Continuity and Change in Neighborhood Racial/Ethnic Transitions: 1980–2020

Abstract

Since the 1980s, several studies have reported a decline in all-White neighborhoods and a rising number of racially mixed neighborhoods, including what have been called multiethnic “global” neighborhoods. Previous research has shown that these changes between 1980 and 2010 partly reflected the rapidly rising shares of Hispanics and Asians in urban areas. However, they also showed that there had been a substantial change in the pattern of settlement, resulting in many transitions to greater diversity than could have been expected from this demographic shift. We update their analysis to 2020, comparing transitions in the 1980–2000 period to those in 2000–2020, to test whether the earlier observed trends have continued, intensified, or weakened. We also quantify the impact of residential changes on the numbers of persons in each racial/ethnic group who live in each type of neighborhood.

1. Introduction

This study addresses how the rapid growth of America’s Hispanic and Asian populations has affected patterns of segregation and neighborhood racial change in urban areas. From a nearly exclusive concern with Black–White segregation, researchers are increasingly paying attention to a more complex multiethnic metropolis. Growing diversity is becoming a core theme of research (Frey 2018; Lichter 2013). An early hypothesis was that the historic separation of Blacks from Whites would be undermined by the presence of Hispanic and Asians, who more easily find homes in White neighborhoods and whose presence as a sort of buffer may lower the barriers to subsequent Black entry (Farley and Frey 1994; see also Denton and Massey 1991; Ellen 1998; Fong and Shibuya 2005). Logan and Zhang (2010) reframed this idea around the concept of global neighborhoods. They posited that the well-known patterns of neighborhood invasion and succession, which supported near-apartheid levels of segregation, was being replaced by a new pattern that would make possible racially mixed—global—neighborhoods with a substantial and stable presence of Whites along with Blacks, Hispanics, and Asians.

In this study, we update the trajectory of changes in neighborhood composition through to 2020. We also examine to what extent changes in neighborhood composition may be the direct consequence of the continuing growth of non-White, mainly Hispanic and Asian, populations in metropolitan areas, or alternatively to changes in locational processes that may have begun to increase the options available to non-White groups. The development of neighborhoods with a greater mix of residents from different racial/ethnic groups, which has been widely observed, could be a direct consequence of the changing composition of residents in the whole metropolitan region. As Zhang and Logan (2016) pointed out, this demographic shift could have a major impact on diversity. Because Hispanic and Asian populations continue to grow quickly, while the White population is stable or declining, diversity in local neighborhoods would necessarily rise even if their growing numbers were distributed among neighborhoods in the same way as in the past. Increasing diversity in that scenario would be compatible with no change in segregation among groups as usually measured (i.e., by the Index of Dissimilarity). For this reason, in order to interpret why diversity is changing and what it tells us about intergroup relations, it is necessary to distinguish between changes attributable to overall compositional change and changes that reflect new processes of inclusion or exclusion, a process shift.

In the following sections we describe the scale and regional location of changes in the racial/ethnic composition of metropolitan populations, then describe a methodology to evaluate its impact on neighborhood racial/ethnic transitions. We present results for all metropolitan regions in the nation for 1980–2000, then for 2000–2020 in order to assess whether the patterns of change are continuing into the present time. We find that the overall demographic shift accounts for almost all of the decline in the number of all-White neighborhoods, many of which added Hispanic and/or Asian residents as a result of the metropolitan-level growth of those groups. These groups were also increasingly included in what had been White–Black neighborhoods, resulting in what have been referred to as “global” and “semi-global” neighborhoods. At the same time, it appears that a process shift has opened opportunities for Blacks to live in neighborhoods from which they previously had been absent.

1.1. Past Research on Changing Composition

This project builds on a series of studies that seek to understand how overall changes in segregation, such as the slow but persistent decline in Black–White segregation, are associated with patterns of change at the neighborhood scale. Rather than measure changes in segregation directly, they have assessed changes in neighborhoods’ racial/ethnic composition that represent increasing diversity. Early examples are Lee and Wood (1991), Denton and Massey (1991), and Alba et al. (1995). These studies all created categories for the racial/ethnic composition of census tracts, adding Hispanics and Asians to the more commonly studied Blacks and Whites. They based their analyses of neighborhood change on tracts’ movements across categories of composition, summarized in the now-familiar form of a transition matrix. An important observation was the reduction in all-White neighborhoods along with the growth in the number of neighborhoods with more complex combinations of residents, which was taken as a sign of declining segregation. More recently, Ellen et al. (2012) updated these findings to 2010, showing that the share of neighborhoods that they identified as “integrated” (i.e., racially mixed) had increased from 20% to 30% in the 1990–2010 period. This was possible because most mixed neighborhoods remained mixed, while many single-race neighborhoods added a substantial presence of one or more other groups. Wright et al. (2020) focused on what they considered to be “highly diverse” tracts in 2010, showing that most of them had initially been all-White in 1990. Zhang and Logan (2016) studied the 30-year period from 1980 to 2010, showing that similar trends were occurring in metropolitan areas across the nation that had very different overall compositions. The predominant trend was a substantial decline in the number of all-White neighborhoods and a large increase in the number of what they described as “global neighborhoods,” which had a substantial presence of all four major racial/ethnic groups.

A further step has been taken by researchers who rely on measures of multigroup diversity (such as the entropy-based measure E) to reflect segregation. Several studies (Farrell and Lee 2011, 2016; Parisi et al. 2011, 2015; Hall et al. 2016; Kye and Halpern-Manners 2023) have examined transitions among categories of neighborhood or community racial/ethnic composition, and then reported how each was associated with increasing or decreasing neighborhood diversity. The more recent studies all report patterns through 2010.

1.2. Changing Metro Composition, 1980–2020

The current study describes the patterns of neighborhood change through another decade. Some researchers have hypothesized that diverse neighborhoods are inherently unstable, as Whites retain a potential for White flight (Kye and Halpern-Manners 2023). Indeed, Logan and Zhang (2010) and subsequent studies have pointed to the fact that there are countervailing trends in racial change, such that at the same time as all-White neighborhoods are dwindling and increasing shares of tracts can be identified as “global neighborhoods,” all-minority neighborhoods are persistent or rising. While there is evidence that very mixed neighborhoods can be stably integrated, Kye and Halpern-Manners (2023) conjecture that White flight could eventually lead to future patterns of the “balkanization” of neighborhoods. By comparing trends in the earlier period (1980–2000) to the more recent 2000–2020 period, we can assess to what extent this is beginning to occur.

Our larger purpose is to distinguish between trends in increasing diversity that are driven by overall changes in the heterogeneity of the population and trends that represent a shift in the residential opportunities of non-Whites. Of the studies cited above, only Zhang and Logan (2016) directly address this question. Yet it bears strongly on how to interpret the pattern of increasing diversity at the local scale as either a demographic shift or as a change in the residential choices and options for persons of different racial/ethnic groups.

To explain the context of this question, we report here how the composition of the metropolitan population has changed in the last forty years.

Table 1. Metropolitan population by race/ethnicity and region, 1980–2020.

		White		Black		Hispanic		Asian
	Population	Count (Millions)	Share	Count (Millions)	Share	Count (Millions)	Share	Count (Millions)	Share
National
1980	163.4	125.7	76.9%	21.2	13.0%	12.5	7.6%	3.1	1.9%
2000	206.4	133.8	64.8%	28.0	13.6%	30.1	14.6%	9.6	4.7%
2020	248.1	131.6	53.0%	33.6	13.5%	52.0	21.0%	18.7	7.5%
Northeast
1980	40.2	32.5	80.8%	4.5	11.2%	2.5	6.2%	0.5	1.3%
2000	43.3	30.2	69.7%	5.4	12.5%	4.9	11.2%	1.9	4.4%
2020	46.8	27.2	58.2%	5.8	12.3%	7.9	16.9%	3.7	7.9%
Midwest
1980	40.7	34.1	83.6%	5.0	12.3%	1.1	2.7%	0.3	0.8%
2000	45.3	34.8	76.7%	6.0	13.3%	2.6	5.8%	1.1	2.4%
2020	49.5	33.5	67.7%	6.6	13.3%	4.9	9.8%	2.2	4.4%
West
1980	46.3	32.5	70.3%	3.7	7.9%	7.7	16.7%	2.0	4.3%
2000	68.6	37.1	54.1%	5.0	7.3%	18.8	27.4%	5.5	8.0%
2020	89.2	37.9	42.4%	6.6	7.4%	30.1	33.7%	10.0	11.2%
South
1980	36.3	26.6	73.4%	8.1	22.3%	1.2	3.3%	0.3	0.8%
2000	49.2	31.8	64.6%	11.6	23.5%	3.8	7.7%	1.2	2.3%
2020	62.5	33.0	52.7%	14.7	23.4%	9.2	14.7%	2.8	4.4%

relies on published tract data for 1980, 2000, and 2020, counting residents as non-Hispanic White, Black, Hispanic, and non-Hispanic Asian, including all census tracts within the 2020 boundaries of census-defined metropolitan areas. The table reports national totals, and also totals for each of the four major regional divisions of the country.

The national totals show a very small increase in the total non-Hispanic White population in metropolitan areas, rising slightly from 1980 to 2000 before falling back by 2020. However, the White share of the population dropped precipitously, from 76.9% in 1980 to barely a majority (53.0%) in 2020. The Black total increased by about 50%, but its share remained steady at about 13%. In contrast, the Hispanic total grew by about 40 million, increasing its share from less than 8% to over 20%. The Asian total was much smaller than the other groups in 1980 (3.1 million), but it grew six-fold to 18.7 million. These are the changes that so radically shifted the relative balance among groups at the metro level, and which necessarily had impacts at the neighborhood level.

The patterns differ somewhat across regions. The most distinctive region is the West, which always had the largest Hispanic and Asian populations and population shares. This distinction continues today, despite the considerable growth of these groups in other regions. Their large Hispanic and Asian shares were matched by a much smaller share of Black residents than found in other regions. The South, in contrast, always had and still has the largest Black share of metropolitan residents. These regional differences reflect much greater variations among individual metropolitan areas. This is why, in assessing how the demographic shift affected neighborhood composition, we must develop estimates independently for each metro.

2. Research Design and Measurement

This analysis is based on all census tracts in 342 metropolitan areas as defined in 2010, which we treat as proxies for neighborhoods. Because one purpose is to examine how individual tracts changed their compositions across decades, it is necessary to use a data file in which tract boundaries remain constant over time. We use the Longitudinal Tract Data Base (LTDB) from Brown University, which provides estimates of census tract data for 1980, 2000, and 2020 within consistent 2010 boundaries. The 2000 estimates are known to be very close to the true values, because they were created from confidential individual-level records from the 2000 census, which have been reaggregated within 2010 boundaries. These estimates have been approved for public disclosure after the infusion of a small amount of random error. The 2020 estimates are based on interpolation from the publicly available 100% counts of persons in each racial/ethnic group at the block level, which have been shown to be a considerable improvement on previous estimates using tract-level data (Logan et al. 2024). The LTDB estimates for 1980 are based on a standard dasymetric population and areal interpolation approach using 1980 tract data (see Logan et al. 2014). They are unbiased but subject to error that is inherent in the interpolation procedure. In this application, the effect of estimation error is reduced because the distribution of group shares is highly polarized—most tracts have very few or a highly disproportionate share of Black, Hispanic, or Asian residents. In these cases, estimation error is unlikely to result in an incorrect coding of whether a group is “substantially present” in the tract. However, there are other cases where the group share is close to the cutoff point that we use, and in these cases any error could move the tract either up or down across the threshold level.

It is necessary to use data with consistent areal boundaries over time in order to conduct this longitudinal analysis. We use the LTDB in part because it makes our results for 1980 comparable to those reported by Zhang and Logan and many other studies. We also prefer it because it is the only option that does not rely on interpolation methods to estimate characteristics in 2000.

There are several other methodological choices that affect our results. These include how to define racial/ethnic groups for the analysis, how to establish categories of neighborhoods based on their racial/ethnic compositions, and how to distinguish between neighborhood changes attributable to the demographic shift versus process shift.

2.1. Racial/Ethnic Categories

Consistent with all past studies on neighborhood racial/ethnic transition, we classify people into four racial/ethnic categories. One is Hispanics (who can be persons of any race). Among persons who are not Hispanic, those who are White alone are treated as White. Those who are Black alone or Black in combination with any other race are treated as Black. Asians are those who are Asian alone or in combination with any other race except Black. Some studies of neighborhood diversity, e.g., Lee et al. (2014), where a single statistic (the entropy-based measure E) is used to summarize diversity, add a fifth category of “Native Americans and others.”

We note that some more specific national-origin categories, such as Dominican, Cuban, Vietnamese, or Afro-Caribbean, could be important to study in more detail. However, these groups tend to be found in large numbers in specific regions (e.g., Dominicans in New York and Cubans in parts of Florida). In a study presenting national averages, results that include the many different national-origin subgroups would not provide useful information on most of them. We should emphasize that limiting the analysis to four racial/ethnic categories does not imply that residential processes for Mexicans are the same as for Puerto Ricans, or for South Asians and Chinese. A robust body of literature has demonstrated the usefulness of treating each group independently, considering its background, socioeconomic position, and social preferences.

There are also limitations on the feasibility of defining neighborhood composition in terms of more than four groups. In a study based on a diversity measure, there is no practical limit on the number of groups; the statistic E can be easily calculated for any set of groups. In the literature on neighborhood transitions, however, every additional group adds more than one potential combination of groups who may be present in the neighborhood. With four groups there are potentially fifteen combinations, and the full transition matrix that needs to be interpreted has a rank of 15 × 15. Zhang and Logan reduce the number of combinations to seven that they consider to be meaningfully different. Farrell and Lee (2011) and Ellen et al. (2012), both also starting with four groups, reduce the number of combinations to nine.

2.2. Categories of Neighborhood Racial/Ethnic Composition

We apply a categorization of neighborhoods in terms of which groups are “substantially present” in a neighborhood. Such categorizations have a long history in research on White–Black segregation. Taeuber and Taeuber (1965), for example, identified neighborhoods in categories such as “established Negro” and “stable interracial” areas. Studies in the 1990s (Denton and Massey 1991; Alba et al. 1995) extended the analysis to consider combinations of White, Black, Hispanic, and Asian residents. Many different kinds of combinations have been employed in recent years. We follow Zhang and Logan (2016) to define seven combinations. Using their notation, where NW stands for non-White, W for White, B for Black, H for Hispanic, and A for Asian, these are as follows:

• NW: Non-White neighborhoods (no significant White presence);
• W: Predominantly White neighborhoods;
• + H/A: White–Hispanic/Asian neighborhoods;
• WHA: White–Hispanic–Asian neighborhoods;
• WB: White–Black neighborhoods (no significant immigrant group presence);
• WB + H/A: Semi-global neighborhoods (White, Black, and one immigrant group);
• WBHA: Global neighborhoods (all four groups present).

Another choice is what to consider to be a “substantial presence” of a group in a neighborhood. Ellen et al. (2012) apply a cutting point of 20% to every group. This method unfortunately means that Asians are rarely treated as present even if they are disproportionately represented in the tract, compared to other tracts in the same metropolitan area. Conversely, Hispanics in an area like Los Angeles, which is nearly 50% Hispanic, are almost always “present,” even if they are considerably under-represented. Parisi et al. (2015) avoid this result by comparing the group’s share in a tract to a standard that is specific to the group and the decade being considered. They consider a group in a central city location to be “present” if its share is equal to at least 25% of the group’s average share across all U.S. central cities in that decade. If in a suburb, the group must comprise at least 25% of the group’s average share across all suburban places. Because we do not study city and suburban neighborhoods separately here, we apply a metropolitan-level standard that was developed by Logan and Zhang (see Logan and Zhang 2010, pp. 1083–84, for a detailed rationale). Their “Quarter Rule” is based on multiple empirical trials to ensure that results were not dependent on the specific cutting points that were applied. It identifies a group as present if its share of the tract population is at least 25% of its average share in multiethnic metropolitan areas in the same decade. We use the composition of multiethnic metropolitan areas as a standard, recognizing that this makes the criterion for minority presence relatively high for metropolitan areas with low shares of some groups. However, setting a high bar means that the group’s presence could actually be seen as substantial in the national context. Another implication of the Quarter Rule is that the criterion changes from decade to decade. This means, for example, that as the Asian population grew in metropolitan areas after 1980, the criterion for Asian presence also rose, and as the non-Hispanic White share fell over time, the criterion for White presence was adjusted downward. By raising the threshold for Hispanic and Asian presence, and lowering it for Whites, this choice results in a modest understating of the extent of increasing diversity. The criterion for Hispanics rose from 3.6% in 1980 to 6.3% in 2000 and 7.8% in 2020, and for Asians from 0.7% to 1.9% and 3.0%, respectively. The criterion for Whites declined from 15.8% to 11.8% and 9.3% in these years. There was little change for Blacks: 4.7% in 1980 and 2000, and 4.4% in 2020.

2.3. Components of Change: Demographic and Process Shift

Distinguishing between demographic shift and process shift involves a standardization procedure. The logic (following Zhang and Logan 2016) is that if the actual observed change in a tract matches what would be expected if there were no change in locational processes, we can treat the change as “attributable” to the demographic shift. To the extent that the change exceeds that expected level, we infer that something must have changed in how groups are being sorted out across neighborhoods. In each of the two time periods (1980–2000 and 2000–2020) that we study here, we calculate an initial residential pattern (measured in terms of the distribution of neighborhoods classified by their racial/ethnic composition). Second, we project an “expected” residential pattern at the end of the period, assuming that the members of every group (White, Black, Hispanic, and Asian) were distributed among neighborhoods in the same proportion as at the initial time. In other words, the difference between the initial and expected patterns is an indicator of changes that could be anticipated solely from the differential rates of metropolitan-level population change for each group. We will refer to this as a “demographic shift.” We then compare the “expected” pattern to the “observed” outcome at the end of the period. The difference between the two can be attributed to shifts in locational processes of separation and inclusion among groups, and we refer to it as a “process shift.” This distinction is analogous to one that is familiar in the occupational mobility literature, where upward mobility attributable to a growing share of persons in higher-level occupations is referred to as “structural mobility,” while upward mobility measured while holding constant the occupational distribution is referred to as “exchange mobility.”

Using this approach, we compare residential changes in the 1980–2000 period to changes in the subsequent 2000–2020 decades. In each period, how much of the observed change is due to changing demographic compositions or to changing locational processes? Our focus is on specific kinds of changes that are substantively meaningful. For example, a key point in the neighborhood diversity literature is that alongside the emergence of more diverse global neighborhoods, White flight continued to replenish the stock of all-minority neighborhoods. Is either of these a persistent or growing phenomenon in the last twenty years compared to the 1980s and 1990s? That is, are neighborhoods without White residents becoming a more permanent feature of the metropolis, an “absorbing state” with little likelihood of ever including a White presence? Or, alternatively, is the trend to reinforce global neighborhoods, suggesting that Whites may “stop leaving mixed neighborhoods, when the experience of growing up in an all-white neighborhood becomes so rare as to change the dynamics of white residential choice?” (Zhang and Logan 2016, p. 1952). Could mixed neighborhoods become a real option for White mobility? Such questions are necessarily speculative, but comparing changes in 1980–2000 with more recent changes may be informative about the current trajectory.

We apply these neighborhood categories in two ways in this analysis, replicating past studies. First, we report how neighborhoods transitioned from the beginning to the end of each period, changing their distributions across types. Second, we report changes in how each racial/ethnic group was distributed across types of neighborhoods. In both analyses we compare expected to observed outcomes, and we compare changes in 1980–2000 with changes in 2000–2020.

3. Results

We begin with a summary of how the number of neighborhoods of each type evolved from 1980 to 2020, which emphasizes the magnitude of the changes that are being experienced in metropolitan areas. Figure 1 confirms the decline in all-White neighborhoods, from 29.3% of all metropolitan tracts in 1980 to 20.9% in 2000 and only 11.2% in 2020. There was a comparable fall in the share of metropolitan population in such tracts, which housed 29.7% of metropolitan residents in 1980 and only 10.3% by 2020 (Table 4 below). The pace of decline could be described as steady from one perspective, because the actual number of White neighborhoods dropped by about 5000 in each twenty-year interval. From another perspective, it quickened, because the decline was about one-third from 1980 to 2000, and it rose to about one-half in the last twenty years. We would describe the loss of White neighborhoods as quickening.

The figure also confirms a jump in the share of global neighborhoods, which doubled from 1980 to 2000 (7.0% to 13.9%), and then nearly doubled again from 2000 to 2020 (ending at 24.2%), now the most numerous type of neighborhood. At the other end of the diversity spectrum, the figure shows nearly a doubling in the share of all-minority (NW) tracts, from 6.8% to 11.4% of all tracts. The increase was larger in the first time period, but continued into the second.

We now consider to what extent these large changes in neighborhood composition could be explained solely through the demographic shift, and how much change occurred beyond what was “expected.”

Table 2. Neighborhood types: observed in 1980 vs. expected in 2000 and expected in 2000 vs. observed in 2000.

	2000 Expected
	NW	W	W + H/A	WHA	WB	WB + H/A	WBHA	Total
1980 observed
NW	3655	1	10	7	69	14	5	3761
W	0	11,897	3206	477	449	194	30	16,253
W + H/A	114	199	9108	2128	9	523	145	12,226
WHA	202	2	322	7794	0	12	415	8747
WB	9	19	1	0	4780	1047	117	5973
WB + H/A	78	0	36	4	141	3465	767	4491
WBHA	125	0	5	188	4	120	3412	3854
Total	4183	12,118	12,688	10,598	5452	5375	4891	55,305
	2000 Observed
	NW	W	W + H/A	WHA	WB	WB + H/A	WBHA	Total
2000 expected
NW	3852	1	74	34	59	75	88	4183
W	2	8574	1498	201	1045	635	163	12,118
W + H/A	209	2168	6320	1171	299	1708	813	12,688
WHA	360	193	1858	4909	21	491	2766	10,598
WB	350	465	88	5	3449	980	115	5452
WB + H/A	383	140	349	34	862	2843	764	5375
WBHA	594	13	145	279	74	780	3006	4891
Total	5750	11,554	10,332	6633	5809	7512	7715	55,305

reports two transition matrices. The upper panel refers to transitions from tracts’ categories in 1980 (observed) and what their 2000 categories would have been if new populations had been allocated to tracts in the same way as groups had been in 1980 (expected). This is what Zhang and Logan (2016) referred to as the “standard” matrix. Most tracts fall along the main diagonal (colored). They may have changed their compositions to some extent, but they would have remained in the same categories. There are not many below the diagonal. The cells with more cases above the diagonal, and those with the largest numbers of tracts, are neighborhoods that had a White presence in 1980 (possibly including Blacks) but were expected to add Hispanics and/or Asians by 2000. The most common examples were from W to W + H/A (n = 3206), from W+HA to WHA (n = 2128), or from WB to WB + H/A (n = 1047) or WB + H/A to WBHA (767). One striking result is that White neighborhoods (W) were projected to drop from over 16,000 to just over 12,000. A much less common pattern, due to the modest growth of Black population in this period, added Blacks to neighborhoods that already had Hispanics and Asians, from WHA to WBHA (n = 415). In studies that rely on standard measures of neighborhood diversity, all of these changes would generally represent growing diversity.

The lower panel of Table 2 compares what was expected in 2000 with what was actually observed. We interpret the transitions in this panel as representative of a shift in locational processes, layered on top of the demographic shift. One theoretically important case is the decline of White (W) tracts. The observed number of White tracts was only modestly fewer than expected, 11,554 compared to the projected 12,118. This means that the dramatic decline in White tracts that has attracted much attention can be mostly attributed to the demographic shift. Although we do not have data on residential mobility, we suspect that it was not so much that Whites were leaving or avoiding moving to White neighborhoods, or that Hispanics and Asians were entering neighborhoods from which they had been excluded. Rather, Hispanics and Asians—simply by continuing to move into neighborhoods in the same proportions as was already the case in 1980—were growing enough to surpass our threshold levels for group presence. Another, very different, important case is the emergence of more global (WBHA) neighborhoods. Their number was expected to grow from 3854 to 4891, mostly because many WB + H/A tracts added Hispanics or Asians (there were 767 such cases). They actually doubled, reaching 7715 in 2000. We interpret the large difference between the projected gain and the observed one as an indication that Blacks were now more able to establish a presence in a new set of locations. As previously shown (Logan and Zhang 2010; Zhang and Logan 2016), the main pathway to this outcome was from WHA to WBHA (2766 more observed than expected). A third set of changes resulted in a large increase in the number of all-minority (NW) neighborhoods, from 3761 observed in 1980 to 5750 observed in 2000. Only a modest number of such shifts (422) were expected. This change represents what has been identified as a countertrend in the segregation literature, a change that manifests growing segregation in a period of generally increasing neighborhood diversity. It results from what could be called “White flight,” a loss of White presence in hundreds of tracts in each of several neighborhood types.

Now consider the transitions in the following two decades, shown in

Table 3. Neighborhood types: observed in 2000 vs. expected in 2020 and expected in 2020 vs. observed in 2020.

	2020 Expected
	NW	W	W + H/A	WHA	WB	WB + H/A	WBHA	Total
2000 observed
NW	5666	0	10	8	35	13	18	5750
W	0	7223	2718	407	637	443	126	11,554
W + H/A	124	1	6741	1828	2	1080	556	10,332
WHA	76	0	106	5452	0	1	998	6633
WB	47	0	1	0	3511	1852	398	5809
WB + H/A	150	0	30	2	8	4750	2572	7512
WBHA	186	0	7	157	0	70	7295	7715
Total	6249	7224	9613	7854	4193	8209	11,963	55,305
	2020 Observed
	NW	W	W + H/A	WHA	WB	WB + H/A	WBHA	Total
2020 expected
NW	5038	0	71	71	207	348	514	6249
W	0	5037	799	111	764	428	85	7224
W + H/A	143	803	4807	1362	225	1660	613	9613
WHA	183	52	765	4687	13	278	1876	7854
WB	174	232	32	4	2748	889	114	4193
WB + H/A	299	67	345	40	927	5086	1445	8209
WBHA	472	18	157	451	130	1996	8739	11,963
Total	6309	6209	6976	6726	5014	10,685	13,386	55,305

. The upper panel of the table reports the observed 2000 distributions and the distributions expected to result from the demographic shift. Because the growth of Hispanic and Asian populations continued unabated in 2000–2020, we would anticipate a continuation of the kinds of expected transitions seen in 1980–2000.What we found is: A declining number of White tracts and a growing number of more diverse types. The largest net gain is in the WBHA category (increasing from 7715 to 11,963), which is more than the expected 1980–2000 gain. Its main source is the addition of Hispanics or Asians in tracts (WB + H/A) where Whites and Blacks were already represented, along with just one of the former groups.

We look to the lower panel of Table 3 for information about changes not attributable to the demographic shift. Similar to 1980–2000, there was a drop of about 5000 White tracts (from 11,554 observed in 2000 to only 6209 in 2020). Also, as in the previous period, most of that fall was expected. So, we regard this change as persistent across the two time periods. There was an increase in global neighborhoods (WBHA) from the 11,963 expected to 13,186. This is a substantial number, and it is due not only to additional transitions from WB + H/A beyond what was projected, but also to about twice the expected number of transitions from WHA to WBHA. Hence, the movement of Blacks into new locales was clearly continuing in 2000–2020. However, it could also be seen to be slowing in comparison to the more dramatic changes in 1980–2000.

Finally, we consider the creation of new NW neighborhoods, which increased from 5750 to 6309. Almost all of this growth was expected (to 6249). This could result from a projection of stable White populations in tracts with strong growth in other groups. Changes beyond what was expected came from the loss of Whites in several kinds of neighborhoods, with the largest losses in WBHA and WB + H/A tracts. We do not have information on individual Whites’ residential mobility, but a likely cause is White flight from diverse neighborhoods, which was also apparent in 1980–2000. What changed is its magnitude and source. In 1980–2000 the growth in all-minority neighborhoods was mostly due to what we interpret as a process shift, while in 2000–2020 it was smaller and mostly due to the demographic shift.

3.1. Changing Neighborhood Contexts for Each Racial/Ethnic Group

The analysis up to now has focused on transitions in neighborhood composition, treating census tracts as the unit of analysis. We now describe how these changes result in shifts in the kind of neighborhood that members of each group—Whites, Blacks, Hispanics, and Asians—live in. These data are reported in

Table 4. Distribution across neighborhood types, by race, 1980–2020.

	1980		2000		2020
Whites
NW	793,104	0.60%	1,018,751	0.70%	1,184,144	0.80%
W	50,432,725	37.40%	42,619,229	29.60%	24,101,766	17.20%
W + H/A	33,607,473	24.90%	33,588,359	23.30%	23,878,514	17.00%
WHA	19,803,276	14.70%	17,418,620	12.10%	18,664,575	13.30%
WB	13,053,033	9.70%	14,687,212	10.20%	13,095,121	9.30%
WB + H/A	9,813,741	7.30%	18,703,677	13.00%	28,009,966	20.00%
WBHA	7,436,540	5.50%	15,910,020	11.10%	31,246,879	22.30%
Total	134,939,892	100.00%	143,947,869	100.00%	140,180,965	100.00%
Blacks
NW	10,354,435	48.30%	11,031,679	37.10%	9,848,791	26.20%
W	423,215	2.00%	571,280	1.90%	490,398	1.30%
W + H/A	454,091	2.10%	776,959	2.60%	792,687	2.10%
WHA	396,978	1.90%	643,498	2.20%	859,937	2.30%
WB	4,790,890	22.40%	5,786,307	19.50%	5,118,382	13.60%
WB + H/A	2,943,980	13.70%	5,773,304	19.40%	9,301,504	24.80%
WBHA	2,057,774	9.60%	5,123,529	17.20%	11,161,636	29.70%
Total	21,421,363	100.00%	29,706,556	100.00%	37,573,335	100.00%
Hispanics
NW	2,899,640	22.20%	8,606,685	27.20%	12,961,373	23.80%
W	496,169	3.80%	800,027	2.50%	884,434	1.60%
W + H/A	2,375,683	18.20%	4,653,647	14.70%	5,544,032	10.20%
WHA	4,056,113	31.10%	6,224,330	19.70%	8,076,541	14.80%
WB	204,688	1.60%	466,333	1.50%	858,541	1.60%
WB + H/A	986,165	7.50%	3,332,488	10.50%	8,515,925	15.60%
WBHA	2,044,103	15.60%	7,552,028	23.90%	17,673,921	32.40%
Total	13,062,561	100.00%	31,635,537	100.00%	54,514,767	100.00%
Asians
NW	334,160	10.40%	1,205,508	10.90%	2,010,260	9.40%
W	155,573	4.80%	388,179	3.50%	376,191	1.80%
W + H/A	594,809	18.50%	2,206,386	20.00%	2,812,619	13.20%
WHA	1,190,996	37.10%	2,806,075	25.50%	4,741,868	22.20%
WB	49,995	1.60%	185,018	1.70%	305,906	1.40%
WB + H/A	215,344	6.70%	1,207,977	11.00%	2,867,060	13.40%
WBHA	672,428	20.90%	3,024,789	27.40%	8,245,370	38.60%
Total	3,213,306	100.00%	11,023,933	100.00%	21,359,274	100.00%

. This table shows, for each group, how many group members lived in each of the seven neighborhood types in 1980, 2000, and 2020. The absolute numbers tend to rise, because every group had a greater or lesser population increase in these years. The change was modest for Whites, and their metropolitan total rose from 135 million to 144 million after 1980 before dropping to 140 million in 2020. Growth in the Black metropolitan population was at a steady pace from 21 million to 30 million in 1980–2000, and reaching 38 million by 2020. The main demographic shift, the trend that had so much impact on neighborhood transitions, was among Hispanics and Asians, whose population recorded multi-fold increases in the forty-year period. Hispanics grew from 13 million to 54 million, and Asians from 3 to 21 million. Because their total populations changed so much, our description of the trends in where they lived is based on their percentage distribution among neighborhood types, also shown in Table 4.

3.1.1. Whites

One of the largest changes for Whites was their declining share in White neighborhoods. The drop was significant in 1980–2000, from 37.4% to 29.6%, and even more rapid after 2000, falling to just 17.2% by 2020. The decline was to be expected, and it parallels the drop in the W share of tracts in the same years. The number of Whites in W neighborhoods could have multiple sources: White exodus from those areas, avoidance of them as a destination for movers, or simply a change in the neighborhoods’ categorization. We cannot distinguish these here.

The other very large change was the jump in the share of Whites who lived in two very diverse neighborhood types, global neighborhoods (WBHA) and what Zhang and Logan called “semi-global” neighborhoods (WB + H/A). If we combine these two, the share of Whites in them grew from just 10.8% in 1980 to 24.1% in 2000 and 42.3% in 2020 (in absolute numbers, from about 17 million to about 59 million). This represents a fundamental transformation in Whites’ neighborhood environment. In many cases, a W neighborhood may have transitioned to another category as a result of a small change in minority representation, from just below to just above their criterion level. But still, these figures show that millions more whites lived in at least moderately diverse neighborhoods in 2020 than in 1980, and millions fewer lived in neighborhoods where every other group was seriously underrepresented.

3.1.2. Blacks

A declining share of Blacks lived in all-minority (NW) neighborhoods, although their absolute number in such tracts remained about the same. Their NW share fell from nearly half in 1980 (48.3%) to 37.1% in 2000 and then to 26.2% in 2020. This trend seems contradictory to the increase in the number and share of NW tracts in the same period. One explanation is that Hispanics and Asians are replacing Blacks in some all-minority neighborhoods. Another possibility is that there has been an exodus of Black residents from historically Black city neighborhoods (these would have been categorized as NW, largely on the basis of their Black population). Despite a Black exodus, these would continue to be all-minority, with few White residents.

At the same time, we find very substantial growth in the share of Blacks who live in global and semi-global neighborhoods. Again, combining these two types, the share rose from 23.3% in 1980 to 36.6% in 2000 and 54.5% in 2020. Hence, a majority of metropolitan Blacks now live in very diverse neighborhoods. This change, as we saw above, could result both from the movement of Blacks into neighborhoods inhabited by Whites, Hispanics, and Asians, and also from Hispanics’ and/or Asians’ growing presence in WB (White–Black) neighborhoods.

3.1.3. Hispanics and Asians

The Hispanic population has been growing rapidly in all types of neighborhoods. Their share has grown most rapidly in the rising number of global and semi-global neighborhoods. Combined, this share has grown from 23.1% in 1980 to 34.4% in 2000 and 48.0% in 2000. There has been a corresponding decline in the share living in neighborhoods with Whites but not Blacks (WHA and W + H/A), from a near majority in 1980 (49.3%) to only 25.0% in 2020. This trend corresponds to the transition of many neighborhoods from WHA to WBHA or from W + H/A to WB + H/A.

The pattern for Asians is similar. Their growth in global and semi-global neighborhoods was especially notable, from 27.6% in 1980 to 38.4% in 2000 and then 52.9% in 2020. At the same time, they declined in WHA and W + H/A neighborhoods, which tended to transition to more diverse types due to the entry of a significant Black population.

4. Summary and Conclusions

This study examines neighborhood racial/ethnic transitions across two consecutive periods. There are strong trends toward increased diversity at the neighborhood level in U.S. metropolitan areas. We identify two sources of this change. Our analysis demonstrates that the shift in overall racial composition of urban residents, a stagnant non-Hispanic White population and a tremendous increase in Hispanics and Asians, plays a central and increasingly important role in stimulating neighborhood transitions over time. Because their growth is so widely spread across all kinds of neighborhoods, they contribute to the substantial loss of all-White neighborhoods and increase in the range of neighborhoods where they are well represented.

The only previous study to which we can compare our results is the 1980–2010 analysis by Zhang and Logan (2016). Their analysis focused on variations between different kinds of metropolitan areas, but it did include one explicit comparison between the expected and observed transitions. Their Figure 3 (p. 1949) categorized transitions as those that were projected to become more diverse (e.g., from all-White to more mixed) or less diverse (e.g., global neighborhoods to all-minority). In every type of metropolitan area, but especially in multiethnic metropolitan areas, the general pattern was for the observed changes in both directions to exceed the expected ones.

A new contribution from our study is to point out that the relative balance of impact from the demographic shift vs. the process shift (i.e., expected vs. observed) depended on what groups were involved. We found that the decline in White neighborhoods could be accounted for almost entirely by the demographic shift. In 1980–2000 the “expected” decline was about 25% of W tracts, and the actual loss was only slightly greater. The pattern in 2000–2020 was similar.

We also observe the results of changes in residential location processes, which we refer to as a “process shift,” that promote diversity by changing groups’ spatial distributions among neighborhoods. Notably, many neighborhoods that lacked any sizeable Black presence have now opened up to new African American residents. To a large extent, the pathway of neighborhood change matches the model observed by Logan and Zhang (2010) in their study of global neighborhoods. White neighborhoods that had already become more diverse through the entry of Hispanics and/or Asians subsequently experienced Black growth. We now see another pathway toward the global neighborhood, and this one stems from the demographic shift. Many global and semi-global neighborhoods in both periods also have their origin in neighborhoods that were predominantly White and Black. These were common in regions of the country that historically had experienced little immigration from Asia or Latin America. The demographic shift at a larger regional scale was bringing these groups to new destinations. White–Black neighborhoods transitioned to global and semi-global neighborhoods very much as expected due to the growth of Hispanic and Asian populations in these metropolitan areas. In this way, the demographic shift supplemented the process shift that opened up new residential options for Blacks.

At the same time as these changes were promoting increased neighborhood diversity, we found a continuation of other familiar residential processes that pressed in the opposite direction. Chief among these was the loss of Whites from diverse neighborhoods. In part this loss was expected as a result of the declining White population, but the transitions from other categories to all-White were greater in number than predicted by the demographic shift. Although we cannot directly measure White residential changes at the level of individual households, it is clear that White flight was an important part of these transitions.

On balance, though, the combined effect of the demographic shift, enlarging the Hispanic and Asian populations in all kinds of neighborhoods, and changing residential opportunities for Blacks, was a dramatic rise in the diversity of people’s residential environments. The population counts in Table 4 show a stark contrast between the situations in 1980 and 2020. In 1980, the largest number of Whites, 50 million, lived in all-White neighborhoods. By 2020, this number had fallen by half to 24 million, and more Whites than this were living in global and semi-global neighborhoods (31 million and 28 million, respectively). Almost half of African Americans in 1980 (10 million) lived in all-minority neighborhoods. By 2020, the largest number lived in global neighborhoods (11 million), and another 9 million lived in semi-global neighborhoods, while the number in all-minority neighborhoods (still nearly 10 million) accounted for only about a quarter of the Black population. In a country where the color line between Whites and Blacks once reached near-apartheid intensity, the change is striking.

The portrait of change presented here differs from the changes that one might infer from two kinds of studies that have become prominent in segregation research. One is research on the diversity of neighborhoods, which has been shown to be increasing steadily (Parisi et al. 2015). An implication that has not been directly tested is that increasing diversity is a sign of lowered barriers separating Whites from non-Whites. This implication needs to be tested in light of our results. To the extent that increasing diversity simply reflects the demographic shift that is occurring nationwide and in most metropolitan areas, it is consistent with persistent levels of residential segregation. Another approach is represented by studies of individual-level mobility among neighborhoods. To a large extent, neighborhood racial/ethnic change is a function of the racially differential retention of current residents and recruitment of new residents. On balance, studies based on longitudinal survey data have shown that White residential choices—White flight and White avoidance of diverse neighborhoods—tends to reinforce racial divisions (Crowder 2000; Crowder and South 2008; Parisi et al. 2019). Our findings suggest that the implications of these studies also need to be reconsidered, considering how strongly the influx of new non-White populations into metropolitan areas may counteract the segregating effect of White mobility choices. The longitudinal studies are typically limited to movements across neighborhoods within a single metropolis, and their samples necessarily include only the original residents of those neighborhoods at the beginning of the period. How can mobility patterns of current residents be dovetailed with locational patterns of new residents to account more fully for how neighborhoods change?

An important limitation of this study is that the analysis relies solely on census data and is carried out entirely at the national level. We consider it to be a step toward more definitive findings and interpretations. It is possible to examine more closely what is happening to the locational outcomes for specific groups in specific kinds of metropolitan areas. A study of the variations in outcomes across metropolitan areas could identify aspects of the context that influence neighborhood transitions—not only the extent of change in a region’s racial/ethnic composition, which we studied here, but other characteristics, such as the level of segregation, the relative socioeconomic positions of different groups, and the overall rate of growth. In what kinds of metropolitan areas and for which groups’ locational pattern is the demographic shift the dominant force, and where and for what groups is there greater change in the boundaries between groups that keep them separated? As another example, in a study of a single metropolitan area, it would be possible to move beyond a broad four-category coding of groups to reveal the situation of a particular national-origin subgroup that is well represented there, such as Vietnamese in Southern California or Afro-Caribbean Blacks in New York. The necessary publicly available census data could distinguish between the earlier-generation members of these groups and those that arrived more recently (i.e., the first and second generations). In a single locale it would also be possible to incorporate fieldwork and historical research into a multi-pronged effort to explain how people are making decisions that break with established patterns versus choices that replicate the existing distribution. These are opportunities for future work that are encouraged by the present study.