Polls got the US election really wrong again.

Here is an early take from Tim Hartford (with reference to Andrew Gelman).

"The pre-election numbers suggest that in a typical poll with 500 positive responses, 255 went for Mr Biden and 243 for Mr Trump. But typical response rates are 5 in 100 — often lower, says Andrew Gelman, a statistician and prominent election modeller. He says that only around 1 in 100 people respond to opinion polls. So now picture 10,000 people, 255 who called their vote for Mr Biden, 243 for Mr Trump and 9,500 who never responded. How confident are we feeling now?"

I think the key issue is selective non-response. For some reason, Trump supporters are less likely to speak to pollsters. If some characteristic (say suspicion of the 'establishment') makes people more likely to vote for Trump and also makes them less likely to respond to surveys, then it is almost impossible to get polls right. Expect more dodgy polls in the future!