the handful of suburban areas that continue to experience large increases in their black enrollment shares are particularly likely to experience meaningful increases in blackwhite segregation.
Fourth, we found that localities were particularly sensitive to increases in the enrollment share of black and Asian students. As examination of the figure reveals, the slope of the relationship between racial diversity and racial imbalance was particularly steep for blackwhite and Asianwhite imbalance. For example, a suburban locality with an average proportion of black students in 2002 that experienced a 5-percentage-point increase in its black population between 2002 and 2012 had an expected increase in blackwhite unevenness of 0.012. Likewise, a suburb with an average proportion of Asian students in 2002 that experienced a 5-percentage-point increase in its proportion of Asians had an expected increase in Asianwhite unevenness of 0.037. Importantly, however, while these effects were statistically significant, neither resulted in meaningfully significant increases in segregation according to the 0.05 criterion. By contrast, 5-percentage-point increases in the proportion of focal nonwhites in suburbs were associated with small declines in multiracial, nonwhite, and Hispanicwhite racial imbalance. Similar patterns were observed for urban and exurban localities.
Fifth, we found that the relationship between changes in diversity and racial imbalance varied across localities, with exurban areas exhibiting particular sensitivity to increases in diversity. As examination of the figure reveals, for four of the five racial/ethnic dimensions studied, the slope of the relationship between racial diversity and racial imbalance was steepest for exurban areas. For example, an exurban locality with an average proportion of black students in 2002 that experienced a 5-percentage-point increase in its black population between 2002 and 2012 had an expected increase in blackwhite unevenness of 0.055. By contrast, an urban and suburban locality experiencing similar growth in their black population had expected increases in blackwhite unevenness of 0.032 and 0.012, respectively.
Scholars have increasingly expressed concern that rapid demographic shifts in metropolitan suburbs are precipitating increases in racial/ethnic school segregation (Frankenberg, 2012; Frankenberg & Orfield, 2012). Despite the profusion of interest in suburban segregation, our empirical knowledge of the trends in and determinants of suburban segregation is still evolving. In this study, we provide initial evidence linking changes in diversity with changes in racial imbalance in suburban, exurban, and urban localities, attending to changes in the geographic distribution of segregation as well as changes along different racial/ethnic dimensions of segregation. Our findings, which highlight areas of concern as well as areas of progress, underscore the increasingly complex nature of the racial dynamics within increasingly diverse and geographically differentiated metropolitan areas.
Partially consistent with the concerns echoed above, we found that changes in racial/ethnic diversity were positively associated with changes in racial imbalance in suburban and exurban, as well as urban, schools. This was particularly true in the rapidly-growing but still predominantly white exurban areas on the metropolitan fringe, which were highly sensitive to increases in nonwhite enrollment. For example, a 5-percentage-point increase in the black enrollment share in an exurban area over the study decade was associated with an increase in racial imbalance of roughly 1.7 times that experienced by a comparable urban area and 4.6 times that experienced by a comparable suburban area. Even more troubling, perhaps, we found that localities are particularly sensitive to growth in black and Asian populations. Indeed, increases in the share of Asian students of 0.3 to 0.4 percentage points were sufficient to precipitate worsening racial imbalance in suburbs and exurbs, respectively. Likewise, increases in the proportion of black students of 0.7 percentage points in suburbs and decreases of 0.1 percentage points in exurbs were sufficient to produce increases in segregation.
Despite the positive association between increases in diversity and racial imbalance, however, we did not generally find that increases in racial/ethnic diversity were associated with worsening racial imbalance in suburbs or other localities. Indeed, the actual increases in diversity experienced by most localities were generally insufficient to counteract the secular trend towards declining racial imbalance. Descriptive analyses corroborated these inferential estimates: Despite rapid increases in racial/ethnic diversity in the suburbs and exurbs, we found that they were not generally experiencing worsening overall racial imbalance. Rather, a typical student residing in a suburb in 2012 generally experienced a level of overall multiracial imbalance that was 6.1% lower than that experienced by a typical suburban student in 2002. Likewise, a typical student residing in an exurb in 2012 generally experienced a level of overall multiracial imbalance that was 6.8% lower than that experienced by a typical exurban student in 2002.
In interpreting these findings, it should be emphasized that none of the declines met the 0.05 threshold to be deemed meaningful according to Reardon and Yun (2001). As such, they reflect incremental progress towards racial equity at best. In addition, as with all aggregate measures, it should be noted that these results capture the typical (i.e., population-weighted average) level of racial imbalance to which students are exposed. A nontrivial minority of localities experienced increases in imbalance over the study period; however, few were sufficiently large to be deemed meaningful. For example, just 2.4% of suburbs and 4.3% of exurbs experienced increases in multiracial imbalance of 0.05 or greater. Moreover, across most racial/ethnic dimensions, a much larger proportion of localities experienced meaningful declines in imbalance than meaningful increases.
Although our findings do not suggest widespread resegregation, they underscore three particular areas of concern. First, although blacks and whites became more evenly distributed in urban areas, this progress was not shared by exurban and, to a lesser extent, suburban areas. While blackwhite imbalance did not increase overall in suburban areas, declining blackwhite imbalance across district boundaries masked increases in blackwhite imbalance within district boundaries, highlighting the shifting nature of blackwhite segregation in suburbs. Although, on average, the increases in suburban blackwhite imbalance did not exceed the 0.05 threshold, they still present a troubling counter to the declining racial imbalance on other indicators. Moreover, a relatively high proportion of suburbs experienced meaningful increases in blackwhite imbalanceslightly more than experienced meaningful declines. In the context of the particularly pernicious historical legacy of blackwhite segregation in America, the finding that blacks and whites have become more unevenly distributed in exurbs and, to a lesser extent, suburbs is cause for concern despite continued progress in urban areas.
Although the cause of troubling trends in blackwhite racial imbalance in exurbs and suburbs is somewhat unclear, the relative intractability of the income gap between blacks and whites vis-à-vis other racial/ethnic groups may be an important contributing factor. Indeed, the median income for black households is only 58% of that of whites ($32,229 vs. $55,412), suggesting that black families residential choices are considerably more constrained than those of their white counterparts. By contrast, the median household income of Hispanics is roughly 70% of that of whites ($38,624 vs. $55,412), suggesting greater potential for mobility into neighborhoods occupied by whites (Author calculations of Census 2010 data). Although income disparities may be an exacerbating factor in blackwhite segregation, it is important to emphasize that race ipso facto is likely still an important stratifying force beyond economics. Indeed, recent research by Jargowsky (2013) modeling the relative contributions of income and race to residential segregation suggests that income disparities accounts for less than 10% of the variance in racial segregation.
Second, consistent with the results of inferential models that revealed that localities were particularly sensitive to growth in Asian populations, we found that across all localities, Asians became more unevenly distributed from whites over the study period. While on average, these increases did not reach the 0.05 threshold to be considered meaningful, they are somewhat troubling given the low initial level of Asianwhite segregation. Indeed, a 0.05-unit increase in Asianwhite racial imbalance between 2002 and 2012 would amount to a 53% increase in segregation above the initial level of 0.094. Segregation between blacks and whites and, to a lesser extent, between Hispanics and whites, which have been the traditional focus of scholars and policymakers, certainly still warrant attention. However, educators and policymakers must broaden the discourse on race in schools to address the emerging problem of segregation between whites and Asians (Orfield & Lee, 2006; Reardon & Yun, 2001; Stroub & Richards, 2013).
Again, it is not immediately apparent why the segregation between Asians and whites has worsened despite improvements for other racial/ethnic groups. However, it is possible that the changing socioeconomic profile of Asian immigrants may be relevant. While Asian immigrants tend to be more affluent than other immigrant groups, the past decades have witnessed rapid increases in the immigration of relatively poor Southeast Asians, such as Cambodian, Laotian, Vietnamese, and Hmong groups (Pew Research Center, 2013). Thus, worsening Asianwhite segregation may be attributable in part to the influx of these less advantaged groups into suburban areas with strong ethnic enclaves (Reardon & Yun, 2001). Alternately, given that Asian households generally have the highest median income of any racial/ethnic groupin 2011, the median household income for Asians was $65,129 vs. $55,412 for whitesincreases in Asianwhite segregation may be attributable to increased concentration of affluent Asian communities (Author calculations of Census 2010 data; DeNavas-Walt, Proctor, & Smith, 2012).
Finally, although suburban and exurban areas generally experienced declining levels of segregation, they were generally smaller in magnitude than those in urban areas. At the same time, the proportion of students residing in suburban and exurban areas increased as the urban share of the population declined. Together, the larger declines in urban segregation in the context of smaller declines in suburban and exurban segregation and continued population growth in suburbs and exurbs have precipitated an important shift in the geographic distribution of segregation: suburban areas, not urban areas, now account for the plurality of metropolitan segregation. Thus, although we do not consistently find that suburbs are segregating, we conclude that segregation is suburbanizing, underscoring the importance of efforts by scholars to broaden the discourse on segregation and racial/ethnic diversity from its traditionally urban geographic focus.
In addition, changes in the geography of racial imbalance across and within suburban and exurban districts also highlights the challenges to desegregating metropolitan schools. Consistent with trends documented by previous researchers (Clotfelter, 2001; Reardon et al., 2000), our results demonstrate that over three-quarters of the segregation in suburban and exurban localities in 2012 lay between district boundaries rather than within district boundaries. This means that, ceteris paribus, traditional intradistrict desegregation efforts would only be able to reduce the segregation of suburban and exurban areas by 25%, even if they were able to optimize the evenness with which students were distributed by race/ethnicity across schools. As others have noted (Reardon et al., 2000), this suggests that more traditional desegregation efforts within districts (e.g., affirmative attendance zoning, magnets, open enrollment, socioeconomic student assignment practices) are increasingly unlikely to achieve substantial declines in segregation on their own. For suburban and exurban districts seeking to meaningfully reduce levels of segregation, partnering with neighboring districts to implement interdistrict desegregation plans such as those operating in Boston, Milwaukee, Minneapolis, and St. Louis may be the most viable, if organizationally more challenging, alternatives (Tefera, Frankenberg, Siegel-Hawley, & Chirichigno, 2011). By contrast, we found that a much larger share of urban segregation occurred within district boundaries. Indeed, in 2012, 57% of urban school segregation occurred within districts. This suggests that more traditional within-district plans may be somewhat more successful in urban districts than in their suburban and exurban counterparts.
Our findings provide new empirical evidence linking changes in racial/ethnic diversity to changes in the metropolitan structure of racial imbalance. However, several limitations of the studys findings must be acknowledged. First, it is important to emphasize that our studys findings are limited to the most recent decade of data from 2002 to 2012. As such, while our findings provide a picture of current trends, our study period may not capture the full range of changes in segregation precipitated by the earlier phases of suburbanization and exurbanization. For example, we find that most metropolitan localities experienced insufficient racial/ethnic change to completely counteract secular declines in segregation between 2002 and 2012. To the extent that racial change was more pronounced prior to 2002 (Fry, 2009), it is possible that the impact of changes prior to 2002 were associated with larger increases or smaller declines in segregation than over the past decade. However, the past decade was a period of pronounced, large-scale changes in racial/ethnic diversity, particularly in suburbs and exurbs. Indeed, between 2002 and 2012, the racial/ethnic diversity of suburbs (as measured via Simpsons index) increased by 37%, while the diversity of exurbs increased by over 40%.
Second, our analyses held constant the boundaries of each metropolitan area and the locality of each school district based on their most recent data. As we discuss at length above, this was necessary to ensure that our estimates of change in segregation isolated changes in the distribution of students across metropolitan areas from changes in metropolitan boundaries or the classification of localities in which students reside. However, it must be emphasized that these localities are not fixed in reality, as districts do change designations over time. Thus, some of the districts that were classified as suburban in 2012 may have been exurban in 2002, and some of the districts that are currently classified as exurban may not have been part of metropolitan areas in 2002. As a result, our analyses may misclassify a small portion of the suburban segregation as exurban segregation, some of the urban segregation as suburban, and so on. It should be noted that the probability of a districts locality code changing increases according to the length of the time span being studied. As such, there is an analytical tradeoff between the length of the time period studied and the validity of the locality classifications being studied. By selecting a shorter time period, we were able to limit the effect of these classification changes on our study findings: For the decade studied, just 5.4% of the districts studied changed locality.
Finally, this study included exurban areas and thereby the full range of districts present within metropolitan areas, where prior research has reclassified these as suburban or excluded them from analysis. While this studys tripartite categorization, which is derived from official U.S. Census designations, broadly captured important differences among metropolitan districts, it should be noted that significant variability between districts of the same locality designation may remain. Frankenberg (2012) has taken important first steps in classifying typologies of suburban districts with shared demographic patterns and trends. Focusing on the 25 largest metropolitan areas, she categorizes suburban districts into six different types, ranging from inner-ring transitioning suburbs characterized by rapid racial/ethnic change to exclusive enclaves that are highly resistant to racial/ethnic change. Future research may seek to understand how trends in racial/ethnic segregation vary among these heterogeneous suburban localities.
Armor, D. J. (1995). Forced justice: School desegregation and the law. Oxford, UK: Oxford University Press.
Berube, A., Singer, A., Wilson, J., & Frey, W. (2006). Finding exurbia: Americas fast-growing communities at the metropolitan fringe. Washington, DC: Brookings Institution.
Clotfelter, C. T. (2001). Are whites still fleeing? Racial patterns and enrollment shifts in urban public schools, 19871996. Journal of Policy Analysis and Management, 20 (2), 199221.
DeNavas-Walt, C., Proctor, B., & Smith, J. Income, poverty, and health insurance coverage in the United States: 2011 [U.S. Census Bureau current populations report]. Washington, DC: U.S. Census Bureau.
Farley, R., Schuman, H., Bianchi, S., Colasanto, D., & Hatchett, S. (1978). Chocolate city, vanilla suburbs: Will the trend toward racially separate communities continue? Social Science Research, 7(4), 319344.
Fiel, J. (2013). Decomposing school resegregation: Social closure, racial imbalance, and racial isolation. American Sociological Review, 78(5), 828848.
Fischer, C. S., Stockmayer, G., Stiles, J., & Hout, M. (2004). Distinguishing the geographic levels and social dimensions of U.S. metropolitan segregation, 19602000. Demography, 41(1), 3759.
Fischer, M. J. (2003). The relative importance of income and race in determining residential outcomes in U.S. urban areas, 19702000. Urban Affairs Review, 38(5), 66996.
Frankenberg, E. (2012). Understanding suburban school district transformation: A typology of suburban districts. In E. Frankenberg & G. Orfield (Eds.), The resegregation of suburban schools: A hidden crisis in American education (pp. 2743). Cambridge, MA: Harvard Education Press.
Frankenberg, E., & Orfield, G. (2012). The resegregation of suburban schools: A hidden crisis in American education. Cambridge, MA: Harvard University Press.
Frey, R. (2011). Melting pot cities and suburbs: Racial and ethnic change in metro America in the 2000s. Washington, DC: The Brookings Institution.
Fry, R. (2009). The rapid growth and changing complexion of suburban public schools. Washington, DC: Pew Research Center.
Green v. County School Board of New Kent, 391 U.S. 430 (1968).
Iceland, J. (2004). The multigroup entropy index. Washington, DC: U.S. Census Bureau.
Jargowsky, P. A. (2013). Segregation, neighborhoods, and schools. In A. Lareau & K. Goyette (Eds.), Choosing homes, choosing schools: Residential segregation and the search for a good school (pp. 97136). New York, NY: Russell Sage Foundation.
Littman, M. (Ed.). (1998). A statistical portrait of the United States: Social conditions and trends. Lanham, MD: Bernan Press.
Logan, J. R. (2004). Resegregation in American public schools? Not in the 1990s. Albany: University at Albany, State University of New York, Lewis Mumford Center for Comparative Urban and Regional Research.
Logan, J. R., Oakley, D., & Stowell, J. (2008). School segregation in metropolitan regions, 19702000: The impacts of policy choice on public education. American Journal of Sociology, 113(6), 16111644.
Massey, D. S., & Denton, N. A., (1988). The dimensions of residential segregation. Social Forces, 67(2), 281315.
Mather, M., Pollard, K., & Jacobsen, L. (2011). Reports on America: First results from the 2010 census. Washington, DC: Population Research Bureau.
National Center for Education Statistics (NCES) (2006). Identification of Rural Locales. Retrieved from https://nces.ed.gov/ccd/rural_locales.asp
Orfield, G., & Lee, C. (2006). Racial transformation and the changing nature of segregation. Cambridge, MA: Harvard University Civil Rights Project.
Pew Research Center. (2013). The rise of Asian Americans. Washington, DC: Author.
Raffel, J. A. (1998). Historical dictionary of school segregation and desegregation: The American experience. Westport, CT: Greenwood.
Reardon, S. F., Farrell, C. R., Mathews, S. A., OSullivan, D., Bischoff, K., & Firebaugh, G. (2009). Race and space in the 1990s: Changes in the geographic scale of racial residential segregation, 19902000. Social Science Research, 38(1), 5570.
Reardon, S. F., & Firebaugh, G. (2002). Measures of multigroup segregation. Sociological Methodology, 32(1), 3367.
Reardon, S. F., Mathews, S., OSullivan, D., Lee, B., Firebaugh, G., Farrell, C., & Bischoff, K. (2008). The geographic scale of metropolitan racial segregation. Demography, 45(3), 489514.
Reardon, S. F., & Owens, A. (2014). 60 years after Brown: Trends and consequences of school segregation. Annual Review of Sociology, 40, 199218.
Reardon, S. F., & Yun, J. T. (2001). Suburban racial change and suburban school segregation, 198795. Sociology of Education, 74(2), 79101.
Reardon, S. F., Yun, J. T., & Eitle, T. M. (2000). The changing structure of school segregation: Measurement and evidence of multiracial metropolitan-area school segregation, 19891995. Demography, 37(3), 351364.
Richards, M. P., & Stroub, K. J. (2014). The fragmentation of metropolitan school districts and the segregation of American schools. Teachers College Record, 116(12), 130.
Rossell, C. (1990). The carrot or the stick for school desegregation policy: Magnet schools or forced busing? Philadelphia, PA: Temple University Press.
Stroub, K. J., & Richards, M. P. (2013). From resegregation to reintegration: Trends in metropolitan school segregation, 19932010. American Educational Research Journal, 50(3), 497531.
Swann v. Charlotte-Mecklenburb Board of Education, 402 U.S. 1 (1971).
Tefera, A., Frankenberg, E., Siegel-Hawley, G., & Chirichigno, G. (2011). Integrating suburban schools: How to benefit from growing diversity and avoid segregation. Los Angeles, CA: The Civil Rights Project.
Theil, H. (1972). Statistical decomposition analysis. Amsterdam, The Netherlands: North Holland.
United States Census Bureau. (2010). Summary File 1, American Fact Finder.
United States Office of Management and Budget. (2008, November 20). Update on statistical area definitions and guidance on their use (OMB Bulletin No. 09-01). Washington, DC: Executive Office of the President.