The great war between people who write about politics for a living is not between liberals and conservatives, but between humanities majors and math nerds, and their battleground is currently the validity of Nate Silver's FiveThirtyEight election prediction model.The English majors have been writing that the presidential race is a total tossup-down-to-the-wire-nail-biter, and the math nerds, most prominently Silver, have been writing that actually, the race has been pretty stable and President Obama has a solid chance of winning. A 72.9 percent chance, in Silver's estimation. The week before Election Day, the English majors launched a major offensive against the math nerds, who pose a threat to fierce urgency of their headlines. "Nate Silver: One-term celebrity?" Politico's Dylan Byers asked Monday, chronicling the complaints of David Brooks, Joe Scarborough, and others whose livelihoods depend on people caring about their subjective feelings about elections saying they think Silver's overrated. On Tuesday, nerds rallied to the defense of their own.
For a long time, most discussion of who will win the election has been dominated by pundits, who depend "heavily on qualitative impressions and hazy narratives," Dartmouth professor Brendan Nyhan writes at Columbia Journalism Review, "But as the audience for quantitative analysis of politics has grown" -- analysis like Silver's -- "the establishment analysts have become increasingly defensive about their status." You see that in articles like the one by Politico's Byers, who suggests that if Romney wins the election, Silver's career is toast. That seems to show a lack of understanding of the difference between analyzing the probability something will happen and predicting that for sure that thing will happen. A 25 percent chance of something happening is still a pretty solid chance. And Silver's model is actually one of the more conservative, Nyhan demonstrates in some handy charts, seen at left, comparing FiveThirtyEight to other models. "Whatever objection pundits or conservatives may have is with the state of the publicly available evidence or the way in which forecasters and betters translate that evidence into probabilities, not with Silver or his methods," Nyhan says.
It's worth noting that it's Politico's writers, as The Washington Post's Ezra Klein points out, who have made some of the most "downright weird" arguments for why Silver is wrong. For example, Politico’s Jonathan Martin tweeted, "Avert your gaze, liberals: Nate Silver admits he’s simply averaging public polls and there is no secret sauce." (The lack of "secret sauce" is not a bug, it's a feature.) Sure, you can criticize Silver for not making his proprietary model public, or for leaning too heavily on polls, or that he adds unnecessary factors, Klein says. But the idea that Silver is wrong because it shows Obama winning is silly. So do most other models. But it makes sense that "a lot of the odder critiques of Silvers have been coming out of Politico," Klein writes. He's a threat to horserace journalism, which Politico specializes in. So what seems like a debate over the value of Silver's work is really a debate over the value of Politico.
No comments:
Post a Comment