Nate Silver: “Chastened by their errors in 2016 and 2020, many pollsters are applying what until now had been a relatively uncommon technique: weighting their samples based on how people say they voted in the last election. Nate Cohn of the New York Times has a terrific explainer on this topic that I’d strongly recommend you read in detail.”
“The reason pollsters traditionally haven’t liked to do this is because people often misremember or misstate who they voted for, in particular being more likely to say that they voted for the previous winner — in this case, Biden.”
“This is complicated by the fact that Biden is now quite unpopular — so people may not be so eager to say they voted for him — and that many Republican voters mistakenly believe Trump won.”
“In theory, however, this could be one reason that polls underestimate Harris. If someone mistakenly says they voted for Biden in 2020 but had actually voted for Trump, and then says they plan to vote for Trump in 2024, it will look like there’s been a shift toward Trump when there actually hasn’t been.”
