Lately, pollsters and pundits have been nervously pondering the following question: “If Trump is behind in the polls, why do most voters say, in the same surveys, that he will win the upcoming election?” As Harry Enten recently noted at CNN, “An average of recent polls finds that a majority of voters (about 55%) believe that Trump will defeat Biden in the election. Trump’s edge on this question has remained fairly consistent over time.” This is far more than mere statistical curiosity by number nerds. Several peer-reviewed studies have shown that surveys of voter expectations are far more predictive of election outcomes than polls of voter intentions.
The polls that appear to portend a one-term presidency for Trump actually predict that the president will trounce Biden badly this November.
According to Professor Andreas Graefe of Ludwig-Maximilians-Universität München (LSU Munich), these citizen forecasts are “the most accurate method that we have to predict election outcomes”:
Across the last 100 days prior to the seven elections from 1988 to 2012, vote expectation surveys provided more accurate forecasts of election winners and vote shares than four established methods (vote intention polls, prediction markets, quantitative models, and expert judgment). Gains in accuracy were particularly large compared to polls. On average, the error of expectation-based vote-share forecasts was 51 percent lower than the error of polls published the same day.
There's more at the link.