Thread by @gelliottmorris

#NEW: Polls are exhibiting concerning patterns of geographic bias and partisan non-response that have maybe even gotten a bit _worse_ since 2016. But it is not doomsday for them—or the models. In fact, quantifying uncertainty is more important than ever.

economist.com/united-states/…

Our models said that Biden had so large a lead, and across many key states, that his position would very likely persist even under even a large systematic polling error. With Biden heading for 290 or maybe even 306 electoral votes, that is... probably exactly what happened.

We have learned a few early lessons about our model.

First, we underrated the chance for error with polls, particularly re: turnout models. We'll use fatter tails next time.

Second, we might think about using a prior of geographic polling error as per 2016-2020.

Third, our fundamentals model was... really pretty good. Our national prior had Biden up 5 and winning closer to 300 electoral votes with 75-80% odds. It's a shame we didn't shrink toward it more! ;)

Ehhh.... I think we were "right" about polarization making the electorate less variable and definitely right about the way we specified our fundamentals. But we said a few weeks ago that our tails could have been fatter, closer to a 93% p(Biden)

https://twitter.com/mattyglesias/status/1324130092675727361

Anyway, back on track:

The election-night narrative about polls and forecasts being cataclysmically wrong has evidently turned out to be an overreaction. There are serious problems with the industry, but it's not as bad as your Twitter feed or elections make it seem.

/End that's it for the thread, except to note that @mattyglesias's tweet made me think to check in on the brier scores for our model v 538. here they are: