Racial and Ethnic Residential Segregation Across the United States

Front matter

i on Demographic Methods and Population ia fisis s 54

Amber R. Crowell Mark A. Fossett

Racial and Ethnic Residential Segregation Across the United States

Preface

A full draft of the main text of this monograph was completed in April 2023. In June 2023, just a day after we received a preview of the book cover, Mark A. Fossett passed away in an intensive care unit in Houston, TX, due to complications following heart surgery. Before he passed, he spent weeks in the ICU battling infections and fighting to recover his strength. During that time, I did what I could to make edits, touch up tables and figures, proofread, and have the monograph ready for Mark to take one last look before we sent it off. I also needed him to write his acknowledgments, something he had talked about doing but did not start. I never stopped believing that he would recover, return home, and eventually get back to puzzling over his favorite demographic problems. That did not happen, and Mark is no longer on this earth to see this book published. But this book is a product of 15 years of working closely together and, more importantly, it is a product of the decades of work that Mark devoted himself to in order to solve problems in segregation research that would help all of us better understand the landscape of residential segregation across the United States. The technical solutions to those problems can be found in his 2017 monograph, New Methods for Measuring and Analyzing Segregation, which was published by Springer and is open access. Those solutions are also integral to this book. So, in the way that his work lives on, so too he lives on and is still here.

When Mark and I started working together on the analyses that would eventually become this book, I was a PhD student and he was a professor and my mentor in the Department of Sociology at Texas A&M University. I had at that point been working with him for four years and over that time had spent countless hours listening to him explain, with both enthusiasm and careful attention to detail, the challenges of measuring residential segregation, and how they can be overcome. He was at the time working on his monograph that covered these topics. He spent hours every day meticulously refining the formulas, simulations, graphical demonstrations, Stata code, and language he used to explain his innovations in segregation measurement.

viii Preface

Sometimes all I could be was a sounding board as he did this work, but over the months and years, “through osmosis” as he would say, I began to deeply understand what he had accomplished and why it was so important. So much so that I was beyond eager to write this book with him, write papers with him, and give as many presentations as possible to teach others how to apply these methods and to appreciate how much it mattered.

The main problem with segregation measurement that he focused on was known, but obscure in its origins. Popularly used measures of segregation were prone to a sometimes very problematic upward bias that would inflate scores and make it challenging to understand segregation at a single point in time, follow segregation patterns over time, and compare segregation patterns across groups and communities. This problem was worse under conditions where small spatial units were needed to observe segregation, when the groups in the analysis were severely imbalanced in size, and when dealing with small population counts. Thus, much of what we knew about segregation was centered on urban metropolitan contexts, where conditions were most ideal for conventional segregation measurements. In the preface to his 2017 monograph, Mark named four concerns that troubled him and motivated him to do the methodological work that is featured in this book. First, there was no clear understanding of how segregation scores represent group differences on a residential outcome. Second, nobody could draw a direct quantitative link between household locational attainments and the overall segregation patterns that are produced. Third was the long-standing issue of segregation index bias, which limited studies of segregation to metropolitan contexts to the detriment of any research focusing on nonmetropolitan communities or smaller populations. And finally, during his work he developed a fourth concern, which is that the most popular measure of segregation, the dissimilarity index, is not always a trustworthy indicator of when prototypical segregation is occurring — that is, the pattern where two groups are living apart and are having very little residential contact with one another, and location-based inequalities are possible. For two decades, he confronted these problems and solved all of them with precision.

The primary purpose of this book, then, is not to focus on rehashing the technical details of the methods that solve these segregation measurement issues, but rather to put forward empirical applications of those methods to demonstrate how important it is to have accurate and trustworthy measurement tools for so many critical research questions about residential segregation across the United States. We spent years constructing datasets, conducting analyses, testing different ways to write about the discrepancies between what we found and what others had found using conventional and flawed methods, and sharing what we could with the academic community in the meantime. This book traveled a long way with us. It saw me advance to PhD candidacy, complete my PhD, begin a postdoctoral position, join the faculty at Fresno State, and earn tenure and promotion. It saw Mark reach the heights of his incredible career which included establishing the Texas Federal Statistical Research

Preface ix

Data Center and its consortium of universities at Texas A&M, being named a Cornerstone Faculty Fellow in the College of Liberal Arts, supervising a number of very excellent students, and eventually talking about retirement, which he was prepared to begin just before he passed. This book is, hopefully, a very helpful resource for benchmarking segregation patterns and trends and for improving research designs in the study of residential segregation moving forward. For me, it is also the manifestation of a long and deeply meaningful collaboration with a truly brilliant scholar.

This book would not have been possible without the groundbreaking and brilliant work that Mark did to bring solutions to segregation measurement that had eluded researchers for decades. His work over those decades was driven by an unshakable commitment to do good research. Mark was determined to uncover true answers, and he would not accept results that were not entirely “bullet-proof.” Sometimes this led to projects carrying on for years as he tested a measurement or a model from hundreds of different angles, but the result was a quality of work that laid a solid foundation for others to build on. Those efforts are what culminated in his 2017 book, New Methods for Measuring and Analyzing Segregation, which provided the methodological framework for this book. His book laid out in great technical detail the flaws in most segregation index scores, and more importantly, the perfect solutions that he developed. Mark was an impeccable scholar and a brilliant sociologist. He was also generous. He never boasted, but I imagine what he would be most pleased about was not his own incredible intellect but rather how his work helped others. Nothing excited him more than an idea that would make research better for everybody else. This included his elegant solution to the problem of segregation index bias, his techniques for agent-based modeling of segregation, the establishment of the Texas Federal Statistical Research Data Center at Texas A&M University which he spearheaded, co-directing multiple cycles of an undergraduate summer research institute to build a pipeline to bring students of color and first-generation students into graduate school (first funded by the American Sociological Association and later by the NSF-Research Experience for Undergraduates program), and his extraordinary energy for students, who got to be an integral part of it all. Including me.

This book has also received a tremendous amount of support from many individuals, institutions, and organizations. We are grateful to the Department of Sociology at Texas A&M University, the academic home for this work for many years and which supported Mark as a faculty member and me as a graduate student. We also recognize the tremendous role that the Texas Federal Statistical Research Data Center had in supporting this work, where Mark was Executive Director and later Associate Director and I was first a graduate research assistant and later a postdoctoral research associate. Some of the analyses in this book were conducted at the Texas RDC and would not have been possible without the energy and competence of Dr. Bethany DeSalvo, who was the RDC Administrator of the Texas RDC at the x Preface time. We are also grateful to the Department of Sociology at California State University, Fresno, where I joined the faculty and completed writing this book. I am especially thankful for my good friends and colleagues at Fresno State who over many years offered encouragement and advice as I worked on this book with Mark. Dr. Jennifer Randles in particular took on the role of mentoring me, and I am fortunate to have been able to learn from one of the brightest and most thoughtful sociologists I have ever known. Additionally, Dr. Randles along with my other friend and colleague Dr. Justin Myers provided a wonderful support network for simultaneously accomplishing the equally arduous tasks of writing books and raising small children. It is no small thing to have that space as a new academic parent. I am also grateful to their partners, Craig Bailey and Shay Myers, for being a part of that village. Finally, I am thankful for the friendship and collegial support of Dr. Cristina Herrera at Portland State University, who has served as a model for how to do excellent scholarship and maintain a work-life balance.

There are specific people who were at Texas A&M University during periods of time that this work was being completed who we recognize for their support. Dr. Jane Sell, Dr. Dudley Poston, Dr. Walter Gillis Peacock, Dr. Rogelio Saenz, and Dr. Mary Campbell deserve mention for their extended support and companionship over the years. They were part of a community of friendship and scholarship that for some dates back to Mark’s early career days and who remained his friends until the end. We must give special mention to Dr. Wenquan (Charles) Zhang, now at University of Wisconsin — Whitewater, who collaborated with Mark for many years on segregation measurement issues and later worked with both of us on empirical analyses employing the methods featured in this book.

We have been the benefactors of the amazing support of Evelien Bakker, Bernadette Deelen-Mas, Prasad Gurunadham, and Corina van der Giessen at Springer, who have very gently and patiently encouraged us to complete this book and were always responsive when we needed more time and more flexibility. In the acknowledgments that Mark wrote in his previous book, he said, “Apparently, it is impossible to exhaust their patience and goodwill.” I imagine he would say something very similar again today, and I would agree.

Other individuals whose contributions to this work we are grateful for include Luna Chandna and Nereyda Ortiz, who are currently graduate students at Texas A&M University and have been supporting an analysis that did not make it into this book but will be published as an article. Mark would also want to thank his many current and former students who have shared ideas with him, learned from him, and co-created space for him to refine the way he described and explained the methodological challenges that he was taking on. These individuals include, in no particular order, Dr. Warren Waren, Dr. Lindsay Howden, Dr. Gabriel Amaro, Warner Henson II, Dr. Bianca Manago, Dr. Jennifer Davis, Dr. Jessica Barron, Dr. Nicole Jones, Bo Hee Yoon, Brittany Rico, Dr. Marisa Sanchez, Dr. Xuanren Wang, Chiying Huang, Katelyn Polk, Bridget Clark, Dr. Danielle Deng, Dr. Nathanael Rosenheim, Cassidy Castiglione, Dr. Xinyuan Zou, Mary Jalufka, Megan Bodily, and Dr. Michael

Preface xi

Upchurch. If there are others that I left out, I sincerely apologize. I know Mark would have named every single student who gave him even a minute of their time to indulge him as he shared his latest thoughts on segregation measurement, because talking with students about research always delighted him. Mark would also want to thank his dissertation advisor, Dr. Omer Galle, for being a lifelong mentor and friend. There are many other colleagues across the discipline who have given us feedback, served as discussants, asked us questions, and spent time with us as we did this work. In no particular order, we recognize Dr. Michael White, Dr. Daniel Lichter, Dr. Matthew Hall, Dr. Ann Owens, Dr. Peter Rich, Dr. Daniel Powers, Dr. Tod Hamilton, Dr. Clark Gray, and Dr. Elizabeth Korver-Glenn, Dr. Eric Jensen, Dr. Angelica Menchaca, and Dr. Bryce Hannibal. We also extend gratitude to the Cornell Population Center for inviting us to give a workshop on segregation measurement in 2019, and to the U.S. Census Bureau for inviting us to present our work during the Summer at Census series in 2022.

We also recognize and appreciate the funding support we have received over the years from the National Science Foundation and the National Institutes of Health. The National Science Foundation has through multiple grants supported Mark’s work on segregation measurement (SES #1024390), my training as an undergraduate and graduate student (SES #0649277), and the open access publishing of this book (SES #2222573). The National Institutes of Health supported Mark’s work on agent-based modeling to simulate segregation dynamics (R43HD038199 and R44HD038199). In addition, we are thankful to the journal editors at Sociology of Race and Ethnicity, Spatial Demography, and Demographic Research for publishing work that came out of this project.

Finally, and most importantly, we give immeasurable gratitude and recognition to our families. I am fortunate to have the unwavering support of my husband, Zach, who kept on encouraging me as I worked on this book over the years, including during our very difficult 16-month isolation period of the Covid-19 pandemic, when we were both trying to work and care for our two small children at home at the same time. This is when Mark and I inexplicably made the biggest push on writing this book. We could not have done any of that without the selfless help of my mother-in-law, Debra, who flies out to California from Texas every few weeks and lived with us during our entire pandemic isolation period to help with our children. My fatherin-law, Frank, has also been a major help over the years in this regard. I am also so very grateful to my parents, Gabriela and Brian, and my brothers, Marcelo, Garrett, Brandon, and Benjamin, for encouraging me through my academic career and cheering on every accomplishment, including the completion of this book. Finally, I am lucky to be mom to Ash and Peter, two amazing little humans who keep me focused on the adventure of life and remind me that there is more to be gained if I prioritize balance.

I am not capable of writing Mark’s words to his wife, Betsy, or their children, Lane, Tyler, and Kate, and their families. I would not dare try. But I know he loved them completely and endlessly and would thank them for bringing him joy. I am also thankful, as Mark’s friend, that he was surrounded by such a beautiful family.

xii Preface

A few days after Mark passed, our dear friend Dr. Walter Gillis Peacock at Texas A&M University told me that I am not publishing this book without Mark, but rather I have the honor of publishing it with him. He is very correct. I send all my gratitude out into the universe to Dr. Mark Fossett for giving me the opportunity to learn from him and to share credit with him on this very good and important work. It was the greatest honor, and one of the greatest experiences of my life, to be Mark’s student, colleague, coauthor, collaborator, and friend. We shall declare a victory.

Fresno, CA, USA Amber R. Crowell June 2023

Abstract

This monograph builds on innovations in segregation measurement and analysis, previously developed by one of the authors of this book, by conducting empirical analyses of racial and ethnic residential segregation across a wide and comprehensive selection of communities in the United States. Past studies of residential segregation have been limited by a well-known and difficult challenge, which is that most segregation indices are prone to a sometimes very problematic upward bias that inflates segregation scores and makes it difficult to measure segregation at a single point in time, follow segregation patterns over time, and compare segregation across groups and communities. These problems are worse when using small spatial units such as census blocks, when the groups in the analysis are extremely imbalanced in size, and when population counts are small. This has resulted in a literature that is heavily focused on segregation in a selection of the largest urban metropolitan environments, with only limited studies focused on nonmetropolitan communities or small racial and ethnic populations. Even so, restrictive case selections do not directly solve the problem of index bias. Fortunately, we have the solution to index bias, in addition to other solutions that address related problems with segregation measurement, which allow us to reanalyze residential segregation patterns and include more communities and contexts. In this book, we examine White-Black, White-Latino, and White-Asian residential segregation across metropolitan, micropolitan, and noncore county communities from 1990 to 2010, giving special attention to how our findings may differ from what previous studies have found with measures that were not corrected for index bias and other related issues. We find that under the conditions where index bias is less likely to be a problem, our results track those from previous studies. But these communities do not make up the majority of cases, and in most communities our findings deviate in substantial ways from previous findings. We also employ new methods for linking micro-level processes xiv Abstract of locational attainments to overall segregation patterns and develop a more complex understanding of residential segregation dynamics. This leads us to conclude that it is important to use our findings as benchmarks for residential segregation patterns over this time period and to adopt the methods of measurement and analysis that we endorse throughout this book for residential segregation research.

Contents

L Tronoh. g35 hon hb SEEK KESSEL SOS RE HE SESS Vl OYENTE cnet phe se SREB Re Re eRE ORES ERE 1.2 Brief Note on Measurement and Implications for Future

21 OUNEN nde Soke tees Eeee RA bA enatee eet aae es 2.2 What Is Residential Segregation and What Motivates Us

tO SMG IT, sc cia eke Rue e eRe eR RE LEE A Eee RES KEES 24- Details Gl Study Desi. 2 vse ripiter deed wes bed Gea oo 2.4.1 Measuring Segregation in Metropolitan Areas, 2A Gipup COMPASS... heehee eee bb de ee he eb 84 2.4.4 Combinations of Group Comparisons Across Communities and Times 25 220226436 .eedde bbs ons

Chapter 1 A Introduction e

The purpose of this book is to describe and analyze patterns and trends in racial and ethnic residential segregation across the United States over time and across communities. With new methods to expand our scope of analysis beyond what has been done before, we cover recent decades in a variety of settings including metropolitan and micropolitan areas and rural communities (i.e., noncore counties). We direct our primary focus to residential segregation between major panethnic racial groups — Non-Hispanic White, Black, Latino, and Asian households in 2010 — and to broad changes in segregation from 1990 through 2010. But we also give attention to several more detailed aspects of trends and patterns in residential segregation. While the literature in sociology, demography, urban planning, and geography is rich with studies of residential segregation patterns, we believe this book establishes an important baseline for placing recent segregation research in a new context and for informing segregation research going forward. The basis for this is that we apply new methods for measuring and analyzing segregation that can at times drastically alter results obtained using more traditional approaches. In particular, we argue that these new methods of measurement and analysis address and overcome important methodological problems that have limited past research and, as a result, allow us to expand the scope of segregation studies and the quality of measurement to obtain improved findings that more accurately capture and reflect the demographic reality we are seeking to document.

With the exception of Chap. 2, which reviews and explains our methodology and study design, the chapters in this book give attention to a set of important substantive concerns addressed in the broader sociological and demographic research literature on residential segregation. Even as we describe and analyze patterns of residential segregation between panethnic racial and ethnic groups, an area that has been heavily researched already, we bring significant improvements in strategies of

© The Author(s) 2023 1 A. R. Crowell, M. A. Fossett, Racial and Ethnic Residential Segregation Across

the United States, The Springer Series on Demographic Methods and Population measurement and analysis when covering this familiar territory. Additionally, we expand the analysis to give close attention to segregation trends in nonmetropolitan and rural settings that are less frequently studied. We also give special attention to segregation patterns in communities that are seeing new and increasing presence of racially and ethnically minoritized populations as the population of the United States steadily becomes more diverse, not only in immigrant gateway cities and areas with established minoritized group presence, but also in metropolitan and nonmetropolitan settings across all regions of the country. Finally, we take the analysis of segregation to a further stage of innovation, where we use new methods to conduct community-specific analyses of micro-level segregation dynamics that shape overall segregation patterns.

The common theme connecting all of the empirical chapters is that we are able to delineate the levels, patterns, and trends in segregation more clearly and accurately than has been possible in previous research by drawing on attractive new options for measuring and analyzing segregation. We necessarily provide an overview of these new methods and note the advantages we gain by using them in Chap. 2. But we do not intend this work to be primarily a study of segregation methodology. Indeed, this is not necessary because the methods we use have been previously introduced and reviewed in depth in a recent work by one of the authors of this book (Fossett, 2017) and we have previously empirically demonstrated the advantages of these new methods of segregation measurement and analysis in earlier work by both authors (Crowell & Fossett, 2018, 2020, 2022). Instead, we intend the main contribution of this study to be to demonstrate the value of applying new methods to help obtain improved answers both to basic questions that have been addressed in the empirical literature for decades and also to questions that have been neglected in past research due to the limitations of earlier methodological practices.

In some cases, as in the study of nonmetropolitan segregation, the previous research literature has been extremely limited in scope and in the conclusions that are drawn because conventional methods of segregation measurement are known to be untrustworthy and potentially misleading for the analysis of interest. This issue is crucially relevant, for example, when investigating segregation in nonmetropolitan settings where it is necessary to use units that are smaller in spatial scale and population size (e.g., census blocks) to measure segregation (Fossett, 2017; Lichter et al., 2010, 2016) and also when the groups in the analysis are imbalanced in size as is certain to be the case in new destination communities and in other communities where new groups are taking on an increasing presence in the population (Hall, 2013; Lichter et al., 2010; Saenz, 2010; Frey, 2018; Winkler & Johnson, 2016; Vasquez et al., 2008). The methodologies we use overcome the challenges that rendered previous measurement strategies untrustworthy and potentially misleading in these situations. This is all to say that we are not necessarily exploring or identifying new areas of segregation research, but rather we are revisiting established areas of research with new and improved methods for understanding the dynamics of residential segregation in a variety of demographic contexts.

Ultimately, this book provides a comprehensive overview of residential segregation in the United States from 1990 up to and through 2010. In addition to describing

1.2 Brief Note on Measurement and Implications for Future Research 3 residential segregation patterns in all areas of the United States, including Latino and Asian new destinations, we test major sociological hypotheses about the mechanisms and dimensions of segregation, highlight new methodological approaches to measuring and analyzing segregation, and offer suggested paths to continue the work of understanding residential segregation in the United States in the twenty-first century. The phenomenon of racial residential segregation shows little sign of abating or becoming an object of backward-looking historical interest. To the contrary, it appears the study of residential segregation unfortunately will be a priority issue in demographic and social science research well into the future. We hope to help improve the efforts in this field by demonstrating the advantages of new methods of measurement and analysis and showing how they make it possible to expand the scope of feasible research on segregation. These efforts will make segregation research more comprehensive and inclusive of a broader range of group comparisons and community settings. Finally, we believe that improving the quality of measurement and the scope of analysis that is possible in segregation research will not only clarify patterns and trends in residential segregation but also contribute to better evaluation and refinement of existing theories that stimulate new insights into the social dynamics that produce residential segregation.

1.2 Brief Note on Measurement and Implications for Future Research

While we review the value of the methods we use in this study in detail in Chap. 2, we want to emphasize here that our approach to measuring segregation is the basis for one of the major contributions of this book. As far back as Winship (1977) it has been acknowledged that commonly used measures of segregation have inherent flaws that lead to upward bias of segregation index scores that can be concerning in general and deeply troubling under certain conditions, particularly when analyses involve small areas or groups that are vastly disproportionate in size. The most common approach in the literature since Winship (1977) has been to simply restrict the scope of segregation studies to avoid conditions where inherent bias in index scores is most worrisome. This has resulted in certain populations and communities being neglected in the broader literature, either directly through outright exclusion from analysis or indirectly by down-weighting segregation comparisons in empirical analyses, and it has resulted in foregoing research in smaller communities where it is necessary to operationalize neighborhoods at small spatial scales. However, the need to adopt broad restrictions on study design to avoid examining segregation in settings that pose challenges for conventional methodological approaches is no longer warranted. Fossett (2017) has introduced refined formulations of all popularly used segregation indices that eliminate the upward bias inherent in their original formulations. With these new formulas, we proceed in this book to reexamine segregation across the United States free of concern for those particular problems that plagued segregation research in the past.

Additionally, we pick up the conversation begun by Fossett (2017) on segregation indices that diverge from one another and demonstrate in each empirical chapter the care that is needed in deciding which segregation index is most appropriate for capturing the most important aspects of the dimension of uneven distribution in any particular scenario. We also take advantage of an innovation to Fossett’s (2017) segregation index reformulations which is that they can now be easily disaggregated and understood as measuring the difference in group means on individual residential outcomes. Placing segregation indices in a conceptual framework where index scores correspond to group-level aggregations of individual-level residential outcomes opens up new and exciting opportunities to analyze segregation as a group-level outcome driven by micro-level dynamics and use the toolkit of methods popular in inequality studies. This in particular builds a bridge between two traditions in the segregation literature that are described in more detail in a later section of this chapter.

The final methodological innovation that we mention here is that we build on the work of Fossett (2017) to call attention to the finding that problems associated with measuring residential segregation using data for persons — which is by far the most common approach used in the research literature — are greater than is generally realized. We first establish that the problems are substantial, and we then review methods that deal with them successfully to permit more accurate and trustworthy measurements of segregation. Our findings on this point have important implications for future research including research using the newly released data files from the 2020 Census of Population and Housing. Specifically, our findings show that analysis of levels and trends in segregation based on index scores computed using tabulations of persons in combination with conventional formulas will consistently overstate levels of segregation by greater amounts than is currently appreciated.

Furthermore, since the impact of index bias, the technical problem that inflates levels of segregation above their true value, varies across cases in complicated ways, there are no easy ways to address the problems when following conventional practices of measuring segregation using data for persons. Happily, we show here that superior, trustworthy measurements of segregation can be achieved. But we also show that it requires using both different measurement approaches and different data. Specifically, it requires using methods for unbiased measurement outlined in Fossett (2017), and these measurement methods must be applied either in combination with data for households or with detailed data for persons by size of household. We review these and other methodological choices we adopt and encourage other researchers to consider in Chap. 2.