Political engineering: optimizing a U.S. Presidential candidate’s platform

Abstract

We explore the application of operations research to the problem of defining/refining the political strategy for a candidate in a U.S. Presidential election. We use Hierarchical Bayesian techniques to model criteria used by a stratified random sample of registered voters to evaluate a candidate/platform. We then use the estimated utility parameters as inputs to a model that finds the positions a candidate can take on the salient issues of the election that will optimize expected Electoral College votes conditional on the positions respondents perceive to have been taken by the opposing party’s nominee. This approach is unique in that it (i) considers the value that individual voters associate with various positions the candidates can take on various issues, (ii) considers the chronicity of the electorate’s perceptions of a candidate’s positions on the salient issues, and (iii) yields a solution that will optimize expected Electoral College votes. We demonstrate this model on data collected immediately prior to the 2004 U.S. Presidential election (the most recent U.S. Presidential election not involving any potential candidate for the upcoming 2012 U.S. Presidential election), and we show how these data and the model can also be used to assess the perceived clarity of a candidate’s positions, the sensitivity of a candidate’s support to her/his perceived positions, and the viability of a third party candidate.