So...Allen Lichtman had a livestream recently where he was talking some major crap about Nate Silver. They've actually had quite a beef this election cycle, representing opposite extremes as far as predicting elections. We discussed Lichtman and his keys and suffice to say, I'm less than impressed by his model. I mean, I'm on Nate's side. I dont always agree with Nate Silver on things, I think both him and Lichtman are people who get way too much attention and are too full of themselves, but I can at least somewhat respect silver here, while Lichtman's model borders on pseudoscience to me.
Lichtman's model is all fundamentals. He has these 13 things he measures, and if the incumbent party is favored on 8+ things, he expects them to win the election. His keys are overly broad, very subjective, and not necessarily in line with how voters think. They have a decent amount of correlative success, if only because presidents' popularity is correlated with the things he attempts to measure, but I think the trick of his keys is that they are often kind of subjective, and could be interpreted in retrospect in ways that are subjective. Like, IS RFK a substantive third party threat? Does Biden handing the torch to Harris represent a shift in incumbency. What IS a charismatic leader? If the economy is good do people feel it? And honestly, anyone could just retroactively apply this model to say just about anything.
With Biden in the race, I think that Lichtman's model was heading toward being discredited. He was all gung ho on Biden. He thought Biden could win, and that he would win, and if Harris replaced him it would be all over, but then Harris shifted, the polls went the opposite way, and then Lichtman suddenly didn't change his prediction.
Honestly, from a polling perspective, Biden vs Harris is night vs day. Biden NEVER was ahead. not even close. Harris has represented a MASSIVE trend, and honestly, I think it goes to show that keys don't mean much. What matters is VOTING, and POLLING attempts to measure HOW PEOPLE WILL VOTE.
I admit, polling isnt perfect, it has error involved, due to the problem with sampling and sample sizes. And it can change, but that's the thing. I'm not trying to predict an event months from now based on existing data, I'm saying "this is the state of the race now, it can change, but if the election were held today, this is what we can expect to happen", and we will come awfully close to the actual outcome.
Now, to be fair, Silver and others with their models DO try to predict ahead, and that's kind of where Silver gets dragged by Lichtman. He was ripping silver's 52% prediction and how it doesnt tell you anything, and idk, I kind of agree. Silver has a more convoluted model than I do, and more complexity isn't always a good thing. For example, despite polling improving for Harris in the past few days, Silver's prediction got worse. This is because of a post convention bounce being anticipated for Harris after the DNC last week, where any improvements in polling are assumed to be a result of the convention, and if the numbers dont move as much as anticipated, his model will actually have Trump do better.
Basically, Silver is trying to compensate the probability increase from polling my predicting a polling increase and when the polls didnt increase as much as he thought, Harris started going down instead of up. In reality, the post convention bounce didnt seem to happen for either candidate, and the latest round of polls just seemed to follow the overall trend we've been seeing since late July where democratic odds are gradually improving as the polls keep shifting toward Harris. And this is the problem with a lot of others who follow polls. Again, polls are a snapshot, but what Silver and others do with their models is try to predict the future. if you want that, get a crystal ball. I'm not here trying to say what things will look like a few months from now, and in fairness to Lichtman, the problem with a lot of the model and stats nerds is they keep trying to make these overly complex models that dont tell us anything. I mean, 50% coming from me means something much different than 50% from Silver, or 538 (two different entities now), or some other model. Because with these "predictive" models, they're gonna be bogged down by so many unknowns and variables they probably wont strongly go one way or the other until near election day, at which point they end up converging with me anyway. A model that is regularly at 50% on the basis of unknowns based on the future is a model that is literally useless. Will it rain tomorrow or won't it? I don't know, flip a coin.
At least with me, 50% means "hey the polls are so close it literally could flip either way." I mean, we see my charts. We went from around 50% as the PA (tipping point state polls) hovered around 0% in the aggregates and now that we're up to almost 1% in Harris's favor, Harris's odds went up. Simple, right? But what if I had no data at all? Then we get 50% because it's a matter of "we don't know". It's not that the data is conflicting, it's that we don't have any data to really make a decent judgment call.
So, that's my view on silver, 538, and their "models." If anything, my issue with them is they're STILL leaning too hard into fundamentals. I say screw the fundamentals, go by them if we dont have reliable polling data, but if we do....go with the polling data.
As I see it, most people get overconfident and arrogant mostly on the basis of these models. They think they know oh so much better than what the data tells them. We saw it with Biden in the race. I admit I was for Biden staying because at the time the polling indicated he was the best candidate, but still, a lot of people end up using these fundamentals as ways to just arrogantly dismiss voter concerns ("the economy is good you idiot, blah blah blah") or to hype up their own probabilities of winning ("yes yes the polls show us down, but these keys tell us biden should win."). It's just copium.
Honestly, Allen Lichtman is being really petty this election cycle. he acts like a big shot, like he's never wrong and look at how many credentials and awards he has, and honestly, I just see him as an arrogant grifter trying to push his model to sell his new book. And his attacks on numbers guys is obnoxious. Again, not a huge fan of silver, or his work, but polling is probably better at predicting elections than his model. And unlike him, numbers guys will own up to being wrong, either due to probability (see 2016), or a methodological flaw (see my 2020 prediction, which i have since admitted was screwed up because i discounted polls from certain right wing pollsters). Lichtman? "Well it was the electoral college" or something. Whatever dude, you were still wrong. Always moving the goalposts. And that's what i consider the real secret behind lichtman's successes. Get the election wrong? Move the goalposts! 2000, he predicted the popular vote not the electoral college. 2016, he predicted the electoral college, not the popular vote. Blah blah blah spin spin spin.
The fact that anyone takes this guy seriously is laughable to me.
No comments:
Post a Comment