There are some valid differences:
--538 shifted from & #39;polls only& #39; to & #39;polls plus& #39; as default. polls plus in & #39;16 was clinton 73.5% on 8/26
--polling and priors now support a Trump E.C. edge, not yet true in & #39;16
--Biden penalized post DNC https://twitter.com/gelliottmorris/status/1298797791523868673?s=20">https://twitter.com/gelliottm...
So, for instance, the 538 polls plus estimate of PA/MI/FL on 8/26 in & #39;16, v. today. Are they very different? No, not really.
There& #39;s a second question: how much uncertainty is there today v. 16? Here, FiveThirtyEight doesn& #39;t see much of a difference. TLDR: uncertain fundamentals cancels out fewer undecided voters.
Is that right? IDK, and tbh they don& #39;t either (one might say that& #39;s the uncertainty)
I think the strongest substantive critique of the FiveThirtyEight model is that it& #39;s leaving info on the table by omitting Trump approval. If you don& #39;t know what the fundamentals are--and I don& #39;t think we do, empirically--the approval at -10 might help sort it out.
That said, they didn& #39;t use approval in & #39;16 and presidential approval is not exactly the same as the & #39;fundamentals,& #39; either. It& #39;s so correlated with vote choice that you& #39;re getting close to modeling off of the dependent variable here
You can follow @Nate_Cohn.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: