Thursday, November 29, 2012

Debunking Nate Silver

Despite what you might read on the internet, its actually quite easy to demonstrate that Nate Silver's election models where very possibly flawed.

According to Silver, September was a disastrous month for Romney. He started September down 47.9% to 50.9% to in the popular vote and a 26.9% chances of winning, and he ended it down 47.4% to 51.5%, and only 14.3%. Yet according to exit polls September was a fantastic month for Romney. He won the 8.7% of voters who decided in September 53% - 45.3%. In contrast, Obama won the 68.8% of voters who decided before 52.6% - 46.3%, and the 20.7% who decided after 49.7% - 46.3%. The last cohort, of course, includes some number of people who thought they had decided in September and then changed their minds, increasing Romney's implied September surge.

It may be that exit polls should be taken with a grain of salt or that masses of people mistakenly remembered the first debate as having taken place in September. But, at the very least, they call Silver's model into question.

No comments:

Post a Comment