US House Reset: #ScrimshawSix And A Model Launch
I Built A Thing
Do I have the utmost confidence in what I’m about to write? Not really, because my spotty track record predicting US elections should have me not particularly confident, but back to the well I go – I have a US House model, and it’s kind of cool, I think.
US House models require two different things to be even remotely close to the right answer – a good model of how all the districts will vote in relation to the House popular vote, and then, a good sense of what the House popular vote will be. In 2018, everyone got the House results broadly right because everyone got the House Popular vote right, and everyone’s House projections were wrong in 2020 because the polls were wrong. In aggregate, the House vote is more important, and in specific seats, getting them to look right is harder.
One of the things that fucked my Virginia coverage was the fact that Biden’s approval was tanking but the Generic Ballot was still mildly Democratic, so I believed the disapproval numbers were just a bunch of angry Democrats who were pissed, but would vote Democratic in 2021 anyways. Turns out, I’m an idiot. That said, trying to discern signal from the approval polling, and figure out how to handle it for November, is really hard, so I’ve been racking my brains for an answer, and I got one. Am I sure it’ll work? No, but it’s gonna be what I use this year, so might as well put it out there. That’s what this site is, at its core.
The seat by seat projection is just a basic trends model, like I used in Canada last year to great success. The more a seat trended left from 2016 --> 2020, the less it’ll swing in 2022, for every point of national swing, and the more seats that trended right in that time will. This is a fairly basic concept that suggests that South Texas will swing more than Southlake, given the GOP’s willingness to appeal to culturally conservative voters than rich, well-off white social liberals. You could even say it’s a way to account for a Global Fucking Realignment, as it were. Incumbency’s not worth much, which helps Democrats in California, Staten Island, and Cincinnati, but hurts them elsewhere. In terms of the seat by seat stuff, this is fairly basic, at least for now.
What’s a little bit intricate is my Generic Ballot calculator, which uses no Generic Ballot polling, only approval. It’s Joe Biden’s White approval numbers, plus Biden’s 2020 Margin with Black voters and Asian voters, and a healthy, but not-quite 2020 lead with Hispanics. Am I positive it will work? No, but in my view, it’s a better system than just uncritically trusting Generic Ballot polling where the GOP win 27% of the Black vote, or unskewing some polls and not others. This, in my view, gives me the best chance of not falling for shitty polling or doing things in a piecemeal, biased way.
So, without further nonsense, let’s go – six questions about the Scrimshaw House Model
What Does It Project?
Right now, 224 Republicans, 211 Democrats, on a House Popular Vote of R+3.4%.
To be clear, this isn’t an attempt at a forecast of where this model will be in November – this is an attempt at where things are right now, and for those curious, the House Popular Vote was R+5.7 last week. A couple of the tracking polls had more favourable Biden approval numbers with whites this week, so it’s a little higher. It’s gonna be bouncy, but this is the range right now - the GOP have a meaningful advantage, they will probably win the House, and there is at least as good a shot of Democrats doing even worse than this in November as some miracle recovery.
What Are Some Of The Notable Seats That The GOP Flip?
I was wrong about Colorado 8th, that one is GOP right now by about 2%, Elaine Luria is down a point in Virginia, Henry Cuellar is down 2% in Texas 28th (and Cisneros would lose by 12% if she beats him on primary day, which she won’t), the GOP take Conor Lamb’s seat in large part because Lamb was (is?) too much of an idiot to abandon a no-hope Senate bid, and the GOP get an 8-5 in Michigan and a 6-3 in Arizona this year.
Yes, Cuellar will lose, because the thing about a trends model which will help Democrats in suburban Minneapolis and suburban Seattle is that it means they’ll get fucked in South Texas – that’s the necessary component of trends, and a lesson I really didn’t learn in 2020.
Any Good News For Democrats?
I have them beating Steve Chabot in Cincinnati’s OH-01 by a quarter of a percent, Max Rose gets back into Congress if the New York map stays the same by three quarters of one, I have two Californias flipping for them, and they hold a lot of their suburban turf – Minnesota 2nd, Washington 8th, Virginia 7th – but all three of those are held on narrow margins and could easily flip if the environment flips at all.
If Democrats Get More Popular, Can They Win The House?
I have them losing the current tipping point seat by 3.68%, so it’s not particularly likely. Do they have a chance? Sure, but it would require them winning the court case in New York and then either praying they get the generic ballot around a tie again or some court case falls out of the sky to save them – particularly, either the GOP courts in Florida draw Democrats 12+ safe seats or the Ohio Supreme Court finds some way to impose a 9-6 map on the state. I doubt either of those courts will do such things, and while I suspect the New York courts will save their map, that’s not a guarantee.
Any Interesting Upset Predictions?
For now, no, and that’s by design.
In the next two months, France, Australia, and Ontario will go to elections, and I will be covering all three of these campaigns, and so I’ll get more into campaign money and things once we get into the summer and fall. At this point, I know some Democrats in seats they “should” win will lose, and some Democrats who have no business holding on will win again. I just don’t have any opinion in April who those will be, and right now, I don’t really care to find out.
Chance Democrats Hold The House?
Gut says 10%, head says 20%, so, split the difference at 15%.