A lot of other models were saying something ridiculous like Clinton had 95% chance to win or something. Nate Silver’s model seems better than others based on this, if anything.
The constant attacks on how 538’s model performed in 2016 says more about statistics literacy than it does about the model.
There is plenty to criticize Nate Silver for. Take your pick. Personally, the political nihilism that’s increasingly flirted with “anti-woke” sentiment is good enough for me. Some people might prefer taking issue with the degenerate gambling. The guy has pumped out plenty of really dumb hot takes over the years, so you have your options.
But his models, historically, have performed relatively well if you understand that they’re models and not absolute predictors.
People forget that Clinton lost because of Comey’s October revelation that the FBI was reopening the investigation into her emails.
Looking at the historical election wins where president with lower popular vote won, trump clearly is outlier and either had outrageous luck (I doubt it) or help to push things just enough to get enough EC votes.
Of course this help, that he got in 2016 he still is getting right now so we should still assume odds will be in his favor and make won’t get suspended and vote (the more people vote, the harder is to artificially affect the results).
I agree with your take on the old 538 model, but if you read Nate’s new substack it become pretty clear that he’s been ‘captured’. Almost all of his post seem to fairly anti-Harris in their biases and it feels like all of his writings are really meant for one person, that person being the owner of Polymarket who he has a very large consulting contract. What these biases are doing to the Model I don’t know but the new model at 538 which was built from the ground up by other statisticians consistently trends about 10-20% higher odds for Harris taking the election.
Oh, I don’t disagree at all.
Like I said, Nate’s definitely increasingly treaded into questionable territory in the past few years and I don’t have a sense for whether it’s impacted the model since I’ve honestly not been paying close attention to the horse race this cycle.
I was mostly pointing out that while the dude has almost always been a bad take generator, the 2016 model very arguably outperformed its contemporaries despite the popular view that they blew it. I wouldn’t be shocked if Nate’s sponsors and general ideological drift has impacted the model this cycle (*especially given Peter Thiel’s involvement), but I don’t have a strong sense for whether that’s the case either. I also wouldn’t be particularly surprised if he sufficiently separated the stats from the dumb ideas to produce a reasonable model either. I just don’t have enough info to have formed an opinion there.
