Online Program

Saturday, February 20
CS23 Data Synthesis and Meta-Analysis Sat, Feb 20, 11:00 AM - 12:30 PM

Political Engineering: Optimal Political Platform Design and the 2004 U.S. Presidential Election (303096)

*James J. Cochran, University of Alabama 

Keywords: Conjoint Analysis, Resampling, Bias Estimation, Combinatorial Optimization, Political Science

We demonstrate how to treat identification of the optimal political platform for a U.S. presidential candidate as a special case of the share of choices problem. We formulate the problem as an optimal covering problem with special ordered sets representing various issues. We explain how to use this formulation to project the winner overall and in each state, estimate the variance and bias of the estimated Electoral College votes for each candidate, assess the sensitivity of each candidate's support to changes in her/his position on various issues, find the optimal platform for each candidate relative to the positions assumed by her/his opponent(s), gauge the potential viability of a third-party candidate, and identify key issues of a U.S. presidential election. Results of this analysis for the 2004 election will be highlighted. This work demonstrates a unique case of product design, and the presentation will immediately help participants better understand the basic product design problem, approaches for finding the optimal design of a product, and ways to deal with the statistical implications (variation and bias) of finding the optimal product design over sample data.