FLOTSAM & JETSAM: ON ELECTION POLLS

Monday, November 08, 2010

ON ELECTION POLLS

Sam Smith

Nate Silver seems like a nice guy but, like many guiding our way, the NY Times’ election poll analyst is so absorbed in his numbers that he seems to forget that life is at best an approximation. On the other hand, when I told my urologist that I had figured out the medicine was one half science and one half gambling, he said, “I think you’re being kind to us.” Now there's a man who understands life.

Silver would have you believe that if you just work hard enough at getting the numbers right, you’ll get the right answer. Not having the time or money for that, the Progressive Review has tried a different approach over the past decade: we have simply averaged the last three polls and this year, as with most of the times we’ve tried it, the results have been surprisingly good. In the key races for both the Senate and governorship we came within three points of calling them on average – within the polls’ margin of error. You note that I didn’t say 3.1% or 3.2% because, unlike Silver, I had Alice Darnell as my high school math teacher and one of the things she taught us was that an mathematical answer can be no more accurate than the least accurate number used to create it. So if you have a polling error of 3 or 4 points, you can’t honestly end up with 3.1%. Yet pollsters, Washington analysts, and the media do this all the time.

The other problem is that life is real. As such, it doesn’t move in a neat analytic sequence. For example, one place where the polls went astray was here in Maine. The polls generally had a good percentage for the winner – Paul LePage – and the Democrat, Libby Mitchell, but blew it on independent Elliot Cutler who clearly had a last minute surge among the undecideds and almost beat LePage. How does a pollster check the Portland bars the night before an election to find out how many people have finally decided how to vote after a conversation over a few beers?

So we’re going to stick with our three poll moving average which did about as well as the best of the individual results, bearing in mind that life is real and that, hell, I didn’t even make up my own mind until the last minute.