Ruminations on Challenges to Prediction with Rational Choice Models

Abstract

The ultimate test of any social science theory lies in its ability to predict and explain real behavior. Converting formal theory into a form that can be tested against real-time predictions, however, is a difficult and demanding task. The growing literature on applied formal models provides an opportunity to investigate how robust predictions are to the relaxation of specific assumptions. Here I investigate the robustness of applied theories in the light of multiple equilibria and the chaos theorems of McKelvey and Schofield, alternative dimensionality assumptions, modifications of conditions for preference aggregation, and bounded rationality - all with an eye toward their implications for the accuracy of predictions regarding outcomes and decision dynamics. We come to a better understanding of the challenges that remain in converting formal models into practical tools to guide the assessment and making of complex decisions.