Andrew Gelman — who, as I noted earlier in the week, helped design The Economist’s election-predictor but admitted it was misfiring before Tuesday even began — takes stock of what happened. Two charts in particular illustrate the basics of the 2020 election.
First, the polls undercounted Trump supporters, on average by 2.5 points (the dashed line is where a state would fall if the polls had predicted it perfectly):
But second, in the vast majority of states, Biden did better than Clinton had, on average by two points:
There’s a lot more, including a comparison of 2016 and 2020 polling errors, in Gelman’s post.