Monday, November 05, 2012

I've seen this movie before

It is starting to look like history is repeating itself.

About 35 years ago, a night watchman in Kansas started spending his spare time between rounds analyzing baseball. He knew he had piles of data, but wondered about what people were doing with it. He wondered when a player hit his prime, pondered how good a catcher was at throwing out baserunners, and asked why no one considered the ability to draw walks as a valuable offensive tool.

Eventually, those techniques caught on, spawned bunches of imitators and copycats, and changed the game of baseball. Some day, Bill James - as unlikely a choice as possible based on apparent athletic ability - may get into the Baseball Hall of Fame for his revolutionary work.

I'm starting to believe Nate Silver is headed down that path, even if there's no Hall of Fame for political forecasters.

Silver was one of the geniuses who worked on the early editions of Baseball Prospectus, something of a successor to James' Baseball Abstracts. He also was a professional poker player some years ago. He decided to use the techniques he'd picked up elsewhere, due to his interest in probability, and apply them to politics. His fivethirtyeight.com blog gained a large niche audience in 2008, and his reputation grew due to the fact that he picked 49 out of 50 states in the Presidential race and every single Senate race correctly.

Silver's blog was picked up by the New York Times, which means readers have to go through the old link (see above) or follow him on Twitter to avoid the paywall. Sounds like the audience has figured that out. The number of readers has to be huge.

Silver has reminded us that the Presidential race is not one election, but 51 - one for each state plus D.C. So all of those national polls that take a snapshot of Candidate A vs. B can essentially be thrown out because they are meaningless. He's concluded that even in an election that is tied nationally, President Obama has a much better chance of winning.

Silver has paid attention to the state polls, and there have been no shortage of them in the past two months. The analyst also has created some large complicated model, which even includes the impact of economic factors such as unemployment, for coming up with probabilities. He's also checked the history books, and made what should be a simple discovery: a three-point lead in January is much different than a three-point lead on November 5 (at least this year). Get enough polls pointing in one direction, even within a margin of error, and the probability is good that such a leading candidate will win.

Silver's done his figuring this year, and he points out a few facts along the way -- such as the fact that Barack Obama has never trailed at any point in Ohio when the polls are factored together. So Obama certainly figured to be the favorite there tomorrow, right? And he is, at least as of my last glance at the site - an 87 percent chance of taking the state. (Florida, by the way, is virtually tied but Silver gives a 56 percent chance to Governor Romney.)

Take that and apply it to the other 50 elections, and as of Monday afternoon Silver gave Obama an 86 percent chance of winning the election. As he explained in an essay, it would take a lot of bad polls for the election to come out differently. That's possible, but in Silver's view not likely.

There's a lot of noise out there about the election. Someone at work the other day mentioned how close the national polls were and predicted a long night of election-watching. The television networks, who have an obvious interest in driving up their audiences, are emphasizing those national polls. And it's fun to hear some of the pundits make predictions; people such as Dick Morris and Jim Cramer are busy predicting landslides (although not the same landslide). Speaker John Boehner supposedly guaranteed Ohio would go to the Republicans yesterday.

That puts Silver on a bit of the ledge, and he's heard about it. The political insiders say in private and public that the election really is too close to call. And that's fine. James heard the same sort of criticism in his first decade of writing. The smart guys usually turn out fine in the end, and Silver is one of the smart guys.

(P.S. It turned out that Mr. Silver had a good night. He got the first 49 states plus D.C. right, and said Florida came out to a virtual tie - with Obama holding the lead on the fourth or fifth decimal point. This was called "the Moneyball election" by one pundit, and Silver was a reason why.)

Be notified of new posts via Twitter @WDX2BB.

1 comment:

Glenn Locke, The Tall Thin Guy said...

Funny thing is how revolutionary the one main, obvious part of Silver's model: the larger the sample size, the more meaningful the data.
By the way, you can also get into NYT for free by typing the headline of an article into google, the clicking the link.