Wednesday, September 17, 2008

A New Way to Predict Election Outcomes

I don't have time to go into why public opinions polls are so problematic at election time.

But, there's a new predictive model out there, that's worth some attention. is the brain child of Nate Silver. He takes a large swatch of polls, their current results of the "if the election were held to today, whom would you vote for" question, then plugs it into a statistical model that is then re-run dozens of times to establish the likely outcome of the election. His poll was an excellent predictor of actual outcomes during the primary.


brooklynkevin said...

Hey, Jenny.

I've been watching since the primaries. His mechanism for calculating and calibrating are "above my pay grade" (sorry, I just love quoting that line). The one thing I appreciate is his willingness to be conservative (read: patient) before posting his results. It may not be a winning marketing strategy for the poll-addicts out here, and I'm sure it must be infuriating for those who are trying to make decisions based on polling, but it gives me confidence that the numbers are right. I also appreciate that he is willing to include trending predictions, and while he always includes caveats, he does take the time to put forth future scenarios.

BTW, I loved THINGS SHOULDN'T BE SO HARD. Perfect timing. Thanks for sharing it.

Jenny Stromer-Galley said...

I agree with you. He's a cautious analyst, which always is a good sign.

brooklynkevin said...

I'm wondering how much of the increased political interest indicated in the Gallup poll (see link below) is due to the internet and social networks. If we just look at issues or historical uniqueness, I guess people would be more interested in general, but this spike is so large it seems there could be some other factor propelling the communication.