Fantasy politics?

A baseball fan looks more likely to call the outcome of the US presidential election correctly than the pollsters.  

Nate Silver, a stat freak best known for inventing PECOTA, a 'sabermetric' system for forecasting the performance and career development of Major League Baseball players before the beginning of the season, is gaining attention for applying similar statistical techniques to the presidential election, by projecting the breakdown of the Electoral College.

''Poblano', as Silver calls himself, has had some notable scalps on his blog, FiveThirtyEight.com: he predicted that Barack Obama would win the North Carolina primary by 17 points when most expected him to win by a single-digit margin. The following day, Obama took the state by 14 points. His Indiana projection was spot-on. He predicted a 51-49 Clinton victory. Pollster Mark Blumenthal devoted a National Journal column to the 'Poblano Model,' noting that it had outperformed five major national polling operations in predicting certain results. At last check, the site projects Obama defeating John McCain in the Electoral College, 272 to 266. Had Hillary Clinton taken the Democratic nomination, Silver predicts she would have beaten McCain by a wider margin, 291 to 247.

Quantifying reality?

Silver's methodology is a mixture of art and science, working to incorporate mean, modal and median averages. He weights polls according to the pollster's historical track record, sample size, and the recentness of the poll. He also simulates the election 10,000 times for each site update in order to provide a probabilistic assessment of outcomes.

Top 5 pollsters
Rank	Pollster	Error
Source: FiveThirtyEight.com: Electoral Projections Done Right
1	Selzer & Co.	+0.75
2	SurveyUSA	+1.02
3	Rasmussen	+1.30
4	U. of New Hampshire	+1.41
5	Chicago Tribune / Market Shares	+1.42

Bottom 5 pollsters
Rank	Pollster	Error
Source: FiveThirtyEight.com: Electoral Projections Done Right
28	Marist (NY)	+2.49
29	Franklin & Marshall / Keystone (PA)	+2.59
30	CBS / New York Times	+2.64
31	Zogby Interactive	+5.73
32	Columbus Dispatch (OH)	+8.00

Yet his secret ingredient to boil out error comes straight from fantasy baseball. Just as Silver factors in players' phenotypic attributes -- such as handedness, height, weight, and career length -- he also considers the political, religious, ethnic, and economic demographics in each state.

It can add an extra layer of sophistication and common sense to a poll. For example, the February Survey USA polls had Obama four points of McCain in North Dakota, but behind by four points in South Dakota. Yet Silver correctly saw the states as cultural blood brothers, and thought it unlikely there was a true eight-point differential in how Obama polled in each state. Applying demographics sniffed out the discrepancy. 

Beer, wine and lies

Silver claims his objective is to prevent the misuse of data. In his blog, he writes that stats are massaged to make stories. In baseball, that may mean too much focus on a batting average when statistics like on-base percentage and slugging percentage actually win ballgames. In politics, that might mean hijacking a certain polling result to weave together a narrative that is not supported by the evidence.

Blumenthal believes the fundamental insight captured by Silver's model is the remarkable consistency of vote preference in the Obama-Clinton race among key demographic subgroups. He writes that Clinton consistently prevailed among a 'beer track' coalition of blue-collar whites, Latinos and seniors. Obama consistently dominated Clinton among blacks and younger white voters, and he draws additional strength from a 'wine track' coalition of independents and well-educated white voters. If Silver can identify similar patterns -- drinking habits or otherwise -- in time for November, his predictions this autumn may be worth serious study. 

Whether or not he calls November correctly, Silver may have succeeded in bringing objectivity to the polls. "There's been a societal bias against objective analysis for so long -- we're just supposed to be reactive, without trying to figure out what anything really means. The 'Sabermetricians' have broken that bias down in baseball just a little bit. Now maybe people like Nate can apply the same standards to some areas of greater importance," Steven Goldman tells The Wall Street Journal.