Thread by @gelliottmorris

Here are my estimates of Joe Biden's net approval rating in each state *among all adults*.

Note the wide margins of error, tho this is roughly what I'd expect based on uniform swing from a +4 election to +10-15 national net approval.

Controlling for 2020 and 2016 vote here. 1/

2/ Note the model is much closer to expectations when making estimates for the voting population, but still produces some odd numbers in GA and FL. Regularization in the south among col-educated whites is often an issue here, so again, emphasizing the MOE.

3/ These state-level estimates from national polls are good when we don't have state-level polls but aren't perfect. The point is to try to get a better picture of attitudes at the geographic level since we don't have national political institutions.

https://twitter.com/gelliottmorris/status/1367252613419446274

4/ One note is that, in my experience, MRP models of polls that weight by past vote still have trouble figuring out attitudes among non-voters — a big source of _potential_ bias here. Our hope is that the educated guess will still be useful for readers/policymakers. More to come.

5/5 Since Congress is supposed to represent "all persons" in the US, I think it's important we look at opinion among both all adults and the estimated electorate. The different data answer two different questions about politics, neither necessarily more important than the other.