The real showdown of election 2012: math vs. instinct?

Brendan James
The Ticket
November 9, 2012

Alongside this year's race for the presidency and myriad battles for Congress, another contest for power quietly unfolded, within the news media. Could the winner of the White House be accurately predicted by analyzing public opinion polls using mathematical formulas, or could some experts divine the future using more ephemeral measurements, like momentum and enthusiasm and what they felt in their guts?

The statisticians won.

Nate Silver of the New York Times has gotten much of the attention for predicting that President Barack Obama would defeat Mitt Romney by winning 313 electoral votes (Obama will likely end up with 332). But many others made accurate forecasts as well, including Yahoo's The Signal, Talking Point Memo's PollTracker and Votamatic.

Evidence of their triumph has cropped up in the form of a new hashtag on Twitter: #DrunkNateSilver. According to the meme, Silver is partying (and further prophesying).

View photos

Many notable cable news pundits and columnists made projections for a Romney win that overestimated his electoral votes by the dozens.

"We're very happy with what our model produced," says David Rothschild, the economist behind The Signal's model, which predicted Obama's victory as early as this past February. He sees commentators' outbursts about Silver as a reaction to the growing reliability and popularity of statisticians.

"In general it came down to Nate being the poster child of those looking at data vs. those talking from the gut," he said.

Math didn't always triumph, however. One analysis by the Colorado political science professors Kenneth Bickers and Michael Berry—strongly favoring a Romney victory in 2012—plugged in the level of unemployment as a means to measure people's attitude toward the incumbent president. In a University of Colorado press release, the duo said, "The apparent advantage of being a Democratic candidate and holding the White House disappears when the national unemployment rate hits 5.6 percent."

"The model was wrong," Bickers told the Denver Post, simply.

Rothschild's team took a different approach. "We focused only on trends, not on levels—when you measured the movement of things like the jobless rate from quarter to quarter, that trumped any inference you got from the unemployment level."

All of the more successful models owe their accuracy to the very accurate polls taken throughout this year. Each of the poll trackers, from FiveThirtyEight to Votamatic, averaged the state polls conducted by companies like Gallup and Public Policy Polling. The polling data teased out of each state was remarkably consistent—with the exception of the polls conducted by Rasmussen, which tilted toward Romney throughout.

While analogies have cropped up about the future of political analysis going the way of professional baseball as depicted in the book and the film "Moneyball," this burgeoning class of political statisticians can't replace the political press without considerably narrowing its function. There will always be a place for reporters and pundits to dissect everything from campaign strategy to political language to—gasp—public policy.

What to Read Next