,
6 tweets,
1 min read
Read on Twitter
Caution to analysts who use exit polls: many big shifts in this year's exit polls are due to methodological changes, not just actual changes among voters
As I've said a million times, the exit polls have long been biased towards having too many well-educated and young voters. This year, they've made changes to try and fix this.
As a result of the methodological change, the college educated share of the electorate in the exit poll is 10 pts lower than 2014. In reality, this was a better educated electorate, but the exit polls were just that biased by this measure before.
The overrepresentation of college educated voters in the pre-18 exit poll (including 16) basically skewed all of their demographic numbers, since the exits were weighted to match the result, even though the underlying demographics were off
So take 2016. They had too many college educated voters, but they had to match the reality of Clinton+2. So what happened? A severe underestimation of Clinton's strength among col+ voters to cancel out the overrepresentation of col+ voters
These biases should be significantly lower now. But if you compare this year's numbers to 2016, you see a huge surge in Dem support among col+ voters. This is a phantom shift. The old exits were underestimating Clinton in '16, and now they're not
