So, I've had a few days to reflect on the lessons of my previous post reviewing the accuracy of my past data, and while I won't change my official methodology for now, keep in mind, even with a sample size of like 80, or even over 100 with the 2004-2012 data, it's not enough to draw any massive conclusions. However, this did give me some insight into how electoral probability actually seems to work, and how I should approach 2024 polling wise.
I think the idea that any state up to 8 points is in play technically is a good one. My worst error was Wisconsin in 2016, I had Hillary at 6.5% estimating a 5% chance of flipping and it flipped. I also think it should be noted that Wisconsin and Michigan did have wild swings in the 6-8 point range in 2020 that could have made them flip if they werent literally hard blue for me. To be fair that's in part error on my part, remember my 2020 prediction was just off in general, but yeah, its worth considering.
Then we have 2016, where a lot of those states in the rust belt were off by up to 6-8 points. Just because WI/MI/PA were the only ones to flip didn't mean that states like Ohio, Iowa, and Minnesota didn't have similar fluctuations. They just happened to go the direction they were supposed to. My success and error rates were completely based on whether I guessed the correct outcome. In a lot of cases, I did predict the right outcome, but I was wildly off by the margins. I only recorded flips from the predicted outcome as being wrong.
Still, in the official predictions on that front, I did notice a weird trend. States up to 4 points or so, all seemed to perform roughly the same. I had this roughly 70% probability no matter what I did. Now, normally, I'd expect like a 55% probability for "tilts", 65% or so with 1-2% leans, 70something with 2-3% leans, and around 80% with 3-4% leans, but yeah, it was actually all over the place and I did do more analysis on the leans that I didn't post. Basically I find that 0-2% predictions are typically all about 70% accurate. Then 2-3% leans will actually GO DOWN for some reason, and then I'd go back up to 3-4%. It's weird. Really weird. Normally, I view election probability as a bell curve, but based on my actual results, I get something kinda like a top hat, where stuff within 4% has a roughly 30% chance of flipping, and stuff beyond 4% has a 4% chance of flipping so far. it's weird.
Part of this could be margin of error being a bit too generous, maybe reducing things to a 3 point MOE would make things tighter and more accurate, but again, I feel like if I did that I'd be wrong too much, given how there were tons of fluctuations in that range that just happened to not flip anything. So idk.
Either way, the fact that the overwhelming majority of flips are within 4 points does tell me something. It tells me that most people are reading the election wrong. For the most part, the media and most people online are framing the big swing states as the following: Wisconsin, Michigan, Pennsylvania, Arizona, Nevada, Georgia, and North Carolina.
I think this is wrong. Wisconsin, Michigan, and Pennsylvania, yes, but again, the other four have been consistently above 4 points. They're not really in play. They're kinda lost causes for the democrats. I think people include them since Biden won 3/4 of them in 2020, but again, if polling is accurate, public opinion has shifted roughly 6 points to the right since then. Wisconsin, Michigan, and Pennsylvania seem to be in that zone where things can flip, but the other four arent.
On the flip side, there are a lot of states in play that no one are really talking about. We're starting to see discussions now on them, but there's a lot of skeptics saying "no way will they flip" because they typically went Biden +6.5-9.5 last time. BUT...if public opinion has shifted 6 points since then, they're now in that 0.5-3.5% range, and that's what polling confirms. Of course, we're talking about NE2/Omaha, Minnesota, Virginia, and Maine. TAKE THESE STATES SERIOUSLY. They are statistically as much in play as the rust belt. They literally have the same chance of flipping as the rust belt. if Trump overperforms the current polls THEY CAN AND WILL FLIP. Everyone is just assuming the polling is overestimating Trump and underestimating Biden. And that can be true. And if I had to predict which direction any purported error will go, I would be inclined to say Biden will overperform and the rust belt is where the election will be won or lost for him. I do think him eeking out a 270-268 win is very well possible if he overperforms like Trump did in 2016, or like how the dems did in 2022.
But at the same time, to be intellectually honest, i gotta view it the other way too. Because statistically, the opposite outcome is literally just as likely. We could see those other 4 battlegrounds flip as well. And any of them doing so means Biden pretty much isn't gonna win. He needs EVERYTHING. He needs to run the board here.
So yeah. I'll say this. Biden CAN win. I do think it's possible for him to win, and if I had to guess it going either way, I do think that the outcome where he DOES win is very well possible. BUT....I do think the idea that his chances are roughly 1 in 3 are accurate. I would favor trump, he simply has more paths to the white house here and the statistics favor him. And considering that, yeah, I would say that Biden should shift some attention to MN/VA/NE2/ME. I feel like these are being taken for granted like the rust belt was in 2016, and yeah, Biden CAN lose these to Trump.
No comments:
Post a Comment