During the course of this research, we've received a variety of questions about the scope of the work, the methodologies we utilize, and our proposed solutions to the problem of segregation. To this end, we compiled this list of frequently asked questions to address some of the burning questions readers have had.
If you have any further questions or requests, do not hesitate to reach out to us at housingOBI@berkeley.edu.
Your research has reported change in segregation in CBSAs, and your webmap displays consistent tract boundaries, across different census years. How did you ensure data across different census years is comparable?
You have persuaded me that segregation is a major cause of group-based inequality (given the variety of examples in Part II of your main report). Is it always the case, however, that segregation is harmful? I see some ethnic enclaves are considered "segregated" but I doubt all residents would feel oppressed (as in New York City's Koreatown or Little Odessa, for example).
QUESTION: How do you define "segregation"?
ANSWER: Segregation is a term that is sometimes used or understood in different or varying ways, often depending upon context. Acknowledging this, we define this term in three steps in our Technical Appendix as follows:
Segregation is the separation across space of one or more groups of people from each other on the basis of their group identity. Racial segregation is the separation of people from each other on the basis of race. Racial residential segregation is the separation of people on the basis of race in terms of residence.
There are, of course, other forms of segregation, many of which we mention throughout our project, such as occupational segregation, public accommodations segregation, and educational segregation. Courts also distinguish between "de jure" and "de facto" segregation, segregation caused by law and policy or segregation caused by other forces, such as private preferences or discrimination.
Clearly defining segregation does not mean that a measure for segregation can be easily operationalized. For example, measures of isolation are sometimes described in the literature as separate and distinct from segregation, and others describe it as a form or expression of segregation. That is why this project uses multiple measures, each of which are described in the Technical Appendix, to provide a fulsome perspective on this problem.
QUESTION: How can I download or access the data behind your report and mapping tool?
ANSWER: Please fill out this form to request data from our mapping tool. Our mapping team should get back to you in a timely manner.
QUESTION: Your research has reported change in segregation in CBSAs, and your webmap displays consistent tract boundaries, across different census years. How did you ensure data across different census years is comparable?
ANSWER: We utilized Social Explorer's demographic data from 1980, 1990, and 2000 which has been normalized to 2010 census geography, making it easier to compare data from different census years while maintaining geographic consistency. Boundaries of block, block groups and census tracts change every ten years based on the census counts in each decennial year. Normalization of data means that the demographic data from previous census years are aggregated, disaggregated or apportioned to calculate the numbers for geographic boundaries in 2010.
QUESTION: Why is the ‘segregation/integration’ measure the default index on the interactive map when it seems like the Divergence Index is the measure of choice throughout the report?
ANSWER: The goal with the web map was to make the visual representation of segregation as intuitive as possible at first glance. We believe the ‘segregation/integration’ scale we created, which is largely based on the Divergence Index, best achieves that goal. We explored different options for default maps and decided against using a simple, unmodified version of the Divergence Index as the default as this would have required additional interpretation and would have been mystifying to most users. Bear in mind, however, that our map allows users to select different measures of segregation, including the Divergence Index unmodified, as well as measures such as the Dissimilarity and Exposure Indices.
QUESTION: How is the interactive map’s ‘segregation/integration’ index calculated?
ANSWER: Our default map is calculated primarily with the Divergence Index, as explained in the technical appendix. The main exception is the category "integrated," which is based upon a combination of three criterion described in the technical appendix: "We define “integrated” not as a low level of observed segregation, but as any place that meets all of the following conditions: 1) falls in the bottom third of the Divergence Index nationally, 2) has an entropy score in the top 50 percent nationally, and 3) has at least 20 percent Black and/or Latino population. We believe this combination of characteristics helps us identify places that are meaningfully integrated, not just the apparent absence of segregation." Further elaboration can be found in the appendix.
QUESTION. The ‘segregation/integration-
ANSWER: These are the four categories of census tracts we use in our analysis in the main report, and the definitions were provided in endnote 65. They are defined as follows:
- High White Segregation: census tracts with a Divergence Index score or value in the top third nationally, are majority white, and have a white Location Quotient above 1.25.
- High POC Segregation: census tracts with a Divergence Index in the top third nationally, but do not have a majority white or have a white Location Quotient below 1.25.
- Racially Integrated: census tracts with a Divergence Index score in the bottom third nationally, an Entropy score in the top 50% nationally, and are at least 20% Black and Latino.
- Low-Medium Segregation: All other census tracts not covered by the other three categories. In practice, this means they are neither highly segregated nor integrated by our definition.
QUESTION: What is the reference geography used in calculating divergence, location quotient, and entropy values?
ANSWER: The base geographic units used in this study are presented and described in the technical appendix. The primary reference geography is Core-Based Statistical Areas (CBSAs). The main exception are instances in which a tract falls outside of a CBSA, in which case the County is used as the reference geography.
QUESTION: What causes racial residential segregation? Your report claims that segregation is, in many cases, worse or more extensive today than in decades past. Why is this?
ANSWER: Our report is specifically focused on presenting 1) the extent of racial residential segregation in the United States and 2) the harmful effects. We deliberately avoided trying to explain these patterns (as noted in endnote 18) because it would have required a far more extensive presentation of the theories and data, and which was beyond the scope of our purpose and would have distracted from our focus.
Another reason to avoid delving into this question is because there are so many theories in the academic literature about the causes of racial residential segregation, and so little consensus regarding which are correct.
Nonetheless, because we have received so many inquiries about this, we sketch a few of the prevailing theories, and then present our best guess as to which is probably most explanatory. Among the prevailing theories for the existence of racial residential segregation are:
That is to say that discrimination in housing markets (sale, rental, credit, etc.) by lenders, sellers, landlords, real estate agents, etc. past and present may explain some portion of the observed level of racial residential segregation today. There are intense debates about exactly how much of observed segregation is attributable to discrimination, but this is likely a significant factor, particularly in creating patterns of racial residential segregation in our metropolitan areas prior to the federal Fair Housing Act. There is evidence that housing discrimination has significantly declined in the last 50 years, although there are intense debates about the particular forms this discrimination now takes, and whether this type of discrimination is sufficient to have a causal effect on racial residential segregation.
2. Compositional Preferences
Some scholars believe that different compositional preferences between racial groups, even slight, can result in large differences in neighborhood composition for those groups. (There is even a neat website that illustrates how this can work).
For example, if Black Americans prefer to live in a neighborhood that is only 40 percent white, but white Americans prefer to live in a neighborhood that is 60 percent white, then it is theoretically possible that those preferences can result in large differences in the racial composition of resulting neighborhoods.
This theory, too, is heavily debated (and disputed) in the academic literature. The main critique of this theory is that neighborhood preference by race shows that people of all races would prefer to live in more diverse neighborhoods than they generally do, such that structural forces likely play a significant role, even if this is a contributory factor.
3. Economic Differences Between Racial Groups
This theory suggests that differences in wealth and incomes between racial groups means that different race families have different capacity to buy into certain neighborhoods. In a context of growing neighborhood stratification (meaning, economic differences between neighborhoods within metropolitan regions), this means that small differences in wealth or income can result in large compositional differences in resulting neighborhoods. Economists tend to lean more heavily on this explanation, and it may well be playing a critical role in resulting patterns of racial residential segregation.
4. Differences in Background Knowledge of Neighborhoods
Another theory very recently developed by Maria Krysan and Kyle Crowder argues that different race people have different background knowledge of neighborhoods in their regions. For example, Black people, in their account, have very different mental maps of the attributes and qualities of neighborhoods relative to white people, and that this knowledge shapes what they call the “pre-search” process. They provide compelling evidence that this contributes to patterns of racial residential segregation. In a sense, this theory suggests that past cycles of racial residential segregation cause new ones.
5. White Avoidance of Black & Latino Neighborhoods
There are several different versions of this theory, but the main idea is that white homeseekers use the percentage of Black (and to a lesser extent, Latino) neighbors as proxies for indicators such as neighborhood and school quality, home values and home value appreciation potential, among other factors. Thus, white families will try to move into whiter neighborhoods, not because of racial animus per se, but because they are making economic judgments about racial composition and neighborhood quality.
Unfortunately, there is no consensus that we discern regarding which theories are best supported by the evidence connected to explaining either the initial formation of racial residential segregation or its persistence, which is another reason we did not try to resolve this debate.
These are briefly sketched explanations that skirt many nuances but should suffice to provide insight as to why we avoided trying to explain our findings. We suspect that each of these explanations likely plays a contributory role (among other theories) but refrain from trying to attribute any particular theory or subset as playing a more central role. Our purpose was not to explain why but to demonstrate the what and the harmful effects.
6. Rising Latino and Asian Segregation
One other factor, however, we believe is contributing to rising levels of segregation, especially as measured by our preferred measure of segregation, the Divergence Index, is rising Latino and Asian segregation. We alluded to this in our report and appendix, but let us make it explicit: in many parts of the country, Asian and Latino segregation has risen in the last 40 or so years.
In our 5-part study of racial residential segregation in the San Francisco Bay Area, we found significant evidence of this. Specifically, we found that, although that Black-white Dissimilarity (see our appendix for an explanation of this measure) declined in the Bay Area between 1970 and 2010, white-Asian and white-Hispanic Dissimilarity rose (as this chart illustrates). Specifically, the white-Asian dissimilarity score rose from 0.4182 in 1980 to 0.4648 in 2010, about five points. Similarly, we also found that white-Latino dissimilarity rose from 0.4082 in 1980 to 0.4670 in 2010, an even larger rise.
This trend is not restricted to the Bay Area. Using our interactive mapping tool, you can select different dissimilarity indices (under measures of segregation in the navigation bar) and select different years to see changes over time to see for yourself how your city has evolved in this regard. For example, our mapping tool shows that New York City’s white-Asian dissimilarity has risen from 0.4882 in 1980 to 0.519 in 2010.
Dissimilarity scores are also readily available by city and metropolitan region from the “Diversity and Disparities” project at Brown (just select the city or region you want to view), and that website also provides nice data visualizations. Three cities selected from that website illustrate this trend of increasing Latino and/or Asian segregation, as measured by the more traditional Dissimilarity Index.
The city of Milwaukee shows that white-Hispanic dissimilarity rising from 55 to 60.7 from 1980 to 2010, and white-Asian dissimilarity rising from 28.5 to 45.2 in the same period, a huge increase in that period. The city of Miami, FL shows white-Hispanic dissimilarity rising from 40.9 in 1980 to 50.3 in 2010, a rather substantial increase. And the city of Raleigh, NC shows an enormous increase in white-Hispanic dissimilarity, rising from 25.3 in 1980 to 48.9 in 2010, and a substantial increase in white-Asian dissimilarity, rising from 24.2 to 34.4 in the same time period.
We have not comprehensively studied this matter to be able to show whether this is a general trend, but it has occurred in enough regions for us to feel comfortable asserting that increasing white-Asian and white-Latino segregation is likely contributing to, and probably helps explain, the increased levels of segregation we observed using our preferred measure of segregation, the Divergence Index. It also underscores a point we made in the main report and technical appendix regarding the limitations of the Dissimilarity Index: by focusing so heavily on Black-white dissimilarity, segregation studies are masking patterns of segregation that are obscured by reliance on that particular measure. This is why it is so important to use multiple measures as well as to employ measures that can reflect levels of segregation for multiple groups simultaneously.
QUESTION: You’ve convinced me that racial residential segregation is a problem, and a bigger one that I imagined. But what can we do about it?
ANSWER: This, also, was beyond the scope of our report, for the same reasons presented in endnote 18. Similarly, we felt that trying to provide solutions to the problem while also explicating on the harms would have been distracting from the main focus. Nonetheless, there is plenty we can offer in the way of solutions.
Broadly speaking, we believe that the solutions are to Integrate and Invest.
By integrate, we mean policies that a) extend people more neighborhood choices by breaking down or overcoming exclusionary barriers, and b) to affirmatively support them in executing those choices, with subsidies, counseling, etc. Surveys show that large majorities of Americans would prefer to live in more diverse neighborhoods than they currently do. Carefully designed mobility strategies with proper supports, along the lines of those conducted in Seattle (the Creating Moves to Opportunity Program), but which is more race-conscious, could be a big help to realizing those preferences. Richard Sander and his co-authors have sketched out such a program in Chapter 22 of their book “Moving to Integration.”
By invest, we mean in policies a) that support public goods in disadvantaged neighborhoods and communities to try to make them more equitable and b) subsidies for individual families that would enable them to make integrative housing decisions, along the lines sketched immediately above.
For a more extensive presentation of five policy solutions geared toward state and local governments, see our report “Racial Segregation in the San Francisco Bay Area, Part 5: Remedies, Solutions, and Targets” which presents five broad policy solutions to the problem of racial residential segregation, including 1) the necessity of curtailing restrictive land use policies and regulations and opening up exclusionary neighborhoods and communities to different-race peoples; 2) the potential of rent control and rent stabilization policies to prevent displacement from integrated or integrating communities or to keep integration sustainable where it exists; 3) mobility strategies can support people who wish to move to neighborhoods where members of a different race predominate, and thereby reduce the segregation of those communities; 4) inclusionary zoning ordinances and statewide fair share laws that mandate a specific level of economic integration; and 5) affordable housing policies and other direct subsidies that permit a larger range of housing options for pro-integrative effects.
QUESTION: Does your study account for Asians and Native Americans? If so, how? I noticed that some of your analysis does not include them.
ANSWER: Yes, absolutely!
Our study is a comprehensive analysis of racial residential segregation in the United States, and that includes Asians and Native Americans. The Divergence Index calculations, most prominently, include both Asians and Native Americans at the tract and city/metro levels. Thus, if you use our interactive mapping tool, for example, the tract-level Divergence scores given in the data bar include both Asians and Native Americans, and also display the percentage of Asians and/or Native Americans in those tracts and/or cities or regions. Therefore, all of our city rankings, metro rankings, and change tables fully include Asians and Native Americans to the same extent as any other racial group.
We also provide Asian and Native American specific segregation measures in our interactive mapping tool, such as Asian Exposure or Asian Location Quotient and Native American Exposure and Native American Location Quotient, among other measures. Researchers focused on these racial groups can use our mapping tool to better understand Asian and Native American segregation and isolation.
That said, there are a few places in our analysis where Asians and Native Americans were not included as part of the analysis. For example, for the purposes of Tables 4 and 5 in the main report (as explained in endnote 63), we focused only on Black and/or Latino neighborhoods in identifying “highly segregated neighborhoods of color.” Similarly, the third criteria in our definition of “integrated” tracts (as explained in the technical appendix) is whether a tract is at least 20 percent Black and/or Latino. Thus, although the other two criteria include Asians and/or Native Americans, the third criteria does not.
QUESTION: You have persuaded me that segregation is a major cause of group-based inequality (given the variety of examples in Part II of your main report). Is it always the case, however, that segregation is harmful? I see some ethnic enclaves are considered "segregated" but I doubt all residents would feel oppressed (as in New York City's Koreatown or Little Odessa, for example).
ANSWER: Our claim is nuanced: we believe that residential segregation is a primary cause of group-based inequality, as noted in the question, but we do not claim that segregation is always harmful, just that it has a general and broad tendency to produce harm by driving group-based disparities. There are some obvious cases in which the benefits of segregation may outweigh the harms. Without being exhaustive, here are a few examples:
First, many recent immigrants to new countries or societies can benefit more than they are harmed by moving into "segregated" enclaves. This is because many recent immigrants have limited language proficiency with the predominant language or familiarity with society norms or behaviors. Thus, immigrant enclaves may offer recent immigrants the most opportunity: the most accessible set of employment opportunities, the most social capital to learn about such opportunities and make connections, and cultural comfort. This is known as the "port of entry" effect. While such segregation may be harmful for second or third generation immigrant children, these enclaves may be the first step on the ladder of opportunity for first generation immigrants.
Second, members of deaf communities may be best served by moving into segregated deaf communities. As the vast majority of hearing people do not know sign language, they cannot easily communicate with or understand people who communicate primarily through sign language. This ability to communicate is more than simply a vehicle for opportunities (as is the case for recent immigrants), it is also a cultural difference, with nuances and inflections that may be missed even by hearing people who know sign language but do not use it regularly. There may nonetheless be some costs or harms that accrue from being separated from the broader society, but these costs may be counterbalanced by other benefits which are difficult to weigh.
Segregation imposes costs, whether it is by choice or not, and it is important that we understand and take cognizance of those costs, even if there are benefits to it in some cases.
QUESTION: Does your preferred measure of segregation simply reflect changing diversity? It is misleading in that regard?
ANSWER: We do not believe so, but it is possible that increases in diversity are likely to co-occur with increases in measured levels of segregation under our preferred measure of segregation, the Divergence Index. Before explaining why this is the case, we first provide some caveats.
First, there is no perfect measure of segregation. Our technical appendix explains, in plain terms, the advantages and limitations of every major measure of segregation. We believe that the Divergence Index is the “least bad” measure, and often the best for particular purposes, especially those which are trying to assess neighborhood segregation levels or segregation in highly diverse contexts, where there are already non-trivial percentages of Asians and Latinos, in addition to Black and white people (more on this below).
Second, we use a separate measure for diversity (the Entropy Score) and have found that some places that have a high degree of diversity have relatively low levels of segregation, and vice versa, suggesting that the two are not always related. Because we use separate measures of diversity and segregation, we have tried to emphasize the difference throughout our study.
But to put the matter directly, since these two ideas may be too frequently conflated, let us state unequivocally that a place can be 1) diverse and segregated, 2) diverse and integrated, or 3) neither diverse nor integrated. Racial diversity widens the potential for integration or segregation, but whether a place is integrated or not depends upon the degree of racial separation between people of different races in that place.
This potential for greater segregation that accompanies increases in diversity helps explain why there appears to be such a strong relationship between diversity and segregation. As an analogy that helps illustrate the underlying idea, consider the relationship between wealth and wealth inequality. Places that have very little wealth have much less potential for wealth inequality compared to places that have immense wealth because there is so little wealth to distribute. This is why places that have more wealth tend to have more wealth inequality (and, further, why “blue” states, which tend to have more wealth, appear to be more unequal). You need very strong policies to prevent wealth inequality from increasing wherever the overall stock of wealth increases. The same is true of the relationship between diversity and segregation.
Regions with zero diversity by definition can have zero segregation (although this is not true of neighborhoods or cities within regions). And, in the absence of very strong forces or policies ensuring that this diversity is more evenly distributed, any increase in diversity is very likely to increase segregation. This is not an inevitable result, any more than an increase in wealth must inevitably be distributed in such a way as to increase wealth inequality. (After all, there were places in our study that had increases in diversity but did not become more segregated.) It just tends to be the case because our society is one of segregation, not integration (and wealth inequality, not equality), which channels increases in diversity into uneven residential distribution. This is the logical flaw in both the argument that “blue states” have more wealth inequality because of their “blue” policies, and in the critique that the Divergence Index is simply reflecting greater diversity.
To be a bit more technical, the Divergence Index is highly sensitive to uneven distributions of people across space. This means that any increase in diversity that is even slightly more concentrated or unevenly spread across a city or region than the pre-existing distribution of the predominant groups (say, white people) will show up as increases in segregation. So, for example, if the percentage of Asians or Latinos in a city increases by a small amount, and if that increase is more unevenly concentrated than the existing distribution of white people, then that increase in “diversity” may also show up as increasing segregation. This is both accurate but also counterintuitive, and part of what has frustrated some recent critics of our use of the Divergence Index. We see no reason to abandon our preferred measure because of this result, except to note that it may prove counterintuitive in some contexts. If the reader feels that way, then we have no quarrel with the use of an alternative measure that better fits or suits your definition of segregation.
To that point, thirdly, the measure that is “best” depends on the purposes for which you are looking, and the context in which you are trying, to understand the phenomenon of racial residential segregation. For example, in contexts where the population is almost entirely Black and/or white, then the Dissimilarity Index, which we strongly critique, may be serviceable. In contexts in which white people are a large percentage of the population and not in relative decline, then the Exposure Index may be quite insightful.
This is also why our mapping tool provides layers for each of the measures discussed: we want to be helpful to anyone investigating this issue.
Finally, the "best measure" for your purposes may depend on which geography you are trying to understand: neighborhoods (tracts) or cities and regions? For tracts, there is no problem in this regard. The Divergence value provides an excellent and holistic sense of whether a tract is segregated or not, irrespective of relative diversity. For cities and/or regions, there is some evidence that diversity is correlated with the Divergence Index value.
We hope that our study has contributed to the literature on segregated indices so that researchers can weigh the pros and cons of each measure going forward. We look forward to the advance in research and theory in this field.
QUESTION: Table 5 in the main report shares the 1990 outcome of segregation based on data provided by the Opportunity Atlas. Are the future income values adjusted for inflation?
ANSWER: The results are presented in the values given by the Opportunity Atlas: in 2015 dollars, and are not adjusted for inflation. Note: These results control for income by focusing only on children born to parents at the bottom 25th percentile of income. Also, Please refer to endnote #69 and #70 for more information and visit Opportunity Atlas for more information on their research.