Agent-based Models in Public Choice

Richard Wallick

Advisor: Charles K. Rowley, Department of Economics

Committee Members: Omar Al-Ubaydli, Robert Axtell

Carow Hall, Conference Room
May 02, 2012, 10:00 AM to 07:00 AM

Abstract:

Public choice is the study of political science using the tools of economics.  The future of modeling in public choice may be glimpsed by examining its evolution in economics.  For problems that are influenced by heterogeneity of actors, social networks, or emergence—the arising of a complex system from simple phenomena, such as Adam Smith’s “invisible hand”—economists increasingly are turning to agent-based modeling.  This dissertation argues that public choice scholars can benefit from following the same path.

Chapter 1 introduces agent-based modeling, describing its advantages and its disadvantages.  By definition, the agent-based approach emphasizes methodological individualism.  Since methodological individualism is a cornerstone of public choice, many public choice models are amenable to agent-based implementations.  This is best exemplified by a review of Gordon Tullock’s approach to modeling.  The agent-based approach features prominently in Tullock’s writings.  Had he been armed with the technology of the 21st century instead of the 20th, Tullock would have been an agent-based modeler.  Agent-based modeling offers a new vehicle for exploring Tullock’s classic works.

Chapter 2 demonstrates the agent-based approach by transforming a familiar public choice model, Samuel Peltzman’s 1976 model of regulatory capture, into an agent-based model.  The agent-based version can explore situations that Peltzman’s formulation cannot, such as the presence of bilateral transfers and the impact of heterogeneity among citizens.  Different types of heterogeneity affect the model’s predictions in different ways.  Taken as a supplement to Peltzman’s model, the agent-based version sheds new light on an old problem.

Chapter 3 then develops an extended agent-based model from scratch:  a model of Condorcet’s famous jury theorem.  Condorcet’s theorem rests on three assumptions, each of which may be questioned:  voter homogeneity, voter independence, and voter sincerity.  The agent-based version explores the impact of relaxing each of these assumptions.  There is no shortage of rational choice models that relax these assumptions independently, but only the agent-based version relaxes all three simultaneously. The model is calibrated to represent a population like that of the United States, and used to explore policy proposals.