An Analysis of Daily Predictions for the 2008 United States Presidential Election

Abstract

We describe a Bayesian model for analyses of the 2008 presidential election polls that incorporates third party candidates as well as undecided voters. The state-by-state results were used in a recursive formula for the total electoral votes for Barack Obama and for John McCain. A web site was updated almost daily with the state-by-state projections and the posterior distribution for the number of electoral votes for each candidate. The probability of Obama winning was nearly one late in the summer before dropping below one half during mid-September, and finally recovering to essentially one five weeks before the election. The presentation is accessible to readers with an intermediate level of statistics.