“So I think you have a story that would start with the fact that maybe pollsters were not prepared for this early voting surge. You have likely voters in polls. That’s based on some combination of their vote history and responses. Well, some of those people won’t vote. Their car won’t start on Election Day, or they will have a Covid outbreak in their area. However, if you’ve actually already voted, then you’re 100 percent likely to vote. So it may be that Democrats weren’t given enough extra credit for early voting. It’s also worth thinking about incentives here.”
“Imagine you’re a pollster and you have a choice between two turnout models. One is a newfangled turnout model that accounts for early voting. The other is a more traditional, conservative model. One of them has Biden up 6 points in Wisconsin. And one has him up 10 points. There’s not much incentive to publish the 10-point lead. If Biden wins by 10 when you had him up by 6, people will say it was a pretty good poll nevertheless. But if Trump wins, you’re going to look that much worse. So I think there are a lot of incentives to be sure that you’re not missing the white working-class voters that may not apply to Hispanic voters in Arizona or to younger voters who have not been reliable voters in the past but are evidently turning out this year” – FiveThirtyEight’s Nate Silver as interviewed by Ezra Klien at Vox.