Nate Silver acknowledged that he was doing something a little unusual in a Sept. 17 blog post when he called out fellow forecaster Sam Wang of Princeton University. But it also appears to have been the culmination of a long-simmering — if largely under-the-radar — feud.
“I don’t like to call out other forecasters by name unless I have something positive to say about them — and we think most of the other models out there are pretty great,” Silver wrote. But he then labeled Wang’s model “wrong” and provided a detailed argument (with footnotes) to explain why he thought so.
And it didn’t stop there. Periodically over the last week or so, Silver has continued to take shots on Twitter at Wang’s forecasting model, which has consistently been more optimistic about Democratic odds of keeping the Senate than Silver’s (or any other forecaster).
That led to a lot of buzz in the tiny world of poll nerds and a series of pained responses on Twitter from Wang. In separate interviews with TPM, Silver declined to say what exactly provoked him but said Wang had been “deceptive” in characterizing their disagreement while, for his part, Wang continued to chide Silver, particularly for refusing to engage with him directly.
Here’s a sampling of some recent Silver tweets knocking Wang:
Problem with overconfident models: Yesterday, Sam Wang’s snapshot had Begich as a 99% favorite in Alaska. Today it gives him a 23% chance.
— Nate Silver (@NateSilver538) September 26, 2014
— Nate Silver (@NateSilver538) September 29, 2014
@NateSilver538 Nate, I am so glad you consider PEC a must-read! Such a pleasure. Now…read more than the headline to learn a little more
— Sam Wang (@SamWangPhD) September 29, 2014
The target of those critiques seems a bit confused — and a little annoyed that Silver has not responded to him personally. In a phone interview with TPM this week, Wang said that he had emailed Silver since the flare-up but has not heard back from him yet. He referenced more than once his relatively meager 6,500 Twitter followers versus Silver’s 959,000.
“He’s the Kim Kardashian here,” Wang said. “Certainly anything he says is impossible to ignore because of his 950,000 Twitter followers. It’s just right there for anybody to see. I actually tried to ignore it for a while, but it got hard because he just didn’t give up.”
Silver declined to comment to TPM about why he had critiqued Wang publicly, when it is his self-described habit not to, or on why he had not engaged with Wang directly. But Silver did tell TPM that he believed Wang had misrepresented their disagreement. It is not a “polls vs. fundamentals” debate, he said, and he praised polls-only models like the Huffington Post’s Pollster and Real Clear Politics as good at what they do.
“First of all, I think it’s very misleading and I would use the word deceptive for him to claim that this is all about polls vs. fundamentals,” Silver told TPM in a phone interview. “The other big problem: He is not going through the rigor that most other people do in being empirical, meaning trying to justify things based on past election results.”
Silver described what he sees as the problem with how Wang averages his polls. “The way he does it is he looks back at average snapshots (of polls) since June,” Silver said. “That’s like looking at the average score of the football game, instead of the current score… It’s a very strange assumption.”
“He should provide evidence that it’s a good sound empirical way to do it and he doesn’t,” he continued. “And I think he’s not aware of how much difference that makes.”
Wang and Silver have interacted personally only once in 2012 when they both appeared on National Public Radio, Wang said. But there is a history here of Wang taking jabs at Silver and perhaps vice versa, long before Silver was one of the most well-known U.S. media figures with his own website and Wang had his own perch at the New Yorker. It dates back at least as far as 2008, when Wang wrote an article on his blog called “A weakness in fivethirtyeight.com.” Silver in turn commented adversarially on the blog back in those days, Wang said.
Wang wrote about “the War of the Senate Models” for Politico magazine this May and, last month, in one of his blog posts for the New Yorker, Wang took some issue with the inclusion of fundamentals by other forecasters, including Silver.
“All of these predictions have one thing in common: they are based on more than pure polling data,” Wang wrote on Sept. 9, about a week before Silver called him out, referencing Silver’s Five Thirty Eight as well as the New York Times’s Upshot and other forecasts. “By considering other factors, these data journalists are putting their thumbs on the scale — lightly, but with consequential effects.”
He later noted in the same post that Silver’s model had missed two Senate races in 2012 — Democratic wins in Montana and North Dakota — that Wang had correctly predicted. A few days later, Silver was openly and uncharacteristically criticizing Wang from his position at Five Thirty Eight.
Wang seems aware that he might have brought this on himself. But he contrasted his experience in academia, where peer criticism is a prominent feature, with Silver’s largely unquestioned reign as the “nerd king” — as Wang called him in that New Yorker post.
“I think part of it is that generally speaking nerds are pretty genteel, right? It’s pretty uncommon for nerds to say anything remotely critical about one another,” he told TPM. “The fact that I’ve been critical at all, I can easily imagined that being annoying to him. I’m an academic — and in academic life, we’re used to sticking each other with pins.”
The technical merits of the Silver-Wang debate is beyond the reach of most laypeople. But in addition to his lengthy blog post and comments to TPM, Silver questioned how Wang weighs different pollsters. “I think all that stuff doesn’t matter very much,” Wang told TPM.
But Wang said (to TPM and on his blog) he also believes that Silver has mischaracterized his work. Wang admits he was “loose in my terminology” in the past when describing the two kind of probabilities that his forecast offers, but he contends that Silver has “selectively used” those past statements as evidence against him.
“In peer review, it’s generally agreed that if a criticism is made, and the criticism turns out to be factually false, then the critic is required to withdraw the criticism,” he said. “And I think that he may have made points that are valid, but he’s made a number of untrue statements about what I do. It’s a really one-sided discussion.”
Silver, by contrast, said he has a firm grasp of Wang’s methodology. He just doesn’t believe it works.
“I think he’s taking a model that is flawed in its design in some very profound ways,” Silver told TPM, “and mistaking that as a feature.”