“Pollsters risk simply weighting the result back to the Biden-Trump contest.”

Whit Ayres, the president of North Star Research, told TMD that any adjustments, whether based on class, party, or previous vote, that try to account for these voters are necessarily unproven. “There’s no question that a nonresponse bias, which is simply a fancy term for ‘some hardcore Trumpies won’t talk to us,’ was one of the reasons why polls underestimated the support for Trump, especially in 2016 but also in 2020,” he said, but “the fact is that you can’t adjust for people who refuse to participate and talk to you.”

Ayres also cautioned against weighing a respondent’s previous vote. “Who knows if the dynamics of a Biden-Trump race will be the same as the dynamics of a Harris-Trump race,” he told TMD. Pollsters risk “simply weighting results back to the Biden-Trump race when Biden isn’t even in the contest.”