Friday, November 2, 2012

Prediction Models

So, for my usual pre-election amusement I was perusing various prediction models and ran across this one after a friend pointed me to this article talking about having it updated with current data (rather than data from back in May/Jun). What surprises me about it is that after looking through the other models presented this year the two professors (Kenneth Bickers and Michael Berry of the University of Colorado) actually did what should have been done with every model. Instead of using polls (which are volatile and varied depending on how skewed they are toward one side or the other) they used a number of economic issues and looked at each state individually along with a number of historical data points.

I won't bother explaining the entire thing, they did an excellent job without me butchering their work. The end result after the recent updates however has Romney winning in a landslide 330 to 208 in the electoral college, and their model even predicted the Al Gore popular vote vs GW electoral vote win (well, showed the same results using data from that election). They've gone all the way back to 1980 and been correct, so I'd say even with their admitted +/- scale of 28 electoral votes that still puts Romney WAY ahead of President Obama.

It's a good read, and they explain how their system works, I recommend it.

No comments:

Post a Comment