Okay, so here’s Dave Wasserman’s estimate of the final House result for Tuesday:
Pretty grim stuff, no?
The Cook Report (where Wasserman calls home) has spent the entire election season breathlessly telling us that this or that 10 races have been moved from Tossup to Lean R or from Solid R to Likely R or whatever. Surely they must be getting a constant influx of new data, right? Well, not so much.
Here’s a question for you. Including even BS pollsters like Trafalgar (who, side note, has just farted out a poll giving Vance a 10 point lead in Ohio. I wouldn’t be shocked to lose the Senate race in OH, but if it’s by more than 5, I’d eat my hat), how many Congressional district level polls do you think have been taken in the last 3 weeks?
Answer: 52
52 polls of varying quality, bias, sample size and political affiliation, to model 435 seats.
Next question: how many high quality polls do you suppose have been taken in the last 3 weeks? That is, by a recognized, independent pollster working for a major media outlet.
Answer: 6, all by Siena.
Finally, using DK’s own excellent dataset, we can look at presidential results in the new congressional districts. Assuming a modest incumbency advantage (5 points is typical), we can even estimate the district lean in a neutral national environment. I’ve done that above.
If you compare those estimates to the average district level polls, how closely do you suppose they compare? That is, do you suppose the polls predict a very pessimistic or optimistic environment.
Answer: The polls average 1 point more R than a neutral environment, but with huge scatter.
District level polls give you almost no information because they are so rare and what little information they do give you is consistent with an R+1 environment (more or less what we get from national polls). And even more than that. With the exception of NV 01 (which Siena gives as a tie) all of the nonpartisan, high quality polls have the Dems significantly outperforming the naive “even national environment” model.
But supposing the average in the larger pool (or the GCB) is correction and it’s an R+1 environment. What does that look like? Well, Dems lose the majority, but with about 213 seats. The only way you get down to 201 is either through some very, very unlucky close races or by assuming (as apparently everyone does now) that after 2020 all swans are suddenly and permanently black and that even though polling averages and the highest quality and most recent polls and early voting suggest that we are at or near parity for popular vote that it’s going to be R+4 for … reasons.
Look, because of gerrymandering, it’s going to be tough to keep the House, but not impossible. It requires us to win the national popular vote by close to 2. But we’re closer to that than to the scenario that Cook is describing. They can take their made-up prognostications based on feelings and shove ‘em where the sun don’t shine.