With only about 45 days until the 2020 presidential election, we might as well offer our quadrennial explanation of what Pauline Kael actually said in 1972.

Apocryphally, Kael, a well-known film critic, expressed amazement at Richard Nixon’s massive win that year since she herself didn’t know anyone who supported the incumbent president. This anecdote is useful each time Americans are heading to vote because of its utility in expressing the extent to which elites — Kael was speaking in Manhattan! — fail to recognize the breadth of support for their political opponents. It was particularly useful in 2016, given that Donald Trump’s was decidedly not centered in the Manhattan elites and because, it seemed, he won as a result of an unrecognized groundswell of support.

Anyway, Kael didn’t actually say that. And, perhaps more important to the subject at hand, that blindness about the opposition works both ways.

Consider new data from Pew Research Center published on Friday. The pollsters asked Americans who they supported in the 2020 contest, Trump or former vice president Joe Biden, and then how many of their friends supported either the same candidate or the other.

Nearly 9-in-10 Trump and Biden supporters said that a lot or some of their friends supported the same candidate as they did. About three-quarters said that they knew at most a few people who supported the candidate that they didn’t — and about 40 percent of each group knew no one at all who supported the other candidate.

It is by no means uncommon for pollsters to hear from people who fit into the dark-purple sections of the above charts. One of the more common complaints when a poll is released (beyond “no one ever polls me”) is “this can’t be true; no one I know supports Trump/Biden/whatever thing.” We are a nation of Apocryphal Pauline Kaels.

But it is true that many Americans actually don’t know anyone who supports a political candidate they oppose. After the 2016 election, The Washington Post and its partners at the Schar School asked voters how many people in their social circles voted for the presidential candidate they themselves had opposed. Seven-in-10 supporters of both Trump and Hillary Clinton said that a lot of their friends voted for the same candidate. Three-in-10 Clinton voters and a quarter of Trump voters said they knew no one who’d supported the other candidate.

What was particularly interesting in the wake of the 2016 election was to look at how people actually voted. Precinct-level data showed that, on average, a quarter of voters in each state lived in a precinct where one candidate or the other won by 50 percentage points.

It varied widely from state to state; nearly no one in Hawaii lived in such a precinct while more than 3-in-5 voters in Louisiana did. But in most places, a substantial portion of the electorate lived in broadly homogenous political bubbles. It was more common in states Trump won, where an average of 32 percent of voters lived in precincts Trump won by at least 50 points. In states Clinton won, the average was 18 percent of voters.

A precinct is as refined a political geography as can be measured, meaning we’re missing pockets of voters in even smaller areas that might better demonstrate the Kaelesque findings from Pew’s new data. Of course, friendship isn’t predicated on geographic proximity these days, what with our “Internet” and so on. We can be friends with anyone we want, anywhere in the world, and we’re still mostly friends with people who share our politics.

Why? In part because of the tension partisanship introduces. More than 9-in-10 respondents in another Pew poll identified the conflict between the two parties as strong or very strong — more than any other demographic divide in American culture. It makes that 10-plus percent of Trump and Biden supporters in Pew’s new poll whose friends mostly support the other candidate all the more remarkable.

If you hear someone describing that phenomenon, knowing mostly only people who support the other candidate, you’ve got a much more hopeful anecdote to elevate every four years.