Modeling Behavior in Public Goods Experiments

Paul H. Bennett

Advisor: David M Levy, PhD, Department of Economics

Committee Members: Peter J. Boettke, Daniel E. Houser

Buchanan Hall, #D180
April 25, 2016, 10:00 AM to 07:00 AM


This dissertation analyzes the behavior of participants in the class of economics experiments known as public goods games. It is well established that most participants in public goods games do not play the Nash equilibrium, of zero contributed to the public good, initially. By initially I mean either the first round of a repeated game or the only round in a one-shot game. The goal of this dissertation is to explain the observed behavior and to develop models that predict it.

In chapter 1, I describe a theory that models rational reciprocity. I develop a mathematical model based on expectation maximization that predicts that players will donate more than zero to the public good. The key factor is a player’s level of belief that other players are like them. If I believe that other players are like me, I can expect other humans to arrive at the same decision with some probability.  I show that if my belief level in the similarity of the other players to myself is high enough, I should logically play contribute-everything-to-the-public-good.  The above theory, combined with Bayesian belief adjustment, models several observed results: decay in cooperation; cheap talk and leadership increase contributions.

In chapter 2 I explore game ambiguity. I show that differences in what players perceive the game to be, explain a lot of the behavior seen in economics experiments.  Most economics experiments define the payoffs, but rarely do they define what it means to win. This and other ambiguities in the definition of the experiment can lead to a wide variety of different behaviors. I argue that most participants in economics experiments perceive their situation as a game. In such situations altruism is an illogical strategy. They adopt a behavior appropriate for the known game which is most similar to the experiment rules. I tie this into Wicksteed’s Common Sense about Political Economy, arguing that he sees life as a collection of overlapping games. I follow Wicksteed’s logic that people behave differently depending on what game they think they are playing and apply it to the behavior of people in economics experiments as well as to real life situations.

In chapter 3 I build on the literature on bet-hedging that invokes the Kelly criterion. Bet-hedging has been adopted by evolutionary biologists to show that evolutionary advantage belongs to those species or subspecies with the highest geometric mean number of offspring. I argue that humans see every decision in the context of long-term growth and that they instinctively apply the Kelly criterion to decide how much of their endowment to put at risk on any bet. Using this understanding I can explain why people in public goods games donate less than everything and more than nothing to the public good. I show mathematically a direct relationship between the amount contributed to the public good and the level of belief they have that other people are like them.