In the Unlikely Event

5/27: Updated version at the Substack

It's something from the Other Nate (Cohn, at The Times) that I'm taking a personal interest in, because I've been asking somebody with the resources to do it for a long time: going in search of the unlikely voter. Except he doesn't know that's what he's doing.

That is, when he was looking into the factors that might be associated with voter preference for Biden, he found that people who were otherwise likely to vote for Biden—self-identified Democrats, people who voted for Biden in 2020, minority members, the relatively young—it was those who didn't vote at all in the 2022 midterms, who were more likely to express a preference for Trump or no preference at all:

Overall, self-identified Democrats who stayed home in the midterms back Mr. Biden by just 67-15, compared with Mr. Biden’s 93-3 edge among those who turned out last November. Similarly, Mr. Biden holds just a 79-6 lead among self-reported Biden ’20 voters who did not vote in the midterms, compared with a 91-3 edge among those who did turn out.

Perhaps surprisingly, the pattern even extends to college graduates. Mr. Biden leads by just 11 points among college graduates who skipped the midterms, compared with a 19-point lead among those who turned out. College-educated Democrats who skipped the midterms back Mr. Biden by just 70-9, while college-educated Democrats who turned out in the midterms back him, 98-0.

Which Cohn interprets—this is The Times, after all—in the Pitchbot paradigm of "bad news for Biden", as a significant place in the cross-tabs where Biden is in trouble; under the assumption, I guess, that the 2024 electorate will consist of the 2022 electorate plus this wavery group for the Democrats and the corresponding group for the Republicans, who are solidly for Trump, and therefore the 2024 electorate will be less Democratic overall than the 2022 one was.

These less engaged voters might just be the single biggest problem facing Mr. Biden in his pursuit of re-election, the Times/Siena data suggests. If there’s any good news for Mr. Biden, it’s that his challenge is concentrated among voters who still consider themselves Democrats — a group that, in theory, ought to be open to returning to the president’s side.

But the assumption is obviously dumb. You might as well assume that the registered voter poll is the same as the likely voter poll. What Cohn has actually done here is a way of approximating the "less engaged", to use his phrase, that is the less likely. They're less likely to be following the news, they're less clear on the differences between Trump and Biden, or between Republicans and Democrats, they're less likely to have made up their minds at all, and will remain so up until the last couple of weeks in October-November, and more likely not to show up in November at all, even though it's a presidential year.

And it hasn't occurred to Cohn to consider those who didn't show up in 2020, but Pew knows that too, and there are interesting partisan differences there:

The plurality of voters who always vote is equally divided between the parties; Republicans dominated the group who didn't vote at all from 2018 to 2022, Democrats the group that only voted in direct reaction to Trump, in 2018-20 or 2020 alone; in 2020 the stay-at-homes were overwhelmingly Republican, and that's a more important number than the Republican stay-at-homes of 2022 (though these defeated the Nates' predictions of the Red Wave).

I'm not saying this is a Good Sign for Biden's chances either, though it might well be, especially in conjunction with the well-known defects of the Times-Siena operation this year, in particular in the representation of Latinos and youth. 

What it is, and this is what has me kind of excited, is an occasion for a testable hypothesis on what gets the unlikely voters going. Is it inflation ("It's teh stupid, economy!"), as the legacy press seems determined it should be? It certainly wasn't in 2022, the year of the missing Red Wave, when inflation was much worse than it is now. Is it Trump fear and anxiety over abortion? That would be the overall lesson from 2018-22 (Democrats doing less brilliantly in 2022, when Trump was personally off the ballot).

I'm saying it's a Good Sign for political science that it should get busy studying unlikely voters, I'm saying, as I was saying in November 2022, that

there's something defective about the pollsters' traditional collection of data for adults, registered voters, and likely voters—they really need to be collecting data for unlikely voters, because they're the ones who make the difference.

This is just another way of putting a claim I've made before. What I normally say is that the normal pollsters' picture, where the electorate is made up of partisan voters who know what they want and swing voters who have trouble making up their minds whom to vote for, is wrong; the unpredictable part is really the people who have trouble making up their minds whether to vote at all.

Cross-posted at The Rectification of Names.