White¹ Americans have comprised the majority of the U.S. population since the nation’s founding. This majority has steadily declined over the past 50 or so years, with non-white Americans expected to comprise more than 50% of the U.S. population in 25 years.
In an ideal world, this milestone would be a little-known piece of trivia, akin to when white Protestants stopped being the majority of the U.S. population in 1993.² In fact, for six individual states this milestone has already been passed: white residents comprise less than 50% of the population of Hawaii, New Mexico, California, Texas, Nevada, and Maryland.³ But we do not live in an ideal world. A long and ugly history of systemic racism has produced deep inequities along racial lines that permeate every facet of American society, and I’ve seen disproportionate attention paid to this milestone for its symbolic importance.
On an interpersonal level, this milestone is meaningless. Residential segregation has been an essential component of the aforementioned ugly history of systemic racism,⁴ which you can see for yourself in nearly every U.S. City.
One’s neighborhood does not go from 60% white to 59.5% white from one year to the next in perfect accordance with national statistics. It is easy to imagine Americans continuing to see little-to-no diversity in their neighborhoods on a day-to-day basis even while this demographic shift chugs along.
So how has this transition played out on the level of lived experience? How does the increasing racial diversity of the nation play out at the level of neighborhoods and blocks? To answer this and more, I pulled some Census data and got cooking.
The U.S. Census has been held every 10 years since 1790. It’s right there in the Constitution itself, intertwined with the infamous 3/5 compromise.⁵ Originally a count of individuals, the Census has expanded a bit since then⁶ and now features demographic and housing data along with an annual “American Community Survey” (ACS) that fills in the gaps between the decades. The ACS is cheaper, less thorough, and less precise, but still good enough to give us an idea of how things are trending.
The Census Bureau organizes data along dozens of different geographies, such as states, counties, metropolitan areas, etc. But there is no official Census geography for “neighborhood”. The closest we have is the Census tract, a subdivision of counties smaller than a zip code.⁷ Below you will find a tract map of roughly the same cross-section of St. Louis pictured above. If you have trouble aligning the two, note the large highlighted rectangle on the left in both (Forest Park, site of the 1904 World’s Fair).
Next on down we have block groups, smaller subdivisions of tracts:
Last, we have blocks, the smallest unit of national geography in the Census. Below is a zoom-in on the above maps (note Forest Park on the left and I-64 at the bottom) to show the granularity of Census blocks.
So let’s first look at the number of all U.S. census tracts by percentage of white residents in 2019:
The far left of this graph shows the number of census tracts in 2019 with 0%–5% white residents, the far right of this graph shows the number with 95%-100% white residents, and the middle shows tracts with a 50%–50% split. You can clearly see segregation at work, with peaks near the right-most and left-most categories.
But how does this compare to historical data? Let’s see what this same graph looks like starting with Census data from 1990, the earliest Census with tracts covering the entire country:
The right-hand peak in 2019 looks utterly diminished compared to the one in 1990! There has been noticeable, consistent decline in the number of all-white or nearly all-white census tracts over the past 30 years, with more tracts becoming more diverse - note the middle of the chart’s increasing level of chonk.⁸
Tracts tend to have a few thousand people each and are fairly large, but they’re the best we have until the 2020 Census data gets published. The American Community Survey mentioned above, where the 2019 numbers came from, is not precise for block data. That being said, we can still look at decennial Census data up to 2010, which includes block groups and blocks.⁹
Here’s the above chart but with block groups instead of tracts:
The results look very similar, albeit showing even more segregation. The ends are spikier and the middle portion flatter, although we observe the same trend of declining “all-white” areas. Now let’s look at the chart with blocks, the smallest unit of geography in the Census:
Yowza! This implies fairly extreme racial segregation on the level of individual blocks. However, blocks are very small, some having as few as a dozen residents. An individual family being all the same race doesn’t indicate residential segregation the way an entire census tract does.
This speaks more generally to the problem of comparing on the basis of geographic unit: there are thousands of census tracts, but there are millions of census blocks. A better measurement of segregation would be how many people live in these geographic units as opposed to just counting the number of unit themselves.
Think of it this way: if you have thousands of different packages of cookies, and each package could contain between 12 cookies and 3,000 cookies, you can only learn so much by counting the packages. Let’s open up those packages and look at the total number of cookies themselves.
Here’s the chart that shows the number of white residents of census tracts in 2019, not the count of the number of tracts, ordered in the same manner as the graphs above:
By definition, census tracts with zero white residents have zero white residents, so we wouldn’t expect to see many in the left-most category. This graph tracks with residential segregation: as the percentage of white residents increases from left to right, the number of White people living in that proportion category increases. Think of it this way: if white Americans tended to live in 50% white census tracts, the graph would look more like a bell curve.
On the far right of the graph, though, we see something interesting. Once we get to the final proportion category, covering tracts that are 95–100% white, the number of white residents actually declines. In 2019, there were more people living in 90–95% white census tracts than in 95–100% white census tracts.
Has this always been true? Let’s find out:
It has not! In fact, a dramatic decline seems to have taken place over the past 30 years in this category. Let’s now zoom in on block groups and see if it matches:
While more white Americans in 2010 lived in block groups that were 95–100%, that number is only slightly greater than the number living in block groups that were 90–95%. And as with tracts, it’s declined substantially since 1990.
Finally, let’s see what this looks like at our smallest geography, the block:
The higher levels of segregation at smaller units of geography implied by the earlier graphs seems to hold up here. However, this trend does seem to have affected census blocks as well. For ease of comparison, here’s all of them in a single graphic with the same Y axis:
Segregation is higher the further back in time you go and the smaller your unit of geography, but the most extreme segregation appears to have diminished.
But this raises more questions: where exactly is this happening? Who is being affected by this? Why is this the case, and what implications does this have for American society going forward? Americans use the Census to shape our society, and as a result it’s been inseparable from questions of political power, taxation, and race from the beginning. But the data itself can only tell us so much: we have to be the ones to interpret this data and apply it. Because if we don’t, we yield the floor to morons who think that “majority-minority” means the end of the world.
Thank you for reading! You can access the code used to produce this article here. American Community Survey data acquired via RStudio and the “tidycensus” package. Citation for decennial census data is below:
Steven Manson, Jonathan Schroeder, David Van Riper, Tracy Kugler, and Steven Ruggles. IPUMS National Historical Geographic Information System: Version 15.0 [dataset]. Minneapolis, MN: IPUMS. 2020. http://doi.org/10.18128/D050.V15.0
 The definition of “white” has changed substantially over time: a century ago an Italian immigrant wouldn’t have been considered “white”. For the purposes of this article I am using the contemporary Census designation of non-Hispanic white.
 And white Christians stopped being the majority sometime around 2015–2016. There’s a book about it, The End of White Christian America.
 If any outright racists are reading this and genuinely frightened at the prospect of White Americans being less than 50% of the population, I would urge you to visit one or more of these states. They’re not such bad places, and I know because I live in one.
 The Color of Law is a fascinating (and enraging!) history of U.S. housing segregation practices.
 It’s a little gross that the official Congressional commentary on this site makes zero mention of the 3/5 compromise despite the fact that it’s right there. They’re perpetuating the same elision of slavery in 2021 that the Framers were in 1787 when they used the euphemism “other persons” to refer to slaves.
 The Census has its own surprisingly interesting history, and you can get started with this timeline.
 If you really want to geek on out census geographies, you can watch this webinar.
 That’s a technical term. Trust me.
 Blocks and block groups are almost completely reset every 10 years. While they’re still about the same size, the Census Bureau does not attempt to maintain consistent boundaries from decade to decade due to how impossible that would be (migration, movement, etc.). Because of this, the following analyses aren’t as reliable as tracts to the decennial map redraw.