What is the probability that your vote determines the outcome of the Presidential election?

Andrew Gelman (Dept. of Statistics, UC Berkeley)

It's typically about 1 in 10 million.

To answer this question, you need a probability model of election outcomes. We first discuss some obvious but wrong ideas and then present the right general framework for answering this sort of question.

Our models for elections are extensions of well known results in political science. The state-by-state outcome of the Presidential election can be predicted with fair accuracy using information available before Labor Day. We briefly discuss the political implication of this fact and then describe a forecasting method based on a random components linear regression model. As a side issue, we also address the question of whether the electoral college is biased in favor of the Democrats or Republicans.

We also discuss the statistical difficulties in creating the model and compare to models for Congressional elections. This work is joint with Gary King, Department of Government, Harvard University.