With the New York State primaries coming up September 14 and the general election on November 2, I predict that as soon as summer turns the corner into September, we’ll start hearing lots and lots about polls that predict election outcomes. To find out if there was any early discussion of polls, polling, and outcomes, I returned to my favorite election forecast site from the 2008 presidential elections, FiveThirtyEight: Politics Done Right.
Sure enough, there it was, a comparative rating
of pollsters. This will give people like me, who tend to believe any poll just because it’s covered in the news, a way to assess the poll reliability. FiveThirtyEight is the brainchild of Nate Silver, and 538 is the number of members of the Electoral College. Silver’s primary business is Baseball Prospectus
, which is also fueled by Monte Carlo simulation and other risk analysis techniques, but FiveThirtyEight has done well enough for the New York Times to want incorporate it in its online coverage during the coming elections.
Silver’s grasp of statistical analysis becomes immediately evident when you go to his page on the pollsters, and he’s more than happy to discuss the statistical methods he uses to rate the pollsters–regression analysis of raw data, Monte Carlo software in an Excel spreadsheet, weighting of poll performance data, and so forth. His take on these matters may be of practical interest to any of you who use these techniques in financial risk analysis.
Elections are all about decision making under uncertainty, especially voter decisions under uncertainty, and according to Nate Silver, only polls taken within 21 days of an election are reasonably reliable. So when the national campaigns are ramping up in October, keep one eye on the polls and one on FiveThirtyEight.