Diversity and Segregation

Introduction

Segregation is a complicated, nuanced concept, and one that has eluded simple definition and universal measure despite decades of intellectual effort. Some scholars believe that segregation is really a composite of five concepts, while others compress it into two dimensions.1 Some ideas are used to characterize segregation or integration, but they may not be the same thing. Isolation, for example, is a characteristic of segregation, just as diversity may be a characteristic of integration, but that doesn’t mean they are definitionally the same, although both ideas are frequently conflated. Segregation and integration may not even be conceptual opposites.2

The conceptual underpinning regarding segregation has long been and remains contested, as does the attempts to operationalize those concepts in spatial representations (maps) or measurement (metrics).3 Maps or other diagrams can help us by showing how different people are separated in space and in different patterns, but even there, complexity quickly arises depending on the scale or the measure used to assess the pattern.4 Certain patterns may show up as more or less segregated, depending on the measure and the formula used to calculate that measure.

Measures of segregation have been developed and applied to provide a more precise assessment of the degree of segregation, and there have been ongoing, unresolved debates about which may be best and in which context. 

Maps may be most illuminating in contexts like the mid-20th Century United States or Northern Ireland – contexts in which racial, ethnic, or religious differences are essentially binary, relating to Group A and Group B, and in which there are clear demarcations of spatial separation. Thus, "segregation" maps tend to show major metropolitan areas or neighborhoods at a particular moment in time, and clearly indicate the concentrated and separate presence of the social groups being depicted. 

In the Northern Ireland context, segregation maps may show neighborhoods of Catholics and Protestants, and vividly illustrate the spatial separation and pattern of residency between these groups.5 In the American context, segregation maps might show Black neighborhoods in a tightly bound pattern in a part of the city surrounded by white neighborhoods and white suburbs.6

Similarly, the most popular measures of segregation, like the Dissimilarity Index, are binary: they measure the relative distribution of one group (often a racial group) relative to another, and indicate how many members of one group would have to move to a different-race neighborhood to integrate.  

Segregation is relatively easy to understand and map when dealing with only two clearly defined groups and clearly defined boundaries. But the depictions, measurements, as well as conceptual frameworks for segregation begin to fray or at least become more intricate and complex as the number of groups or dimensions of social concern grow or even multiply. 

The Problem of Diversity 

As a region or geographic area becomes more diverse, the depictions and measurements used to capture the concept of residential segregation become more complicated to render. To begin with, as you move from contexts that focus on two social groups to those that contain three or more groups, the question of segregation takes on a different set of meanings. Consider a few hypotheticals:

Hypothetical 1: Suppose a city is strictly divided between two racial groups, Group A and Group B. We would easily recognize this as racial segregation. But suppose that a third racial group, Group C, moves into the city in large numbers, but only moves into neighborhoods where members of Group B live. As a result, group B and C are well integrated relative to each other, and have low “B-C” Dissimilarity scores, but they both remain highly segregated from Group A, and Group A remains highly segregated from Group B and C, individually and collectively. 

How should this pattern be described, depicted or characterized? Is the city considered diverse? Are neighborhoods where members of Groups B and C reside considered diverse?  

This is not a particularly difficult case, but it already illustrates the problem between measuring segregation “holistically,” for all groups or for the entire geography, and for specific neighborhoods or for specific groups.  

We could and should say that Group A is segregated, and that neighborhoods where Group A lives are segregated. Characterizing Groups B and C and their respective neighborhoods is more difficult, as would be attempts to characterize the entire city. The ultimate answer would probably depend upon the overall proportions of the groups, and our focus. If Group A was a relatively small portion of the population and city, say, just 5 percent, then we would probably be justified in describing both the city and Groups B and C as relatively integrated.

But, on the other hand, if Group A was 50 percent or more of the population, and Groups B and C were relatively small, then the city would probably be better characterized in terms of the segregation experienced by Group A. 

Hypothetical 2: Suppose, instead, that there were four racial groups, A, B, C, and D, and that members of Groups B, C, and D lived in close proximity and relative balance, while members of Group A lived in completely different neighborhoods where no members of any other group lived. 

The introduction of the fourth group probably has the effect of making us more confident that Groups B, C, and D are integrated, and that Group A is segregated. But, again, without knowing proportions, characterizations regarding the city are difficult, if not impossible. Let’s add a twist, though:

Hypothetical 3: Suppose, again, that a city is composed of four racial groups: A, B, C, and D.  But in this case, suppose that Groups A and B live in close proximity and relative balance, and Groups C and D live in close proximity and relative balance, but that neighborhoods where Groups A and B live are strictly divided from and separate from neighborhoods where Groups C and D live, and that no members of Groups A or B live in neighborhoods where members of Group C and D live, and vice versa.  

Are any of these groups segregated? Are they isolated? Are these neighborhoods diverse? Are these neighborhoods segregated? Is the city diverse? It is difficult to answer these questions in the abstract, and may depend upon the level of geographic focus, the formula used to analyze the data, or change depending upon the group we are focused on. But even with specific settlement maps, a clear focus, and granular data, it may be difficult to answer these questions. Increasing diversity, even within a single category, makes it harder to gauge segregation, but that problem is compounded when we add more categories. Consider:

Hypothetical 4: Suppose a city is made up of two groups who live in separate and distinct neighborhoods and different parts of the city: Group A and Group B, but that these are religious groups. Suppose that Group A is composed of 10 different ethnic groups, and that these groups are in relative balance across and within the neighborhoods and communities that Group A inhabits. Suppose, however, that Group B is composed of a single ethnic and racial group as well as a single religion. 

In this hypothetical, are the neighborhoods where group A lives segregated? Are they diverse? Are they isolated? The answer depends on what we are focused on, and what questions we are trying to answer. 

As you can see, diversity introduces considerable complexity into the already-fraught question of segregation. The increase in the number of groups creates a number of complications in trying to assess the level of both diversity and segregation. Already, however, a few things should be clear. 

To begin with, although diversity can be a gradient along a single dimension (as we saw when increasing the number of groups), or diversity can be a multi-dimensional concept. It encompasses race and religion, but also ethnicity, sexual orientation, caste, socio-economic status, familial background, and much more. It is possible – in fact it is not unlikely – to have a high level of diversity and simultaneously a high level of segregation. The multi-dimensional nature of diversity as an institutional goal, as in some affirmative action practices, can be confounding. 

Twentieth Century Chicago was infamously diverse, with a plethora of white ethnic settlement (Italians, Polish, Irish, etc.) as documented by Robert Park and others in his 1915 study, but also highly segregated by race.7 This is paradoxical but true: diversity and segregation often co-occur. Neighborhoods in the Bay Area that appear deeply diverse, with Pakistanis, Israelis, Russians, Armenians, etc. may be incredibly ethnically diverse, but highly racially isolated and even segregated by race.  

This means that when assessing these questions in diverse contexts, clarity is required around the dimension (and geography) that is the focus: is it race? Is it ethnicity? Is it religion? Whatever is being studied, the specific dimension(s) must be specified, and the assessment of whether segregation (or diversity) exists must be assessed along that particular dimension(s), and not some other set of dimensions. Otherwise, the question will produce paradoxical results. A place can be both segregated and non-segregated or diverse and not-diverse, but in different dimensions. It has to be one or the other along a single dimension or set of dimensions. 

The Relationship Between Diversity and Segregation

Diversity is a condition of having many unlike things exist in a community, institution, or place. And in the human context, it is the condition of having a variety of people with salient social differences co-exist in a place or location. Segregation is the spatial separation of people along any possible dimension of human difference. Racial segregation is the separation of people on the basis of race. And racial residential segregation is the separation of people on the basis of race in terms of residential location.  So what is their relationship?

Diversity – human difference – is the raw material out of which segregation can occur. Without human difference, segregation is impossible to observe or measure. Racially homogeneous environments, for example, are by definition not racially segregated, because there are no racial groups to segregate.8 They are also not diverse (at least, along the dimension of race).  

Analogically, diversity and segregation have the same relationship as the stock of wealth does to wealth inequality. The stock of wealth is the sum total of wealth held by a community or place. Wealth inequality measures the distribution of the ownership of that wealth within that community.9 One logical extreme would be if one person in a large community owned all of the wealth held by that community, and everyone else owned none of it (perfect inequality). At the other extreme, that wealth might be evenly held by every member of the community (perfect wealth equality).  

Whether the case is one extreme or the other or anything in between, the stock of wealth is the basis out of which wealth inequality can arise. Logically, a community with zero wealth (as was the case in some traditional, preagricultural, nomadic human bands) can have no wealth inequality. There would be nothing to distribute, so in that case both the level of wealth and level of inequality is zero. Only a society with at least some wealth can have wealth inequality. Similarly, only a society with diversity can have segregation (along the same dimension of difference).  

Experience suggests that increases in wealth are typically accompanied by increases in wealth inequality. There may be many reasons for this, but the simplest explanation is that it is very difficult to evenly distribute stocks of capital wealth, at least in the short-run, because of how wealth tends to accumulate and grow. 

First, the mass sales products or other activities of corporations are a prominent way that wealth is generated. Consider, for example, the pharmaceutical company Novo Nordisk, which manufactures and sells Ozempic and Wegovy. These products are so enormously demanded that it has reshaped Denmark’s economy and generated enormous tax revenue for the country.10 The shareholders and owners of critical corporations, especially those making products that are important or most popular (think Apple) generate wealth that is not necessarily shared by non-owners. Eventually, the broader economy may benefit, but the owners of those companies will see their fortunes balloon first. 

Another feature of modern life is that economic activity tends to be greater in more urbanized places with larger populations. Thus, wealth is much more likely to grow at a faster rate in urban places than rural places simply because there is more economic activity. This is the reason for the paradox that “blue” places tend to have more wealth inequality than “red” places. Conservative critics will sometimes point out that cities or states led by Democratic Party mayors, governors, or legislatures tend to have more wealth inequality.11 But what these critiques elide is the fact that more populous places tend to be more liberal everywhere, and that these places also tend to have more economic activity and generate more wealth. The greater amount of wealth means that there is more potential for income inequality. It is not inevitable that this inequality will arise or widen, but it is likely to do so, absent extremely powerful countervailing forces. 

Another way that wealth grows is through return on investments. As people buy property or invest in companies during the course of their lifetime, they will naturally hold more wealth than younger people who have not yet had a chance to do the same. Therefore, older people tend to hold more wealth than younger people. Again, it is not inevitable that older people hold more wealth than younger people, but the advantage of time makes it much more likely that they will. 

This principle is also true of various social groups. Newcomers to a place, such as recent immigrants or refugees, will likely hold less wealth than families that have long been settled in a place. Politically powerful groups or more educated members of a community are also likely to have or hold greater relative wealth. Discrimination and prejudice also affect accumulated wealth in various ways, such as by diminishing income that can be used to acquire property and grow wealth. 

In any case, the important principle is this: increases in a stock of wealth are likely to increase wealth inequality. And the implied analogical corollary is also true: increases in diversity are also likely to increase segregation. Experience also tends to bear this out: new immigrants or migrants into a city tend to move to places where members of their group already predominant or are otherwise residing in non-trivial numbers. This is called the “port of entry” effect: recent immigrants will move into these communities to find linguistic and cultural familiarity, and to use social networks to find housing and jobs.12 This tendency generally increases both diversity and segregation. Let us illustrate with a thought experiment:

Suppose that a city is 65 percent white, 25 percent Black, 8 percent Latino, and 2 percent Asian and Native American at a particular point in time. Suppose that this city experiences a population boom spurred by economic growth, such that, ten years later, its population has doubled, and it is now 25 percent white, 25 percent Black, 25 percent Latino, and 25 percent Asian, Pacific Islander and Native American. 

The city is clearly more diverse than it was ten years prior, but it may also be more segregated than it was before. Even if the city was highly segregated before, the fact that most of the new Latino and Asian immigrants to the city have probably moved into previously segregated and isolated Latino and Asian neighborhoods means that increases in diversity are likely accompanied by increases in segregation rather than into different race neighborhoods. This, for example, was what happened to Boston during the first wave of the Great Migration. Boston’s Black population was relatively well-integrated for decades, but newcomers from the south moved into increasingly identifiable “Black” neighborhoods.13

The only way that the hypothetical city in our thought experiment did not experience an increase in segregation is if the newcomers moved into their new neighborhoods at exactly the same rate across the entire city – evenly, or moved into non-Asian and non-Latino neighborhoods at a greater rate than they did same-race neighborhoods.  

This condition is extremely unlikely. No human settlement pattern is perfectly even, unless there is a coercive external force compelling it to be so (such as Stalin’s population transfers).14 While it is not always true that newcomers to a region or city will move into same-race neighborhoods, there is a tendency for this to be the case. And it is certainly enough of a tendency to explain why increases in diversity tend to increase segregation.  

Only a strong and proactive policy or cultural influence to integrate is likely to translate greater diversity into reduced segregation. The same is true of wealth and wealth inequality: only a strong and proactive policy is likely to translate a greater stock of wealth into reduced wealth inequality. And, similarly, “blue” places will tend to have more wealth inequality, simply because more populous places are more politically liberal, and those places tend to have more economic activity that results in wealth creation. None of these outcomes are inevitable, but they are powerful tendencies. 

Observations on Diversity and Segregation 

A scatterplot analysis that measures and represents the relationship between diversity and segregation supports and illustrates the relationship suggested earlier by the analogy to wealth and wealth inequality. Diversity is a stock; and segregation is the spatial distribution of that stock. 

Figure 1 below illustrates, using 2020 census data, the relationship between the Divergence Index (a measure of segregation) and the Entropy Score (a measure of diversity) for cities and metropolitan areas in the United States. 

Figure 1: Diversity and Segregation in the United States in 2020

The scatterplots above validate the essential insight: As the level of diversity approaches zero, the observed level of segregation also approaches zero. If these were independent variables, we would expect there to be at least some possibility of higher levels of segregation where there is a low level of diversity. This is never the case for the conceptual reasons already covered.  

Moreover, as the level of diversity increases, the observed range of segregation increases correspondingly. It is not exactly a step pattern, but the range of outcomes widens considerably. This validates the insight that an increased level of diversity creates the conditions for greater segregation.

The mosaic plots in Figures 2 and 3 illustrate the same relationship, but perhaps more clearly. The most segregated places are essentially absent in the low-diversity contexts, and vastly overrepresented in the most diverse.

Figure 2

Figure 3

These mosaic plots, or quadrants, show the relationship between diversity and segregation clearly. The groups < Q1, Q1-Q2, Q2-Q3, and > Q3 refer to values lower than first quartile, between first Divergence Index quartile and median quartile, between the median and third quartile, and greater than third quartile, respectively. Q1 refers to the values lower than 25 percentile, Q2 for values between 25 to 50 percentile, Q3 for values between 50 to 75 percentile, and Q4 for values higher than 75 percentile for each measure. 

As you can see, these mosaics show that as you increase the level of diversity (entropy), you get more values for observed levels of segregation).15 As suggested by our theory, there are no metros that are in the highest divergence but lowest entropy group. 

These observations are not intended to suggest that policies aimed at reducing segregation should curb or limit diversity. Rather, it is an observation that diversity tends to generate segregation, absent countervailing forces.

This dynamic is not an artifact of the Divergence Index. This relationship holds for other multi-group measures of segregation as well. 

Figure 4 below illustrates, using 2020 census data, the relationship between the Theil’s Index (H) and the Entropy Score for cities and metropolitan areas in the United States. Similar to the scatterplot between Entropy Score and Divergence Index, this plot confirms the relationship between diversity and segregation.

Figure 4

Another implication, in addition to necessary clarity around the dimension being studied, is that as diversity increases, different measures of segregation may need to be considered, especially those that examine multiple groups at once. This is indeed one of the fundamental problems with most traditional measures of segregation. In the United States, in particular, segregation research is often motivated to understand conditions that either caused or produce racialized inequality, especially in connection with African-Americans. Thus, researchers and scholars seek to identify segregated Black neighborhoods in order to determine the degree to which the condition of segregation may be contributing to inequitable outcomes in that neighborhood.

But as racial diversity increases nationally and in particular regions, measures of segregation that are attuned to the conditions or experiences of a particular racial group, including African-Americans, may mask, obscure or even obfuscate critical dynamics of segregation. For example, it is now generally well known that a focus on the Black-white dissimilarity index masked non-trivial increases in the segregation of both Asian and Latino people since 1980.16

Conclusion

Increases in diversity tend to produce increases in segregation just as increases in the stock of wealth tend to produce more unequal distributions in that stock of wealth.  

When looking at patterns of wealth inequality across nations, it is probably wise to compare nations of comparable levels of wealth, rather than nations with vastly different levels of wealth stock. But controlling for the stock of wealth in calculating the level of wealth inequality would result in a tendency to underestimate the relative magnitude of wealth inequality that exists in higher-wealth countries and therefore present a less accurate measure of wealth inequality. Controlling or normalizing for the level of diversity may make it easier to compare cities, but it doesn’t actually help us understand the level of segregation that exists in a place more accurately or on a relative basis.