Before the Census was mailed to every household in America, it was gathered by counters walking door-to-door who collected along the way an inadvertent historical record not just of who we were in 1880 or 1940, but who we lived next to.

Buried in that record is a fascinating picture of race and residential segregation both more detailed and more sweeping than how we've typically looked at it: The hand-drawn enumerator forms of a century ago show where black families lived right next to white ones, or where entire streets were lined with households of a single race. They show, in other words, where Americans were integrated not across entire cities, but across fence posts and brick walls. They show patterns of fine-grained racial segregation that are impossible to see in public Census data today where privacy concerns override precision.

"It’s sort of shocking how much you can know about these historical patterns," says John Parman, an economist at the College of William and Mary, "especially relative to these modern studies."

Parman and Ohio State economist Trevon Logan have cleverly leveraged this historic data in a new working paper that re-examines this period, when many of the housing patterns that persist today originally took root. Their results turned up some new findings about an era when historians have long recognized a rapid rise in racial segregation. Between 1880 and 1940, they found that segregation doubled nationwide, touching not just big Northern cities, but every kind of community.

During that time, the chances that a black household in America would have a non-black neighbor declined by more than 15 percentage points.

Back when the Census was manually collected door-to-door, household entries that appear in the historical record right after each other were next-door neighbors (when residents weren't home, Census counters typically left their entry empty to later complete).

This data doesn't tell us much about people who lived across the street from one another, or opposite each other on a back alley. But it gives an intimate picture of those neighbors with whom most of us come in constant contact, the immediate families on either side. Logan and Parman used this record — since digitized — for the entire national Census in 1880 and 1940 to measure patterns of segregation in a new way, based not on the racial makeup of communities, but on the prevalence of opposite-race neighbors within them.

Contrary to previous research, this data show that cities in the South remained the most segregated over time, even as segregation grew worse everywhere. Nearly all of the cities below, except for a few in the Northeast, were more segregated by 1940 than they had been in 1880:

This rise in segregation over time also happened just about everywhere in America: in rural communities and urban ones, in cities with small black populations and large ones, in places that saw a large influx of blacks during the Great Migration and in places that were barely impacted by that great demographic shift.

This implies, Logan and Parman argue, that rising segregation in America in the early 20th century can't simply be explained by the sources often given by history: by urbanization around industrial jobs, by the Great Migration of blacks to Northern cities, where they settled in urban ghettos, by white flight out of those cities in response.

"We’re seeing a similar rise in segregation in these rural areas that can’t be explained by those same mechanisms," Parman says. "Maybe there’s some more general underlying change in American society happening here."

He demurs on the question of what that change might be, what else might have caused a sweeping nationwide rise in segregation — shifting culture influences? housing policies? unseen economic factors? — even in rural towns that weren't feeders for the Great Migration. But it does seem clear that this earlier chapter, whatever else contributed to it, remains deeply relevant to how we live today. Parman and Logan have found that levels of county segregation as far back as 1880 strongly correlate with economic mobility rates a century later.

"There’s this extraordinary persistence of the effect of either segregation, or the underlying causes of segregation," Parman says.