Predicting Presidential and Other Multistage Election Outcomes Using State-Level Pre-Election Polls

Abstract

Although much of the media attention during presidential election years focuses on polls tracking popular support for the major candidates, the complicated role played by the Electoral College in this multistage election process must be accounted for in order to address the issue of winning the presidency. State-level pre-election polls are used in a manner that allows the structure of multistage election processes to be addressed directly. We consider frequentist and Bayesian approaches for predicting election outcomes and discuss ways to incorporate such analyses in a course project suitable for undergraduates or graduate students studying statistics. Using state-level pre-election polling data, we consider the U.S. presidential election of 2004 and we also apply this approach to predict the control of the U.S. Senate in 2006. This class exercise has proved to be a useful “capstone project” which requires students to address a complicated problem by synthesizing multiple sources of available data and applying a combination of statistical methods. Using simulation-based approaches for addressing the multistage nature of presidential elections and control-of-Congress processes can be valuable and instructive for students of statistics and political science, and can be beneficial to the media in providing consumers with political news.