Nate Silver, the statistical wizard who correctly called the 2012 presidential election in all 50 states by averaging poll results, says what he does isn't that complicated.
"There is some art to this because we give a poll more weight if it has a better track record, but the principle is simple," Silver told about 1,000 members of the Minnesota Interactive Marketing Association at its annual summit meeting at the Hilton Minneapolis on Tuesday. "It got a lot of attention in the media last year because it goes against the way political campaigns are covered."
With political polls, the news media often cover what's exciting instead of what's significant, said Silver, whose FiveThirtyEight blog was carried on the New York Times website last year.
"If there are 19 polls that say one thing, and the 20th poll shows something different, it's probably an outlier poll," Silver said. "But because people are looking for an exciting story, the 20th poll gets a lot of attention."
Silver prefers to study all the polling data, and makes a point of saying he makes forecasts, which have an element of uncertainty, not predictions, which he considers more concrete. His model is the weather forecast.
"Weather forecasts are honest about what they know and don't know," Silver said. "TV pundits, not so much."
Silver's correct call of the 2012 presidential election was considered a big win for the science of statistical analysis. In the final days of the campaign, when some national polls favored Mitt Romney while many state polls favored President Obama, Silver's analysis showed that Obama's chances of winning were well over 90 percent.
Mitt Romney's forces, who apparently didn't put much faith in Silver's statistical analysis, were caught by surprise on Election Day. Silver said he has some sympathy for Romney.