5.1 Overview

Beginning in the 1980s and 1990s, hundreds of large and small communities across the United States underwent significant demographic changes as immigrant households began moving away from initial destinations in traditional “gateway” cities and regions and migrating to areas of the United States where the population had been predominately native-born and often predominately White (Lichter & Johnson, 2006; Hall, 2013; Massey & Capoferro, 2008). As their numbers grew, they were in many cases also joined by immigrants who migrated directly to these “new destination” communities. This new migration pattern captured researchers’ attention for a multitude of reasons. First, it has largely been driven by Latino migrants (Kandel & Cromartie, 2004), and to a lesser extent by Asian migrants, who have historically been concentrated along the U.S. borders and in select large metropolitan areas that have been and remain popular gateway destinations for immigrants like Los Angeles, Chicago, and New York City. Consequentially, the areas that are now receiving new but steady streams of migrants and immigrants are now commonly referred to as new destinations because they are communities where a co-ethnic population was not already in place. This fact alone has sparked significant scholarly interest, with researchers seeking to understand how new destinations are receiving and reacting to new migrants and how these communities are transformed socially, demographically, and economically as a result of growing racial and ethnic diversity (Lichter & Johnson, 2020; Ludwig-Dehm & Iceland, 2017; Waters & Jiménez, 2005).

A second reason for interest in these communities is more specific to residential segregation research. It is that new destinations can serve as empirical testing grounds for longstanding theories about how segregation patterns for new groups emerge and change over time. The segregation literature is dominated by studies of large metropolitan areas where segregation patterns in many cases were established many decades ago before comprehensive data for neighborhoods was available and have been durable over time. The new destinations of the contemporary era can sustain studies of segregation more readily because neighborhood-level data is more detailed and comprehensive and, by their very nature, they offer a more dynamic setting for observing spatial population distributions and potentially identifying the causal factors that underlie contemporary residential segregation patterns. As a result, over the past decade we have seen a growing number of studies of segregation in new destination settings, particularly studies comparing patterns of White-Latino, White-Asian, and immigrant segregation across new destinations and established areas of settlement (e.g. Hall, 2013; Lichter et al., 2010; Park & Iceland, 2011). Some of these studies have also examined how segregation patterns in new destinations have changed over time as newly arrived populations become more settled and more visible in the area (e.g. in the labor force, in schools, and in community settings) and how nonmetropolitan communities in particular are affected.

In this chapter we contribute to this developing body of research by providing a comprehensive analysis of White-Latino and White-Asian residential segregation in new destinations that takes advantage of superior methods for measuring residential segregation (as described previously in Chap. 2). In doing so we are able to address and overcome some of the significant obstacles that have limited this empirical literature, including the fact that index bias often renders conventional measures of segregation problematic for communities where segregation involves groups that are small in absolute and/or relative size and where segregation must be assessed using data for small spatial units. The problems are similar to those affecting studies of segregation in nonmetropolitan communities more generally (as discussed in Chap. 4) – namely, researchers rightly worry that segregation scores are distorted by bias and that ad hoc strategies for dealing with the problem result in smaller and less representative samples of communities. On top of that, research on segregation in new destinations includes the added complication that new destination communities are undergoing rapid demographic changes in racial-ethnic composition, often starting from minimal or no minoritized group presence and transitioning to significant and rapidly growing minoritized group presence. As discussed in more detail earlier, the challenges of measuring segregation under these circumstances pose major problems for drawing general conclusions about trends in segregation in new destination communities and comparisons with segregation in communities with established minoritized group presence based on minoritized group settlement in earlier decades. New methods for measuring segregation without bias makes it possible to address these problems more completely and effectively than has been possible in previous research.

By also focusing on Asian new destinations, we contribute to the literature by expanding our knowledge of White-Asian residential segregation in emerging Asian settlement areas. This topic has been neglected and understudied and, thus, is not well-understood in the residential segregation literature (Flippen & Farrell-Bryan, 2021). Finally, we conduct a more exploratory parallel analysis of White-Black segregation in areas that might be designated as Black new destinations, where increasing presence of Black households is less related to immigration and more related to regional and spatial diffusion of this primarily native-born population. As we did in our previous chapter on nonmetropolitan residential segregation, we use our new methods of measurement and analysis to broaden the possibilities for understanding the nature of residential segregation in communities beyond the highly populated, diverse metropolitan areas that dominate the literature.

5.2 New Destinations: An Overview of Changes and Potential Trajectories

New migration across the United States to the Midwest and South, primarily driven by Latino immigrants, has transformed areas that had previously and over generations been predominately native-born White. This has produced the demographic trend of emerging ethnic diversity in new destination communities in locations far removed from the traditional gateway communities where Latino and Asian immigrants have historically settled upon arrival. When immigration to the United States from countries in Asia and Latin America surged following immigration reforms in the 1960s, immigrants initially tended to settle primarily in major cities along the East and West coasts such as Los Angeles, New York City, and San Francisco, and also in a few other large metropolitan areas including Chicago and Houston. In general, the largest Latino and Asian populations are found in major metropolitan areas on the West coast, along the U.S.-Mexico border, in some areas of the Upper Midwest, and in the Northeast. The attractiveness of these cities is easy to understand. Migrants seeking economic opportunity and the “American dream” found the best opportunities in the welcoming labor markets of rapidly growing metropolitan areas in the 1960s and early 1970s. Immigrants who came later would be drawn by inertia of pre-existing migration patterns reinforced by expanding migration networks to many of the same communities and often to ethnic neighborhoods that were established by their predecessors in response to a combination of constraints on residential options, co-ethnic settlement based on personal ties involved in chain migration of kin, friends, and co-ethnic networks, and the attractiveness of enclave areas to many first-generation immigrants.

Given this history, it is not surprising that much of segregation research focusing specifically on Latino or Asian segregation has mostly given attention to the major metropolitan areas where many Latino and Asian communities are located in racially and ethnically segregated neighborhoods. This has shaped how scholars theorize about segregation formation. Traditional theories of assimilation, ethnic disadvantage, and racial conflict that emerged from studies focusing on the experiences of White immigrant groups in the early twentieth century and White-Black segregation since the twentieth century are being continually revisited and used to build a lens for understanding Asian and Latino segregation in large and mostly urban metropolitan settings. For the most part, the research focusing on new destination communities that has emerged in recent decades has drawn from this foundation to build theoretical frameworks for understanding patterns of segregation in new destination communities.

The features of Latino immigration and migration that have led to the trend of emerging new destination communities have already been well-documented by demographers and examined by social scientists in the sociological, economic, geographic, and demographic literature. We note in particular an article by Daniel Lichter in 2012 that provides a thorough review of Latino settlement in new destinations and several important reports and books that have examined both quantitative and qualitative aspects of the shifting social landscape of new destinations, including William Kandel and John Cromartie’s 2004 report New Patterns of Hispanic Settlement in Rural America, Douglas Massey’s edited volume New Faces in New Places: The Changing Geography of American Immigration (2008) and Victor Zúñiga and Rubén Hernández-León’s edited volume New Destinations: Mexican Immigration in the United States (2005). What is less well-documented but is also contributing to the changing racial demography of these previously homogenous communities is the migration of Asian households to new destinations. The factors that draw Asian immigrants and migrants away from traditional, urban areas and into smaller, sometimes rural, communities in the interior are not as well established. New destination emergence for Asian migrants is in some ways a different dynamic as immigration from Asian nations over the past several decades has been driven not only by economic “pull” factors but also by “push” factors due to conflict in home countries, especially in Southeast Asia. Some of this migration is driven by refugee resettlement, as in the case of communities in the upper Midwest known for welcoming the largest numbers of Hmong refugees (Singer & Wilson, 2006). Other push and pull factors that are bringing Asian migrants to new destinations are likely playing a role, including economic drivers.

One reason why Asian new destinations are less well understood, as Chenoa Flippen and Eunbi Kim (2015) argue, is that the distinction between new and established destinations is less clear for the Asian population because such a large portion of the Asian population in the U.S. is foreign-born, which means that many destinations are, in a sense, “new” destinations. Indeed, many communities with established Asian presence, in comparison to communities with established Latino or Black presence, tend to be at lower levels on absolute and relative size of the Asian population and often are experiencing increases in Asian presence that are comparable to those seen in new destinations. A secondary and more practical reason why Asian new destinations receive less attention, particularly in studies of residential segregation, is due to the technical challenges researchers encounter when measuring segregation in new destinations. The concerns we previously noted as relevant in studies of nonmetropolitan communities are equally if not more relevant here; namely, conventional approaches to measuring segregation yield flawed index scores when groups are small in absolute and/or relative size and when the scope of analysis extends beyond the largest metropolitan areas. Despite all of this, Flippen and Kim (2015) insist that it is important to expand our understanding of new destination migration as a phenomenon to include Asian migration as well, especially as immigration overall to the United States is now dominated by migrants arriving from Asian countries and there is enough preliminary evidence to justify studying Asian residential segregation in new destinations (Hall, 2013; Park & Iceland, 2011).

The changing demography of new destination communities is of interest here for substantive as well as methodological reasons. By definition, new destinations are communities where a specific population initially had a small or perhaps no presence but then grew rapidly over a short period of time due primarily to migration. Theory provides two contrasting scenarios for how inter-group relations may play out in these scenarios. The competitive group relations perspective includes a scenario of initially low segregation that will later transition into high segregation. In the earliest stages of this sequence, hierarchical relations among the groups may not yet exist, partly because the minimal presence of the minoritized group carries no practical consequences for the established majority group. In this situation, the minoritized group population in question may settle based primarily on housing availability and affordability and is not necessarily likely to be segregated except under certain special conditions, such as when dedicated housing for migrant workers concentrates the group into a single neighborhood. Later, as the minoritized group population grows rapidly in absolute and relative size, they can become more visible as a distinctive group in the community, potentially triggering a negative reaction on the part of the majority group. Blalock’s racial threat theory (1967) and theories of competitive ethnic relations (Olzak & Nagel, 1986; Fossett & Cready, 1998) emphasize ethnic composition of the population as a factor shaping the extent to which members of the majority group will come to recognize the presence of the new group, increasingly view competition for scarce resources in group terms, and, ultimately, become less tolerant of the minoritized group’s presence and engage in discrimination against the minoritized group out of motivation to preserve the majority group’s advantaged position in the community.

An alternative perspective outlines a possible assimilation scenario involving a sequence of initially high segregation at time of settlement later transitioning to lower segregation over time as the new group is incorporated more broadly into the life of the community. In this scenario, initial segregation is high because members of the new group tend to congregate in one or a few neighborhoods based on strong ties of kinship and mutual support and the attractions of ethnic community institutions that serve the specialized linguistic and cultural needs of the minoritized group population (e.g., particular religious services, establishments conducting business and providing services in the group’s language, stores and restaurants providing familiar products and cuisine, etc.). In the next stages of group relations, the new minoritized group population undergoes relatively rapid cultural and linguistic assimilation over time within generations and also across generations. They additionally are expected to experience increasing assimilation in social and economic spheres (e.g., education, employment, etc.). Segregation then falls over time as cultural assimilation reduces the attraction of the enclave and socioeconomic assimilation brings increasing wherewithal to move to better housing in neighborhoods outside the enclave. Under this scenario, the level of segregation will be a balance of two processes. Continuing arrival of new immigrants can serve to maintain segregation by replenishing and sustaining ethnic enclave neighborhoods even as settlement patterns of later generations of the group lead segregation to decline.

It is important to note that these two scenarios are not mutually exclusive. In a given community both hypothesized dynamics can play out along the lines just described, or in a wide variety of hybrid combinations. White communities in areas that have received a large and rapid influx of Latino and Asian migrants may develop a sense of threat and competition and begin engaging in behaviors that hoard resources and opportunities along racial and ethnic lines. One of the most effective ways to restrict group access to resources is to create conditions of residential segregation through economic and social exclusion, which can cut off a minoritized group not only from housing but also to a variety of other resources and amenities tied to neighborhoods. At the same time, however, new migrants could also assimilate rapidly on language and other aspects of culture, as well as on socioeconomic outcomes, and this may reduce their level of social distance from White households. This would manifest spatially as greater integration with White residents over time.

These two dynamics together represent the dominant hypotheses of segregation research: place stratification and spatial assimilation. Therefore, an important question to ask about these communities is: do these communities respond to rapidly changing population composition by segregating? If so, to what extent is segregation offset by reductions in social distance? As Flippen and Farrell-Bryan (2021) point out in their recent review article, new destinations have provided the opportunity to refine and revise theories of incorporation that lay at the heart of migration research. However, these authors also note that even after three decades of research the literature has hardly reached a consensus regarding the degree to which these different hypothesized dynamics shape group relations in new destination communities. This conclusion is equally true for studies focusing on trends in residential segregation in new destinations.

5.3 Residential Segregation Studies of New Destinations: Findings and Limitations

The state of the research literature focusing specifically on residential segregation in new destinations is similar to that of the broader research literature focusing on segregation in nonmetropolitan communities (which we reviewed previously in Chap. 4) in the fact that findings vary across studies and a consensus has yet to emerge regarding the exact nature of patterns and levels of segregation across communities and the trajectory of segregation over time. Flippen and Farrell-Bryan (2021) rightly point out that this is partly because findings from studies of segregation in new destinations are sensitive to multiple methodological choices including which groups are considered, which communities are included in the analyses, how communities are categorized (i.e., as “new,” “established,” and “other”), and which spatial units are used to capture group distributions across neighborhoods. In this section we review these and related issues, noting relevant limitations affecting past research and how we address them in this study.

The literature on segregation patterns and trends in new destinations has overwhelmingly focused on Latino and immigrant segregation. This is unsurprising in light of the fact that the largest proportion of new destination situations involves Latino migrants and immigrants settling in previously racially homogeneous communities in the Midwest and South and particularly, but not exclusively, in nonmetropolitan communities. But we argue Asian new destinations should receive greater attention than in the past both because these communities represent an increasingly important demographic phenomenon and also because Asian new destinations add value for understanding segregation since they involve a minoritized racial population that is similar to the Latino population in some respects and also different in other respects. For example, until recently, the Asian population nationally was concentrated in a small number of coastal metropolitan areas and a small number of nonmetropolitan communities in the western United States, particularly on the West Coast. Consequently, in most of the country, Asian populations are not only new in Asian new destinations, they also are new to the broader region, especially in nonmetropolitan settings. Thus, since there is no prior sizable Asian presence, there are no co-ethnic communities or histories of group relations to give shape to Asian residential settlement patterns. In communities newly receiving Asian migrants, the driving questions are the same: Where do they live?, Who do they live among?, and How do these outcomes change over time? Answers to these questions give us important knowledge for understanding the reception of new racial and ethnic groups in a community in a contemporary context, allowing us to reevaluate predominant theories of segregation and incorporation. With recent attention now given to anti-Asian racism in the wake of the COVID19 pandemic, the segregation literature must also be called upon to give greater priority to documenting and better understanding the experiences of Asian populations in the United States.

As we mentioned earlier, the literature is not guided by a consensus view on patterns and trends of residential segregation in new destinations and what implications they hold, due partly because findings vary with variations in methodology and also likely due to the limitations of available neighborhood-level data for Asian and Latino groups (Flippen & Farrell-Bryan, 2021; Hall, 2013). Some studies suggest a developing process of minoritized group exclusion and racialization in new destinations leading to higher levels of segregation than in established areas (e.g. Lichter et al., 2010) and which may be pronounced when undocumented immigration is driving the formation of a new destination (Hall & Stringfield, 2014). Other studies report that segregation is lower in new destinations than in established areas (Park & Iceland, 2011) and yet others report that levels of segregation are at similar levels in both area types (Hall, 2013).

Two basic choices concern how new destinations are defined and what communities are included in the analysis. For example, Lichter et al. (2010) examined census-designated places, drawing distinctions between urban, suburban, and rural communities. Others have focused exclusively on metropolitan or micropolitan areas (e.g., Fischer & Tienda, 2006; Park & Iceland, 2011). In addition, studies adopt different practices for identifying new destinations based on factors such as minimum population thresholds, population changes over time, and how population changes compare to larger trends. The challenge here is to identify areas where a certain group’s presence in the community has risen from demographically small to sizable, where the group is growing rapidly in absolute and relative size and is most likely, but not exclusively, driven by migration, and where these demographic trends are fundamentally changing the racial or ethnic composition of the receiving community.

Some of the more well-known work over the past decade, particularly studies by Daniel Lichter and colleagues (e.g. 2010) and Matthew Hall (2013), have presented evidence that segregation is an emerging outcome in new destinations. All studies that examine segregation in new destinations encounter challenges in measuring segregation for small groups. Studies focusing on nonmetropolitan settings necessarily must be less restrictive because it is necessary to use data for small spatial units to measure segregation in nonmetropolitan communities. Non-trivial index bias is certain to be present in these situations. Thus, studies adopt multiple strategies starting with using sample restrictions to try to exclude the most egregiously concerning situations and using various ad hoc practices such as differential case weighting to mitigate the unwanted impact of index bias on cases in the analysis sample. Research focusing on metropolitan areas can adopt study designs aimed at providing greater protection from bias, but at the expense of limiting analysis to a smaller, less representative set of communities. For example, Matthew Hall’s (2013) study provides one of the more detailed analyses on Asian segregation in new destinations by examining immigrant segregation in new destinations by specific ethnic groups, including Asian ethnic subgroups. Like many studies, this analysis was restricted to cases where scores for conventional segregation measures could be deemed more trustworthy, and on this basis limited the study to large metropolitan areas.

Given the salience of these measurement concerns and the impact of methodological choices on study findings, we contribute to the literature by using new methods that allow for superior measurement of segregation and larger, more representative samples of communities. New destinations, by their definition, are initially overwhelmingly White with an emerging minoritized group population which, while growing rapidly, still comprises a small fraction (e.g., less than 10 percent) of the population in most cases. Additionally, many new destinations have emerged in nonmetropolitan settings which present challenges for measuring segregation due to the need to use data for small spatial units (e.g., census blocks). These are the very conditions that make conventional segregation measurement untrustworthy first because index scores are certain to be inflated by index bias and second because the magnitude of the impact of bias varies from one community to the next. The problems are so concerning and so variable across measurement situations, one cannot safely perform close analysis of segregation scores and cannot even assess changes in scores over time in the same community.

We overcome these problems and limitations on analysis by using new methods to obtain scores for segregation indices that are unbiased across all group comparisons, even when segregation is measured using small spatial units (as is necessary when investigating segregation in nonmetropolitan settings) and when the groups are imbalanced in size and/or one or both groups are small in absolute size. Accordingly, the scores we use can sustain close case analysis including, for example, directly comparing White-Latino segregation in a given community with segregation in another community, or in the same community over time. And, because the scores are unbiased even under extreme conditions where standard index scores cannot be trusted, we can examine segregation in large, representative samples of communities. Thus, the analyses we present in this chapter make a valuable contribution to the literature on segregation in new destinations by examining Latino, Asian, and Black segregation in new destinations in a larger, more representative analysis sample than has been possible in past research, and we are able to trust the segregation scores in our analysis as giving an accurate representation of how residential patterns have emerged and changed in new destination communities over the past few decades.

Our first task in this chapter will be to summarize levels and trends in segregation scores for metropolitan, micropolitan, and noncore communities from 1990 to 2010. When doing so, we will compare scores across new destination communities and communities with established minoritized group presence as defined based on the absolute and relative size of group populations in 1990 and the rate of growth of the minoritized racial group over time. Our second task in this chapter will be to analyze how levels and patterns of segregation in new destinations vary with the relative presence and rate of growth of the minoritized racial population, region, community type, industrial composition of the labor force, and other community-level covariates. Our third task is to review aspects of segregation measurement that have been neglected in past research including what insights are gained by contrasting results obtained using the dissimilarity index and the separation index. The former can and does take high scores when uneven distribution is dispersed, while the latter takes high scores only when uneven distribution is polarized, and one can safely conclude groups are separated across different spatial units and can potentially experience inequality on location-based outcomes.

5.4 Data and Measurement

As we did in the previous chapters of this book, we use block-level tabulations of householder by race-ethnicity from the 1990, 2000, and 2010 decennial censuses to measure residential segregation between groups. This is a carefully considered departure from previous segregation research on new destinations which uses data for persons instead of data for households. Having reviewed the basis for this decision in detail earlier in Chap. 2, we limit comment here to noting that the choice makes it possible for us to obtain unbiased scores for segregation indices based on eliminating the impact of fixed levels of same-group contact that are incorporated into standard index calculations which implicitly and incorrectly treat all individuals as locating independently, when in fact most individuals locate as part of a household composed of multiple, often same-race individuals. Acknowledging this social fact brings two problems with standard index scores into clear focus.

The first problem is that the standard of exact even distribution cannot be achieved when the integrity of households and individuals is respected. That is, standard versions of indices of uneven distribution will take non-zero values when households and individuals are assigned to neighborhoods “intact”, not in fractional parts. As a matter of measurement theory, this problem originates with the decision to define integration as exact even distribution instead of as statistical independence of race and neighborhood. The consequence is that, even under “optimal” or “strategic” assignment, standard versions of indices will take positive scores (not zero) if households and individuals are assigned to neighborhoods intact. The issue generally escapes notice in the broader segregation literature because the practical consequences for index scores typically is negligible when segregation is measured for large groups using large spatial units such as census tracts. Unfortunately, the issue becomes consequential and cannot be ignored when segregation is measured using small spatial units such as census blocks. As our discussion in Chap. 2 explains in more detail, if households are distributed intact across blocks, standard versions of indices of uneven distribution will have two non-negligible sources of bias: a random component that can be eliminated in principle by assigning households to neighborhoods in arrangements that are “optimal” for reducing index scores and a “floor” component that can only be eliminated by assigning first households and then individuals to neighborhoods in fractional parts.

The crucial point is that the impact of bias on index scores calculated using block-level data is much larger and consequential than is generally appreciated. The problem is even more concerning when measuring White-Latino segregation in micropolitan and noncore new destination communities because the average number of households per block is smaller in micropolitan areas compared to metropolitan areas and smaller still in noncore counties and, additionally, because the average size of households is larger for Latino households than for White, Black, or Asian households. Both factors lead to higher levels of index bias for White-Latino segregation in micropolitan areas and noncore counties compared to, for example, White-Black or White-Asian segregation in metropolitan areas. Finally, these problems exacerbate the initial problem that segregation index values are inflated by bias to a greater degree in new destination communities because, by the nature of these communities, the groups are more imbalanced in size than is the case in communities with established minoritized group presence. This set of problems poses major challenges for studying segregation in new destinations using standard versions of index scores computed using data for persons. We address and overcome these problems by measuring segregation using unbiased versions of indices and data for households. The resulting index scores we obtain are superior to those reported in previous research. Put simply, we can describe every index score we obtain as being valid and free of distortion from bias and this description cannot be applied to index scores reported in previous studies of segregation in nonmetropolitan new destination communities.

Methodological studies establish that unbiased versions of index scores perform as desired and are trustworthy across a much wider range of circumstances than is the case for standard versions of index scores (Fossett, 2017). Consequently, we can include many more cases in our analysis than is typical in previous studies. Specifically, we include cases where both groups in the comparison have at least 50 households and the pairwise percentage of either group in the comparison is at least 0.5 percent.Footnote 1 As we described in previous chapters, these criteria are to ensure that we are only measuring segregation in areas where the notion of segregation between groups is meaningful and unbiased index scores are reliable.Footnote 2

We use the race and ethnicity tabulations from the 1990 to 2010 censuses to calculate racial composition measures and measures of group-specific population change over time from 1990 to 2010. We then use the results of these calculations to categorize metropolitan areas, micropolitan areas, and noncore counties as either new destinations, communities with established group presence, or some other type of community based on a protocol similar to that used by Lichter et al. (2010). Specifically, we define new destinations as those communities where the referenced minoritized racial population was less than 10 percent of the total population in 1990 and experienced a rate of growth above group-specific thresholds based on absolute percentage growth and relative percentage growth (noted below). Communities with established group presence are defined as those where the minoritized racial population was at least 10 percent of the area population in 1990. We also identify a subset of cases as communities of “highly established” presence when the minoritized racial population was at least 30 percent of the community population in 1990.

In addition to these two primary community types, we identify communities of low minoritized group presence where the minoritized racial population was less than 10 percent of the population in the community and did not grow at a high enough rate over time. The rate-of-growth thresholds were adjusted by group to take into account that expected growth rates cannot be applied uniformly across groups. Thus, the minimum growth rate for the Latino population is higher than it is for the Asian population based on their higher rate of growth overall in areas outside of established gateways. For an area to be designated as a Latino new destination, the absolute Latino growth rate had to be at least 3 percentage points, or the relative growth rate had to be at least a 50 percent increase. For Asian new destinations, the absolute Asian growth rate had to be at least 2.5 percentage points, or the relative growth rate had to be at least a 50 percent increase. For exploring the possibility of Black new destinations (discussed more below), we applied the same criteria as we used for Asian new destinations.

We measure segregation for group comparisons by applying the formulas for obtaining both unbiased and standard versions of the dissimilarity index (D) and the separation index (S) as reviewed in Chap. 2 (and Fossett, 2017). We calculated standard scores primarily to document problems of index bias. Our discussion of findings focuses exclusively on unbiased scores and, in general, assign priority to scores for the separation index (S). Our justification for these choices follows the reasoning we outlined in our analysis of segregation in nonmetropolitan communities (see Chap. 4). The justification is even stronger when focusing on segregation in new destination communities because they are prime candidates for index scores, particularly scores for the dissimilarity index (D), to be distorted due to the problems that arise when segregation indices are not corrected to eliminate the impact of inherent upward bias. Our measurement methods provide the most accurate and trustworthy measurements of segregation of any similar study to date and thus make it possible to provide better descriptions and reach better conclusions regarding the reality of segregation in new destination communities over time and how it compares with segregation in established areas of group presence.

We conclude our discussion of case selection, methodology, and measurement by noting that this chapter focuses primarily on Latino and Asian new destinations because the growth and spatial diffusion of these two populations are the primary drivers of the emergence of new destination communities across the United States. However, we recognize that Black migration and spatial diffusion trends deserve attention as well, particularly Black migration to the South. Most of the receiving areas would likely not fall under the conceptual definition of “new destination” based on population history, although there is evidence that more of Black migration to the South is driven by “primary” migration (as opposed to “return” migration), which refers to Black migrants to the South who were not born in the South (Hunt et al., 2008). However, for the sake of exploration we chose to also consider areas where Black population growth is substantial using our schema for defining new destinations and established areas of group presence. In Table 5.1, we document the number of new destinations and established areas of group presence by group in 2010 based on our criteria. A distinct finding is that the new destination phenomenon is more common for Latino populations than for other groups with 368 metropolitan areas, micropolitan areas, and noncore counties identified as Latino new destinations. In contrast, we found only 37 Asian new destinations and 21 Black new destinations, despite using more liberal growth rate criteria for these two groups. Many areas, consisting mostly of noncore counties and micropolitan areas, remain classified as areas of low settlement for each of these groups.

Table 5.1 Frequency of communities by destination type by group, 2010
Full size table

5.5 Residential Segregation in Latino New Destinations and Established Areas of Settlement

We begin by summarizing how levels of White-Latino segregation vary across communities categorized on Latino presence over the period 1990 to 2010 in Table 5.2. In 1990, at the onset of significant Latino population growth for most new destination communities that emerged over the following decades, White-Latino segregation in new destination communities was very low, particularly in comparison to communities with established Latino presence where the average level of segregation was in the medium range and even more so in comparison to communities where Latino presence is very high. At the same time, the average level for White-Latino segregation in new destination communities is only slightly higher than the average level seen in communities with low levels of Latino presence. The major contrast, however, is that while average levels of segregation were steady or declining over time in all other categories of Latino presence, the average level of segregation in new destinations communities was rising and even doubling from 1990 to 2010, at which point it was close to the average level of segregation seen in areas of established Latino presence.

Table 5.2 Mean separation index scores by Latino community types, 1990–2010
Full size table

There are two ways in which these findings in whole or partially depart from what has been posited in the Latino new destination literature. First, average levels of White-Latino segregation are in the low-to-medium range from 1990 to 2010 across all categories of Latino presence. Second, although White-Latino segregation was rising in new destination communities, the average levels remained lower than the average levels observed for communities of established Latino presence.

The divergent patterns across categories of Latino presence help us understand general patterns and trends for White-Latino segregation across the United States. Communities that have seen sudden and significant Latino population growth are experiencing rising segregation, but communities with established Latino presence have had stable levels of segregation and communities with highly established Latino presence have experienced declines in average levels of segregation. Breaking trends down this way yields a more nuanced and dynamic picture of White-Latino segregation in the United States. Segregation is generally higher in communities with established Latino presence but has been stable or declining over recent decades. In contrast, average levels of segregation in new destination communities are initially much lower, but as Latino presence increases, average levels of segregation rise in the direction of converging on levels observed in communities with established Latino presence. Interestingly, the rising average level of segregation in Latino new destination communities results in the average level of segregation for these communities in 2010 matching the average level of segregation observed for areas of established Latino presence in 1990.

5.6 Residential Segregation in Asian New Destinations and Established Areas of Settlement

Less is known about Asian new destinations because it is less widespread and has not been the object of many studies. Consequently, we cannot draw on perspectives from prior research focusing on the origins and trajectories of new areas of Asian settlement as we were able to do for Latino new destinations. However, since the Asian population is growing rapidly and is diffusing out from historical areas of settlement, it is appropriate to apply a similar schema for identifying new destinations and communities of established group presence based on initial levels of Asian population presence and rates of increase over time and compare levels of segregation in categories of Asian presence over time. As we showed earlier in Table 5.1, our schema does not identify as many Asian new destination communities as Latino new destination communities. But we identify more than enough to sustain a preliminary review of patterns and trends based on the average scores for White-Asian segregation for communities classified by category of Asian presence presented in Table 5.3.

Table 5.3 Mean separation index scores by Asian community types, 1990–2010
Full size table

We find that in 1990 Asian new destination communities have lower levels of segregation than communities where Asian presence is established and much lower than in communities where Asian presence is highly established. Indeed, the average level for White-Asian segregation in new destination communities is not much higher than the average level seen in communities with low levels of Asian presence. This changes in later decades as the average level of segregation in Asian new destination communities increases in each decade and more than doubles by 2010 while the average level of segregation in communities with low Asian presence is stable. Echoing the patterns for White-Latino segregation in Latino new destinations, the rising average level of segregation in Asian new destination communities leads the average level for these communities to match the average level of segregation observed for communities with established Asian presence in 1990. A major difference here, however, is that, where average levels of White-Latino segregation were stable or declining over time in communities with established Latino presence, average levels of segregation are rising over time in communities with established Asian presence as well.

Before turning next to consider patterns for Black new destination communities, we first note that the patterns of segregation across categories of Latino and Asian presence in communities are surprisingly similar. Average levels of segregation are very low in the low group presence category, initially very low but rising over time in the new destinations category, at the low end of the medium range in areas of established group presence, and solidly in the medium range in areas where group presence is well established. The main difference across the patterns for Latino and Asian segregation is that average levels of segregation are rising over time in communities with established Asian presence but stable or slightly declining in communities with established Latino presence. While we do note differences between patterns and trends for White-Latino and White-Asian segregation, they are much more similar to each other than they are to the patterns and trends seen for White-Black segregation, especially with regard to average levels of segregation in communities with established Black presence.

5.7 Residential Segregation in Black New Destinations and Established Areas of Settlement

As is the case for Asian new destinations, communities that have only recently experienced initial Black settlement and population growth have not received as much attention as Latino new destinations in prior research and thus these communities are not as well documented or as well understood. Indeed, because sustained post-1965 immigration is often a major factor in emerging Latino and Asian new destinations, it is not surprising that less thought and attention has been given to assessing the existence or prevalence of Black new destination communities and the levels and trends in segregation that may be present in them. Certainly, it is true that immigration is a lesser, albeit not negligible factor, for the growth of the Black population. Nevertheless, it is still possible for Black new destinations to emerge when migration and spatial diffusion of the Black population leads Black households to settle in communities where previously Black presence has been minimal. We address this gap in the literature by applying the same schema and measurement approach we used for Asian new destinations to first identify Black new destination communities based on Black population presence and growth over time and then to compare segregation patterns and trends in White-Black segregation across the resulting categories of Black population presence. Not surprisingly, we identified far fewer Black new destination communities that experienced Black settlement around 1990 and subsequent population growth from 1990 to 2010 at a level that would elevate the community from low Black presence to Black new destination. In contrast, we identified hundreds of communities that had established Black populations, far outnumbering communities with established Latino or Asian populations.

This reflects fundamental differences in the demographic history of each of these populations, with the Latino and Asian populations seeing rapid growth and spatial diffusion in recent decades whereas the Black population has seen slower growth and only modest spatial diffusion during this same period. The large number of communities with established Black presence reflects the demographic legacy of the historical forced migration of enslaved people primarily to Southern states, and then the Great Migration of Black households in the early part of the twentieth century fleeing racial oppression and limited economic opportunities in the post-Reconstruction, Jim Crow era in the South and gravitating toward relatively better economic opportunities in the growing metropolitan centers in the North and Midwest. The Black population remains disproportionately concentrated in these regions and communities where high levels of Black presence were established many decades ago. This leaves open the possibility that migration and spatial diffusion of the Black population could create Black new destination communities. But the actual occurrence of this demographic sequence is limited to a small number of communities. Of course, this does not mean that the spatial distribution of the Black population has been stagnant. But it does mean that it has been more evident within metropolitan and nonmetropolitan communities (e.g., movement toward suburban settings in metropolitan areas) than across communities on the scale seen for the Latino and Asian populations.

Despite the fact that the number of Black new destination communities is not large, we still see value in comparing levels and trends in White-Black segregation across communities classified on the basis of Black population and also with levels and trends in White-Latino and White-Asian segregation across communities classified on the basis of Latino and Asian presence, respectively. One reason for this is that, while we know a great deal about how de jure Black segregation developed and intensified in northern U.S. metropolitan areas to the highest levels in the nation in response to the Black Great Migration, we know less about what happens to spatial patterns in communities where Black populations are emergent in the post-Fair Housing era. A second reason is that Black segregation in the United States is one of the more entrenched spatial patterns that characterizes urban areas, and it is implicit that we often understand Latino and Asian segregation patterns and trends by how they compare to Black segregation patterns and trends. The broad consensus in the literature is that White-Latino segregation and White-Asian segregation by comparison are more moderate and more fluid than White-Black segregation in part because factors relevant in spatial assimilation theory such as acculturation and socioeconomic assimilation within and across generations appear to play a greater role in Latino and Asian residential segregation – a pattern we document in detail in Chap. 6. However, evidence for these group differences, including the evidence we review later in this monograph, is primarily based on studies of large metropolitan areas which could not be considered new destinations for any group (e.g. Los Angeles, Houston, Chicago, etc.). So, we take the present opportunity to address the narrower question of how segregation patterns for these minoritized groups compare across communities that vary on group presence.

To serve this goal, we document how average levels of White-Black segregation as measured by the separation index (S) vary across communities classified on level of Black presence in Table 5.4. Based on reviewing the results presented in this table we can draw some tentative conclusions about the nature of segregation in Black new destinations and how these outcomes compare with patterns observed for other minoritized racial groups. First, we must note that the most important overall finding is that White-Black segregation is higher than White-Latino and White-Asian segregation across all four categories of group presence, ranging from low settlement to highly established settlement. Even though segregation is declining over time within all categories, it is always the highest level in every category in every time period. The highest average levels of White-Black segregation are seen in highly established areas of Black settlement where the Black population is at or above 30 percent. The average score for the separation index is at the high level of 61 in 1990 and, despite declining to 53 by 2010 it remains more than 20 points higher than White-Latino and White-Asian segregation in the same category of group presence. A similar pattern is seen for communities of established Black settlement where the Black population comprises 10 to 30 percent of the total population. These communities have a slightly lower average score for S in 1990 of 56, and despite having a larger absolute and relative decline to an average score of 44 in 2010, it also remains about 20–22 points higher than average scores for White-Latino and White-Asian segregation. As we just described, the main pattern in communities with established group presence is that White-Black segregation is much higher than White-Latino and White-Asian segregation. But we also should note a secondary pattern of convergence because White-Black segregation is declining from 1990 through 2010 while White-Latino segregation is mostly stable over this time and White-Asian segregation is increasing.

Table 5.4 Mean separation index scores by Black community types, 1990–2010
Full size table

White-Black segregation is also clearly higher than White-Latino and White-Asian segregation in communities of low group presence. Average levels of White-Latino and White-Asian segregation in these communities are around 6–7 points, which is very low, across all decades. In contrast, the average for White-Black segregation in 1990 is over 20 points higher at 29 points which is well into the medium range and is at or above the average levels of segregation observed for White-Latino and White-Asian segregation in communities where the minoritized racial group in the comparison has an established presence. We do observe that White-Black segregation in communities with low Black presence declines significantly over time in both absolute (down 12.5 points) and relative (down 44 percent) terms. But even so, it remains 10 points higher than White-Latino and White-Asian segregation in 2010.

Findings for White-Black segregation in new destinations are both similar to and different from White-Latino and White-Asian segregation in new destinations. The main points of similarity are that, across all three groups, the average levels of segregation in new destination communities in 1990 are much lower than the average levels of segregation seen in communities of established group presence and are not very different from the average levels in communities of low group presence. The most important difference is that the average level of White-Black segregation in new destinations in 1990 is much higher than the average levels for White-Latino and White-Asian segregation, with White-Black segregation in new destinations in 1990 exceeding the average levels of White-Latino and White Asian segregation in communities of established Latino and Asian presence. However, a second point of difference is that White-Black segregation in new destinations is declining over time while White-Latino and White-Asian segregation in new destinations is increasing over time. By 2010, the average levels of segregation substantially converged across the three groups. The level of White-Black segregation remains higher by 3–5 points in 2010 but at a much smaller margin than the 16–17-point gap in 1990.

We conclude by commenting that it is intriguing to observe that the average level of White-Black segregation in new destination communities is much lower than the levels of segregation observed in communities with established Black presence, and this lower average level of segregation is declining over time. The very high levels of White-Black segregation in areas of established Black presence first emerged in Black new destinations of the North and Midwest nearly a century ago when segregation increased rapidly as the Black population changed from being a low presence to an established presence in a short period of time. Those high levels of segregation persisted for decades and have only begun to decline in recent years. This trend, plus the downward trend we see in average levels of White-Black segregation in new destination communities provides a possible basis for anticipating that segregation in Black new destinations of the present era will not follow the trajectory seen a century ago. However, we must temper this hope based on the fact that the number of Black new destination communities is small, and also that the averages for White-Latino and White-Asian segregation in new destinations are increasing over time and are based on a larger number of cases.

5.8 Understanding Segregation Patterns Across New Destinations and Established Areas of Settlement

In this section of the chapter we move beyond the review of broad summary statistics in the previous section to try to gain a better understanding of how segregation varies across communities that differ on categories of group presence. We do this by also taking account of other characteristics of communities that may be relevant. To serve this goal, we report results for a set of fractional regression models by group comparison where the outcome is the separation index and the covariates are temporal and cross-sectional characteristics of the group comparison and community. In addition to including variables for established areas of settlement and new destinations, we also draw on data from the 2010 census and 2012 American Community Survey 5-year estimates to develop several predictor variables suggested in previous research including: overall population growth, type of community (metropolitan area, micropolitan area, or noncore county), and region. We include measures of industry employment profiles and percent enrolled in college to take account of situations where minoritized group population growth is potentially correlated with broader patterns in the growth and composition of the labor force and higher education opportunities. We present descriptive statistics for the relevant variables in the analysis in Table 5.5, and in Table 5.6 we present the regression coefficients and standard errors for White-Latino, White-Asian, and White-Black segregation.

Table 5.5 Descriptive statistics for communities in 2010
Full size table
Table 5.6 Fractional regression analysis of White-Latino, White-Asian, and White-Black segregation, 2010
Full size table

First, we find that White-Latino segregation is significantly higher in communities with high established Latino presence compared to new destinations as well as areas with established but lower Latino presence and areas with stable and minimal Latino presence. This pattern is consistent with the broad patterns we reported earlier in this chapter and also with the general hypothesis that segregation varies as a positive function of relative group size. Relatedly, we also find that absolute Latino population growth from 1990 to 2010 has a large and significant positive impact on White-Latino segregation.

Regarding the effects of other community characteristics, we find segregation is higher in larger communities (based on natural log of total population) and that White-Latino segregation is significantly higher in communities located in the South and Northeast than in the West. We also find that communities with larger percentages of workers employed in the manufacturing sector have higher levels of White-Latino segregation while communities with higher percentages of workers in the retail sector have lower levels of White-Latino segregation. Finally, we find no significant effect of the percent enrolled in college on levels of White-Latino segregation. Overall, we find that the largest effects on White-Latino segregation are the rate of Latino population growth and categories of group presence, with the contrasts between new destinations and communities of established Latino presence being of special interest here. While segregation is on average lower in new destinations compared to established areas of settlement, the large positive effect of Latino population growth signals that new destinations may be likely to experience increasing segregation over time and, potentially, eventually converge on levels of segregation seen in communities with established Latino presence.

The results for White-Asian segregation closely follow the patterns we just described for White-Latino segregation. White-Asian segregation in communities with a highly established Asian presence is significantly higher than in Asian new destination communities and in communities with low Asian presence. The magnitude of the differences is even larger than those seen for White-Latino segregation. In addition, community population size has a positive relationship with White-Asian segregation, but here the magnitude of the effect is smaller. Where variables relating to region and industry composition of labor force had modest, but statistically significant, associations with White-Latino segregation, we found that only the percentage of the labor force specializing in retail had a significant, and negative, association with White-Asian segregation. However, the percentage of the population enrolled in college is positively associated with White-Asian segregation. As with White-Latino segregation, the most important finding relative to our interests is that Asian new destinations have lower average levels of White-Asian segregation than communities with established Asian presence.

We find that the effects of greatest relevance to this chapter are not only evident in the case of White-Black segregation, but, if anything, are stronger. Specifically, we find that for White-Black segregation, communities with established Black presence have much higher levels of White-Black segregation than Black new destinations and communities with low Black presence. We also find that Black population growth has a strong positive effect on White-Black segregation. The patterns here are thus consistent with the idea that Black households in new destination communities experience lower levels of segregation in comparison to Black households in communities of established Black presence but, due to the positive effect of growing Black presence, sustained Black growth in new destinations would likely lead segregation to converge on the higher levels seen in communities with an established Black presence. Regarding the secondary independent variables, we find a pattern of results that is distinct from the findings for White-Latino and White-Asian segregation. We find larger regional differences where, in comparison with average levels of White-Black segregation in communities in the West, communities in the Northeast and Midwest have higher levels of White-Black segregation. As seen for White-Latino and White-Asian segregation, community specialization in retail has a negative association with White-Black segregation. But unique to White-Black segregation, specialization in the government and manufacturing sectors also has a negative association with segregation. Finally, and uniquely across the three group-specific analyses, the percent of the population enrolled in college has a significant negative effect on White-Black segregation.

The primary focus of the analyses we reviewed in this section is to gain better insight into how White-nonwhite segregation in new destination communities compares with segregation in other communities. The first takeaway point is that the basic findings from the initial review of variation in mean levels of segregation across communities categorized by level of minoritized group presence persist net of controls for a variety of community characteristics that are also associated with White-nonwhite segregation. The findings here are consistent with the general view that minoritized groups experience lower levels of segregation in new destination communities in comparison with communities with established minoritized group presence. The contrast is with another possibility, for which there is scant evidence, that when first arriving in sizable numbers in a community that previously had low group presence, minoritized racial groups would experience a high level of initial segregation out of a combination of one or more dynamics that foster this outcome. One is encountering discrimination and constrained opportunities in the local housing market when the local population does not accept the new group due to social distance tracing to prejudice and/or group differences in culture and socioeconomic characteristics. Another is the rapid formation of ethnic enclave neighborhoods as households in the group co-locate with kin and other co-ethnics connected via social networks of migration. Subsequently, these high levels of segregation may potentially decline as the new group assimilates on culture and/or socioeconomic characteristics and intolerance declines and acceptance grows in the broader community.

The findings here suggest a different trajectory may be more common. White-nonwhite segregation is lowest of all in communities where the minoritized group has a non-negligible, but low and stable, presence. Possibly this is because the group’s presence in the community does not register with the much larger White majority population, thus leaving ethnic relations inchoate and allowing minoritized group households to locate opportunistically where housing is available. Because the size of the minoritized group population is small, enclave formation is not strong because the group cannot support minority-serving institutions (Breton, 1964) that could make enclave neighborhoods attractive. Segregation is then higher, but still generally lower, in new destination communities as the minoritized group’s presence begins to rise rapidly (by definition). This can result from one or more factors such as the co-location of members of the same ethnic group arriving in larger numbers during the surge of immigration/migration, the first beginnings of enclave formation, and the onset of awareness by the White majority that a new ethnic presence is emerging in the community. Subsequently, as minoritized group population growth continues and the minoritized group population becomes established in the community, White-nonwhite segregation also steadily grows as the White population’s initially, possibly benign, low-level awareness of minoritized group presence turns into concern that White dominant position in the community may be threatened by the minoritized group population’s growing presence, thus leading to greater racial intolerance and discrimination on the part of White individuals and institutions (Blalock, 1967; Olzak & Nagel, 1986; Fossett & Kiecolt, 1989; Fossett & Cready, 1998). Under this scenario, ethnic enclave neighborhoods can emerge and persist for multiple reasons including as a response to discrimination and blocked housing opportunities in the broader community as well as having well-developed institutions that independently can attract and retain many minoritized group households.

5.9 Differences in Segregation Measurement When Studying New Destinations

In Chap. 2 we reviewed an important distinction between types of uneven distribution: polarized and dispersed displacement from even distribution. White-Black segregation is especially likely to involve polarized unevenness, where White and Black households live apart from each other in different neighborhoods and thus have little residential contact with one another. We refer to this pattern as prototypical segregation because it is invariably the form of uneven distribution depicted in didactic discussions of segregation measurement (Fossett, 2017) and it also is the form of segregation observed in the best known examples of hypersegregated metropolitan areas such as Chicago, Cleveland, Detroit, and Milwaukee. It also is found in more moderate expressions in many other metropolitan areas across the United States. In cases such as these, where segregation involves a high degree of polarized unevenness, index choice is not very consequential as all segregation index scores will be high regardless of whether one chooses to use the popular dissimilarity index or our preferred, albeit less widely used, separation index.

The situation is very different when segregation involves dispersed unevenness. Index choice matters and researchers who rely solely on the dissimilarity index run the risk of misinterpreting a high score on D as indicating groups are separated across residential space when the separation index would correctly take a low score and indicate that this aspect of segregation is in fact low. Situations involving D-S discordance associated with dispersed unevenness are likely to occur in new destination communities due to the fact that, by definition, the segregation comparison involves an emerging group that is much smaller in size than the other group in the analysis. In general, our primary focus is on the separation index because it signals the presence of polarized unevenness, the pattern that characterizes prototypical segregation and creates the necessary preconditions for group inequality on location-based outcomes. Here we also take special interest in identifying situations involving dispersed unevenness because they are common in new destination communities and this fact is not widely appreciated.

5.9.1 Myth or Fact: Low Minoritized Group Size Necessarily Leads to Low Values for the Separation Index

We take a moment here to forcefully debunk a mistaken belief regarding the separation index. The belief in question is that S somehow necessarily takes low values when groups are imbalanced in size, as will always be the case for emerging minoritized group populations in new destination communities. Simply put, this view is unfounded and should be discarded. Detailed review of the issue in Fossett (2017) points out several relevant findings including: there is no formal basis for this view, it is easy to construct simple examples that refute the view, and one can also find empirical examples that refute the view. Thus, we offer the following statement:

The separation index is more reliable than any other index in being able to indicate when uneven distribution is polarized, such that both groups in the segregation comparison disproportionately reside in mostly homogeneous (same-group) neighborhoods.

Segregation involving polarized unevenness may or may not occur when groups are imbalanced in size. Whether it does occur is a matter of social process; not an artifact of index choice. When uneven distribution is polarized, groups live apart from each other and occupy different neighborhoods. The separation index will always correctly take a high value in this situation and, equally importantly, the separation index will always correctly take a low value when polarized unevenness is absent. In this regard, the separation index differs from other indices – particularly the dissimilarity index and the Gini index – that give ambiguous signals about segregation because they will take high values when the separation index does but also will take high values when uneven distribution is dispersed and does not involve group separation.

The primary methodological concern when groups are imbalanced in size is a concern that applies with equal force to all indices of uneven distribution, not specifically the separation index. It is the concern of whether the spatial units used in the analysis are adequate for the task of measuring segregation. The issue is especially important in nonmetropolitan communities where larger spatial units such as census tracts cannot reliably register the full extent of any aspect of segregation and are especially incapable of revealing when a small group is concentrated in homogeneous neighborhoods. We address this concern in our study by using data for census blocks. The separation index will reliably detect and signal the presence of polarized unevenness involving small groups when using block data. Equally important, a low value on the separation index obtained using block data will reliably indicate that polarized unevenness and group separation is not present.

As a final technical side point, we note that the dissimilarity index will be more likely than the separation index to take intermediate and even high values when larger spatial units such as census tracts are used to measure segregation in nonmetropolitan communities. But this should not be construed as suggesting D is valid and reliable for measuring group separation. To state it bluntly, D is never reliable for measuring group separation. When polarized unevenness is manifest at the block level but not at the tract level, the separation index will correctly yield a high score using block-level data and a low score using tract-level data. In this circumstance, D will yield a high score using block-level data, a necessary result when the value of S is high, and D may also yield an intermediate or high score using tract-level data, even when the value of S is low, when polarized blocks associated with predominately minoritized group neighborhoods are found only in one or two tracts – as would be likely in a nonmetropolitan community. This is because the polarized unevenness at the block level will be manifest as dispersed unevenness at the tract level and D, but not S, will respond strongly to this pattern of unevenness. The problem of course is that the value of D cannot sustain an unambiguous conclusion regarding the nature of segregation.

As just described, an intermediate score could occur because group separation across small spatial units is registered as dispersed unevenness across larger units. Or, it could occur simply because segregation involves only dispersed unevenness across both smaller and larger spatial units. What then can safely be inferred about the pattern of segregation? As always, discordant values of D and S obtained using larger spatial units definitively signal dispersed unevenness at that level of spatial resolution. After that, nothing more can be inferred with confidence. The fact that this result is compatible with multiple, distinctly different patterns of segregation across smaller spatial units may be intriguing. But ultimately, nothing specific can be safely inferred. The only way to clarify the situation is to use smaller spatial units that are well-suited for the task of measuring segregation in new destinations, particularly those that are also nonmetropolitan communities.

5.10 Findings for Dispersed and Polarized Unevenness in New Destinations

In this section we distinguish between patterns of polarized and dispersed unevenness in new destinations by comparing values of the separation index and the dissimilarity index. Concordant values of D and S indicate polarized unevenness, while discordant values on D and S indicate dispersed unevenness. In Table 5.7 we report average values of both the separation index, the measure we assign priority to throughout this book, and average values of the dissimilarity index for Latino, Asian, and Black new destination communities. Comparing the values of these two indices provides insight into the nature of the pattern of segregation in new destinations. The results document clear patterns of systematic discordance between the separation index and dissimilarity index in Latino new destinations with, as is necessarily the case, values of D being higher than values of S. In 1990 the average value of D for White-Latino segregation was 30.9 which, in comparison with the average value of S of 9.5, is higher by a factor of three and a margin of more than 20 points. These results suggest, and close review of individual cases confirms, that the pattern of White-Latino segregation in the typical Latino new destination community is one of dispersed unevenness. In this situation, most Latino households live in neighborhoods where the representation of White households among their neighbors, while often technically below parity on proportion White with the community overall, is consistently high and quantitatively close to parity. At the same time, most Latino households live in neighborhoods where the representation of Latino households among their neighbors is also close to parity, which in new destination communities means that the level of Latino presence among neighbors is low. Under this pattern, White and Latino households generally share similar neighborhood contexts, and this minimizes the potential for segregation to produce White-Latino inequality on location-based outcomes.

Table 5.7 Comparing segregation indices in Latino, Asian, and Black new destinations, 2010
Full size table

From 1990 to 2010 average values of D and S for White-Latino segregation both rise by approximately 10 points to stand at 39.6 and 19.3, respectively. D-S discordance remains high in absolute terms as the average value of D remains 20 points higher than the average value of S. But D-S discordance is falling in relative terms as the average value of D in 2010 is higher than S by a factor of two, down from a factor of three in 1990. The rising average value of S relative to D indicates a transition from highly dispersed unevenness to moderately dispersed unevenness, but not fully polarized unevenness. In terms of Latino residential experiences, this means that a larger percentage of Latino households in Latino new destination communities are residing in neighborhoods where the presence of White and Latino households is further from parity than was initially the case. These changes indicate that, on average, White and Latino households are less likely than before to share neighborhood contexts and thus the potential for segregation to produce White-Latino inequality on location-based outcomes increased over the two decades. In other words, from 1990 to 2010, White-Latino segregation moved away from dispersed unevenness and toward polarized unevenness and prototypical segregation, the pattern typically seen in communities with high levels of established Latino presence.

Importantly, while we observe that values of S for White-Latino segregation are rising at a faster rate than values of D, thus indicating a transition from dispersed unevenness toward more polarized unevenness, this was not a foregone conclusion, as other trends were possible. This is why crucial aspects of trends in White-Latino segregation in new destinations cannot be established by examining only the value of the dissimilarity index. By conventional interpretation, the average value of D was already at a medium level in 1990 and stayed in the medium range when rising by almost 10 points from 30.9 in 1990 to 39.6 in 2010. But the value of D by itself cannot reveal whether the pattern of segregation at any point in time involves the relatively benign pattern of dispersed unevenness with a high level of group co-residence and shared residential experiences or the more consequential pattern of polarized unevenness where both groups are separated in space based on being disproportionately concentrated in neighborhoods where they do not share location-based outcomes with members of the other group.

To definitively establish what is happening for this aspect of segregation, one must examine values of S. The reason for this is that values of D and S can not only be discordant at a point in time, they can and sometimes do vary independently over time, including potentially moving in opposite directions. Thus, when D is increasing by over 10 points from 1990 to 2010, the value of S could be falling, stable, or increasing and each result would indicate something distinct about the segregation pattern. If the value of S is rising at roughly the same rate as the value of D and is not rising rapidly, it indicates the pattern of segregation is holding steady at the initial level of dispersed unevenness. If the value of S is falling or stable, it indicates the pattern of segregation is moving toward even greater dispersion in uneven distribution wherein the fraction of Latino households living in below-parity neighborhoods is increasing (leading to rising D) but the below-parity neighborhoods are generally moving closer to parity (leading to falling S). Finally, if the value of S is rising at a faster rate than the value of D, it indicates the pattern of segregation is moving toward more polarized unevenness.

We find the broad pattern for White-Asian segregation in new destinations is generally similar to the pattern just described for White-Latino segregation in new destinations with White-Asian segregation in new destinations initially taking a form of dispersed unevenness in all three decades but moving toward higher levels of polarized unevenness between 1990 and 2010. We also find two notable points of difference for White-Asian segregation in new destinations. The first is that the initial pattern of dispersed unevenness in 1990 was more pronounced for White-Asian segregation and the second is that movement toward more polarized unevenness over time was weaker for White-Asian segregation. In 1990, the separation index indicated low levels of White-Asian segregation while at the same time the dissimilarity index indicated moderately high levels. This is a clear sign that while there was uneven distribution occurring in 1990 between White and Asian households, it was dispersed unevenness and Asian households were not living in fundamentally different neighborhoods. The discordance between the two scores drops only slightly in 2010 as the separation index shows increasing polarized unevenness between White and Asian households in new destinations over time. However even in 2010, the separation index shows only medium levels of segregation while the dissimilarity index indicates even higher levels of segregation, which means that White-Asian uneven distribution in new destinations is still more dispersed than polarized.

Regarding the first point, systematic discordance between the separation index and dissimilarity index for White-Asian segregation in Asian new destinations is even more pronounced in 1990 than was the case for White-Latino segregation in Latino new destinations. The average value of D for White-Asian segregation was 39.9, which is higher than the average value of S by a factor of more than four and a difference of more than 31 points. Thus, White-Asian segregation in a typical Asian new destination community is characterized by highly dispersed unevenness in which a large fraction of Asian households live in below-parity neighborhoods where White households are technically underrepresented among their neighbors but with shortfalls from parity that are quantitatively small. Accordingly, in Asian new destinations, most Asian households live in neighborhoods where the presence of White and Asian households among neighbors is close to levels expected under random distribution, which would consist of a very high presence of White households and a very low presence of Asian households. Therefore, Asian households generally reside in the same neighborhood contexts as White households and the logical potential for White-Asian inequality on location-based outcomes is very low.

From 1990 to 2010 average values of D and S for White-Asian segregation both increased, reaching the values of 17.5 and 46.0, respectively. The separation index increased more both in absolute terms (8.6 points compared to 6.1 points for D) and in relative terms (up over 96 percent compared to about 15 percent for D). However, because the initial absolute D-S discordance was more than 31 points, the D-S difference in 2010 remains extremely high at more than 28 points even though the average value of S increased by 2.5 points more than the average value of D. The change in the relative discordance of D and S was more noticeable as the average value of D was larger than the average value of S by a factor of more than four in 1990, and this dropped to less than a factor of three in 2010. The increase in the average value of S from 8.9 in 1990 to 17.5 in 2010 does indicate appreciable movement in the direction of greater polarization in White and Asian residential distributions. But the main finding is that the level of polarization remains low in 2010 and is lower for White-Asian segregation in new destinations than for White-Latino segregation in new destinations.

The results for Black new destinations differ from the results for Latino and Asian new destinations on multiple points. Our earlier discussion of variation in White-nonwhite segregation across communities by level of minoritized group presence established that average levels of segregation in new destinations were lower than in communities with highly established minoritized group presence. The differences were large across all group comparisons but the difference of over 33 points for White-Black segregation was easily the largest. However, at the same time the average level of the separation index of 26.7 for Black new destinations in 1990 was much higher than the averages of 9.5 and 8.9 for Latino and Asian new destinations, respectively. In this regard, White-Black segregation in new destinations follows the earlier finding that White-Black segregation is higher than White-Latino and White-Asian segregation across all categories of minoritized group presence.

Another point of difference is that the discordance between the dissimilarity index and the separation index in new destination communities is appreciably lower for White-Black segregation than for White-Latino and White-Asian segregation. The magnitude of the difference in average levels of D and S of 27.7 points is not especially small since it falls between the levels seen for White-Latino segregation (21.4) and White-Asian segregation (31.1). But the relative comparison of average values for D and S is the smallest across the three White-nonwhite comparisons with the average value of D for White-Black segregation being higher than the average value of S by a factor of just over two (2.0) compared to a factor of over three (3.3) for White-Latino segregation and a factor of over four (4.5) for White-Asian segregation. The low ratio of D to S in combination with the much higher level of S helps clarify that White-Black segregation in new destinations involves uneven distribution that is more polarized and less dispersed than White-Latino and White-Asian segregation and so creates more potential for White-Black inequality on location-based outcomes. We place this last point in further perspective by noting that, while White-Black segregation in new destinations is more polarized and less dispersed in comparison to White-Latino and White-Asian segregation in new destinations, White-Black uneven distribution is still much more dispersed and less polarized than in major metropolitan areas such as Los Angeles and Houston, where S is in the low 60s. Compared to Chicago, the definitive example of a maximally polarized city with S at 79 and, with D only 6 points higher, a D-S ratio of 1.1, White-Black segregation in new destinations is hardly comparable. So, in contrast to these examples, Black new destination communities have unevenness that is much more dispersed and thus a large portion of Black households that live in below-parity neighborhoods experience a presence of White households among neighbors at levels that rarely occur in prototypically segregated areas.

The final point of contrast for White-Black segregation in new destinations is that it is stable or even declining slightly from 1990 to 2010 while White-Latino and White-Asian segregation in new destinations is rising. However, because the initial level for White-Black segregation is higher, it is trending toward convergence on the patterns seen for White-Latino and White-Asian segregation, not decreasing to levels below them.

5.11 Highlighting Measurement Issues: The Case of Worthington, Minnesota

Many of the findings we report in this chapter cast a new and different perspective on segregation in new destinations in comparison to findings previously reported in the literature. There are two main reasons for this. One is that we focus attention on aspects of uneven distribution – namely, polarization and group separation – that have been neglected in previous research. The other is that our study is the first to use new methods for measuring segregation. We do so because the task of measuring segregation of small groups in new destination communities presents difficult problems that can easily distort index scores and raise concerns regarding whether findings based on them are fully trustworthy. The new methods for measuring segregation we use in this book were developed to specifically address these problems. One of their attractive characteristics is that they only yield different results in situations where conventional measurement practices yield questionable and potentially misleading results. Thus, when past measurement practices yield trustworthy results, the new methods we use will replicate these results. But, when the new methods we use yield different results, these results will be superior and should be preferred over results obtained using past measurement practices. Given the importance of this issue, we use this section of the chapter to present an in-depth technical review of the case of White-Latino segregation in Worthington, MN, a micropolitan area and Latino new destination in the Midwest with patterns of White-Latino segregation that are typical across many Latino new destination communities.

The Latino population in the Worthington, MN micropolitan area grew rapidly in both absolute and relative terms from 1990 to 2010. In 1990, the number of Latino households in the community was just 70 and comprised slightly less than 1 percent of the 7,682 total households in the community. By 2000 the community had a net gain of an additional 471 Latino households to stand at 541, a more than sevenfold increase, leading Latino households to increase to 6.5 percent of all households. By 2010 the community had a similar net gain of an additional 481 Latino households and nearly doubled to stand at 1,022 households, comprising 12.9 percent of all households. Throughout this period, the average size of Latino households was larger than the size of White households and the disparity grew in absolute and relative magnitude with each decade. By 2010 the average Latino household size was more than double that of White households. Consequently, the Latino percentage representation among persons was larger, and growing faster than the Latino percentage representation among households. The demographic trends and patterns observed for Worthington are not in any way unique. To the contrary, they are typical of Latino demographic trends in new destination communities. Moreover, these demographic circumstances – an initially small, rapidly growing population in a nonmetropolitan setting – exactly embody the sort of scenario where we should be concerned by how analyses of patterns and trends in segregation are affected by the following two issues relating to segregation measurement: the distorting impact of index bias and the related consequences of measuring segregation of persons rather than households, and neglecting to consider whether displacement from even distribution is dispersed or polarized.

In Table 5.8 we present results for scores of the separation index and the dissimilarity index obtained using data for census blocks under selected combinations of methodological choices regarding microunit (segregation of persons versus households) and whether or not the index score has been corrected for index bias. Our point of reference for discussing the impact of methodological choices on the value of each index score is the unbiased separation index score obtained using data for households, as we hold scores obtained under this particular combination of choices as the benchmarks against which other results should be evaluated. The unbiased version of the separation index calculated using data for households in Worthington starts at a low value of 3.7 in 1990, increases to 22.5 in 2000, and increases further to 27.7 in 2010. The unbiased version of the dissimilarity index calculated using data for households starts at a value of 30.2 in 1990, increases to 55.3 in 2000, and decreases slightly to 51.4 in 2010.

Table 5.8 White-Latino segregation in Worthington, MN, 1990–2010
Full size table

The first thing we note regarding how these scores compare to scores obtained using other combinations of practices is that index bias is an important issue regardless of what index is used and whether the micro-level units are persons or households. The impact of bias is never negligible, but it varies dramatically in magnitude. The smallest impact of bias is seen for the standard version of the separation index calculated using data for households. It runs 6–7 points higher than the unbiased version of the separation index. Shifting to using person data increases the impact of bias to much higher levels because the scores register “lumpy” spatial distributions due to the fact that persons typically locate as members of households that are homogeneous on race. Person-level adjustments for bias partially reduce bias but are insufficient because they do not take account of a person’s co-location with other members of their household. Bias in the separation index is unaffected by imbalance in group size, so bias levels remain relatively stable across time even though Latino group size is changing rapidly. However, bias in the separation index is affected by changes in “effective neighborhood size,” the size of the combined count of the two groups in a given spatial unit. Thus, to the extent that bias in S changes over time, it is either random or possibly reflects changes in the number of households per block over time and/or changes in the relative presence of other groups in the population.

The dissimilarity index is much more susceptible to index bias than is the separation index, and it is well known that bias in D can reach extreme levels when groups are imbalanced in size and segregation is measured using data for small spatial units (Winship, 1977; Fossett, 2017). Thus, while it is alarming, it should not be surprising to see that the impact of bias for D is extremely high in 1990 with values in the range of 40–55 points across the three alternative options for measurement. It is also alarming, but again should not be surprising, to see that the impact of bias on D differs dramatically over time. The reason for this is that bias in D is sensitive to imbalance in group size, and imbalance in group size declines substantially over time due to the rapid growth of the Latino population. As a result, bias falls to still high, but far less extreme, values in the range of about 13–15 points in 2010.

These findings have several implications for how our findings may differ from findings reported in previous studies focusing on new destination communities in nonmetropolitan settings. First, and most obviously, previous studies overwhelmingly use the dissimilarity index and the analysis here shows that not only are values of D highly distorted by index bias, but the magnitude of the impact of bias on values of D is changing dramatically over time. Segregation scores are at their highest for White-Latino segregation in Worthington when using the dissimilarity index to measure segregation of persons or households without correcting for bias, with scores of 81.3 and 85.3, respectively, in 1990. If accepted at face value, these scores indicate a level of segregation comparable to that observed for White-Black segregation in large hypersegregated metropolitan areas like Chicago, Detroit, or Milwaukee, all of which are consistently found at or near the top of lists of the most segregated metropolitan areas in the United States. The problem with these scores, which correspond to the raw data used in almost all analyses of segregation in new destinations, is that they absolutely cannot be accepted at face value. To the contrary, these scores are fatally compromised by index bias and taking them at face value will lead to grossly incorrect conclusions about the nature of White-Latino segregation in communities like Worthington.

Close inspection of GIS-based mapping of group distributions across census blocks in Worthington shows no evidence that White-Latino segregation in 1990 (or in 2000 or 2010) is in any way comparable to White-Black segregation in Chicago. In Chicago, most Black households reside in neighborhoods where 80–100 percent of neighboring households are Black and no or almost no neighboring households are White. In Worthington in 1990, almost no Latino household resides in a neighborhood where 80–100 percent of neighboring households are Latino and most live in neighborhoods where 80–100 percent of neighboring households are White. The case of White-Black segregation in Chicago is known for the large and expansive region of predominantly Black neighborhoods in the southern region of the city, the existence of which is the demographic foundation for the concept of hypersegregation. Nothing like this exists for Latino households in Worthington. The only blocks that are predominantly Latino in population are occupied by one or two Latino households. There are no significant clusters of contiguous blocks that are predominantly Latino. In every meaningful way, White-Latino segregation in Worthington is different from White-Black segregation in Chicago. This drives home the point that scores of the dissimilarity index obtained via conventional methodological practices used in previous research cannot be taken at face value. Instead, they must be called into question and considered carefully to avoid reaching unfounded conclusions about White-Latino segregation in new destination communities.

The unbiased scores for the dissimilarity index show clear improvement over the standard versions of the same index. The value of the unbiased version of D calculated using data for households in 1990 is 30.2, some 55 points lower than the value of D obtained using the standard version with data for households. The source of the astounding impact of bias on D is surprisingly easy to explain. D registers the White-Latino difference in percentage of households that attain parity contact with White households at the neighborhood level. Standard versions of D assess percent White in the spatial unit based on the combined count of all White and Latino households. But in new destinations, Latino households are by definition a small fraction of the population so the presence of even a single Latino household in a census block will in most cases cause percent White for the block to fall below parity (i.e., below percent White for the city overall). In fact, in Worthington in 1990 pairwise percent Latino is about 1 percent, so the standard calculation would result in below-parity status being assigned to any block with fewer than 100 White and Latino households combined with a single Latino household. That accounts for all but 2 of the 1,120 blocks in Worthington in 1990. There are 70 Latino households in Worthington in 1990. Of these, 37 reside in a block where they constitute the only Latino household on the block and the block has fewer than 100 households. Under the standard calculation of D, all of these Latino households are designated as residing in below-parity neighborhoods despite the fact that all of their neighbors are White and their only contact with Latino households stems from self-contact. This is the source of the extreme level of bias in the result for the standard calculation of D for Worthington in 1990.

The unbiased version of D is 55 points lower than the standard version of D because it eliminates bias by calculating contact based on neighbors and excluding self-contact, the sole source of index bias. Thus, the unbiased version of D registers the difference between the percentage of White and Latino households that attain parity-level contact with White households among neighboring households. In these calculations, the 37 Latino households that reside on blocks where they are the only Latino households and have only White neighbors are correctly treated as having parity-level contact with White households (since 100 percent of their neighbors are White). There is no way to accept the standard scores for D as trustworthy for measuring segregation in new destinations in nonmetropolitan settings where group size is small and it is crucial to use block-level data. The standard scores for D are fatally flawed and egregiously misleading. In contrast, the unbiased scores for D are correct and trustworthy and their difference from the standard scores is easy to explain. Accordingly, our analyses are based only on scores obtained using the unbiased versions of segregation indices.

The decision to focus on unbiased index scores is consequential for more than just correctly assessing the level of segregation in new destinations. It also is important for correctly assessing how segregation is changing over time in new destinations. The case of Worthington is useful for illustrating how conclusions about trends in segregation can vary dramatically when using standard and unbiased versions of indices. The scores for D calculated using standard formulas with data for persons suggest a large decline of 14.4 points in White-Latino segregation from 81.3 in 1990 to 66.9 in 2010. The scores for D calculated using standard formulas with data for households suggest an even larger decline of 19.9 points in White-Latino segregation from 85.3 in 1990 to 65.4 in 2010. In contrast, scores for D calculated using the unbiased formulas with data for households suggest the exact opposite trend with the value of D increasing 21.2 points from 30.2 in 1990 to 51.4 in 2010. The unbiased scores can be trusted to reveal the true trend in segregation. The dramatic reversal in findings for the time trend occurs because the standard score for D in 1990 is inflated by 55.1 points by index bias but is “only” inflated by 14 points in 2010. Eliminating the impact of bias on the value of D thus results in a massive 40.8-point change in findings from a trend of a 19.9-point decline in segregation using the standard version of D to a 21.2 increase in segregation using the unbiased version of D.

We justify using scores for the unbiased formulation of the separation index over scores for the standard version based on the same logic. Index bias can distort standard scores and raises concerns that findings based on them are questionable and possibly misleading. That said, we should note a very big difference between the separation index and the dissimilarity index. It is that, in general, the standard version of the separation index is not as susceptible to index bias as the standard version of the dissimilarity index. And, equally importantly, the impact of bias on the separation index is more uniform across cases while the impact of bias on the dissimilarity index can vary greatly across cases. Consequently, when segregation is measured using the separation index, one is less likely to encounter the egregious, pathological results along the lines just discussed for trends in the dissimilarity index in Worthington. This is evident in the results for the separation index for Worthington. The scores for the standard version of the separation index are inflated by bias and are about 6–7 points higher than the scores for the unbiased versions of the index. But, in decided contrast with the dissimilarity index, the impact of bias on scores for the separation index is fairly stable over time even as the Latino percentage in the population is changing rapidly. So, both versions of the index suggest White-Latino segregation is increasing by about 23–25 points. The same trend also is observed when switching from data for households to data for persons and the reason for this is the same; the bias impact associated with using person data instead of data for households tends to be uniform across cases. Thus, in comparison with the dissimilarity index, bias for the separation index is less likely to impact findings regarding variation in segregation across communities and/or over time.

The discussion we offered above provides both a rationale for why we measure segregation using unbiased versions of segregation indices in combination with data for households rather than persons and reviews an example where the choice has practical consequences. Now we turn to the question of why we assign priority to measuring White-nonwhite segregation in new destinations using the separation index over the dissimilarity index. Here the case of White-Latino segregation in the new destination community of Worthington, MN is again useful. The unbiased dissimilarity index based on households rises from a medium level (30.2) to a high level (51.4) from 1990 to 2010. In contrast, the unbiased separation index begins at a very low level (3.7) in 1990 and rises to a medium level (27.7) by 2010. There is clear discordance between the scores for the two indices; the raw score difference is similar at both points in time, 26.5 points in 1990 and 23.7 points in 2010, but the relative comparison changed dramatically as the dissimilarity index is larger by a factor of 8.2 in 1990 but only by a factor of 1.9 in 2010. We explain below that we gain most of the information we need from the scores for the separation index, but we gain additional interesting information from the contrast between scores for the separation index and the dissimilarity index.

In Worthington in 1990, the value of the unbiased dissimilarity index for White-Latino segregation based on households is 30.2 while the unbiased separation index based on households is only 3.7. This value of the dissimilarity index is in the medium range. But the low value of the separation index indicates that White-Latino segregation in Worthington involves a pattern of highly dispersed unevenness. Because the dissimilarity index is insensitive to whether uneven distribution is dispersed or polarized, its value provides no basis for inferring which pattern prevails. Examining the separation index brings clarity. Under polarized unevenness, the value of the separation index would be in the range of 24–30 (at or above 80 percent of the value of the dissimilarity index). But, instead, its observed value of 3.7 is very close to zero. This provides a definitive signal that White-Latino segregation in Worthington follows a pattern of dispersed unevenness where White and Latino households alike reside in neighborhoods that are quantitatively close to parity regardless of whether they are technically below or above parity. The low value of the separation index also strongly indicates that in the early stages of Latino migration and settlement in Worthington in 1990 Latino households do not reside in homogeneous neighborhoods. This is consistent with the close review of block-level outcomes for Latino households discussed earlier which noted 37 of 70 Latino households lived in blocks where all of their neighboring households were White. We additionally note here that every Latino household that resides in a block with five or more households has more White neighbors than Latino neighbors.

The value of the dissimilarity index increases from 30 to over 50 in the later decades, thus moving into the high range. But just as we could not know from the value of the dissimilarity index whether the initial pattern of uneven distribution in 1990 was polarized or dispersed, we cannot know whether the increase in the value of the dissimilarity index reflects a change in the underlying pattern of segregation. If one only knows the value of the dissimilarity index, the logical possibilities range from group separation declining (i.e., uneven distribution becoming more dispersed) to group separation increasing (i.e., uneven distribution becoming more polarized). To know what is occurring, one must examine values of the separation index. The S index indicates that White and Latino households have become more residentially separated, with the average group difference in neighborhood proportion White rising from 3.7 to 27.7 in just two decades, as directly measured by the separation index.

A key point here is that the value of the separation index is telling the story of primary interest – namely, whether groups are living together and experiencing similar neighborhood contexts or living apart and potentially experiencing unequal neighborhood contexts. For this concern, additionally learning the value of the dissimilarity index adds little to no relevant information because its relationship to this aspect of uneven distribution is inherently ambiguous. Once the value of the separation index is known, the extent of group separation is known. Values of the dissimilarity index can range from being approximately equal to the value of the separation index or higher, possibly much higher. Knowing that the value of the dissimilarity index is on the low end of this range indicates that below-parity neighborhoods skew toward having higher levels of Latino presence (i.e., depart from parity by larger amounts), a hallmark of polarized unevenness. Knowing that the value of the dissimilarity index is on the high end of its possible range indicates that below-parity neighborhoods include a mixture of not only neighborhoods with higher Latino presence but also a larger number of below-parity neighborhoods that are quantitatively close to parity, a hallmark of dispersed unevenness.

Note that once the value of the separation index is known at a single point in time, additionally knowing whether its level occurs under a pattern of dispersed or polarized unevenness does not have important implications for the question of whether segregation is conducive to racial stratification in neighborhood outcomes, it will be the same under either pattern. However, when the level and trend of change in the separation index are known, additionally knowing the level and trend in the dissimilarity index can provide a basis for speculating about the trajectory of segregation. When the value of the separation index is increasing, finding that values of dissimilarity are higher and/or rising faster might be seen as a leading indicator of future progression toward a pattern of neighborhood polarization and group separation. When the value of the separation index is decreasing, the finding that values of the dissimilarity index are higher and/or declining more slowly might be seen as a trailing indicator of a fading pattern of neighborhood polarization and group separation (see also Chap. 4).

We present polarization charts for White-Latino segregation in Worthington in 1990 and 2010 in Fig. 5.1 to visualize the nature of uneven distribution in this new destination community and clarify how it is registered by the separation index and the dissimilarity index. The charts depict the observed distributions of White and Latino households across levels of presence of White households among neighboring households in 1990 and 2010. The chart for 1990 shows that all households, White and Latino alike, had more White neighbors than Latino neighbors (i.e., percent White among neighbors was greater than 50) and that the overwhelming majority of both White and Latino households – specifically, 97.8 percent and 87.1 percent, respectively – were living in neighborhoods where at least 90 percent of the neighboring households were White. As a result, average levels of contact with White households among neighbors was very high; 99.1 percent for White households and 95.4 percent for Latino households. The difference between these two values yields the value of the unbiased separation index of 3.7.

Fig. 5.1
4 dual-line graphs represent white and Latino households. Graphs A and B are relative frequency versus proportion white in 1990 and 2010. Graphs C and D are relative frequency versus contact with whites scaled for D in 1990 and 2010. In Graph A, both lines reach the highest peak of 98 approximately.

Observed distributions of White and Latino households by proportion White among neighbors

Full size image

The unbiased dissimilarity index of 30.2 in 1990 also reflects a simple group difference in contact with White households among neighbors. But it summarizes the patterns in a crude way that exaggerates the underlying quantitative differences on contact with White households. To review from our discussion in Chap. 2, the dissimilarity index registers contact as either 0 or 100 based on whether contact matches or exceeds parity. Since uneven distribution is highly dispersed, not polarized, the contact scores registered by the separation index are generally very close to parity for both groups. So, the typical rescaling of those contact scores to extreme values of 0 and 100 (scaled from 0 to 1) when calculating the dissimilarity index exaggerates the group difference. To the best of our knowledge, no rationale has ever been offered to justify rescaling contact in this way before comparing group differences in contact with White households. In part this is because few researchers were aware that the dissimilarity and separation indices, along with other indices of uneven distribution, reflected group differences in contact with White households with the difference in index scores tracing solely to how original contact scores are registered by the index. For the separation index, contact scores are registered as observed and thus take values over the full logical range of 0 to 100. For the dissimilarity index, contact scores are rescaled with all intermediate scores being assigned to the extreme values of either 0 or 100.

In light of this, the polarization chart for 1990 provides a visual insight into how the dissimilarity index comes to take much higher values than the separation index when uneven distribution is dispersed instead of polarized. Close review of patterns of contact for Latino households provides further insight into how separation registers information about group differences in contact with White households in a way that is most relevant for the potential implications of residential segregation for racial stratification. Of the 70 Latino households in 1990, 37 reside in above-parity neighborhoods; all of these households have only White neighbors. The remaining 33 Latino households reside in below-parity neighborhoods. Not one has more Latino neighbors than White neighbors, as the lowest value for percent White among neighbors for these households is 63 percent. Furthermore, the median value of contact with White households among neighbors for Latino households residing in below-parity neighborhoods is 95 percent. Thus, close inspection of residential outcomes for Latino households shows that even when focusing only on those that reside in below-parity neighborhoods, Latino households in Worthington live alongside White households in neighborhoods that are overwhelmingly White. Consequently, Latino households in Worthington experience essentially the same neighborhood contexts as White households experience and thus White-Latino inequality on location-based outcomes is simply not possible. This aspect of segregation is accurately reflected in the very low value of the separation index of 3.7. The fact that the value of the dissimilarity index of 30.2 is more than 8 times higher does not require any reconsideration of the conclusion. We would characterize the high value of the dissimilarity index as a curious byproduct of its crude construction if not for the fact that so many people rely solely on this measure to evaluate segregation.

A final point about Worthington in 1990 is that the low value of the separation index is not at all a necessary outcome. The median value for number of households on a census block is 13. So, if Latino households in Worthington were distributed in the same way as Black households are distributed in Chicago, 80 percent of Latino households would reside in blocks that are at least 65 percent Latino. That pattern is absolutely feasible. It would involve 56 Latino households residing in 6 blocks of typical size (i.e., 13 households) with 9 Latino households and 4 White households. Discussion in Chap. 2 of this work and also Fossett (2017) review cases where high values of the separation index are in fact observed under similar demographic settings. The fact that this kind of pattern is seen for Black households in Chicago and in some nonmetropolitan settings but not for Latino households in Worthington is due to differences in the social dynamics of residential distribution, not to any technical limitations of the separation index.

The polarization chart for 2010 shows a significant change in the pattern of White-Latino segregation. The Latino presence among all households increased to 12.9 percent and, due to larger Latino household size, to an even larger 22.5 percent of total population. Latino presence in the White-Latino comparison is 13.9 percent for households and 25.1 percent for persons. Thus, the parity threshold for contact with White households using data for households fell from 99.1 in 1990 to 86.1 in 2010. This substantial change in the racial-ethnic composition of the community, characteristic of all new destinations, does carry implications for patterns of contact, but it does not have any implications for changes in how group differences in contact determine values of the separation index and the dissimilarity index. This is illustrated in Fig. 5.2 which presents polarization charts for unbiased contact in 1990 and 2010 when the White and Latino households in Worthington are randomly assigned across the blocks where White and Latino households reside. The upshot of the charts is that patterns of contact with White households shift down from a very high level in 1990 (91.1 percent) to a somewhat lower level in 2010 (86.1 percent) but the distributions of contact with White households that occur under random assignment are the same for White and Latino households in both decades. Therefore, the expected group difference in level of contact with White households is zero (0.0) in both decades for both the separation index and the dissimilarity index.

Fig. 5.2
4 dual-line graphs represent white and Latino households. Graphs A and B are relative frequency versus proportion white in 1990 and 2010. Graphs C and D are relative frequency versus contact with whites scaled for D in 1990 and 2010. In A, both lines reach the highest peak of 100, approximately.

Expected distributions of White and Latino households by proportion White among neighbors under random distribution (per bootstrap simulation)

Full size image

Returning to the polarization chart for observed distributions of White and Latino households in 2010 in Fig. 5.1 we note that the percentage of households living in neighborhoods where 90–100 percent of neighboring households are White dropped significantly for both White and Latino households, but most dramatically for Latino households. The polarization charts in Fig. 5.2 depict expected group distributions for contact with White neighbors under random distribution and clarify that the changing demographic composition of the community accounts for only some of the change. A major part of the change is that observed distributions depart more from expected distributions in 2010 compared to 1990, leading to increasing group separation because highly dispersed unevenness was giving way to a pattern of emerging polarized unevenness. In 1990, the observed distributions of White and Latino households by level of contact with White neighbors closely follows the distributions expected under random assignment, producing a very low value for the separation index and a higher value of the dissimilarity index, which reflects a benign pattern of dispersed unevenness.

In 2010 the expected distributions of contact with White households among neighbors under random distribution in Fig. 5.2 shift left toward lower levels for both White and Latino households. But unlike in 1990 when the percentages of White and Latino households in the 90–100 category are very close to the percentages expected under random distribution, in 2010 the percentage of Latino households is far below the expected level and the percentage of White households is far above the expected level. Additionally, the percentages of Latino households living in neighborhoods with White presence among neighbors at 70 percent or lower (Latino presence at 30 percent or higher) increase beyond expected levels by large margins. This reflects a rapid transition from dispersed unevenness to polarized unevenness, which is reflected in the separation index rising from 3.7 to 27.7. The dissimilarity index also rises from 30.2 to 51.4. But the most notable change is the convergence of values for the dissimilarity index and the separation index from a ratio of over 8 in 1990 to a ratio of only 1.9 in 2010. This indicates that, while the level of segregation in 2010 is in the medium range, it rose rapidly in just two decades and uneven distribution moved strongly in the direction of becoming more polarized. These important changes in the nature of White-Latino segregation in Worthington can only be captured by examining values of the separation index and using unbiased versions of segregation indices.

All of this highlights the value of using new methods to measure segregation and also the value of examining scores for the separation index over the dissimilarity index. Following conventional practices used in previous research, we would measure White-Latino segregation in Worthington in 1990 using person data and obtain a value of 81.3 for the dissimilarity index. Since this high value would be comparable to values seen for exemplars of high-segregation such as White-Black segregation in Chicago, many would be tempted to assume White-Latino segregation in Worthington therefore involves a high level of group separation with Latino households residing in all-Latino neighborhoods and White households residing in all-White neighborhoods. Taking note of the separation index value of 15.3 based on person data provides a first indication that the nature of White-Latino segregation in Worthington is very different from White-Black segregation in Chicago.

Based on understanding the technical properties of the two indices, two factors can potentially explain why the separation index takes a much lower value than the dissimilarity index. One is that, because the separation index is much less susceptible to bias, the value of the dissimilarity index is higher because it is inflated by index bias. The other is that the segregation pattern involves dispersed unevenness instead of polarized unevenness. As it happens, both play a role. Correcting for bias, which includes switching to data for households instead of persons, reduces the score for the dissimilarity index 51.1 points from 81.3 to 30.2. By comparison, correcting for bias reduces the score for the separation index 11.6 points from 15.3 to 3.7. The score of 81.3 for the standard version of the dissimilarity index calculated using person data can only be described as grossly misleading. The score of 30.2 for the unbiased version of the dissimilarity index is correct. But prevailing habits for interpreting values of the dissimilarity index would lead to misleading conclusions about whether Latino households live with or apart from White households. The values of the separation index, especially the value of the unbiased version, make it very clear that in 1990 Latino households live alongside White households in this new destination community and necessarily experience the same neighborhood contexts that White households experience.

5.12 Summary

In their review of the research on new destinations, Flippen and Farrell-Bryan (2021) described new destination migration as one of the most “striking demographic trends of recent decades” (2021: 27.2). Residential patterns can be a telling indicator of emergent racial and ethnic relations, and their trajectories also tell a story of how these relations are changing over time. A handful of scholars have recognized this and contributed research on residential segregation in new destinations, but these efforts have faced multiple challenges that are shared by segregation research on nonmetropolitan communities (see Chap. 4). The demographic conditions of many new destinations make conventional approaches to segregation measurement prone to inflated index bias and can also lead to high scores on the more popular dissimilarity index when in fact the two groups in the analysis are not living in fundamentally different neighborhoods. Additionally, the choice between measuring segregation of persons versus households becomes critical because the index scores are already more prone to upward bias. Finally, it is important to measure segregation at the level of the census block when studying segregation of new destinations because these areas often have smaller populations, especially in the case of the newly emerging group.

While some of these issues are simply underexamined and therefore have gone unaddressed, index bias is a problem that researchers have been aware of but have been unable to fix until Fossett (2017) developed the formula correction that directly removes the source of the bias from the calculation of the segregation index. With this correction for index bias in addition to our empirically-driven choices to rely on the separation index for segregation measurement and to measure segregation of households rather than persons, we contribute to the literature on residential segregation in new destinations with refined and superior analyses of patterns and trends of segregation in new destinations, including how these areas compare to established areas of settlement and how they change over time. We also extended beyond Latino new destinations to further develop our understanding of Asian new destinations and begin asking questions about the possibility of Black new destinations. Even though Latino new destinations are far more common and better understood as a social phenomenon, areas across the United States where Asian and Black populations are newly emergent going into the twenty-first century give us the opportunity to observe how these groups experience new settlement into predominately White communities by the way they are residentially distributed initially and over time.

Our empirical analyses in this chapter can be summarized by three key findings. First, residential segregation is lower in new destinations as compared to established areas of settlement for all three minoritized groups. In the case of both Latino and Asian new destinations, residential segregation is at very low levels in 1990, which is the starting point of the analysis. However, we also find that White-Latino and White-Asian residential segregation is rising in new destinations and reaches moderate levels by 2010. White-Black segregation in new destinations follows a different pattern initially and over time, beginning at moderate levels and fluctuating to slightly lower levels in 2010. These findings contradict some of the conclusions that have been drawn in the literature, as we would have expected based on past studies to find higher levels of segregation in new destinations. But the theoretical arguments on racial conflict and place stratification that frame previous studies may still stand, as we do find segregation increasing over time for Latino and Asian households in new destinations. It is possible that as these groups become more visible and more permanent in their new communities, they could face increasing conflict with and separation from White residents.

We also find discordance between the dissimilarity index and the separation index when measuring residential segregation in new destinations, with the dissimilarity index consistently suggesting higher levels of segregation occurring than the separation index. This is indicative of a form of uneven distribution in new destinations that we refer to as dispersed unevenness, where the two groups in the analysis are living in neighborhoods that have different average levels of proportion White, but the differences are not large enough to produce patterns of segregation that we would think of as prototypical segregation. However, for Latino and Asian new destinations this discordance between the separation index and the dissimilarity index is changing over time as scores on the separation index increase, suggesting that residential segregation in Latino and Asian new destinations is shifting from patterns of dispersed unevenness to patterns of polarized unevenness, where now Latino and Asian households are living in fundamentally different neighborhoods than White households in the community.

We have documented that the impact of bias on segregation index scores can be very high when scores are computed using standard computing formulas and this is true both when indices are calculated using data for households and data for persons. We have documented how researchers can use refined formulas introduced in Fossett (2017) and measure segregation of households rather than persons to obtain index scores that are free of index bias. Happily, it is a relatively simple matter to obtain unbiased index scores when indices are calculated using data for households because the adjustments to calculations are simple and do not require detailed data beyond the basic household counts used in calculating index scores with standard formulas. Unfortunately, the situation is more complicated when index scores are calculated using data for persons. In this situation the sources of index bias are more complex and adjustments to calculations must accordingly be more complicated. Furthermore, additional detailed data on race distributions of households by size across residential areas are needed to implement the adjustments (see Chap. 2 for further discussion).

In conclusion of this chapter, we have accomplished two goals. First, we have provided a sound analysis of residential segregation patterns and trends in new destinations to support future research in this area using methods that produce reliable and trustworthy measures of residential segregation even under conditions that researchers have avoided because of the problems that they presented for conventional methodological approaches. Second, we have demonstrated in this chapter (and described in more detail in Chap. 2) the measurement tools and guidelines needed to successfully study residential segregation in new destinations or in any communities where one group is small or newly emerging. The analyses presented here document that application of the formulas for unbiased index scores can yield scores much lower than the scores obtained when using standard index formulas. Researchers who are accustomed to seeing high scores for segregation indices may wonder if the unbiased scores are in some sense too low and perhaps unnecessary. We are confident in advocating the use of unbiased scores to gain the best understanding of the state of segregation in new destinations and its implications for life chances and race relations. Our approach overcomes many of the limitations that have hindered research in this area. We hope that researchers will adopt these methods and continue to develop our sociological understanding of demographic changes and residential patterns in new destinations.