For some time, I’ve known that there was a little argument going on between election modelers Sam Wang and Nate Silver. I first noted it months back when I saw Wang poking at Silver about the fact that Wang that perfectly predicted the 2012 election and Silver had not. It always seemed playful to me, but with a serious undercurrent. After all, Silver was the man everyone talked about in 2012, but Wang was just this guy that more serious nerds knew about.
But then on 17 September, Nate Silver wrote, How The FiveThirtyEight Senate Forecast Model Works. Don’t let the title fool you; it was mostly just an exercise in calling Sam Wang an amateur. I discussed the situation at that time, The Senate Model Arguments. It’s gotten rather worse since them.
Silver has sent out a number of tweets that are, well, rude. I should be clear about my position. For quite a long time, I’ve considered Nate Silver a jerk. It isn’t surprising to me at all that he would be good at poker — it is a game that does tend to go with that temperament. And Sam Wang’s model has more told me what I want to hear this election cycle. But in general, I believe Silver’s model more. It is the one that I focus on now. But I’ll admit, a lot of it is style. I like how the model has inertia built in; it doesn’t bounce around like The Monkey Cage model, for example.
But recently, the whole thing seems to have turned ugly. For example, on Monday, Silver tweeted this:
Looks like @SamWangPHD pulled the plug on his forecast that had shown Democrats ahead. pic.twitter.com/pMs5U6XzkP
— Nate Silver (@NateSilver538) September 29, 2014
My response to this was like Michael Salerno, “You’re looking petty, Nate. Just do you, my man. Your methods and results speak for themselves.” I was frustrated while the Princeton model was down. But to imply that something was amiss? Also, this tweet goes along with others where Nate Silver’s entire argument is, “The Princeton model is an outlier.” Everyone knows that. It’s what makes it interesting.
If all the Senate election models provided the same results, we would have to wonder what the point was. The truth is, at times, The Monkey Cage model has been a big outlier. But I’ve always been interested in that model because of its basis in (at one time: only) political science fundamentals. What’s interesting about Wang’s model is that it is the simplest. My understanding is that it doesn’t take into account “house effects” — estimates of how good particular polling organizations are. The assumption is that it will all work itself out. This is the wisdom of the crowd approach. Because of this, the Daily Kos model, which is also “polls only,” gives the Republicans a distinctly better chance than the Princeton model.
For whatever reason, Wang has gotten under Silver’s skin. I suspect it is because Silver is very competitive. Wang is a scientist. So regardless of his desire to win (which I don’t doubt), in the end, he can take the high road and say, “Now that the election is over, what have we found out about different approaches to models?” If Silver’s predictions prevail, he will just say he already knew, but Wang will be able to save face. If Wang’s prediction happens to prevail, then everyone else will look very foolish, even though it won’t actually mean that they were wrong or that Wang’s model is better. Silver’s prediction currently gives a 60% chance of the Republicans taking the Senate. That still means the Democrats have quite a reasonable chance of holding the Senate. If I were Silver, it would bug me that I would take a reputational hit if the slightly less likely thing happens.
I wish this would all stop. I think Wang and Silver and everyone else are doing very good and interesting work. I’m glad that we have different approaches to modeling the election. But at this point, I think there is real bad blood between these two men. And “The Ballad of Sam and Nate” may go on long past the election.