Annual Social Science Statistics Lecture
The Department of Statistics
University of Missouri-Columbia

December 5, 3:00 p.m., MDLBH 309

Dr Jeff Gill, Department of Political Science, University of Florida

Empirical Tests of Complex Voting Models with Uncertainty.

For the past two decades, many formal spatial models of voting have explicity included probabilistic assumptions as a means of describing uncertainty about candidates and issues. A great deal of progress has been made in our understanding of the effects of increasing this uncertainty on the way that citizens analyze political information and the predicted effects that this process has on resulting elections.

The next logical step in this literature is to increase the realism of these formalizations by incorporating: many candidates, large numbers of voters, different kinds of uncertainty, and a greater number of evaluated issue dimensions. While these enhancements lead to a great number potentially testable predictions, they also produce some substantial challenges in analytically deriving key results like equilibrium points and optimal candidate strategies in empirical work.

This paper develops a new implementation of a recent idea in Markov chain Monte Carlo theory, the simulated tempering algorithm, which makes it possible to analyze very complex voting models. This approach builds on dramatic developments in Bayesian estimation that have yet to be fully applied to substantive questions in political science. We expect this work to make research progress in both statistics and political science, as well as in related areas.