The most precise and intensive study of residential segregation ever done delivers some surprising conclusions about its origins in the United States. Trevon Logan and John Parman (from Ohio State University and the College of William and Mary, respectively) find that between 1880 and 1940, the odds of a white person living next door to a black one (or vice versa) essentially fell by half, and did so uniformly across all regions:
(Logan & Parman analysis of census data)
Of course people knew about segregation before, but the uniformity here is interesting. The researchers find that it happened in areas where there were few African Americans and in areas where they were many. They find that it happened in places lots of African Americans were moving to and also in places lots of African Americans were fleeing. Last but by no means least, they find that it happened in big cities and in rural areas alike.
This suggests that a lot of research that has focused on the specific dynamics of specific kinds of places — restrictive covenants in certain towns, influxes related to the Great Migration — is a little bit misguided. Local circumstances were important, of course, but the trend was fully nationalized.
A small caveat: they were able to reach this conclusion by delving into the actual manuscript pages of each decennial census and counting, block by block, the race of each respondent. That's an amazingly cool piece of research, letting them see literally how many African Americans had a white next-door neighbor. But this isn't necessarily the last word in what's the best way to characterize the level of segregation in a community. It would be nice to see a few different cuts at how to construct the segregation index before we fully embrace the conclusion.