One of my last pieces with DFP is up- an analysis of polling from the last year (notably including those 15 or so senate polls right before the election)

Tldr DFP did okay, but polling keeps getting harder, and you have to correct for geographic bias now https://www.dataforprogress.org/memos/2020-polling-retrospective
Imo this is the most important graph - we're missing rural Republicans, even with weighting+sampling that accounts for partisanship+rurality
It used to be that you could replace voters who don't take surveys with voters in the same demographic group who *do* take surveys, but that's increasingly not true, as response rates correlate with the things we're trying to measure
Shoutout @ColinJMcAuliffe @coolhannes @lgsanchezconde @jasonkatzbrown and the twitter-less Annie Wang for this work
You can follow @charlotteeffect.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: