Film critic Pauline Kael is often quoted as saying after the 1972 presidential election some variant of “How can [Nixon] have won? Nobody I know voted for him.” The quote is almost certainly apocryphal, paraphrased from a less insular comment.* But that sentiment is no doubt familiar to many Americans today. As politics has become more partisan in recent decades, it gets harder to talk to people across the political divide.

Our research on the 2016 election underscores how common this has become, with three-quarters of voters most often talking about politics only to people who shared their views.

Red feed, blue feed

More and more Americans live in partisan “bubbles,” reinforced by changed news and communications media. A 2014 Pew study found that just over a quarter of Facebook users (including 31 percent of consistent conservatives and 44 percent of consistent liberals) have muted or unfriended someone because of political disagreements. A study of Wisconsin voters also found that nearly a third of respondents said that they had stopped discussing politics with someone after disagreeing about Wisconsin Gov. Scott Walker’s 2012 recall election.

Meanwhile, Americans have become better sorted into politically like-minded networks by geography, occupation and lifestyles. Although recent research suggest that Americans are not choosing where to live because of politics, the resulting clustering naturally limits exposure to those of different political persuasions.

Such sorting into “red” and “blue” regions and communities means that, American politics increasingly feels “tribal.” Party competition routinely antagonizes ideological, cultural, and religious differences among factions, whose suspicion and dislike is exacerbated by ignorance about the other side’s motives.

If getting to know one another as people helps reduce stereotypes of and prejudice toward groups different from our own, then the political homogenization of U.S. society bodes poorly for deliberation and tolerance.

But how homogeneous are Americans’ political networks, really?

How we did our research

To find out how much “crosscutting” discussion — that is, conversation with those across the aisle — ordinary Americans had during the 2016 election season, we asked the 2016 Cooperative Congressional Election Study (CCES) respondents to name the three people with whom they most frequently discuss political matters.

Then, using a format used by Edward Laumann, and later, political scientists Bob Huckfeldt and John Sprague, we asked questions about these three people (called “discussants”): What is their preferred party, who is their preferred candidate, how frequently do the two discuss politics, and so forth.

These questions let us gauge how politically diverse voters’ social networks might be. Specifically, we categorized social networks (the three people with whom the respondent discusses politics) into five groups: composed of only Hillary Clinton voters; both Clinton and not sure/other voters; both Clinton and Donald Trump voters; both Trump and not sure/other voters; and only Trump voters.

So how politically diverse are Americans’ social networks now?

The figure below shows how common each social network type was, based on whether the respondent herself supports Clinton, Trump, or neither.

You can see pretty clearly how polarized voters’ social networks are. Seventy-five percent of Clinton voters do not have a single Trump supporter in their immediate network, and just the reverse is true for Trump voters. More than half of Trump and Clinton voters say they do not regularly discuss politics with someone planning to vote for the other main candidate — or even with someone who wasn’t sure or planned to vote third party.

Only about one-fifth of Clinton and Trump voters had truly mixed close social networks — that is, they regularly discussed politics with both Clinton and Trump supporters. And it’s not much different for voters who were undecided or planned to vote for a third-party candidate; only about 30 percent of those had their most regular conversations about politics with both Clinton and Trump supporters.

Of course, it’s important to note that asking people who they talk politics with might lead to some measurement error. People’s memories might be faulty, or they might more readily recall people with whom they agree. Nonetheless, the approach we use has been scrutinized in previous research and shown to accurately represent people’s communication networks.

Is this any different based on where someone lives?

We were also curious how the patterns we found would vary across different kinds of counties: “red” (Trump received more than 60 percent of the vote), “blue” (Clinton received more than 60 percent of the vote), and “purple” (neither received more than 60 percent of vote).

In the figure below, you can see that this does affect voters’ political discussion networks. Clinton voters in red counties and Trump voters in blue and purple counties are indeed talking politics across the aisle more often. But regional sorting means that there are fewer such counties than there were in the past.

In just 15 years, Americans’ political discussion networks have become even more closed off 

It’s hardly new that birds of a political feather flock together. But as recently as the 2000 presidential election, political scientists Robert Huckfeldt, Paul Johnson, and John Sprague found that about 65 percent of Republicans and Democrats had homogeneous discussion networks — meaning, none of the people with whom they most often discussed politics were voting for the opposing presidential candidate. We find a 10-point increase in such homogeneity since then.

* Here’s the actual Pauline Kael quote: “I live in a rather special world. I only know one person who voted for Nixon. Where they are I don’t know. They’re outside my ken. But sometimes when I’m in a theater, I can feel them.”

Ross Butters is a PhD student in political science at the University of California at Davis, where he studies the impact of social factors on political attitudes and behavior.

Christopher Hare is an assistant professor of political science at the University of California at Davis, where he studies ideology and voting behavior in the mass public, campaign strategy, and political polarization. Find him on Twitter @PeauxliSci.