Bubble, Bubble, Partisan Segregation Spells Trouble

A great new study sheds light on the relative social and spatial isolation of Republicans and Democrats

Sex is great. But have you tried data visualizations of partisan spatial segregation?

Emily Badger, Kevin Quealy and Josh Katz of the New York Times report on a new paper in Nature Human Behavior by Harvard’s Ryan Enos and Jacob Brown titled, “The measurement of partisan sorting for 180 million voters.” These gorgeous maps, courtesy of the Times, are built from their data.

The methods behind Enos and Brown’s paper represent a leap in the measurement of partisan segregation. As they put it:

Despite the association between segregation and important outcomes, and the claims of increasing partisan segregation, the measurement of segregation among partisans, as with the measurement of segregation for most social groups, is severely limited: researchers must usually rely on data aggregations that do not include the actual locations of individuals, and thus measurements are limited to summaries across large geographical areas, and the experience of individual exposure across groups is masked.

Which is to say, until now, the data we’ve used to measure the spatial separation of Democrats and Republicans sucked. But by “using advances in spatial data computation,” Brown and Enos have managed to “measure individual partisan segregation by calculating the local residential segregation of every registered voter in the United States, creating a spatially weighted measure for more than 180 million individuals.” Wow. That’s incredible.

Here’s what they do. For each of these 180 million voters, they identify their geographically nearest one thousand neighbors, draw on various sources of data to identify the party affiliation of all these folks, and then calculate the degree to which the target voter is exposed to or isolated from voters from the other party. This is a monumental task of data analysis. It’s likely that Brown and Enos know how isolated you are from opposite-party Americans (though I assume names aren’t associated with individuals and addresses in their dataset). That’d be fun to know, wouldn’t it? If you live in one of the cities depicted above (larger versions run through the Times article), you might be able to eyeball it.

Turns out we’re very spatially segregated by party. Surprise! And it’s not just an urban/suburban/rural thing. You can zoom in and see it all the way down at the neighborhood level.

Badger, et al report:

Drawn this way, about 25 million voters — urban Democrats especially — live in residential circles where at most only one in ten encounters is likely to be with someone from the opposite party. Democrats in parts of Columbus, Ohio, and Oklahoma City live this way. So do Republicans in the reddest parts of Birmingham, Ala., and Gillette, Wyo.

Even when Democrats and Republicans are more equally represented in the same ZIP code or census tract, Mr. Brown and Mr. Enos still find traces of segregation. That means the two groups don’t appear randomly jumbled together. A Republican in a more mixed Denver suburb is still more likely to live close to other Republicans than mere chance would suggest.

“If we get down to a very low level and we still see this sorting going on,” Mr. Enos said, “it probably means there’s something pretty fundamental going on here.”

This isn’t just a function of mobility. The level of partisan segregation is increasing partly due to the changing composition of party coalitions. Its people changing their minds as well as people changing their addresses. For example, as the “diploma divide” widens, suburbs with a high concentration of college-educated voters have become more Democratic. That’s the big story of the 2018 and 2020 elections. But that’s not because Republicans are moving out and Democrats are moving in. It’s just that the loyalties of rooted residents have shifted. And there’s a correlative shift in areas with relatively high concentrations of non-college whites.

The surprise in all this, to me, is that Enos and Brown’s individual-level approach uncovers some evidence that Americans are making residential decisions based on the perceived party loyalties of their prospective neighbors.

Residential partisan sorting seems to be driven primarily by the fact that party preferences and residential preferences are correlated thanks to an underlying factor, like personality, that accounts for them both. Republicans like single family homes with yards in not-so-diverse areas with lots of churches and ample parking; Democrats like denser walkable diverse neighborhoods close to urban amenities, etc.

Even in relatively compact geographic units, like Census tracts, that have both single-family homes and apartment buildings, you see some partisan segregation. Up until now, the microscope wasn’t powerful enough see this clearly. But the real surprise is that a bit of detectable segregation shows up even in neighborhoods with a single housing type.

Enos and Brown report:

Even at these small geographic levels, the difference between Democrats’ exposure to Republicans and that of Republicans to other Republicans (and vice versa) is 11 percentage points greater than we would expect if there was no partisan sorting within the geographic unit. This disparity indicates that, even after the Democrats and Republicans make the choice to live in similar neighbourhoods, they still live with noticeably different levels of partisan exposure

The Times crew adds some nice detail:

Across 98 percent of census tracts nationwide, Democrats and Republicans live with at least some segregation. That’s true even within suburban neighborhoods southwest of Kansas City, Mo., where the residents are almost all white and homeowners, and the houses are all single-family.

This is exactly the sort of thing that more broad-brush methods can’t pick up. That’s cool. But the fact that it’s happening is … not great. As Enos told the Times, “We know that with groups in general, when they’re separated, bad things happen… The question with political parties is whether those are sufficiently like [racial, ethnic or religious] groups that we should worry about that happening.” This is a theme tackled by Enos’s excellent 2017 book, The Space Between Us: Social Geography and Politics, which I highly recommend.

Overall, Democrats are more isolated/less exposed than Republicans. This is intuitive if you’ve already got a general grasp of the relationship between density and party vote share.

Start with the fact that there are many more Democrats than Republicans. If partisans were randomly distributed around the country, the average Republican would be more exposed to Democrats than vice versa. Of course, partisans are not in the least randomly distributed in space. Democrats like density a lot more than Republicans do. If you pick a random person in the very densest part of a big city, they’ll be a Democrat about eight times out of ten. And you won’t have to go far at all to find their nearest 1000 neighbors, because this is the densest part of the city. So, 800-ish of those 1000 people will be Democrats, too.

If you’re a Republican in the dense urban core, you are very exposed to Democrats. And if you’re a Republican at the density divide suburban/exurban 50/50 vote-share crossover point, then your 1000 nearest neighbors are going to be more than 50 percent Democratic. If that seems counter-intuitive to you, imagine that the notional density divide line runs through your kitchen. Now, let the line bisect a circle with a one-kilometer circumference centered on your refrigerator. There will be more people on the denser half of the circle because that’s the definition of “higher density.” If the percentage of Republicans on the less dense half is 51 percent and the percentage of Democrats on the denser half is 51 percent, more than half your 1000 nearest neighbors are going to be on the 51 percent Democratic side.

To the extent that partisan affiliation is affected by social influence, the fact that Republicans are more exposed to Democrats than Democrats are to Republicans at densities where party loyalties split 50/50 would seem to favor Democrats over the long run. And it may very well be that the blue shift in the suburbs over the last couple of cycles reflects this asymmetry in exposure.

Now, Republicans in sparsely populated rural areas are likely to be about as isolated from Democrats as urban Democrats are from Republicans. However, there aren’t that many of them. Most Republicans live in the suburbs and exurbs of big metros because most everybody lives in or around a big metro. That’s what it means to live in a highly urbanized country. Sure, there are some rural Democrats just as exposed to Republicans as urban Republicans are to Democrats, but there are way more urban Republicans. I’m no mathemagician, but 20 percent of a million is a bigger number than 20 percent of a thousand, right?

[Aside: it would be great for our country if the GOP would get its head around the fact that there are more Republicans in Los Angeles County (1,145,530 votes for Trump) than in Iowa (897,672 votes for Trump). Will no one think of the millions of urban Republicans disenfranchised by the Electoral College! (Actually, by the stability of the state-based winner-take-all elector-allocation equilibrium. But I’ve already digressed.)]

Anyway, it’s a pretty straightforward corollary of the nature of the relationship between population density and party vote share that Democrats are more isolated from Republicans than vice versa.

A fascinating 2019 study on “The Perception Gap” by Daniel Yudkin, Stephen Hawkins, and Tim Dixons shows that, on the whole, Democrats have a more accurate bead on Republican views than the reverse — which is not necessarily what you’d expect given greater Republican social/geographic exposure to Democrats.

But then there is the curious finding that more education increases Democratic misperception of Republicans but not Republican misperception of Democrats.

I think the story above about asymmetrical other-party exposure probably accounts for this. First, universities are highly Democratic settings, so going to college exposes Republicans to opposite-party views in a way that it doesn’t for Democrats. But, probably more importantly, higher education magnetizes you to cities. Members of the small urban Republican minority are very likely to have college degrees. But they’re surrounded by Democrats everywhere they go.

Generally, more educated people are more partisan, and partisanship makes you less accurate about and more hostile to the other party. It seems plausible that the relatively high social exposure of educated Republicans to Democrats would tend to offset the distortionary effects of strong partisanship. But the relatively intense isolation of highly educated Democrats wouldn’t check the distortion of partisanship. Sunstein’s law of group polarization suggests the relatively homogeneous social bubble of educated Democrats should amplify partisan misperception.

Still, Republican misperception is worse than Democratic misperception at every level of partisanship. That Republican misperception of Democrats doesn’t go up as educational attainment goes up doesn’t mean that educated Republicans who are highly socially exposed to Democrats have especially accurate views of Democrats. It just means that they share the same highly inaccurate views of Democrats that less educated, less socially exposed, but probably less partisan Republicans do.

Meanwhile, a preference for urban density is overdetermined by a liberal disposition and a postgraduate degree, so the best-educated Democrats are likely to be disproportionately concentrated in the densest parts of the biggest cities, leaving them very isolated from Republicans. This seems to be supported by the finding that better-educated Republicans have more opposite-party friends, while better-educated Democrats have fewer.

This just makes sense. But it’s interesting because, if you ask them, liberals say that it’s less important to them to live around others with the same ideological inclinations than conservatives do — at least according to this 2014 Pew Survey:

That also makes total sense, since one of the defining differences between conservatives and liberals is comfort with difference.

This is one reason why it’s probably best that the partisan asymmetry in spatio-social exposure runs in the direction it does. Despite greater exposure, Republicans don’t seem to understand Democrats as well as Democrats understand Republicans. Doesn’t it seem likely that if Republicans were less exposed to Democrats rather than vice versa, Republicans would have even less accurate views of Democrats? Given the fact that it has become standard for Republicans to treat Democrats per se as un-American, quasi-traitors whose votes should count less, if they should count at all, it seems like that would be pretty dangerous.

That partisan segregation is visible even in small geographic units and seems to be increasing is bad news. It contributes to the mutual antagonisms of polarization. And I feel pretty certain Trump has made it worse. I wish I could disagree with Harvard political theorist Nancy Rosenblum’s comments in the Times piece, but I can’t.

Interactions between neighbors have long been distinctly nonpartisan, [Rosenblum] said, grounded in values like reciprocity — I’ll lend you my leaf blower, you watch my kid.

But she fears that a more malignant kind of politics is seeping all the way down into neighborhoods: “The most interesting question to ask here is: How deep does it go? And the test for how deep it goes for me is: How do neighbors in neighborhoods behave during disasters?”

That’s when we normally see neighborly reciprocity really come through, she said. “If we look at Covid — and we consider Covid a national disaster — you see something change,” she said. “And this is really very discouraging. It could make you weep.”

My Republican-majority state has been particularly bad in its Covid response. This has definitely affected my feelings about proximity to Republicans. Indeed, it has done a lot to consolidate and harden my sense of Democratic partisan identity, which, before Trump, was very marginal and vague. For most of last year, I avoided one of several Hy-Vee grocery stores a short drive from my house because I encountered more than a few unmasked people there on a couple different occasions. I read their pointed masklessness (there are free surgical masks at the door) as “Republican” and I probably wasn’t wrong.

Despite the strongly liberal character of my little college town, Democrats here are much less isolated from Republicans than Democrats in big cities in blue states. This fact is widely and often vocally lamented. Many Democrats here feel that Republicans pose a serious risk to our lives and health. And it looks like this sense of partisan danger hasn’t been entirely unfounded.

But it’s been broad Republican acceptance of Trump’s lies about the election, more than Covid denialism, that has left me with a sense of possibly irreconcilable differences with Republicans and a clear and distinct desire to get further away from them. Its hard to feel like shared civic life is possible with Republican who have come to doubt the legitimacy of elections when Democrats win them and are increasingly antagonistic to democracy itself.

This is extremely unhealthy. I’m definitely not proud to feel this way. But I also believe that these sentiments are a perfectly rational response to the realities of Republican governance and civic behavior over the past several years. It’s not unreasonable to feel this way. Still, like most of us, I desperately want to find a way to bridge the partisan divide, reduce polarized contempt, and unify the American people. But I’m definitely not going to do my part by moving my family closer still to Republicans. On the contrary, we eagerly intend to move much, much further away.

I don’t think I’m representative, but I also don’t think I’m special. That’s why I suspect that partisan segregation is already worse than Enos and Brown’s data depict and will continue to get worse for… well, I have no idea how long. I hope to be surprised.


Leave a comment