Reposting two twitter posts here that I think are interesting.
Ok, my *favorite* plots.
— Sixty-Six Wards (@sixtysixwards) November 4, 2020
The results can change either because preferences change or turnout changes. We can decompose the results into changes in each.
If we held turnout constant from 2016, and just changed the county-level preferences, Biden would be up 4.1% from 2016. pic.twitter.com/UfpkoUkZx7
Notice these add up to a +2.4D change from 2016. Trump won by 0.7pp, so that yields the 1.7pp eventual D win that nytimes is showing.
— Sixty-Six Wards (@sixtysixwards) November 4, 2020
If we held preferences constant and just changed turnout, Trump would be up 1.7% from 2016. pic.twitter.com/xtmCyDIeK4
— Sixty-Six Wards (@sixtysixwards) November 4, 2020
And so much of it looks like a student effect (though maybe not all).
— Sixty-Six Wards (@sixtysixwards) November 6, 2020
To be clear, this would just mean those votes were showing up elsewhere. pic.twitter.com/9YIjPzlSe5