Roots of Structural Racism

Introduction

In June of this year, we published the “Roots of Structural Racism” project examining the persistence and harmful effects of racial residential segregation in the United States from a number of angles. We examined trends, ranked cities and metropolitan areas, compiled local histories and narratives, analyzed economic, political, and educational correlates to segregation, and created an interactive mapping tool that allowed users to explore their own communities.

As our project had been long in development (and delayed long beyond our intended release), we elected not to wait until the 2020 US Census data was available to publish our findings. We fully planned, however, to update our study when the census data became available and more readily useable. The formatted 2020 census data became available in September of 2021, and thus we present our updated research findings in this bulletin as planned.

Key Findings:

• We find that 23.6 percent of metropolitan areas were more segregated in 2020 than they were in 2010, and 54 percent of metropolitan areas were more segregated in 2020 than in 1990.
• The top 10 most segregated cities in the country, as of 2020, are Detroit, Newark, Hialeah, Chicago, Milwaukee, Cleveland, New York City, Irving, Miami, and St. Louis.
• The top 10 most segregated metropolitan regions of the country, as of 2020, are Chicago, New York City, Milwaukee, Detroit, Trenton, Miami, Los Angeles, Cleveland, Port Arthur, and Atlanta. Philadelphia holds the eleventh spot in both lists.
• The most segregated regions in the country remain the Midwest, Mid-Atlantic, and West Coast. 
• We find significant racial demographic change in the country between 2010 and 2020, with the Asian and multi-racial demographic segment growing the fastest in percentage points terms (36 and 132 percentage points increase, respectively).
• In addition to Port St. Lucie, FL and Colorado Springs, CO from our analysis using 2019 ACS data, we find that Chico, CA, Palm Bay-Melbourne-Titusville, FL and Jacksonville, NC are the only five "integrated" metropolitan regions in the country as of 2020.   

A Note on the 2020 Census 

The United States Constitution requires an “actual enumeration” of people living in the United States every 10 years. The decennial census is the gold standard for demographic and population data in the United States because a serious effort is made to account for every person in the United States. In contrast, other data sources are either samples or surveys and are not as accurate as the decennial census. For example, in our June report, we relied heavily upon the 2019 American Community Survey, a rolling sample survey. The ACS data, while useful, has a larger margin of error.

That said, the 2020 census was a fraught undertaking under difficult circumstances, and as a result, the data contains more systematic errors than usual. The decennial census is a massive operation that requires about half a million additional workers to conduct. Teams of surveyors search every nook and cranny of the country, knocking on doors to get an accurate headcount. The COVID-19 pandemic, however, hit just as the census was getting underway. This made traditional operations much more difficult. In addition, this was the first “online” census, where people were urged to fill out census surveys by computer, with all the additional challenges this entailed.

As a result, many community advocates were concerned about undercounts, especially for hard-to-reach populations such as undocumented immigrants, homeless and unsheltered people, and the young and transient, among others. Large-scale efforts were devised to raise public awareness of the importance of the census, including the hundreds of millions of dollars in federal funds that are allocated as a result of the census, in addition to the reapportionment of political representation. In particular, many community advocates were concerned that undocumented persons might be reluctant to fill out census results for fear of being tracked by the government. Homelessness has risen since 2016, especially in states like California, making it much more difficult to survey everyone. As a consequence of the pandemic, many of these advocacy efforts were undercut or throttled.

Additionally, the Trump administration proposed last-minute changes to the census, attempting to add a citizenship question to the census, which would be a transparent attempt to further discourage reluctant non-citizens from filling out the census. The obvious political motivation was to try to tilt the results in ways that may, in the long run, prove politically advantageous to that administration. A legal challenge to this effort went all the way to the Supreme Court, and it was ultimately thwarted, but the Trump administration’s management of the count may have affected the count in ways that are difficult to measure.

On top of these problems, the Census Bureau attempted, for the first time, to introduce “noise” into the data to protect privacy. This may affect the accuracy and precision of the data at small levels of geography.

All of this is to say that the 2020 census was a fraught endeavor, and there may be aspects of the operation that affect the results, although in ways we may never fully understand or appreciate.

Racial Demographics in 2020

In 2020, the United States was found to have nearly 331 million residents, with a racial breakdown presented in Table 1 below. In particular, the 2020 census found that there were 19.6 million Asians (5.9 percent of the population), 39.9 million African Americans (12.1 percent), 62 million Latinos (18.7 percent), 2.3 million Native Americans (0.7 percent), and 191.7 million white people (57.8 percent) living in the country.

Table 1: Racial Composition in the United States, 2010-2020

The white population fell in both relative and absolute terms, showing that the United States continues the progression to a “majority-minority” country. Although the African American population grew in absolute terms (by 6 percent), it slightly declined in relative terms. The Latino and Asian populations grew rapidly in the United States, increasing in absolute terms by 23 and 36 percent, respectively. 
The most dramatic change, however, is the large increase in people who identify as multi-racial, a 132 percent increase from 2010 to 2020.1

Racial Compositional Change at the Metropolitan Level

The United States continues to undergo significant racial demographic and compositional change, as it always has. But these changes are not evenly spread out across the United States. Some places have seen rapid growth or decline among particular racial groups. We identified the metropolitan regions (with populations above 200,000 people) with the largest increase or declines in the specific representation of major racial groups in both absolute and relative terms. Here, we highlight the most significant changes.

For African-Americans, the metropolitan areas with the largest increases in racial representation from 2010 to 2020 are St. Cloud, MN (a 5.1 percent increase to 8.3 percent of the population), Fargo, ND-MN (4.4 percent increase), Cedar Rapids, IA (2.7 percent increase), Gulfport-Biloxi, MS (2.3 percent increase), Las Vegas-Paradise, NV (2.2 percent increase), Lafayette, IN (1.9 percent increase), Minneapolis-St. Paul-Bloomington, MN-WI (1.9 percent increase), Rochester, MN (1.8 percent increase), Sioux Falls, SD (1.6 percent increase), and Worcester, MA (1.5 percent increase).

The metropolitan regions with the largest declines in African-American racial representation between 2010 and 2020 are Hilton Head, SC (5.3 percent decline), Charleston-North Charleston-Summerville, SC (4.7 percent decline), Wilmington, NC (3.2 percent decline), Savannah, GA (3.1 percent decline), Durham-Chapel Hill, NC (2.6 percent decline), Richmond, VA (2.5 percent decline), Seaford, DE (2.1 percent decline), Myrtle Beach-North Myrtle Beach-Conway, SC (2.1 percent decline), Raleigh-Cary, NC (1.9 percent decline), and Charlottesville, VA (1.7 percent decline).

For Latinos, the metropolitan with the largest increases in racial representation from 2010 to 2020 are Winter Haven, FL (8 percent increase to 27.9 percent of the population), Merced, CA (6.9 percent increase), Orlando, FL (6.9 percent increase), Reading, PA (6.8 percent increase), Trenton, NJ (6.6 percent increase), Modesto, CA (6.2 percent increase), Scranton, PA (5.8 percent increase), Bakersfield, CA (5.7 percent increase), Yakima, WA (5.6 percent increase), and Allentown, PA (5.3 percent increase). 

The only three metropolitan regions with declines in Latinos as a share of the population are San Jose, CA (a 1.5 percent decline to 26.3 percent), Laredo, TX, and Hilo, Hawaii, both registering a 0.5 percent decline.

For Asians, the metropolitan regions with the largest increases in racial representation from 2010 to 2020 are San Jose-Sunnyvale-Santa Clara, CA (6.9 percent increase), San Francisco-Oakland-Fremont, CA (4.3 percent increase), Seattle-Tacoma-Bellevue, WA (3.9 percent increase), Trenton-Ewing, NJ (3.6 percent increase), Stockton, CA (3.5 percent increase), Sacramento-Arden-Arcade-Roseville, CA (2.9 percent increase), Champaign-Urbana, IL (2.7 percent increase), Raleigh-Cary, NC (2.6 percent increase), Dallas-Fort Worth-Arlington, TX (2.6 percent increase), New York-Northern New Jersey-Long Island, NY-NJ-PA (2.5 percent increase).

Only three metropolitan regions experienced a decline in Asian population share greater than 0.1 percent: Hilo, HI (2.3 percent decline), Honolulu, HI (0.8 percentage point decline), Norwich-New London, CT (0.2 percent decline).

Remarkably, for white Americans, the only three metropolitan regions that experienced a percentage point increase in their racial representation from 2010 to 2020 are Hilton Head Island-Beaufort, SC (2.4 percent increase), Hilo, HI (1.0 percent increase), Laredo, TX (0.2 percent increase).

The metropolitan regions with the largest declines in white population share from 2010 to 2020 are Trenton-Ewing, NJ (11.0 percent decline), Lakeland-Winter Haven, FL (10.1 percent decline), Seattle-Tacoma-Bellevue, WA (10.1 percent decline), Orlando-Kissimmee-Sanford, FL (10.0 percent decline), Scranton-Wilkes-Barre, PA (10.0 percent decline), Worcester, MA (9.8 percent decline), Champaign-Urbana, IL (9.4 percent decline), Poughkeepsie-Newburgh-Middletown, NY (9.3 percent decline), Allentown-Bethlehem-Easton, PA-NJ (9.2 percent decline), Chico, CA (9.2 percent decline).

For Native Americans, the metropolitan regions with the largest increases in racial representation from 2010 to 2020 are Fort Smith, AR-OK (0.6 percentage point increase), Albuquerque, NM (0.4 percent increase), Anchorage, AK (0.3 percent increase), and, Fargo, ND-MN (0.2 percent increase) followed by Lubbock, TX, Prescott, AZ, Sioux Falls, SD, Amarillo, TX, Albany-Schenectady-Troy, NY, Athens-Clarke County, GA; all with an increase of 0.1 percent.

The metropolitan regions with the largest declines in Native American population share from 2010 to 2020 are Oklahoma City, OK, Tulsa, OK, Houma-Bayou Cane-Thibodaux, LA, and Fayetteville, NC, with a 0.3 percent decline, and, Pensacola-Ferry Pass-Brent, FL, Reno-Sparks, NV, Greeley, CO, Fayetteville-Springdale-Rogers, AR-MO, Seaford, DE, Seattle-Tacoma-Bellevue, WA, with a 0.2 percent decline.

Change in Racial Residential Segregation

We mention these changes as a backdrop to observed changes in the levels of racial residential segregation in the United States.

Although the heart of our study was to draw attention to the relationships between racial residential segregation and observed life outcomes, one of the main statistics most frequently cited from our study was the finding that 81 percent of major metropolitan areas in the United States (169 out of 209) were more segregated as of 2019 than they were in 1990. This finding relied upon the 2019 ACS estimate. When using the (hopefully) more precise 2020 census data, we find that this finding is not quite as dramatic.2

We find that 53.80 percent of metropolitan areas were more segregated using our preferred measure of segregation in 2020 than in 1990. 

What explains the difference between the 2019 ACS estimate and the 2020 census data? While we can’t be sure, we suspect it probably is a result of the greater imprecision of the ACS compared to the census. Although this statistic is not as dramatic, it does suggest, in our view, that racial residential segregation continues to be a serious problem, and one that is not fading away. 

Furthermore, we found that 23.6 percent of metropolitan areas were more segregated in 2020 than they were in 2010 using the Divergence Index (described in our technical appendix), our preferred measure of segregation. We believe this is because the increases in the proportion of some racial groups tends to be highly localized, so that increases in racial diversity tend to produce increases in racial residential segregation.

We explained this tendency in greater detail in the FAQ we published in the months after the launch of our project. In short, to reduce the observed level of racial residential segregation, any increase in diversity would have to be more evenly distributed than the existing demographic composition. This is rarely the case. 

Using more traditional measures like the Dissimilarity Index or the Isolation/Exposure Indices, racial residential segregation appears to persist. The national Black-white dissimilarity index fell from 59 in 2010 to 55 in 2020, still in the moderate-high range.  This figure indicates that more than fifty percent of Black or white Americans would have to move to a different race neighborhood to integrate. White-Asian and white-Hispanic dissimilarity scores remain steady.

The Isolation/Exposure scores show even less progress in critical respects.  Black "exposure" to white residents fell from 35 in 2010 to 34 in 2020, the same value for 1990 and 2000. The "typical" Black and white Americans reside in vastly different neighborhood milieus. As of 2020, the average white resident of a metropolitan area resides in a neighborhood that is 69% white, 9% Black, 12% Hispanic, and 6% Asian. In contrast, a typical African-American resident lives in a neighborhood that is 41% Black, 34% white, 17% Hispanic, and 6% Asian.

The Most (and Least) Segregated Cities and Regions of the United States

Using the same methodology in our study, we generated a list of the most to least segregated cities and metropolitan regions in the United States (also available on the right-hand menu table).

Overall, the order of cities in this list is not much different from the one we produced using 2019 data. However, this update has allowed us to report on the levels of segregation of six more cities, bringing the total number of cities we ranked to 119. Most of the cities are listed in the same bracket of segregation as we reported in our previous release. Only Aurora, CO and Sacramento, CA have moved from the “High Segregation” to “Low-Medium Segregation” category.

Port St. Lucia and Colorado Springs remain the only two “integrated” cities in our findings. We suspect this has to do with the presence of military bases in those cities. The military is unique because it has the coercive power to draw people from every corner of society together, and foster integration in ways that generally do not occur otherwise.

Similar to our analysis of cities, the list of most-to-least segregated metropolitan areas is not much different from our 2019 list. However, we report four more metro areas in this list compared to the last one. Although the top 20 most segregated metro areas are the same as reported in our study, there are some notable changes. 

Most importantly, we found that three additional metro areas now meet our definition of “integrated”: Chico, CA, Palm Bay-Melbourne-Titusville, FL and Jacksonville, NC.  The latter is a new addition to our analysis, whereas the other two metros moved from “Low-Medium Segregation” to “Integrated.”

Seven metros improved their ranking by moving from “High Segregation” category to “Low-Medium Segregation” category: Albuquerque, NM, Ann Arbor, MI, Augusta-Richmond County, GA-SC, Daphne-Fairhope-Foley, AL Micropolitan Statistical Area, Rockford, IL, South Bend-Mishawaka, IN-MI and Wichita, KS.

We also found that the Mid-Atlantic, Midwest, and West Coast remain the three most segregated regions of the country (see Figure 1), in that order, an identical pattern to what we found this spring.

Figure 1: U.S. Regions, Ranked by Segregation

Notable Segregated and Integrated Neighborhoods in the United States

In response to our study, some observers raised important questions about the specific relationship between racial diversity and observed levels of racial residential segregation using the Divergence Index. We responded to questions of this nature in our FAQ, although we do not claim to have fully resolved this issue. In our study, the change-over-time comparisons we make are simply comparing the Divergence Index scores at one point in time to another. Because we do not control for diversity in making these comparisons (for reasons we provided in the FAQ), we are cognizant of the fact that some observers may find such comparisons to be less revealing or useful.

For that reason, we also provide static, point-in-time findings regarding the most-to-least segregated cities and regions in the country, using the 2020 census. Even for those who may be critical of the Divergence Index as a measure of segregation for analyzing change over time, these lists of most-to-least segregated cities and regions can provide some sense of the relative degree of the spatial distribution of racial groups, especially for large cities or regions with high levels of diversity or similar racial compositions.

For people nonetheless concerned about compositional dependence using point-in-time rankings, our interactive mapping tool has been updated to include the 2020 census, and it should be able to provide highly insightful figures regarding the observed level of racial residential segregation at the census tract or neighborhood level. This is something that, without special micro-data, the Dissimilarity Index cannot do.

To illustrate this, we present contrasting racially segregated and integrated neighborhoods in three metropolitan regions: San Francisco-Oakland-Fremont, CA, Dallas-Fort Worth-Arlington, TX, and New York-Northern New Jersey-Long Island, NY-NJ-PA. We present this analysis to not only illustrate the dynamics of racial segregation at the neighborhood level, but also to show how to use our powerful mapping tool.

The San Francisco metro area is approximately 27 percent Asian, 7 percent Black, 23 percent Latino, and 36 percent white. The Dallas metro is 8 percent Asian, 16 percent Black, 29 percent Latino, and 43 percent white. The New York metro is approximately 12 percent Asian, 15 percent Black, 25 percent Latino, and 43 percent white. Considering that these metro areas are high in racial and cultural diversity, our analysis found many tracts in the categories of High Segregation, Low-Medium Segregation, and Racially Integrated in each metro.

Let us review the Richmond Annex neighborhood in Richmond, CA, in Figure 2 below. It is racially integrated, with a population that is 20 percent Asian, 8 percent Black, 18 percent Latino, and 39 percent white.

(Click image to enlarge)

In contrast, consider the Parkview, Eastshore, Panhandle Annex and Laurel Park neighborhoods in Richmond, CA which are highly segregated Black and Latino neighborhoods in Figure 3 below.

(Click image to enlarge)

With a 13 percent Asian population, 37 percent Black population, 31 percent Latino population, and 12 percent white population, this is just next to the racially integrated tract described above. This highlights the range of levels of segregation in a highly diverse metro. In particular, these two neighborhoods are virtually adjacent. Similar juxtapositions can be found throughout the Bay Area using our mapping tool.

To consider another contrasting example, take a look at the Hollywood/Santa Monica neighborhood and the residential areas north of Tennison Park in Dallas, TX, in Figure 4 below. This neighborhood is racially integrated with a population that is 4 percent Asian, 16 percent Black, 29 percent Latino, and 43 percent white.

(Click image to enlarge)

In contrast, consider the neighborhood adjacent to and south of Hollywood/Santa Monica in Dallas, TX, shown in Figure 5 below. This neighborhood is highly segregated, and is adjacent to the racially integrated tract in the city described above. With a 1 percent Asian population, 1 percent Black population, 52 percent Latino population, and 42 percent white population, it illustrates how residential segregation plays out in racially diverse areas as well.

(Click image to enlarge)

Finally, take a look at the Hell's Kitchen (or Clinton) neighborhood in New York City, NY shown in Figure 6 below. This neighborhood is racially integrated with a population that is 22 percent Asian, 10 percent Black, 19 percent Latino, and 44 percent white.

(Click image to enlarge)

Similarly, the east end of Bedford-Stuyvesant in New York City, NY shown in Figure 7 is segregated with a high Black population. With a 4 percent Asian population, 52 percent Black population, 17 percent Latino population, and 20 percent white population, it shows the high level of segregation in this tract. 

(Click image to enlarge)

These tracts reflect a substantial amount of racial diversity in regions that are equally diverse. But these also illustrate how well the Divergence Index measures the residential segregation pattern. It is visible from the images above that neighborhoods within the city range from the categories of Highly Segregated to Low-Medium Segregation, and Racially Integrated as well.

We encourage you to use our mapping tool to observe similar dynamics where you live.

If you have any further questions or requests, do not hesitate to reach out to us at housingOBI@berkeley.edu. If you have data requests, please use our data request form here.

Learn more about our analysis of national segregation here