The American Association for Public Opinion Research held its annual conference over the weekend. And they are very nervous about the return of candidate Trump.
Pollsters had a lousy year in 2016, and an even lousier year in 2020. They did well in 2018 and great in 2022. The pattern here is clear: When Donald Trump is on the ballot, the numbers go haywire. It's even reasonably clear why that happens. The former president tends to attract people to the polls (both pro- and anti-Trump) who don't often vote. Further, and even more damaging, is that the distrust-of-the-media stuff that Trump both harnessed and fed into has caused his supporters to decline to participate in polling, or sometimes even to give deliberately false responses.
Various outlets are trying all sorts of things to try to get a proper sample of the electorate. CNN, for example, abandoned random dialing, and shifted over to random... addressing (?). The pick addresses at random, and send a polling request to those folks via USPS. People who open the envelope can send their responses back, postage-paid, or call a phone number. This worked great in 2022, though how it solves the skeptical-Trump-voter problem, we do not know.
As to the deliberately false responses, some readers may recall a story we've recounted once or twice. The first time (V) and (Z) ever communicated was about 10 years ago, when (V) wrote an item about online polling, and how those pollsters were desperate to find certain kinds of voters who tend not to be online. As an experiment, (Z) attempted to persuade YouGov that he was actually an octogenarian Black woman living in Denver. There were a few obvious screener questions (e.g., "Do you prefer the music of The Beatles or of Aretha Franklin?") but they were pretty easily identified. Anyhow, (Z) wrote to (V) about his successful masquerade, and so it began. The pollster Global Strategy Group is trying something like this, using voters' 2020 vote as a variable, but again, if Trump voters are willing to lie, this screen is an obvious one to lie about.
It's not all hopeless, of course. The pollsters are much more aware of the problem than they were in 2020, and certainly than they were in 2016, and are working hard on it. Further, while Trump wasn't on the ballot in 2022, Trumpism was, and so the polling successes of that election are not entirely irrelevant to 2024. Still, when considering polls over the next 18 months, "we must be cautious," to quote Obi-Wan Kenobi. Political polling is what he was talking about, right? (Z)