Thread dropping in a couple hours. D +2.8 today in our model, but one thing to note is that what our model does is *not* forecast. It takes the VBM/EV splits by precinct, takes the November margins, and projects that onto today's electorate with turnout per precinct.
This is to say that I expect our final precinct model to end up at D +2.5 or so, which will probably be a touch more favorable to Democrats than the county model. It's currently D +2.8, but I don't expect that to last because the outstanding ballots should be closer than 54-46.
Should I have included an artificial modifier in the precinct model like in the county model? IMO, no, because while we had days to debug and track the changes in the county model, we don't have that time with the precinct model and couldn't regularize for outstanding ballots.
So one of the things to accept is that the precinct model will be more volatile and likely converge to the county model as time goes on. I think it would eventually match the county model to within couple tenths of a percentage.
Unless, of course, I decide to add regularization to it, but I don't even know if that's worth it at this point given how few days are left in voting.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
