Okay, so, I basically saw a thread on how cook PVI works, and I'm going to try to bring back something I experimented with a while ago. Basically, I'm going to take the cook PVIs directly and compare them to the generic congressional vote. I'm not going to go by individual districts leading to a more compact model. Rather I go by the amount of districts with a certain score and group them in a way where I should be able to get a rough idea of where the house sits.
The Cook PVI is weird. Basically whatever the score is, the actual margin is double, that's why this model failed previously. i didn't account for that properly. Anyway, this should have resolved it, and without further ado, here's another house model based on this methodology:
Okay, so let's go over how this differs. First of all, the number of house seats I get is almost exactly what I get in my other model at the moment. I updated the other too today and after adding a few more republican seats, I get 235-200 with that one. So right now, we're projected to get around 234-235 seats depending on what model you use. Pretty close. Definitely better than the first one I used. This is, btw, an improved version of my 2024 one.
At the same time, let's discuss the differences. The big one is the certainty of the win. Here's the thing. My other model is based on 2024 voting data. However, we've had a lot of gerrymandering since 2024, leading to the median district being R+3 vs R+1, this means we'd literally need a GCV of D+6 rather than D+2 to take the house. Currently, we're at D+8.6, which is good, but still, this level of gerrymandering is gonna hurt us into 2028 and beyond, since this level of gerrymandering means we need a strong margin to actually take the house. The republicans F-ed us on that one, and democrats havent really done enough to gerrymander their own states.
Anyway, this leads to a more Lean D rather than safe D rating for the house. The house probability only slightly exceeds senate probability under these conditions.
Another flaw with this design is how....not granular the results are. Im grouping districts by their rating here. If one of a certain rating goes, they all go. So if the GCV dropped below 8, we would suddenly be down to 225 house seats, whereas in my other model, it would literally require a 7 point drop to get down to 225 under current conditions. So....yeah, something is off between the two models here. Which is probably more trustworthy? Well, I trust cook more than I trust converting 2024 voting margins given how many districts have shifted due to gerrymandering. A lot of the results my other model spits out probably will be wrong in the real world. Like, I know I currently have my other model flipping AK-AL, but then when I see polling data for it, it's R+14. So that other model...isn't necessarily great.
Anyway, this just goes to show that if the democrats want to remain relevant into the future, they HAVE TO use their power in 2028 assuming they get a trifecta and RUN WITH IT. They HAVE to pass reform that attempts to rein in gerrymandering, guarantee voting rights, etc. We can win NOW, but this is a ludacriously liberal voting environment. I mean, texas and alaska are flipping blue in my senate forecast FFS. Even states like Kansas and South Carolina are SWING STATES. So yeah. Democrats better actually use their power to DO SOMETHING to ensure their future political relevance or they're gonna lose it because in a more neutral political environment, the house is gonna go 210-225R. They literally need to overperform somewhere in the ballpark of 4-6 points in the GCV JUST to take the house. It's nuts. Basically, they'd only hold the house in a 2006, 2018, or 2026 style environment here. Gerrymandering is the bane on democracy's existence, but at this point, we either need to get a blue wave big enough to end the practice, or we fight fire with fire. I say we uncap the house and expand it to around 1300 members, reducing each district size to 1/3 of its current residents. That will get rid of the ridiculous built in advantages the GOP seems to have. Either that or do like, proportional representation based on the percentage of how people vote for various candidates. Point is, they cant just keep the house like this. The GOP is attempting to rig it and erode democracy under the principle that they choose their voters before the voters choose them. It's a BS system that has to change.
Anyway, we'll see how these models pan out. I mean, I'm good with senate and presidential forecasts, but the house has always been something I find iffy. When I forecast the house at all, I generally add a forecast last minute because without adequate polling data, it's just so hard to get the methodology down. I'm literally trying to use metrics other than polling and the best I got is cook PVI ratings and previous voter data. Typically, prior voter data might be adequate, but when you DO got so much gerrymandering going on, yeah that other model is gonna be straight up wrong on stuff.
No comments:
Post a Comment