Yes but... the errors are correlated. There was a lot of analysis like this about Clinton, too, in 2016 (not saying the setup is the same) and.. Yeah, the errors are correlated. If PA moves to Trump, so will many other things and it is no longer narrow.
-
This Tweet is unavailable.
-
-
Well, maybe. It depends on what error it is. In 2016, the general contributing error was oversampling of college educated voters in white voter subsamples. Education weighting in 2020 aims to alleviate that. It could happen similarly but also could be an error with no covariance.
2 replies 0 retweets 0 likes -
But the same thing goes the other way too. If Biden is winning Texas then he's probably blowing everything else out in other areas too. Which is why Nate's team applies very large uncertainty bands in his 2020 model compared to 2016.
1 reply 0 retweets 0 likes
Correct. And those uncertainty bands should have been there in 2016, too. All the sources of variation and errors (polling issues; pandemic-specific stuff etc.) are likely correlated too. For example if there is a "shy Trump voter" (no idea!), it will likely affect many states.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.