Wednesday, November 07, 2012

The Biggest Winner for the 2012 Election: Math?

Well, I didn't believe Nate Silver but he shows that a mathematical analysis might be the way to go to handicap election's in the future. All that gut-feeling stuff is not the same as cold-hard math.

This year, according to all projections, Silver’s model has correctly predicted 50 out of 50 states. A last-minute flip for Florida, which finally went blue in Silver’s prediction on Monday night, helped him to a perfect game. 

A caveat: Florida has not yet been called officially, but Obama is in the lead with 98% of precincts reporting. If anything, Silver’s placing of Florida on a knife edge makes him look even more prescient. No wonder one of the night’s more popular tweets suggested that he was actually from the future, working from old newspapers.

I think his idea of taking all the polls and weighting them for historical accuracy is a very sound one. The idea that all the polls were used is what took some of the bias out of the equations. So you can have some polls (CNN) oversampling Democrats but others (Wall Street Journal) may have under-sampled them. Then he weighted them according to accuracy. That seems like a sound plan because the electorate seemed very much the same as in 2008. So if the poll was right in 2008 it should have been right in 2012.

The GOP made the mistake of thinking the electorate had changed from 2008 to 2012 and that did them in. I guess Silver needs to pick stocks next because I would love to see if his algorithm will work on Wall Street. Maybe he can use a mix of research Buy/Sell/Hold calls and newsletter picks as his "polls" for each stock to see if they hold up.

No comments: