An Experimental Comparison of the Fairness and Learning Hypotheses

点击翻译按钮获取中文摘要

The provision of public goods has long been a important problem in microeconomics. Roughly, a public good is any item which can benefit a large number of individuals simultaneously and which it is difficult to exclude individuals from consuming. Examples of public goods might include public parks, clean air, or national defense. The central difficulty with the provision of public goods is the free rider problem; if it is impossible to exclude an individual from the benefits of a public good, he/she has little incentive to pay for provision of the public good if others are willing to provide it. Theorists have proposed a number of mechanisms designed to alleviate the free rider problem. While all of these mechanisms work in theory, they are no better than their underlying assumptions about individuals' behavior. Experimental economics has played an invaluable role in the study of public goods by providing controlled empirical evidence on behavior in public goods problems. It is well known that public goods and bargaining experiments yield results which are at odds with standard microeconomic theory. Two classes of theory, based on interpersonal utility functions (`fairness`) or on boundedly rational learning models, have been proposed to explain these results. Neither explains all of the experimental data or clearly outperforms the other. Our experiments explore whether subjects' behavior in a public goods game can best be explained by the fairness hypothesis, the learning hypothesis, or a hybrid model which combines them. Our experiments are based on the Minimum Contributing Set public goods game developed by van de Kragt, Orbell, and Dawes. We consider three basic variations. One is designed to push inexperienced players towards the subgame perfect equilibrium, one is designed to push players toward the `fair` equilibrium, and one is designed with either outcome as a plausible outcome. By having subjects participate in one of the first two games before playing the third, we control the subjects' prior experience entering the third game. Examining differences in subject behavior between sessions with differing prior experience allows us to clarify the roles of learning and fairness in generating subjects' behavior.