So, this was a clusterfudge. It's such a clusterfudge, I don't know how I can get to all of it in a single article. But trust me, I will be deep diving into this, both from a statistics perspective, and from my own personal perspective of what went wrong. This first part, I'm gonna do the autopsy from the statistical and forecasting perspective. Not all results are in, but I've seen enough to have an idea of what went on.
7 possible outcomes for the 2024 election
I didn't get to do this part of my election forecast because the polls kept changing on monday and on tuesday i was too tired to get around to posting it before the polls closed, but I did plan on doing an article discussing the 7 main scenarios for the election. This is a statistical exercise. As I always say, my model of elections is statistical, and carries some form of error with it. I was going to discuss various scenarios at very parts of the statistical bell curve that my prediction makes up. Posting this now might help shed light on how I see 2024 right now.
The median projection (50th percentile outcome)
D+1 projection (60th percentile)
This is the outcome I expected. I was cautiously optimistic that Harris would overperform and pull off a win. I thought the MSG rally really clinched it. I wanted to be optimistic for Harris, I really did, and I wanted to be optimistic for her chances. Anyway I knew I'd get 48 states give or take right, it was just a matter of which way the winds blew. I guessed wrong.D+4 projection (84th percentile)
This is the optimistic Harris outcome, akin to what happened. 7 state sweep, start threatening and applying pressure to Iowa and maybe Kansas (if accurate). I wasn't bold enough to predict this, but I considered this "the good" harris outcome.
D+8 projection (98th percentile)
This is the most optimistic outcome within the margin of error. Massive sweep, Iowa and Kansas taken, Florida is a tossup, and even Blue Texas and Ohio are on the table. This is basically the Ann Selzer special. It didn't happen. I never saw it as realistic, but it's technically within the realm of possibility...
Now, the republican outcomes:
R+1 projection (40th percentile)
Trump flips Michigan and Wisconsin, sweeps all 7 swing states. It's a little mild given the actual margins we saw, but it probably has the right electoral count.
R+4 projection (16th percentile)
Trump not just flips all 7 swing states, runs up the margins, and starts applying pressure on the blue states like Virginia, Minnesota, and New Hampshire. This is kind of accurate, given virginia, minnesota, and new hampshire were toss ups way longer than they had any business being. Also, the margins in red states like Ohio, Texas, Iowa, and Florida are reminiscient of this one. Still, it's just a little too aggressive. Still, if the 7, this is the closest one to what we got.
R+8 projection (2nd percentile)
The the Ann Selzer special for republicans. Apocalpytic red wave. At least we didn't get this one, not that it helps that much...
So yeah. As you can see, what we just witnessed wasn't really outside of the realm of possibility. The real outcome is between the R+1 and R+4 projection. Let's call it R+2.5.
That's...actually pretty accurate all things considered. On a bell curve this is approximately the 26th percentile outcome. So about 1/4 of possible scenarios are worse, but 3/4 are better. Keep in mind Harris would need 54th percentile or better to actually win. it was close in the polling, but on a scale of "is this 2016 or is this 2022?", it's 2016 again. This was literally 2016 again. 2016 could've been seen as roughly a R+3 style scenario, give or take. And yeah. Of all of the possible realities, this is what we got.
The current state of the race
As of right now, the race is here.
NV and AZ aren't called, but yeah. Probably the 226-312 Trump scenario.
Congratulations, simulation 61, you presumably won. My simulator did seem to predict the outcome. Both AZ and NV are 5 points ahead for Trump, and I'm not the AP, I've seen enough, they're likely going red.
What about the senate?
Right now, this is the state of the race.
So again, overperformance. Let's look at my prediction again.
YIKES. So basically they performed on par to overperforming. It's hard to put a line on how we're doing here. We're losing Nevada and Pennsylvania (although neither are called yet). But then we're overperforming in Arizona. We carried Michigan and Wisconsin. There's a lot of randomness in how things are performing. This is a weakness of both of my models. The main forecast focuses on national over/under performance, but then theres randomness where some states do better and others do worse. Still, if I were to use my election map and add an offset to the situation, we get a map most resembling what we have with an R+2 offset.
It doesn't fit perfectly. We kept michigan, where here it falls. We're losing nevada, but here it's still relatively safe. PA and AZ seem to be trending about right though.
Honestly, it's a mess. Still, let's focus on the real question: was polling actually off?
The nature of polling in 2024
While the polls were off by normal polling error, on the whole, polls did a good job. Those right wing polls who are constantly accused of bias nailed it again. The herders came close enough to getting it right, and "right leaning RCP" still underestimated the actual electoral margins. 538 got it wrong again. Race to the white house forget it. All of these liberals were huffing copium the whole time, and even I ended up huffing it too in the end.
In all fairness to me, I did promise to report the numbers as they were and I did. It's why I did one prediction and then a second prediction that was more personal. And if the error went the other way, i would've nailed it. But yeah, the official prediction got it more right than my unofficial one. Official prediction for the presidency was off by 2 states, which is my par. Both were tossups within a point. my unofficial got at least 3 and up to 4 wrong. So by changing the data to fit my biases again, i screwed up just like in 2020.
Say it with me guys, LISTEN TO THE POLLS POLLS RIGHT, BIASED PREDICTIONS BASED ON FAULTY ASSUMPTIONS ARE WRONG. LISTEN TO THE POLLS, THEY'RE NOT FRICKING WRONG.
And for all the people saying the polls were off...again. R+2-2.5 scenario wasnt exactly unrealistic. That's well within the margin of error. It's well within the 95% confidence interval. Heck it's technically just inside the 50% confidence interval (25th-75th percentile). Z score was around -0.625. To put things in stats terms, relatively accurate outcome. And my own personal outcome? Well, I made a bet based on my biases and i was wrong. It couldve went in my favor, it had a 50% chance of doing so. And my own projection wasn't outlandish. It was the 60th percentile outcome of all outcomes. So...yeah.
I know a lot of people huffing a lot more copium than I was on the democratic side. I kept it relatively grounded and realistic, understanding the odds and seeing so many election cycles by this point this wasn't my first rodeo, but yeah. Should've known better.
And yeah, to settle the "should I stop using RCP" debate...NO. They got it more right than anyone else. Again, dems were huffing copium and ignoring everything that didnt fit their narrative. Speaking of which...
From a nerd perspective, if there's any upside to trump winning, it's that hopefully we'll never have to hear from that arrogant windbag Allen Lichtman and his 13 keys ever again. Truly he did not have the power to wield the polls.
Conclusion on the election's statistics
So, this is a real basic one. This part I covered the statistics of the election and how I did.
And for people wondering about the house, there are too many races uncalled and i don't even wanna touch that with a 10 foot pole. All i'll say is if I apply a 2-2.5 offset to that, we're gonna lose 207-228 and I now expect the GOP to maintain control. That's how my forecast shifts both with a 2 point and a 2.5 point offset. As such, I expect to be wrong on 11 of the 50 races I predicted. But again, so many of them are still up for grabs I ain't touching that until everything is called.
But yeah. This was basically a butt whooping for us. Democrats underperformed nationally around 2-3 points give or take, although the margins vary by race. Again, despite the national trend thing being real, there also is an element of randomness there. Ya know? Some overperformed by 2, some overperformed by 5-6. It varies. Anyway. That's my statistical analysis of the situation based on the numbers, without me even going into my personal feelings of the election. I'll be getting to that by around part 3 or so. I have A LOT to say about that. And uh...a lot of it isn't gonna be nice, and we're gonna be seeing a return of 2016-2020 populist me here. Just a warning. Miss the old populist me who would rip the dems constantly and tell it like it is and be adversarial to them? Well I'm back, baby, and I'm not gonna hold back on this travesty of an election. Until next time.
No comments:
Post a Comment