Racial Diversity and Change in Metropolitan Neighborhoods

Abstract

This study investigates the changing racial diversity and structure of metropolitan neighborhoods. We consider three alternative perspectives about localized racial change: that neighborhoods are bifurcating along a white/nonwhite color line, fragmenting into homogeneous enclaves, or integrating white, black, Latino, and Asian residents into diverse residential environments. To assess hypotheses drawn from these perspectives, we develop a hybrid methodology (incorporating the entropy index and majority-rule criteria) that offers advantages over previous typological efforts. Our analysis of 1990–2000 census tract data for the 100 largest U.S. metropolitan areas finds that most neighborhoods are becoming more diverse and that members of all groups have experienced increasing exposure to neighborhood diversity. However, white populations tend to diminish rapidly in the presence of multiple minority groups and there has been concomitant white growth in low-diversity neighborhoods. Latino population dynamics have emerged as a primary force driving neighborhood change in a multi-group context.

Changes in neighborhood racial and ethnic composition have long been of scholarly and policy interest because of their implications for racial stratification in the United States. Classic studies of neighborhood change during the post-World War II era focused on white-to-black transitions (Duncan and Duncan 1957; Taeuber and Taeuber 1965), reflecting Du Bois’s ([1903] 1918) assertion that the black-white “color line” was the defining fissure of the twentieth century (see also Myrdal 1944). We investigate whether spatial manifestations of the color line remain salient in the face of increasing metropolitan diversity. In recent decades, immigration from Latin America and Asia has markedly altered the racial and ethnic mix of our largest cities and the settlement patterns of groups within those cities (Fong and Shibuya 2005; Singer 2005).¹

Understanding neighborhood change in the racially diversifying metropolis can illuminate more specific forms of inequality. For example, uneven access to institutional and economic resources often coincides with the uneven spatial distribution of racial groups across neighborhoods (Kain 1968; Orfield and Lee 2007; Wilson 1996), as does exposure to crime, pollution, and other socio-environmental sources of racial health disparities (Acevedo-Garcia et al. 2008; Fitzpatrick and LaGory 2000; Klinenberg 2002; Sampson, Morenoff, and Gannon-Rowley 2002). Neighborhoods also provide a context for race relations. Residents of diverse neighborhoods experience more casual interracial contact, which tends to dampen negative stereotyping and lower the perceived threat of outgroups (Oliver and Wong 2003; Welch et al. 2001). For persons living in more homogeneous environs, however, increasing inter-group contact might have few beneficial effects. In the case of rapid immigration-fueled neighborhood change, the influx of newcomers could spur flight or antagonism on the part of incumbents (Wilson and Taub 2006).

Given the continuing significance of race at the neighborhood level, numerous investigations have sought to describe the complex processes associated with rising neighborhood diversity over the past two decades (see, e.g., Alba et al. 1995; Denton and Massey 1991; Ellen 2000; Logan and Zhang 2010). In general, they show a decline in the number of all-white neighborhoods and a striking increase in neighborhoods with multiple groups present. Patterns vary somewhat across metro areas and among specific groups, but the clear message is that metropolitan residents are increasingly “living together” in racially diverse neighborhoods (Fasenfest, Booza, and Metzger 2006).

These studies are not without shortcomings, however. Diversity is often ill-defined or operationalized using arbitrary or inconsistent criteria. In the present paper, we seek to sharpen the conceptual and methodological treatment of diversity and to develop a simple “majority-rule” criterion to assess its racial structure. We also propose a comparative theoretical framework within which to understand neighborhood change. Using census tract data from 1990 and 2000, we examine whether metropolitan neighborhoods are better characterized as bifurcating along a white/nonwhite divide, fragmenting into homogeneous enclaves, or integrating white, black, Latino, and Asian residents into diverse residential environments. We are guided by four questions: (1) How has multi-group diversity—and the exposure of particular racial groups to diversity—changed in metropolitan neighborhoods? (2) What neighborhood racial structures are associated with stability and change? (3) How are group populations distributed across neighborhoods of differing racial structures? (4) What group-specific population dynamics contribute to neighborhood diversity transitions?

APPROACHES TO NEIGHBORHOOD RACIAL COMPOSITION

Our desire to strengthen the conceptual and methodological foundation of neighborhood racial change research is motivated in part by dramatic demographic transformations occurring over the past half-century. While roughly one-third of the American population—more than 107 million people—now identifies as a member of a nonwhite racial/ethnic group (U.S. Census Bureau 2009), only one in ten residents were nonwhite in 1960 (Gibson and Jung 2002). Much of this shift was spurred by the wave of immigrants from Latin America and Asia subsequent to the 1965 Immigration Act. New immigration has continued to register a profound demographic impact, with net migration and natural increase among immigrant-rich Latino and Asian populations accounting for nearly two-thirds of recent U.S. population growth (Haub 2005). Increasing racial diversity is most conspicuous in the nation’s big cities, many of which are now “minority majority” in composition and less likely to exhibit the traditional black-white bipolar structure common in the past (Bean et al. 2004; Berube 2003).

These large-scale demographic patterns filter downward to affect neighborhoods, rendering them ever more challenging to investigate. Studies of neighborhood racial change typically create typologies that use population thresholds to establish the presence of specific groups (Alba et al. 1995; Denton and Massey 1991; Ellen 2000; Fasenfest et al. 2006; Fong and Gulia 2000; Friedman 2007; Lee and Wood 1991; Logan and Zhang 2010). They then track changes in group size or representation in order to detect whether there is a tendency toward neighborhood homogeneity or diversity. These studies have proven valuable, but many of them lack a summary measure of multi-group diversity, employ relativistic typologies of neighborhood racial structure, or pay inadequate attention to the group-specific population dynamics underlying neighborhood change. Such limitations reveal a need for alternative methods of analyzing how neighborhoods evolve in a multi-group context.

Our first methodological concern, about diversity measurement, stems from the practice of treating neighborhoods as diverse based upon racial population thresholds as small as 30 or 100 group members (Alba et al. 1995; Denton and Massey 1991; Fong and Gulia 2000). Because this approach likely overestimates the magnitude of neighborhood diversity, the conceptualization of diversity requires closer scrutiny. If we define diversity as the demographic richness exhibited by a population, two properties of that population are germane to its level or magnitude of diversity: the number of groups it comprises, and the sizes of the groups in relation to each other (Pielou 1977; White 1986). On intuitive grounds, a population consisting of many groups of equal size would be described as highly diverse. At the opposite extreme lies a completely homogeneous population in which all members belong to one group. Fortunately, there are existing measures which conform to this refined meaning of diversity, even though few recent studies have taken advantage of them (see, e.g., Modarres 2004). We discuss one such measure in the methods section below.

Our second methodological concern—related to the first—involves the description of neighborhood racial structure, i.e., the particular groups represented. A number of investigators have classified racial structures by using a relativistic approach to help account for overall differences in group size (Smith 1998). Specifically, a group’s predominance, presence, or absence in a given neighborhood is assigned based upon that group’s local representation relative to some broader context. An important limitation of this approach is that classification has limited utility outside the designated context. Maly (2000), for instance, develops a summary diversity measure to categorize Chicago neighborhoods but relies on a relative classification scheme that is tailored specifically to that metropolitan area. It is not evident how one could effectively use Maly’s scheme to compare Chicago’s neighborhoods to those in other cities: a relatively large Latino neighborhood presence in predominantly white Scranton is likely to look very different than a relatively large Latino presence in a Chicago neighborhood. We seek to avoid treating such neighborhoods as equivalent and thereby creating the appearance of diversity where little exists.

The comparability issue can be partially ameliorated by expanding the referent context. Rather than relying on a single metropolitan area, Logan and Zhang (2010) classify a group as present if its decade-specific neighborhood share is at least 25 percent of its aggregate share of the overall metropolitan sample. While this is an improvement over a metro-tailored approach, classification continues to be conditioned by sample size and make-up. Logan and Zhang base their neighborhood racial classification on a sample of 24 diverse metropolitan areas. Demographically speaking, a larger (i.e., less diverse) sample of metropolitan areas would lower the bar in establishing groups’ neighborhood presence. Thus, any attempts to widen substantially the scope of inquiry might define neighborhood diversity down to a point that could challenge face validity. The relativistic approach becomes even thornier when assessing decadal changes since it sets the bar higher in recent (i.e., more diverse) decades than in preceding ones, thereby threatening temporal comparisons (Galster 1998).

What is needed, and what we propose, is a way to examine neighborhood racial structure in a consistent and parsimonious manner. We opt for an absolute approach, classifying neighborhoods according to group majorities. In our opinion, the majority-rule criterion is less arbitrary and more meaningful given that, when asked about neighborhood preferences, residents tend to be quite attuned to their group’s numerical majority/minority status in relation to other groups (Charles 2000; Clark 2002). One consequence of our approach is that racial structures will be distributed unequally among metropolitan areas (due to regional concentrations of different groups) and over time (due to the national trend toward greater diversity). Rather than view this as a shortcoming, we contend that a constant criterion is critical for making legitimate cross-sectional and longitudinal comparisons with a large sample of metropolitan neighborhoods and for evaluating the objective demographic diversity that residents experience in their local areas. Simply put, living in a majority Hispanic—or black, white, or Asian—neighborhood should mean the same thing regardless of spatial or temporal context.

A final methodological concern is that much previous scholarship neglects the group population dynamics driving neighborhood change. Increasing neighborhood diversity—or, by extension, increasing neighborhood homogeneity—may be due to any number of group-specific gains or losses. Even the appearance of stable neighborhood diversity could mask swings in subpopulations as different groups experience offsetting growth and decline. Denton and Massey (1991) examine group-specific population changes in neighborhoods with varying racial/ethnic structures (see also Lee and Wood 1991). However, their work covers the 1970s; subsequent studies have largely bypassed such population dynamics.

THREE THEORETICAL PERSPECTIVES

Trajectories of neighborhood change can tell us much about how different racial and ethnic groups are being incorporated into the residential mosaic. As the mosaic becomes more complex, our theories of change—like the methods used to study it—must rise to the occasion. Toward that end, we present three theoretical perspectives on neighborhood racial change in a multi-group context. The first two perspectives, here labeled bifurcation and fragmentation, constitute refinements of the stratification model utilized in residential segregation and locational attainment research; the third, demographic integration, overlaps with the popular spatial assimilation approach (Charles 2003; Logan and Alba 1993; Logan, Zhang, and Alba 2002; Massey 1985; Rosenbaum and Friedman 2007). Together, they provide a comparative framework for generating alternative hypotheses about how neighborhoods change.

Bifurcation

The bifurcation model, as traditionally formulated, predicts the racial composition of metropolitan neighborhoods to fall along a broad, white/nonwhite color line. In a bifurcated setting, diverse neighborhoods are characterized by the presence of black, Latino, and Asian residents while white households reside in separate areas. Indeed, prior research has shown the pace of white population loss to be particularly swift in neighborhoods in which multiple minority groups are present (Crowder 2000; Denton and Massey 1991), with exiting whites often replaced by incoming Asian and Hispanic households (Friedman 2007; Lee and Wood 1991).

A central principle of the bifurcation perspective is that white residential preferences and mobility decisions are due in large part to racial prejudice. Negative racial stereotypes on the part of whites are associated with less acceptance of integration (Krysan et al. 2009), which helps explain why diversifying neighborhoods tip and begin to rapidly lose white inhabitants (Card, Mas, and Rothstein 2007; Pais, South, and Crowder 2009). This is compounded by the fact that many minority residents are uncomfortable moving into white neighborhoods because such neighborhoods are seen as inhospitable (Krysan and Farley 2002). In addition to simple outgroup antagonism, whites may perceive a collective threat to their privileged status in an increasingly multiethnic environment (Blalock 1967; Blumer 1958). Bobo and Zubrinsky (1996), for example, argue that an entrenched racial hierarchy (with whites on top) ensures that whites will “tend to view integration with any of the minority groups as threatening or undermining a previous status relation of superiority” (p. 904).

A more benign (though still stereotype-driven) explanation for whites’ neighborhood preferences and mobility behavior stresses the use of neighborhood racial composition as a cue about nonracial characteristics such as property values, school quality, poverty, and crime (Harris 1999; Laurenti 1960; Taub, Taylor, and Dunham 1984). This race-based neighborhood stereotyping leads white residents to exit integrated areas or, more commonly, to avoid entering them in the first place (Ellen 2000). The composition of such neighborhoods should become more exclusively minority over time, according to bifurcation logic.

As Massey and Denton (1993) contend, the color line in American cities is not merely the culmination of white prejudice and preference. Rather, discriminatory treatment by government agencies, lenders, and real estate agents has ensured a bifurcated pattern of de facto racial apartheid. Massey and Denton emphasize the struggle of African Americans, but audit studies suggest that Latinos and Asians also face significant housing discrimination (Turner and Ross 2005). Ultimately, these practices may push neighborhoods toward either white or nonwhite (i.e., mixed minority) occupancy, as the bifurcation perspective would lead us to expect.

Fragmentation

The fragmentation perspective differs from bifurcation in that it predicts multiple color lines, with racial and ethnic minorities as well as whites seeking out homogeneous neighborhoods. Asian and Latin American immigrants establish their own enclaves that afford social support and cultural familiarity, prompting many non-Hispanic blacks to gravitate toward racially homogeneous niches as a buffer against the newcomers. Under this scenario, Latino and Asian growth should spur residential turnover akin to the black-white transitions of the past.

As a consequence of immigrant-driven enclave formation, Latinos and Asians have become more segregated in most metropolitan areas during the past two decades (Iceland, Weinberg, and Steinmetz 2002; Logan, Stults, and Farley 2004). Neighborhood change underlies these structural shifts: Latino residential succession now dominates in metropolitan areas like Los Angeles and New York (Clark 1996; Lobo, Flores, and Salvo 2002). Afro-Caribbeans (Crowder 1999) and Asian immigrants (Li 1998; Ming, Lauderdale, and Kandula 2009) are also creating ethnically distinct neighborhoods that can further fragment the metropolis.

Racial stereotyping and prejudice are not limited to whites: when they exist among multiple groups, the odds of fragmentation should increase. In Los Angeles, Charles (2000) documents that one-third of Latinos and two-fifths of Asians identify an ideal neighborhood as one completely devoid of black residents. The feelings may be mutual; a different L.A. study finds that substantial proportions of black respondents hold negative views of Hispanics and Asians (Johnson, Farrell, and Guinn 1997; but see Taylor and Schroeder 2010). Stereotyping encourages the rank ordering of racial outgroups as potential neighbors (Charles 2000), which in turn could contribute to neighborhood fragmentation as certain groups seek distance from others. An alternative explanation for minority self-segregation emphasizes ingroup affinity rather than outgroup aversion. Metropolitan residents express strong own-race neighborhood preferences, and their mobility destinations are generally consistent with such preferences, not to mention with the fragmentation perspective (Clark 1992, 2002). Ingroup affinity is particularly salient among foreign-born Latinos and Asians lacking English proficiency (Charles 2007).

Demographic Integration

The demographic integration perspective parts company with the two previous perspectives by predicting growing diversity across metropolitan neighborhoods, including those with largely white populations.² This is due to increasingly tolerant racial attitudes, a decline in housing discrimination, minority socioeconomic mobility, and expanded minority housing options due to new housing construction. The fact that neighborhood interracial stability (including whites) has become more common poses a direct challenge to the bifurcation model (Ellen 2000; Lee and Wood 1990; Nyden, Maly, and Lukeheart 1997). And while immigration is changing metropolitan racial composition, it does not necessarily follow that immigrants are settling into fragmented enclaves (Galster, Metzger, and Waite 1999). Even if some are, their offspring may convert educational attainment and financial capital into neighborhood proximity with Anglos (Iceland 2009; South, Crowder, and Pais 2008).

Much of the recent research on neighborhood racial change supports the demographic integration perspective. One clear generalization is that all-white neighborhoods are becoming rare while metropolitan neighborhoods as a whole are more likely to contain multiple groups (Alba et al. 1995; Denton and Massey 1991; Fasenfest et al. 2006; Logan and Zhang 2010). These changes have coincided with favorable shifts in white attitudes about racial integration, as documented in two Detroit surveys by Farley and colleagues (Farley et al. 1978; Farley et al. 1994).

At the same time, housing discrimination has lessened. Paired-tester audits of rental and real estate agencies from 1989 and 2000 reveal a sharp decrease in adverse treatment against black and Latino homeseekers and a parallel decline in discrimination against black renters (Turner and Ross 2005). Farley and Frey (1994) argue that discrimination is less prevalent in newer housing developments because they lack the history of racial strife burdening many established neighborhoods. Contrary to the fragmentation model, they assert that the growth of Hispanic and Asian “buffer groups” can alter the polarized black-white dynamic in local real estate markets, thereby fostering greater integration (Frey and Farley 1996).

DATA AND METHODS

Our investigation of the bifurcation, fragmentation, and demographic integration perspectives relies on 1990 and 2000 census tract data for the 100 largest U.S. metropolitan areas (hereafter Metro-100). We use census racial and ethnic definitions to identify four mutually exclusive groups: Hispanics or Latinos of any race and the non-Hispanic segments of the white, black, and Asian (including Native Hawaiians and Pacific Islanders) populations. The four groups accounted for 97.4 percent of the Metro-100 population in 2000. Although diversity gains have not been limited to large metropolitan areas (Lichter and Johnson 2006), we achieve impressive coverage with this Metro-100 sample, capturing 55 percent of whites, 69 percent of blacks, 77 percent of Latinos, and 84 percent of Asians residing in the United States in 2000. Our analysis includes 37,951 tracts at both time points, making it one the most comprehensive studies of neighborhood racial change to date.³

Measuring Diversity

We use the entropy index (E) to capture the level of multi-group diversity in metropolitan neighborhoods (White 1986). E is calculated as:

$$E=\sum _{r=1}^n{Q}_{r}ln\frac{1}{Q_r},$$ (1)

where Q_r refers to group r’s proportion of a neighborhood population. The formula sets the maximum value of E to the natural log of the number of groups. Thus, the maximum entropy in our analysis of four broad racial groups is ln(4) = 1.386. Such a value would occur only when all groups are of equal size (i.e., each racial group constitutes one-quarter of the population). At the opposite extreme, an E value of zero signifies complete homogeneity or no diversity, with all population members in the same group (e.g., an all-white neighborhood). We standardize E by dividing it by its four-group maximum and multiplying by 100, thereby setting its range of possible values from zero to 100.⁴

In addition to quantifying overall diversity levels, we measure the degree to which residents are exposed to neighborhood diversity. To capture such exposure, we follow Farrell’s (2005) lead and weight the tract entropy indexes by the share of a group’s total population. Formally,

$$E_e=\sum _{r=1}^n\left[(E_i)\left(\frac{r_i}{R}\right)\right]$$ (2)

where n is the number of neighborhoods, E is the entropy score for neighborhood i, r is a racial group’s population in neighborhood i, and R is the summed group population across n neighborhoods. This entropy exposure index (or E_e), which resembles P* indexes used in segregation research, indicates the level of neighborhood racial diversity experienced by the average group member. We calculate E_e values for each year and racial group to determine how group-specific exposure to diversity has changed across Metro-100 census tracts. Additionally, we use histograms to graph the decadal shares of group populations living in environments of varying diversity.

Classifying Racial Structure

The bifurcation, fragmentation, and demographic integration perspectives offer contrasting predictions not only about the overall level of neighborhood racial diversity but also about its changing compositional makeup. Hence, our entropy-based analysis is supplemented with a classification scheme that categorizes tracts according to their majority racial group (if any) and then assigns them to one of three subcategories. First, in a group-dominant neighborhood (labeled white-dominant, black-dominant, etc.), the majority group constitutes 90 percent or more of the population. These tracts exhibit the lowest levels of diversity in our sample. We use the 90 percent threshold because of its history as an indicator of extreme group presence (Duncan and Duncan 1957; Taeuber and Taeuber 1965). Moreover, the racial composition of such a neighborhood would be immediately obvious to anyone passing through. The cultural markers of public life, embodied in visible street scenes and ethnic institutions, would convey homogeneity to ingroup and outgroup members alike.

The second subcategory of group-majority neighborhood is group-shared (e.g., Latino-shared), meaning that the majority group (Latinos) represents less than 90 percent of the population and none of the other three groups comprises as much as 10 percent. Such neighborhoods are fairly homogeneous but have a more visible mix of outgroups than group-dominant areas. In the third type of group-majority neighborhood, labeled group-other, one or more outgroups each attain at least 10 percent representation. Thus, a Latino-white neighborhood is a Latino-majority neighborhood with a white population share of at least 10 percent.

Apart from group-majority neighborhoods are the no-majority neighborhoods, which we classify according to pluralities rather than majorities. In an Asian-no-majority neighborhood, for instance, no group achieves majority status but Asian residents constitute the largest proportion of the population. We are particularly interested in white population change in these highly diverse tracts and thus use the same 10 percent threshold outlined above to define white presence or absence.

Typology of Change

Finally, we develop a two-dimensional typology of neighborhood change that aligns with the bifurcation, fragmentation, and integration perspectives and that takes group dynamics into account. The first dimension, multi-group diversity, recognizes that a neighborhood can experience increasing, decreasing, or stable levels of overall diversity. We consider tracts increasing or decreasing in diversity if their E score rises or falls by more than five percent between 1990 and 2000. The second dimension, group-specific change, recognizes that the diversity trend in a neighborhood can be due to a unique combination of race-specific population gains and losses (e.g., white population losses could accompany diversity increases or decreases depending on the rates of growth/decline of the other three groups). We combine these two dimensions to determine whether neighborhoods are changing as hypothesized by our three models.

HYPOTHESES

The bifurcation, fragmentation, and integration perspectives should be viewed as complementary rather than mutually exclusive, given that changes in line with two or more of them could occur simultaneously within a single metropolitan area. This is why analyses focusing on smaller-scale neighborhood racial change are needed in addition to studies of larger-scale patterns of metropolitan segregation. The appearance of stability or gradual change in segregation at the metropolitan level might obscure offsetting changes occurring in neighborhoods. For example, Logan, Stults, and Farley (2004) find that segregation levels between whites and Hispanics (as measured by the index of dissimilarity) barely changed between 1990 and 2000. By contrast, our analyses (see below) suggest that the localized experiences of these two groups have been anything but static. In other words, studies of neighborhood change are necessary to flesh out the dynamics underlying gradual shifts (or lack thereof) in segregation. When considering the three perspectives, our goal is to weigh the evidence in order to judge whether most instances of neighborhood change conform to any one of the perspectives. In Table 1 we summarize hypotheses about trends in neighborhood diversity levels and exposure, transitions in neighborhood racial structures, and group-specific population distributions and dynamics occurring across neighborhoods.

Table 1.

Summary of Bifurcation, Fragmentation, and Integration Models of Neighborhood Change

	Bifurcation	Fragmentation	Integration
Diversity trends
Overall	Uneven change	Decreasing diversity	Increasing diversity
Low diversity tracts	Uneven change	Persistence	Instability→high diversity
High diversity tracts	Uneven change	Instability→low diversity	Persistence
White diversity exposure	Low/decreasing	Low/decreasing	High/increasing
Nonwhite diversity exposure	High/increasing	Low/decreasing	High/increasing
Neighborhood transitions
White-dominant	Persistence	Persistence	Instability→W-other
White-shared	Instability→W-dominant	Instability→W-dominant	Instability→W-other
White-other	Instability→no-majority	Instability→BLA-majority	Persistence
BLA-dominant	Instability→BLA-nonwhite	Persistence	Instability→BLA-white
BLA-shared	Instability→BLA-nonwhite	Instability→BLA-dominant	Instability→BLA-white
BLA-white	Instability→BLA-nonwhite	Instability→BLA-dominant	Persistence
BLA-nonwhite	Persistence	Instability→BLA-dominant	Instability→BLA-white
White-no-majority	Instability→BLA-no-majority	Instability→BLA-majority	Persistence
BLA-no-majority	Persistence	Instability→BLA-majority	Persistence
Population distribution
White population shares	Increasing in W-dominant tracts	Increasing in W-dominant tracts	Increasing in group-other tracts
BLA population shares	Increasing in group-other tracts, whites absent	Increasing in BLA-dominant tracts	Increasing in group-other tracts, whites present
Diversity and population dynamics
Diversity levels	Increasing	Decreasing	Increasing
Population change	White decline	BLA-growth	White stability

With respect to diversity trends (top panel), each of the three perspectives predicts different patterns of change. The bifurcation perspective anticipates uneven change as certain tracts become more diverse while other low-diversity tracts (predominantly white) remain unchanged. The fragmentation perspective predicts decreasing overall tract diversity coinciding with persistence among low-diversity tracts. The demographic integration model predicts just the opposite, that there will be diversity increases across the board and that high neighborhood diversity will persist rather than wane. Changes in exposure will be group-specific under the bifurcation scenario, with whites experiencing low and decreasing levels of diversity exposure while black, Latino, and Asian populations experience increasing exposure.

Tract transitions in racial structure are addressed in the second panel of Table 1, which provides hypothesized propensities for change (i.e., persistence, instability) and predicted destination states for unstable neighborhood types. The bifurcation and fragmentation perspectives predict that white-majority tracts with the least diverse racial structures will persist over time while those with nonwhite outgroups present will experience instability and lose their white majorities. The demographic integration perspective predicts that white-majority neighborhoods will retain or acquire a significant outgroup presence. In the case of tracts with black, Latino, or Asian (BLA) majorities, the bifurcation perspective leads us to expect movement toward more diverse racial structures devoid of whites while demographic integration hypothesizes the retention of a visible white presence. The fragmentation model predicts BLA-majority tracts to move toward single-group dominance regardless of outgroup composition. Similarly, no-majority tracts should shift to less diverse BLA majorities if fragmentation is occurring. The bifurcation perspective predicts that no-majority tracts will lose white population but the demographic integration perspective predicts that no-majority tracts with a white presence will keep it over time.

The final parts of our analysis assess group-specific population distributions and dynamics. As shown in Table 1 (third panel), the three models predict that changing population distributions across neighborhood types will vary by racial group. The bifurcation and fragmentation models predict increasing shares of the white population to reside in predominantly white neighborhoods. They differ in terms of the distribution of nonwhite populations, with the bifurcation perspective predicting greater BLA shares in diverse neighborhoods (absent whites) and the fragmentation perspective predicting greater BLA concentration in homogeneous group-dominant tracts. The demographic integration perspective anticipates larger proportions of whites and nonwhites to reside in group-other tracts that have a white presence. The population dynamics accompanying diversity change are summarized in the fourth panel. Tracts adhering to the bifurcation model are expected to lose white population in the face of increasing diversity. For tracts following an integration trajectory, however, white populations will remain stable as diversification proceeds. Fragmenting tracts are those in which a single BLA group increases in size while overall tract diversity declines.

DIVERSITY TRENDS

Consistent with the demographic integration perspective, our analysis indicates that most Metro-100 neighborhoods are shifting toward greater diversity. Figure 1 displays mean entropy scores, broken down by region, to buttress this conclusion. Metro-100 tracts experienced a 20 percent increase in mean diversity from 1990 to 2000. Western and southern tracts are more diverse than their northeastern and midwestern counterparts but diversity gains were pervasive, affecting all four regions.

Figure 1.

Trends in Average Tract Diversity, 1990–2000

While such results point to heightened integration, they might obscure divergent changes across tracts. Table 2 addresses this possibility, presenting a multi-group diversity transition matrix for the 37,951 metropolitan neighborhoods in our sample. The table margins provide year-specific tract counts for each of five 20-point diversity categories. The row headings (origin states) identify 1990 diversity categories and the column headings (destination states) identify 2000 diversity categories. The diagonal (in bold) represents the proportion of tracts achieving stability or persistence between 1990 and 2000. All of the cells below the diagonal identify tracts that dropped one or more categories; the cells above the diagonal identify tracts that climbed one or more categories.

Table 2.

Diversity Transitions in Metro-100 Tracts,1990–2000^a

		2000 diversity (E)					N of tracts
		0–20	20–40	40–60	60–80	80–100
1990 diversity (E)	0–20	64.8%	30.8	4.0			12,801
	20–40	6.1	47.6%	37.6	8.3		10,139
	40–60		9.9	48.2%	37.5	4.0	8,459
	60–80		1.5	17.8	57.1%	23.6	5,034
	80–100			2.0	34.6	63.0%	1,518
N of tracts		8,942	9,688	9,320	7,467	2,534	37,951

^aCell percentages representing less than 1 percent of category or fewer than 10 tracts are omitted.

Like the mean entropy scores, the evidence in the transition matrix supports the demographic integration model. About three-fifths of the tracts in the two highest 1990 diversity categories stayed put over the decade, suggesting that diversity is not as fleeting as the fragmentation perspective predicts. The tracts in the next two categories exhibit lower rates of stability but they were likely to rise rather than drop in diversity, yielding an absolute increase of over 1,000 tracts in the highest diversity category. Nearly two-thirds of the lowest diversity tracts retained their 1990 classification, which is consistent with fragmentation. However, the margins of the table show that, despite this relatively high level of persistence, some 4,000 fewer low-diversity tracts existed in 2000 than in 1990. On the whole, the dominant trend was toward increasing demographic integration.⁵

Changes occurring in metropolitan neighborhoods with differing diversity levels, though informative, do not reveal how groups are spread across neighborhood types or the degree to which group members are exposed to neighborhood diversity. Figure 2 presents four diversity histograms which reflect group-specific population distributions in tracts of differing diversity levels. The vertical axis of each panel is organized from bottom to top into successive 5-point tract diversity categories, and the left and right sides represent 1990 and 2000 distributions, respectively. Each of the panels also includes group-specific entropy exposure indices (E_e) for both time points. These average E scores are weighted by the share of a group’s total population residing in each tract. Thus, they can be interpreted as the level of localized diversity experienced by the average group member.

Figure 2.

Group-Specific Diversity Distributions and Exposure Indices, 1990–2000

A glance at the diversity histograms in Figure 2 reveals notable differences in the diversity distributions of racial groups. White distributions are highly skewed toward lower diversity while the Latino and Asian “pyramids” are inverted, reflecting larger shares in high-diversity tracts. The black metropolitan population is characterized by more of a bimodal pattern. In 2000, about one-fifth of the black population was located in low-diversity tracts (E < 20) while another quarter was clustered in tracts with E scores ranging from 50 to 70. The accompanying exposure indices provide a useful statistical summary for each of the histograms. In 2000, an average white resident experienced much lower levels of neighborhood diversity (E_e = 37.5) than members of other groups, a pattern in keeping with the bifurcation perspective. On the other end of the spectrum, the average Asian resident was exposed to a neighborhood diversity level 20 points higher (E_e = 62.5) than the overall Metro-100 average.⁶

Despite racial differences in diversity exposure, members of all groups found themselves in increasingly diverse tracts over time. Whites experienced the largest absolute and relative gains (21 percent) in diversity exposure, and the share of whites residing in the most homogeneous tracts (i.e., the base of the white diversity histogram) declined drastically between 1990 and 2000. Blacks followed suit with a 15 percent increase in diversity exposure over the decade. Latinos and Asians experienced more modest increases, perhaps due to a ceiling effect because the two groups had such high initial levels of diversity exposure. These findings are contrary to the predictions of the bifurcation model, though whites continue to experience the lowest exposure to diversity. The fragmentation model is also unsupported given the high and increasing levels of diversity exposure for black, Latino, and Asiann residents. Once again, diversity trends are most consistent with demographic integration hypotheses.

NEIGHBORHOODS IN TRANSITION

Our next step is to address the changing racial structure underlying shifts in neighborhood diversity. Using our majority-rule classification scheme, we examine changes in neighborhoods with white, black, Latino, Asian, and no majorities. In 2000, 68 percent of the Metro-100 tracts had white majorities, 12 percent had black majorities, 9 percent had Latino majorities, one percent had Asian majorities, and 10 percent had no majorities. Below we emphasize national trends but devote passing attention to regional patterns.⁷

White-Majority Neighborhoods

Table 3 is a transition matrix for all tracts with white majorities in 1990 and 2000. Similar to Table 2, the row headings identify 1990 categories, the column headings identify 2000 categories, and the diagonal represents the proportion of tracts experiencing no change in racial structure. We also include a column reporting the median percent change in the entropy indices and a row reporting the 2000 mean entropy indices. One of the most striking results can be found in the margins of the table, which provide decade-specific tract counts for each category. There was a net loss of more than 5,000 white-dominant tracts over the decade, and the median white-dominant tract experienced a 57 percent increase in its entropy score. No doubt a floor effect is operating; these tracts have little room to become less diverse. Still, it is clear that tracts with the largest white majorities have become more diverse. Regional analyses (not shown) indicate coast-to-coast declines in white-dominant tracts. The southern, midwestern, and northeastern regions each experienced absolute losses of more than 1,000 such tracts, and there was a 60 percent decline in the western region. By 2000, less than one in ten western tracts could be classified as white-dominant. All told, 95 of 96 eligible metropolitan areas (i.e., those with any white-dominant tracts in 1990) experienced declines during the decade.

Table 3.

Transitions in White-Majority Tracts, 1990–2000^a

		2000 racial structure									E % changeb	N of tracts
		Dominant	Shared	W-B	W-L	W-A	W-B-L	W-B-A	W-L-A	No W maj.
1990 racial structure	Dominant	62.3%	26.8	3.9	4.6	1.4					57.3	14,788
	Shared	2.1	38.0%	10.9	25.0	10.8	5.2		3.8	2.6	34.0	4,751
	W-B		5.6	54.7%	1.1		8.0	1.2		27.4	22.7	3,596
	W-L		3.3		54.1%		3.9		4.9	32.6	19.2	3,688
	W-A		3.4			61.8%		2.1	12.5	17.0	17.2	893
	W-B-L				5.9		16.3%			73.5	11.7	680
	W-B-A							20.8%		56.8	11.1	125
	W-L-A				3.6	3.6			23.8%	68.1	12.4	639
Mean E (2000)		14.7	33.7	52.2	53.8	53.8	71.9	75.0	70.5	59.0
N of tracts		9,356	6,135	3,126	4,008	1,343	873	144	670	12,193

^aCell percentages representing less than 1 percent of category or fewer than 10 tracts are omitted. W-B-L-A tracts are omitted due to small Ns and empty cells. There were 29,259 white-majority tracts in 1990; in 2000, there were 25,758 white-majority tracts.

^bMedian percent change in tract diversity.

Despite these widespread shifts, the diagonal in Table 3 indicates that, among white-majority tracts, white-dominant ones were most likely to remain stable in their racial structure. This stable white homogeneity—even in metropolitan areas with relatively diverse racial compositions—is not consistent with the integration perspective. White-dominant neighborhoods that did change were likely to drop into the white-shared category, which still includes (by definition) fairly homogeneous tracts. White-shared neighborhoods were not particularly stable but, rather than moving toward less diverse white-dominant structures (as predicted by the bifurcation and fragmentation models), they moved toward more diverse dual- or multi-group structures that usually included Latinos (in line with demographic integration).

There is some evidence of stability among white neighborhoods with sizeable outgroup populations, in accord with demographic integration. This was most prominent for white-Asian tracts, three-fifths of which persisted between 1990 and 2000. Over half of the white-black and white-Latino tracts also persisted in their dual group classification but they lost their white majorities at a higher rate. For example, a regional breakdown reveals that white-Latino stability occurred in just one-third of eligible northeastern tracts; in fact, these tracts were more likely to lose their white majorities than to achieve stability. Finally, while white-majority tracts with multiple outgroups increased in number, the cell percentages in Table 3 indicate that there was very little stability in these tracts. For example, nearly three-quarters of white-black-Latino tracts had lost their white majorities by 2000, and a similar trend is apparent for other multi-group combinations. It does not appear that these tracts are rapidly turning into Latino-majority neighborhoods, as the fragmentation perspective predicts. The results of a separate analysis (not shown in Table 3) support the bifurcation model by documenting that nearly two-thirds of these diverse white-majority neighborhoods transitioned to no-majority status in 2000.

Black-Majority Neighborhoods

Like their predominantly white counterparts, black-dominant tracts are the modal black-majority type according to Table 4. Regional analyses reveal that better than eight in ten of these tracts are located in midwestern and southern metros, which often have large black populations and high levels of residential segregation. Overall, the number of Metro-100 black-dominant neighborhoods declined from 1990 to 2000 but there was a strong tendency toward stability, consistent with the fragmentation perspective. Black-shared tracts exhibit a noticeable fragmentation trend in the table, as 30 percent of them moved toward less diverse black-dominant racial structures (see also the median percent decline in diversity). Compare this to the small proportion of white-shared tracts that shifted to white-dominant status (Table 3).

Table 4.

Transitions in Black-Majority Tracts, 1990–2000^a

		2000 racial structure									E % changeb	N of tracts
		Dominant	Shared	B-W	B-L	B-A	B-W-L	B-W-A	B-L-A	No B maj.
1990 racial structure	Dominant	80.7%	12.3	1.6	5.1						19.1	1,820
	Shared	29.6	35.4%	4.6	25.0						−5.3	280
	B-W	9.6	15.6	58.6%	1.9		6.5			6.3	−6.6	1,332
	B-L		5.3		61.6%					29.1	3.1	505
	B-A										7.3	17
	B-W-L				35.3		28.6%			30.4	−0.3	224
	B-W-A										4.3	19
	B-L-A										−3.0	9
Mean E (2000)		11.0	29.2	51.8	50.9	54.3	70.1	72.7	72.2	41.8
N of tracts		1,691	577	1,334	608	26	272	37	10	33,384

^aCell percentages representing less than 1 percent of category or fewer than 10 tracts are omitted. B-W-L-A tracts are omitted due to small Ns and empty cells. There were 4,213 black-majority tracts in 1990; in 2000, there were 4,567 black-majority tracts.

^bMedian percent change in tract diversity.

Nearly three-fifths of all black-white tracts—the second most common class of black neighborhood in 1990—remained stable over the decade, consistent with demographic integration. However, it is notable that the typical (median) black-white tract experienced an entropy decline of more than six percent between 1990 and 2000. The latter finding implies white population loss, so prospects for the future stability of black-white racial structures seem tenuous. Black-Latino stability was also common overall, though we detected regional distinctions in stability levels. In general, northeastern black-Latino tracts demonstrated much more stability than those in the western region. Western black-Latino tracts tended to lose their black majorities, and Latino majorities emerged in their wake. There are simply too few black-Asian tracts for meaningful analysis.

The only multi-group combination of consequence in Table 4 includes a white and Latino presence. These black-white-Latino tracts illustrate the many pathways of change possible in a multi-group context. Over one-quarter of such tracts retained their multi-group classification, about 12 percentage points higher than their white counterparts (W-B-L) in Table 3. Yet nearly two-thirds either lost their white presence and shifted to a black-Latino structure (bifurcation) or lost their black majorities in conjunction with Latino growth and white decline (fragmentation). Both types of change were concentrated in southern and northeastern metros.

Latino-Majority and Asian-Majority Neighborhoods

According to Table 5, the number of Latino-majority neighborhoods grew by more than 50 percent during the 1990s. Once a tract had a Latino majority, it was highly unlikely to relinquish it. Of 2,193 tracts with Latino majorities in 1990, just 74 (3.4 percent) had lost them by 2000.⁸ Latino-dominant tracts increased by nearly half, and the number of Latino-shared tracts doubled. The great majority of Latino-dominant tracts remained stable, and two-fifths of Latino-shared tracts made the transition to Latino dominance. More than 90 percent of the dominant and shared tracts are located in the South and West. Declining diversity was common during the decade, with double-digit percent declines in entropy indices across nearly every type of Latino-majority tract. This stands in stark contrast to the national trend for all tracts and hints at an increasingly fragmented pattern of Latino settlement.

Table 5.

Transitions in Latino-Majority Tracts, 1990–2000^a

		2000 racial structure									E % changeb	N of tracts
		Dominant	Shared	L-W	L-B	L-A	L-W-B	L-W-A	L-B-A	No L maj.
1990 racial structure	Dominant	92.3%	4.7								−12.6	297
	Shared	40.9	42.2%			7.1					−20.7	154
	L-W	9.3	16.3	65.9%		1.1	1.5	1.8		3.1	−16.4	968
	L-B		7.7		84.6%						−9.0	338
	L-A					80.0%					−12.1	65
	L-W-B		10.0	18.9	29.5		30.0%			9.5	−16.2	190
	L-W-A		12.2	17.3		31.4		26.9%		7.7	−14.2	156
	L-B-A										−16.7	13
Mean E (2000)		16.7	34.6	55.9	54.0	56.1	71.3	73.8	76.3	39.1
N of tracts		441	317	1,396	531	154	309	186	47	34,560

^aCell percentages representing less than 1 percent of category or fewer than 10 tracts are omitted. L-W-B-A tracts are omitted due to small Ns and empty cells. There were 2,193 Latino-majority tracts in 1990; in 2000, there were 3,391 Latino-majority tracts.

^bMedian percent change in tract diversity.

The decade also witnessed the emergence of new dual-group Latino neighborhoods, as documented in the margins of Table 5. The number of Latino-white and Latino-black tracts grew by 44 percent (from 968 to 1,396) and 57 percent (from 338 to 531) respectively and the number of Latino-Asian tracts more than doubled (from 65 to 154). These emergent tracts were disproportionately—and almost exclusively in the case of new Latino-Asian tracts—located in the western region. Among established dual-group Latino-majority tracts (i.e., those with a dual designation in 1990), Latino-white tracts experienced impressive stability (66 percent), but this pales in comparison to the Latino-black (85 percent) and Latino-Asian (80 percent) levels of persistence. The multi-group transitions tell a similar story. Multi-group structures initially including whites (i.e., L-W-A, L-W-B) trended toward a diminished white presence, supporting the hypothesis from the bifurcation model.

A pair of limitations must be kept in mind when considering Asian-majority tracts. Most significantly, few such places existed in either 1990 (230) or 2000 (367), reflecting both the modest size of metropolitan Asian populations and their lower levels of segregation relative to blacks and Latinos.⁹ In addition, Asian-majority tracts are disproportionately concentrated in Honolulu, which accounted for 76 percent of them in 1990 and 65 percent in 2000. Despite these limitations, three trends deserve mention. First, there is a dearth of Asian-dominant tracts (just four in 1990 and three in 2000). Second, the 130 neighborhoods classified as Asian-white in 1990 experienced remarkable persistence (70 percent) over the subsequent decade. Finally, the presence of Latinos alongside whites resulted in much less stability, as just 10 of 41 Asian-white-Latino tracts retained a white presence. Thus, the evidence points to integration in the two-group case but to bifurcation in the multi-group case.

No-Majority Neighborhoods

Neighborhoods without a majority racial group are particularly pertinent because our three theoretical perspectives hypothesize very different prospects for them. The bifurcation perspective predicts whites to be increasingly absent from these neighborhoods while the demographic integration perspective predicts a visible white presence to remain in the face of diversity (and to contribute to it). The fragmentation perspective does not hold out much hope for stable diversity, even among minority groups. It predicts that diverse neighborhoods will move toward group-majority structures. The number of no-majority tracts has increased substantially. In 1990, there were 2,056 no-majority tracts, or about one of every 20 tracts in our sample. By 2000, there were 3,868 no-majority tracts, constituting one of every ten Metro-100 tracts. No-majority tracts were concentrated in the western region, though southern and midwestern metros also registered large increases in such tracts.

The transitions of diverse neighborhoods vary according to their initial racial structure. Almost 90 percent of the newly emerging no-majority neighborhoods (N = 2,823) originated as white-majority neighborhoods, a trend consistent with bifurcation. Many became white-no-majority neighborhoods, which were a solid plurality of all no-majority neighborhoods in 1990 (N = 963) and 2000 (N = 1,805). The high levels of instability in these white-no-majority tracts are in line with the predictions of the bifurcation and fragmentation models (see Figure 3). Only 17 percent of the neighborhoods classified as white-no-majority in 1990 persisted in that category ten years later while nearly 44 percent transitioned to a nonwhite plurality (bifurcation) and another third took on nonwhite majorities (fragmentation). Further evidence of fragmentation is found in no-majority neighborhoods with black, Latino, or Asian pluralities in 1990. Over half of these tracts (56 percent) transitioned to a BLA majority by 2000. Group-specific analyses indicate that such transitions were driven by Latino growth. On the whole, roughly one-third of no-majority neighborhoods in 1990 took on Latino majorities by 2000.

Figure 3.

2000 Racial Structures of 1990 No-Majority Tracts.

Population Distributions across Neighborhood Types

Table 6 shows how white, black, Latino, and Asian residents are distributed across neighborhood types with varying racial structures. Over one-third of whites still resided in the most homogeneous white-dominant neighborhoods in 2000, but this represents a substantial decline since 1990 and lends credence to the demographic integration perspective. Nevertheless, much of this redistribution was absorbed by relatively homogeneous white-shared neighborhoods and the proportion of the white population found in highly diverse multi-group neighborhoods remained quite small in 2000. Consistent with the entropy exposure indices presented in Figure 3, it appears that most of the diversification experienced by whites has been modest in magnitude.

Table 6.

Distribution (%) of Racial Groups Across Tract Types, 1990–2000

	White		Black		Latino		Asian
	1990	2000	1990	2000	1990	2000	1990	2000
Population (1000s)	105,985	106,525	20,345	23,571	17,047	27,010	5,892	8,767
% of total	70.6	62.5	13.6	13.8	11.4	15.9	3.9	5.1
Tract type:
Dominanta	51.8	36.4	28.8	21.4	7.7	8.1	0.1	0.1
Shareda	15.7	22.6	4.2	7.0	3.5	5.2	1.5	1.9
Two-groupb	24.9	29.0	43.3	40.7	48.4	46.0	34.4	35.0
White presence	24.7	28.8	34.1	30.6	38.5	35.6	31.6	30.8
White absence	0.1	0.2	9.2	10.1	9.9	10.4	2.8	4.2
Multi-groupc	7.3	11.6	15.9	24.0	28.2	32.1	38.5	44.1
White presence	7.3	11.6	15.0	22.4	27.1	30.2	37.0	41.5
White absence	0.0	0.1	0.9	1.6	1.1	1.9	1.5	2.6

^aGroup-specific shares are limited to those residing in tracts with own-group majorities.

^bGroup-specific shares residing in any group-majority tract with a single outgroup present.

^cGroup-specific shares residing in and group-majority tracts with three or more groups present and no-majority tracts.

Blacks are more likely to reside in group-dominant tracts than other nonwhite groups, but the trend is solidly in the direction of declining residential fragmentation. In fact, by 2000 more black residents lived in multi-group tracts than in black-dominant tracts. While Latinos were more likely to be found in Latino-dominant tracts in 2000 than 1990, this fragmentation trend should not be overemphasized. By 2000 four times as many Latinos inhabited multi-group tracts (32.1 percent) than Latino-dominant tracts (8.1 percent). The overwhelming majority of Asians live outside of homogeneous Asian-dominant or Asian-shared tracts.¹⁰ Consistent with our earlier analyses, by 2000 more than four in ten Metro-100 Asians could be found in a multi-group or no-majority tract. These results run counter to fragmentation predictions and are more consistent with demographic integration. Finally, black, Latino, and Asian residents increasingly find themselves in neighborhoods with a white presence, contrary to the bifurcation perspective. While minority shares declined in dual-group tracts with whites present, this was offset by their increasing shares in multi-group tracts. By 2000 over half of blacks (53 percent), nearly two-thirds of Latinos (65.8 percent), and seven of every ten Asians (72.3 percent) were found in dual or multi-group tracts alongside whites.

DIVERSITY AND GROUP-SPECIFIC POPULATION DYNAMICS

A variety of population dynamics might account for the transitions discussed above. Table 7 displays entropy indices and racial compositions for tracts exhibiting changes congruent with each of the three theoretical perspectives. We define bifurcating neighborhoods as those undergoing 1990–2000 white population decreases in the face of increasing multi-group diversity (as measured by E). A large plurality of tracts went through the racial changes hypothesized by bifurcation. These neighborhoods both diversified and lost white residents at nearly twice the rate of the total Metro-100 sample. Black, Latino, and Asian populations grew rapidly, but the mean racial characteristics in the right-hand portion of the table indicate that bifurcating neighborhoods were still characterized by substantial white majorities in 2000. The aggregate consequences of white population decline were far from trivial, however. Neighborhoods experiencing bifurcation lost nearly 7.7 million white residents during a decade in which the overall Metro-100 white population remained stable. Additional analyses (not shown) reveal that bifurcating tracts were evenly distributed across regions, though seven of the ten metropolitan areas with the highest prevalence of bifurcation were located in the Northeast.

Table 7.

Changing Diversity and Racial Composition in Bifurcating, Fragmenting, and Integrating Metropolitan Tracts

	# of tracts	% of tracts	Mean diversity (E)			Median % change, 1990–2000				Mean 2000 composition (%)
			1990	2000	% change	White	Black	Latino	Asian	White	Black	Latino	Asian
All tracts	37,951	100.0	35.0	42.0	20.0	−6.7	20.7	68.8	39.5	61.9	16.0	14.6	4.9
Bifurcating	15,985	42.1	35.0	48.8	39.6	−12.9	49.4	91.3	51.9	66.4	13.2	12.0	5.4
Fragmenting
Black	560	1.5	40.8	29.5	−27.9	−42.6	23.5	−43.1	−72.1	36.7	51.5	7.0	2.3
Latino	1,231	3.2	47.9	37.3	−22.1	−38.1	−29.1	37.2	−54.3	25.2	24.2	45.1	3.1
Asian	473	1.3	36.7	30.0	−18.3	−32.9	−30.7	−44.2	50.6	39.0	33.6	10.4	14.1
Integrating	10,147	26.7	22.9	33.7	47.1	19.7	63.6	144.9	118.2	75.1	11.9	7.7	3.2
Other
White growth	3,831	10.1	33.6	30.2	−10.1	14.9	−17.7	6.0	−6.7	72.9	10.0	12.1	2.9
White decline	5,724	15.1	53.8	48.8	−9.4	−34.2	−3.9	20.8	−0.3	31.2	28.0	30.3	7.5

The population dynamics predicted by the fragmentation perspective operate in a small number of metropolitan neighborhoods. In the second panel of Table 7, we focus on a “pure” form of fragmentation in which neighborhood diversity declines as a single group grows in size. Fragmentation was most likely to be associated with Latino growth, and twice as many tracts experienced Latino as black fragmentation. Regionally, Latino fragmentation was concentrated in western metropolises, although 96 of the 100 sample metro areas had at least one such tract. There are a number of Asian fragmentation neighborhoods as well, but the small Asian representation in these tracts suggests that low and decreasing levels of diversity are primarily the result of other group dynamics.

We find ample evidence of demographic integration. Over 10,000 tracts fit the integration model during the decade, accounting for more than one-quarter of all tracts. Unlike bifurcating neighborhoods, these integrating neighborhoods retained their white populations while increasing in diversity. However, they remained much less diverse than their bifurcating counterparts. In fact, by 2000 the mean diversity score for integrating neighborhoods (E = 33.7) was still one point lower than the score for bifurcating neighborhoods (E = 35.0) a decade earlier. Mean racial composition bears this out: white residents constituted three-quarters of the 2000 population in an average integrating neighborhood but just two-thirds in the average bifurcating neighborhood. Perhaps, in line with bifurcation, these integrating neighborhoods have not yet reached the tipping point at which whites feel uncomfortable with rising diversity.

One-quarter of the sample tracts do not fit neatly into our typology of neighborhood change. We display results for these tracts at the bottom of the table, broken out by whether their white populations grew or declined. While the fragmentation perspective is focused on nonwhite neighborhood transitions, a sizeable subset of the unclassified tracts experienced declining diversity in conjunction with white growth. Like bifurcation, this dynamic is distributed quite evenly across regions. Overall, these low-diversity neighborhoods registered some modest Latino growth in conjunction with white gains. They nevertheless appeared to be bucking the trend toward greater diversity.

A larger subset of unclassified neighborhoods experienced white population decline, yielding a profile very different from that of the white-gain group. These high-diversity neighborhoods did not take a pure form of group-specific fragmentation, but they did exhibit notable diversity declines. Their dynamics illustrate the complexity involved in assessing change in a multi-group context. A neighborhood might become less diverse even as multiple groups are growing in size if one group increases so rapidly that it outpaces the others. In the case of the unclassified neighborhoods with white losses, this usually entails Latino growth accompanied by the growth of at least one other group, though non-Latino loss appears more typical than gain. As with Latino fragmentation, the white decline type of tract transition occurred disproportionately in the western region.

CONCLUSION

Methodologically, our research demonstrates the value of measuring the overall multi-group diversity of neighborhoods in concert with a majority-rule approach to the classification of their racial structures. Diversity is often characterized in a vague or ad hoc manner, a problem that we address by clearly defining and quantifying the concept. We have also illustrated in objective terms the changing localized circumstances to which metropolitan residents of varying racial identities are exposed. Future work should augment our efforts by covering greater temporal ground, given that processes of neighborhood transition may take multiple decades to play out (Friedman 2007). Ideally, an expanded temporal horizon will be accompanied by increased attention to the spatial nuances of localized change occurring at smaller and larger scales than the typical census tract (Lee et al. 2008). Disaggregating mean patterns at the metropolitan level will allow researchers to determine whether neighborhood bifurcation, fragmentation, or integration predominates in specific metropolitan areas (see, e.g., Farrell 2008). Finally, a more daunting task will be to disentangle regional differences along with the social and economic forces that contribute to or inhibit neighborhood change. A recent study by Logan and Zhang (2010) takes a step in that direction.

Despite rather limited aims, we do manage to generate valuable insights into neighborhood racial transitions in the metropolitan United States. Returning to the summary predictions provided in Table 1, the three models of neighborhood change receive varying levels of support. Our analysis points to three broad generalizations that span the bifurcation, fragmentation, and demographic integration perspectives. First, white residential patterns are particularly sensitive to the presence of multiple minority groups, a fact that lends credence to the bifurcation scenario. Most white neighborhoods with multi-group structures lost their white majorities between 1990 and 2000, and we uncovered substantial white population losses even in the presence of modest diversity. In a concomitant trend, white population gains are occurring in low-diversity white neighborhoods. This latter finding is worthy of further study because it suggests a reshuffling of the white metropolitan population, ensuring that many whites will remain residentially removed from anything more than token diversity.

Our second generalization is that Latino population dynamics have emerged as a primary force driving neighborhood change. Diversity declines were the typical trend in Latino neighborhoods and, though fragmentation proves rare in its “pure” form, the number of predominantly Latino neighborhoods was on the rise during a decade in which black neighborhood homogeneity eroded. However, the Latino population also plays a prominent role in increasing neighborhood diversity, as substantial and growing numbers of Latinos—often found alongside Asians neighbors—can be found in some of the most diverse neighborhoods in the United States. On the whole, it appears that changing Latino residential patterns are blurring existing color lines more often than establishing new ones. In fact, our analysis likely underestimates the tendency toward Latino residential diversity since it has not taken into account the ethnoracial distinctions within this varied group.

Finally, many of our findings are in accord with demographic integration. Most metropolitan neighborhoods are diversifying and the typical metropolitan resident is experiencing this diversification in some fashion. Moreover, multi-group and no-majority residential settings have become much more common as the number of predominantly white neighborhoods has dropped sharply. In this sense, we are “living together” and “sharing neighborhoods” more now than in the past (Ellen 2000; Fasenfest et al. 2006). Such a shift could be a harbinger of fundamental change, helping to narrow existing disparities, to bridge racialized polities, and to stimulate constructive interracial interaction and transracial community involvement. However, any optimism about the future must be tempered by stressing that the picture is very complicated. Because much of the growth in diversity occurs on the margins, profound, inclusive (i.e., including whites), and stable multi-group integration remains in a fragile state. The incremental changes documented here indicate a shifting color line but do not yet foretell a time in which it will be erased.