So, I've been predicting elections for a while. Since 2008. And in the overall result, I got most of them right, outside of 2016. I do tend to get a few states wrong in the details, but I'm often CLOSE, as we'll see. Now, I've shared much of this before, but now I got my fancy charts with my fancy maps, so I'm gonna be posting receipts on this for future reference.
2004
So, 2004 was the first election I took active interest in in my life. I wasn't old enough to vote, but it was the first election where I considered myself politically active and informed enough to have some sort of actual opinion. I may have had the intelligence of your average republican voter in that election cycle, which isn't amazing by current standards, but wasn't bad for a teenager.
Anyway, I didn't make a formal election prediction this race. I did believe Bush would win, but I wasn't informed enough to actually follow polling, I was mostly following vibes based on how my family, teachers, and peers voted. And everyone was conservative and believed democrats were evil, that was my echo chamber back then, so I was passionate about Bush winning too. And he did win, narrowly. Not as narrowly as 2000, but it was still, in retrospect, a very competitive election.
Anyway if I did make a prediction with current methodology, I would've gotten Wisconsin and Hawaii wrong. Wisconsin I would've understood, it's a fairly competitive rust belt state even today, and it came down to less than a point. Hawaii, idk why the polling was like that, I think hawaii and I think a D+30 state or something. Mustve been junk polling. Ah well, young me wouldn't have known better anyway.
2008
So, this actually is an election I had a prediction on. And no, this prediction isn't it. This is based on polling and my more modern methodology. Ironically, my actual prediction was more accurate. How do I know this? Because I was pissed I got north carolina wrong. What caused me to undertake this election prediction thing in the first place was a contest in my political science department in college for who could predict the election best. I spent hours pouring over the data, mostly going by polls, but occasionally making judgment calls based on past history of the states involved. I somehow guessed Indiana correct, but got North Carolina wrong. And I lost the contest to someone who guessed perfectly.
Anyway, in retrospect, going by past trends did help me over this prediction. I did get Indiana right somehow, idk how I did but I do remember I had I think 351-187 Obama that election? Yeah. Makes sense based on the math. Mightve gotten one of those weird districts wrong too. But they had no polling. Oh well.
Funny thing is, I did take stats and research methods around this time and I kind of knew that Obama was polling so well his win was basically statistically significant. I went by a 3 MOE back then, but yeah. Even with my newer, more lax MOE, Obama still had a 94.8% chance of winning. Kind of insane. I miss that at this point, even though I was dooming as hard at the time as I was when Biden was in the running this year. Biden actually had better odds. Let that say something.
2012
In retrospect, 2012 was kind of a boring year for election predictions. it was 2008 2.0 with the GOP standing no chance. And let me just say, I kind of understood why at the time. The GOP wasn't popular. They were advocating for trickle down economics in the middle of a recession and they ran the rich frick who liked firing people and whined about the 47% not paying taxes. So this was a lap for Obama.
I was also almost dead on in my prediction, only getting florida wrong in this methodology.
At the time, I had a different prediction, much like 2008. I was once again irritated about being off by 1 state. I got Florida right, but guessed Virginia wrong, thinking it would go red. It was a tossup, but still. I should've known better. That's when I stopped trying to make judgment calls and just let the data do the talking. Of course,analyzing those past predictions now, maybe my judgment calls were sometimes slightly more accurate. Idk. I seem to be making the same amount of mistakes generally, just different ones, following the data. Randomness be randomness.
2016
This is my infamous 2016 prediction. It's the first election I guessed live on this blog, and the first prediction where I started using something resembling the current prediction chart we all know and love today. Since starting blogging, I started beefing up my methods and by election day, I was getting relatively sophisticated in my predictions. At the time I did the charts manually, and they took up a lot of time. Switching to sheets has basically greatly improved the speed at which I can pop predictions out. And that has allowed me to get even more sophisticated...like adding maps directly into my predictions now.
But yeah. This one I was off. Then again, who wasn't? No one guessed MI/WI/PA would flip, at least no one reputable. As such, the only other state I got wrong was Nevada, and the red call seemed justified at the time based on polling. I mean, it was a tossup. I get tossups wrong. Meh.
Also, on 2016, I like to famously say I was closer than anyone else here. As I said, no one guessed the rust belt would flip like that. And my probability had clinton at a 56% chance. Silver was being praised for having 71% in his model. So if anything I would say i actually beat Nate Silver on this one.
2020
So yeah, 2020, I screwed up. Then again, once again, everyone did. I was way too bullish on Biden. This is because I culled the polling averages and removed certain R leaning pollsters who had junk polls from the averages. This ended up leading to me having a way too D happy forecast. And then I was crapping bricks on election night as North Carolina and Florida went red.
To be fair, Nate Silver messed up too. Same error I made tbqh. This is why I don't weight the polling averages any more. I feel like whenever I go against the data, I make mistakes. In retrospect with 2008 and 2012, maybe I wasnt as bad as I thought, but yeah, this once again confirmed that it's generally not a good idea to go against the data. Hence why this time I'm not weighting anything almost.
Anyway, I was so displeased with the accuracy of this prediction I went back and redid it, making a second retroactive prediction based on what I would have gotten had I just followed the data:
This prediction is better, at least in terms of margins. Far more accurate, and while I made the same number of overall errors, I did so on closer states. My original prediction I got GA right but FL and NC wrong. In my revised prediction, I got NC right, but then I was wrong on GA. Still, I think in terms of margins this is a much better prediction, and 62% is more accurate of Biden's actual chances given the final result than the 89% I got with my original prediction. I basically predicted another 2008 landslide, I got 2016 but slightly more D leaning.
2024 (Biden)
*holds up cross* GET BACK! GET BACK SATAN!
...yeah. Biden was bombing hard. This was the worst electoral prediction map for the dems I've ever seen. You'd need to go back to 1988 with tank man (Dukakis) to get a worse map than this. He was losing everything. NY, NJ, WA, IL, and ME1 were all on the chart, with the first few of those being actively in play. I mean, when do you hear of those states in play? When he pulled out, he was losing virginia. The sun belt is turning red red. The rust belt is still looking like a tomato. This was not a good map. And while I was skeptical replacing Biden with Harris was a good idea given her initial polling seemed even worse, I'm pleased to say that, yes, it was a good decision.
I won't say the modern map looks great. And I won't repost any Harris maps here since that election is still ongoing with the data shifting by the day, but yeah...it looks closer to my 2020 revised map. Maybe just a little worse.
As such, I'm not entirely confident even harris can win, but to be fair, I'd rather have a map that looks just slightly worse than my revised Biden 2020 map than this above map. This was TERRIBLE. And this is why I was kinda freaking out. At least we have a chance now...
Conclusion
So yeah, I just wanted to post these so I could bring them up whenever I need them. I wanted to post a condensed history of my election predictions, and figured given my fancy new format with google sheets that these pictures would provide my predictions in a relatively snazzy format.
No comments:
Post a Comment