March 15, 2009

FiveThirtyEight Founder Speaks On Stats, Elections and Baseball

Even as many voters sat on pins and needles, the 2008 presidential election was among the easiest to predict, according to Nate Silver of, whose site skyrocketed to stardom amid high visibility throughout the primary and general election season last year, as he took a thorough background in statistical analysis and focused on the world of politics, hoping to improve the accuracy of polls and predictive analysis. In his comments on a keynote at the SXSW Interactive conference today, he said media outlets relied too much on most recent news, and could improve their prognoses if they instead turned to historical statistics and trend data.

Nate said he started "out of frustration" with the traditional media properties, including CNN, MSNBC and Fox News. He said their coverage "wasn't empirically valid and correct," adding, "polls were too much of the narrative and they were taken too seriously as they were poorly conducted and interpreted."

Armed with a history of crunching baseball statistics to predict how well players would perform in the future from his time with the Baseball Prospectus, Nate tried to spot irregularities in polling or find variables that indicated how voters in each state were likely to trend which most pollsters were missing. One example included how in the Appalachian region, voters who declined to state their ethnicity, choosing to instead be labeled as "American", were doing so as a badge of pride, but also indicated a level of "redneckness," as he put it.

Like with baseball, Nate called the political process a long season, where largest trends were not swayed by an individual game's data, or by a single primary.

"You don't get that much information at once. A puzzle gets solved a little at a time," Nate said. "People are trained to over-react to these kinds of things, and I would urge patience."

Having successfully predicted the 2008 elections more accurately than practically anybody else out there, Nate is now being courted to try and solve a wide array of other issues, ranging from predicting the Academy Awards (which was partially successful), to predicting economics. As could be expected, he was asked to provide his thoughts on the economic slowdown that has effected everyone, and how long it would be until potential voters started to blame the Obama administration, instead of the Bush administration.

"We haven't had a situation like this in the modern era," Nate said. "People are really scared, and they don't see the light at the end of the tunnel. You have people who are very pessimistic about the future of the country, but Obama has great approval ratings. Usually that doesn't happen. There is a kind of grace period of three to six months, and after this grace period, of about 18 months, people assign him as much blame as they would Bush for the economy, so he needs the economy to turn around sooner, rather than later."

But even armed with as much data and talent as he has, Nate recognizes that other factors are in play. Baseball players can get injured, and "you can only expect a human to do what they do what they were doing so many times," he said, adding, "in baseball, everything in the last 100 years is 99.9% accurate. The real world is not like that."

Speaking of not predicting the real world, the term FiveThirtyEight, which refers to the number of electoral votes available in a presidential election, may need to change, should Washington DC gain representation in Congress, throwing all the numbers off. Should that happen, Nate just may have to change his URL.