I'm a little late to the Catalist report on the demographic makeup of the electorate from earlier this week, but I wanted to note a few quick things
catalist.us/wh-national/
catalist.us/wh-national/
One observation: in the fight between AP/Votecast and the exit poll, the Catalist estimates are a fairly clear vote in favor of Votecast.
And when you put all the various sources together (adding CCES, CPS, live phone), the exit polls are an outlier on several indicators
And when you put all the various sources together (adding CCES, CPS, live phone), the exit polls are an outlier on several indicators
I'm not going to be comprehensive, but the exit polls continue to show--as they have for more than a decade--an erroneously diverse and well educated electorate, with countervailing (and erroneous) Dem weakness among white voters (esp with a degree) 
The Votecast figures aren't exactly like Catalist/CCES/CPS/my own estimates, but they're a lot closer and they come out on Election Day
Second observation: the Catalist figures show, by far, the most Democratic gains among white voters and largest Trump gains among Latinos of all of the various sources.
And I'm inclined to believe, based on the other available evidence, that they're right about it
And I'm inclined to believe, based on the other available evidence, that they're right about it
The Catalist figures show Trump making nearly 3x as many gains among Latino voters as either the CCES or exit polls showed, with Trump gaining *8* points of major party vote share.
That comes close to what you'd guess based on the precinct results, live phone data.
That comes close to what you'd guess based on the precinct results, live phone data.
Similarly, the Catalist figures show Biden making more gains among white voters than the CCES/exits. Here again, the magnitude of the swing is fairly similar to what the precinct results / live polls implied
Third, there are some pretty big differences--across all the estimates--on the attitudes of white voters by education and Latinos.
Honestly, I think a lot of this is in part because the polls were bad this cycle, and the varying techniques that get used to make up for it
Honestly, I think a lot of this is in part because the polls were bad this cycle, and the varying techniques that get used to make up for it
Variance in the composition of the electorate is another factor: if you have more Dem-friendly groups (like more well-educated white or nonwhite voters) then estimates for each group will need to be more Republican in order for everything to add up to the result
Anyway this brings closer to getting to whatever our 'best' answer about 2020 might be in the end. There are still some important things to nail down--magnitude of white educational polarization and shifts in turnout, in particular. More later
• • •
Missing some Tweet in this thread? You can try to
force a refresh
