A look at growing neighborhood diversity, and continued segregation, in the 21st century United States
One of my favorite articles, published by Nate Silver in 2015, is titled “The Most Diverse Cities Are Often the Most Segregated”. It’s a great piece that combines storytelling, visualization and number-crunching to clearly explain most of the key pattens of segregation in major U.S. cities.
Now that the 2020 census is out, I thought I’d do a similar analysis, with my own spin on how to think about segregation and neighborhood-level diversity.
All figures here are based off Census Redistricting data, accessed via tidycensus, an API created by TCU Professor Kyle Walker.
I’ll be showing two main types of figures. Both are based on partitioning the population into 4 race/ethnicity categories (White, Black, Hispanic, Asian), and looking at how these groups are distributed across census tracts (areas of around 3,000 to 5,000 people). It’s not a perfect heuristic, but I think of a census tract as being comparable to one’s immediate neighborhood.
To minimize people yelling at me on Twitter, I’ll list caveats upfront. Like all analyses, this one involves generalizations and heuristics. I am aware that Mexican != Puerto Rican. I am aware that Chinese != Indian. I am aware that a small (but growing) share of the population doesn’t identify with one of these four categories. But I also believe that these are meaningful categories in the context of the U.S. social fabric, and that they capture enough information to be useful, while avoiding overcomplicating the analysis. I’d encourage you to do a more detailed analysis of your favorite city.
The first type of figure I’ll show is a histogram, showing tract-level “diversity index” in a given region. “Diversity index” is the probability that two randomly-selected people from the same census tract are in a different one of the four racial categories.
Here’s an example from Philadelphia, based off 2020 data:
The x-axis runs from 0 to 75. A tract that’s >90% of one racial group would have a diversity index close to 0. A tract that’s about evenly split between white, Black, Hispanic and Asian would have a diversity index close to 75.
The y-axis is a count of how many tracts have a diversity index in a given score range. For Philadelphia, the distribution is pretty even. There are many diverse tracts, many homogenous tracts, and many with midling scores.
Finally, the histogram is partitioned into different colors, based on the plurality race of each tract. Notice the left edge of Philly’s histogram is dominated by the color green. That’s because the most homogenous tracts in Philadelphia tend to be majority-Black. That is not uncommon in urban cores east of the Rockies.
The second type of figure I’ll use is a matrix that looks at how much tract-level interaction there is between each racial group. Again, I use Philadelphia as an example:
The way to read this is: if you pick a Philadelphian at random, and that person happens to be white, and then you pick another person at random from the same tract, there is a:
- 61% they’ll be white
- 17% chance they’ll be Black
- 11% chance they’ll be Hispanic
- 10% chance they’ll be Asian
Citywide, Phildelphia is about 36% white, 40% Black, 16% Hispanic and 9% Asian. So while there is substantial neighborhood-level interaction between white and Black Philadlephians, it’s much lower than if residents were distributed at random throughout the city. Once again, this is roughly par for the course in eastern urban cores.
Now that I’ve explained the two main tools, let’s look at some of the top trends:
Suburbs are diversifying everywhere. But Rust Belt suburbs look very different than Sun Belt suburbs.
If you’ve watched CNN or MSNBC on an election night in the past 10 years, you’ve no-doubt heard that America’s suburbs are diversifying. This is true. But it’s also true that many of them remain extremely white, particularly in the northern half of the country.
Here’s Hamilton County, just north of Indianapolis. In 2000, it had almost no diversity: the median tract was 94% white. Flash forward to 2020, and the middle of the chart finally has some meat to it, as the Black, Hispanic and Asian population shares have all increased.
But Hamilton County is still over 80% white, doesn’t have a single tract with a diversity index above 56, and has a median index of just 28. What diversity it does have is distributed fairly evenly throughout the County, but it remains small, growing at a fairly modest pace.
Suburbs in the largest Southern metros are a different animal. Here’s Gwinnett County, in the northeast Atlanta suburbs. Over the past 20 years, Gwinnett County has not only diversified, it’s become one of the most racially integrated places in America. The vast majority of tracts have an index above 50; the median is 65. And around half of all tracts aren’t plurality-white. Countywide, the white population share has been halved from 68% to 34% over the course of 20 years.
As an aside, while these two suburban counties have seen very different demographic trajectories over the past 20 years, their electoral trajectories have been fairly similar. Hamilton shifted 44% to the left in margin from 2000 to 2020, Gwinnett shifted 50% left. Hamilton’s shift is driven almost entirely by changes in voting behavior among educated whites, with demographic change playing only a supplemental role. In Gwinnett, it’s the opposite, with demographic change as the big driver.
Black isolation continues to define segregation in eastern urban cores, while immigration steadily increases overall neighborhood diversity
As discussed in the Philadelphia example, one of the starkest trends in U.S. segregation is Black isolation. In most of the largest urban counties east of the Rockies, the left edge of the histogram lights up green, indicating a cluster of tracts that are overwhelmingly Black, with relatively few tracts that are overwhelmingly white (or Hispanic or Asian). The below examples from Chicago, DC, and Atlanta all demonstrate this pattern:
But in comparing 2000 to 2020, you can also see a clear increase in neighborhood diversity. In all cases, this is at least partially driven by international migration and fast-growing Asian and Hispanic populations in urban cores.
Where can you find an eastern city that’s diverse at the neighborhood-level? Here’s two very-different examples
Like in many areas, New York City is an outlier with a remarkable level of neighborhood-level diversity.
To be sure, NYC has a ton of people living in homogenous neighborhoods. A decent chunk of the distribution has an index below 30. But in the median NYC tract, it’s more likely than not (53% chance) that two randomly selected people are of a different race. That’s not particularly close to being true of Chicago, Atlanta, Philadelphia or DC.
Here’s a very different city, with a suprisingly similar distribution: Indianapolis, with a median index of 52.
At the city-level, Indianapolis is substantially less diverse than New York City: it’s majority white, and has just a 4% Asian population share, whereas NYC is 33% white, with a population share of at least 16% in each of the four categories.
But the diversity Indianapolis does have is reasonably well spread out, and there’s little of the extreme Black isolation you see in other eastern cities.
Western metros have less segregation (and fewer Black residents)
On average, western metros have more neighborhood-level diversity, but this varies greatly. Clark County and San Francisco aren’t far from Gwinnett County levels of integration, wheras Maricopa and LA Counties both have a lower median index than New York City. And while the extent of Hispanic isolation in LA County doesn’t approach the magnitude of Black isolation in eastern cities, it certainly exists.
One western county that stands out is King County, Washington.
King County used to be very white, and still is very white for an urban county. But as its white population share has fallen from 77% in 2000 to 59% in 2020, it has integrated its growing diversity fairly well. Out of the 12 largest counties, it went from the 2nd lowest median diversity index in 2000, to the second highest in 2020 (behind only Queens).
But the elephant in the room in Seattle, and many other western cities, is the lack of Black residents. King County is just 7% Black. Compare side-by-side the histograms of Cook County and King County:
If you just look at the purple, these distributions aren’t horribly different. The key differentiatior is that the left edge of Cook’s histogram is filled with green, while King’s is mostly empty. The upshot is that the tracts most white Cook County residents live in aren’t *that* much less diverse than the ones white King County residents live in. But Cook County has a bevy of Black residents in highly segregated neighorhoods, while King County has few Black residents at all.
The U.S. will continue to see more neighborhood-level diversity, due to generational change and international migration. But segregation, particularly between Black and White residents, remains a problem. The few areas that have signficantly increased Black-White integration over the past 20 years, such as Gwinnett County, have mostly done so via Black residents moving into historically white neighborhoods. Instances of historically Black neighborhoods seeing a large increase in white residents are much discussed, but remain a rare occurence in the overall context of the metropolitan United States.