An Outlier Poll on Trump vs. Biden That Still Informs

To the extent the usual rule of thumb would mean dismissing the poll result and returning to an assumption that Mr. Trump can’t win, the usual guidance might be counterproductive.

Before I go on, I should acknowledge that I do have a few gripes with this survey. It reported the results among all adults, not registered or likely voters. The question about the presidential matchup explicitly offered respondents the option to say they’re still undecided, which could tend to disadvantage the candidate with less enthusiastic support. For good measure, the matchup was buried 16th in the questionnaire, following other questions about the debt ceiling, abortion, the presidential primary, the allegations against Mr. Trump and so on.

But my various gripes probably don’t “explain” Mr. Trump’s strength. The poll actually did report a result among registered voters and still found Mr. Trump and Mr. DeSantis ahead by six. And just a few months ago, an entirely different ABC/Post survey asked about the presidential matchup among registered voters in the typical way, without offering undecided as an option. What did they find? Mr. Trump still led by three points among registered voters. Similarly, they found him leading by two last September.

Interestingly, the January and September surveys showed all of the same peculiar results by subgroup — Mr. Trump’s lead among young voters (18 to 39), and the staggering Democratic weakness among nonwhite voters. And while this was not included in the most recent poll, Mr. Trump led among voters making less than $50,000 per year, historically a Democratic voting group. No other high-quality survey has consistently shown Mr. Biden performing so poorly, especially among young voters.

All of this means that the ABC/Post poll isn’t quite like the usual outlier. This consistent pattern requires more than just statistical noise and random sampling. Something else is at play, whether that’s something about the ABC/Post methodology, the underlying bias in telephone response patterns nowadays, or some combination of the above. It should be noted that the ABC/Post poll is nearly the last of the traditional, live-interview, random-digit-dialing telephone surveys that dominated public polling for much of the last half-century. So it’s easy to understand why it could produce different results, even if it’s not obvious why it produces them.