Kevin Collins: And credit where credit is due, @dhopkins1776.bsky.social made a very similar point in an article (linked to by Cohn) in FiveThirtyEight back in April abcnews.go.com/538/vote-bac...

This Nate Cohn article on Trump’s over-performance among less frequent voters is worth your time, and has at least two important implications folks should be considering that the article doesn’t directly address, one substantive and one methodological www.nytimes.com/2024/06/15/u...

If Everyone Voted, Would Biden Benefit? Not Anymore.www.nytimes.com Inside the unusual dynamic shaping the 2024 campaign.

And credit where credit is due, @dhopkins1776.bsky.social made a very similar point in an article (linked to by Cohn) in FiveThirtyEight back in April abcnews.go.com/538/vote-bac...

The less you vote, the more you back Trumpabcnews.go.com Joe Biden is polling better than Donald Trump among voters who turn out more often.

Implication 1: Voters may be sorting on their concern about, and dedication to democracy. After all, Trump (who questions the validity of elections he doesn’t win) does best among those who vote least regularly, and democracy-defending Biden is doing best among the most regular voters

Cohn doesn’t address this directly, probably because the Times’ surveys haven’t included a survey item that tries to measure voters’ dedication to democracy. (He does talk about issue prioritization of democracy and abortion among Dems, though that’s not quite the same thing)

So if you’re running a national voter-file sample or -matched survey, let me suggest that you ask people directly how much they value elections as a mechanism for selecting leaders. Something like below

There are other questions you could ask too, like measures of authoritarianism, but I think it would be particularly important to ask directly about commitment to the democratic institutions of elections if we’re trying to assess the attitudinal origins of this support split by past vote turnout

Implication #2: Sample from, or match to, the voter file. What this shows is that support is correlated with past turnout, and we know past turnout to generally be correlated with political interest (and interest in taking political surveys).

This is recipe for partisan non-response bias unless corrected using weighting. Fortunately, one of the things measured most accurately on the voter file is past turnout. The rich frame data that voter files can provide have always been useful for weighting, but it seems now particularly important

Post