Wednesday, September 27, 2023

Responding to criticism of my 2024 political forecast

 So, I ended up posting my map on the yapms subreddit since they seem into that sort of thing and I got a criticism that I think is worth discussing. Basically they said that I can't just shift polls 6 points and have it be accurate because the shifts might manifest differently on the state level. And he's....totally right. Maybe some states like Arizona would shift less than say, Florida. It's possible there are localized effects that this prediction won't pick up.

HOWEVER, I do want to keep a few things in mind. First of all, yesterday's prediction was just a best attempt at approximating things RIGHT NOW. Right now, we have limited to no state wide data on polling. Most polling organizations aren't even polling states yet. So I'm deriving a scenario based on a single national number. And....all things considered, I don't think it's unrealistic. Basically, this IS roughly what a scenario with trump up 1.5% in the national polls would look like. And it seems not all that different than 2016's actual results, minus nevada shifting to the red column. But yeah, this is basically what I would expect to happen roughly.

Keep in mind I internally debated whether trends applied nationally or not, and whether states will respond the same way to a national shift in polling one direction or another or whether states should be treated separately. And I think the "trend" model is roughly accurate. When one state shifts, and the national polling shifts, the rest likely shift too. This is not to say this is uniform. We saw more significant than expected shifts from Pennsylvania, Wisconsin, and Michigan than we did in the rest of the country. So there are some localized effects there, but at the same time, this is really just an approximation based on limited data. I'm not gonna say "well the rust belt is gonna shift more than expected and the sun belt less", that might happen, but until we see polling numbers on the individual state level, I think looking at the previous election's national environment is probably the best approximation. As I said, I plan to make more serious predictions when we have more data in mid 2024. Right now, that's around 9 months away. So this is the best we're doing for now. 

And one last thing. Given the margins that AZ and GA went blue last time, ANY shift in national polling away from the democrats is going to flip those states. GA went blue by a margin of 0.2%. AZ went blue by 0.3%. At that point, you only need like 5% of the projected national shift to apply to those states to flip them. I just don't see AZ or GA going blue this election style as long as the national numbers look as they do. Maybe they won't go red by a massive 5-6% margin like I projected, but if the shift is minimized in those states, maybe by 2-3%? Electoral college result looks about the same. 

No comments:

Post a Comment