I spent a lot of time thinking last week about why it felt like I was still wrong, even though my prediction was right. It felt like we couldn’t trust the polls, which would have meant my whole argument was wrong-headed. As it turns out, the national polls for the presidential race were no worse than usual, but it did seem like the state-level polls were worse than usual. Most analysts seem to have a theory that people who respond to polls nowadays are such a small percentage of voters that they make it very hard to get a representative sample. But I think it probably had just as much to do with the fact that Trump’s campaign went door-to-door, and Biden’s didn’t. I think it’s likely that Trump had a higher turnout of infrequent voters in the swing states, and that made the polls less accurate in those states.
But from another perspective I really was right, because I was arguing that almost all the voters had already made up their minds in February, and there was nothing that Trump could do to catch up. That turned out to be true. Biden’s lead fluctuated between six and ten points all year long, and that amount of fluctuation was mainly explained by a small number of uncommitted voters who could still be influenced by the daily headlines about the pandemic and the BLM protests. The people represented in the February polls never changed their minds either way, and that turned out to be at least 90% of all the voters — enough that the election was already decided back then, even if it ended up being closer than the polls indicated.