Caste, efficiency and fairness with public goods and common pool resources

We report results from public good and common pool experiments that directly reveal economic surplus foregone or destroyed by failure of cooperation. We compare the behavior of caste-uniformed West Bengali villagers to behavior of undergraduates at American universities as well as behavior of caste-informed villagers. By design, the public good game and the common pool game in our study are payoff equivalent. Thus behavior is predicted to be similar across the two games by models of unconditional social preferences, regardless of whether that behavior is spiteful or altruistic.

In contrast, reciprocal preference theory predicts more altruistic (or less spiteful) behavior in public good than common pool games. We test hypotheses derived from the alternative models. When we withhold information about caste of other subjects in a session in India, we find that the efficiency of play in these caste-uninformed treatments with villagers is not significantly different from the efficiency of play observed in similar experiments with student subjects at American universities; this is true in both the public good game and the common pool game. Thus an overall cultural difference in level of cooperation across countries is not observed.

Our results confirm the empirical failure of the isomorphism of public good and common pool games implied by the unconditional social preferences models. The data are consistent with a theory of reciprocal preferences. Different outcomes are obtained when we provide caste information to subjects in India.

Our data indicate that the caste system has a strong effect on how social dilemma situations among Indian villagers are resolved. We find strong within-group favoritism that supports cooperation among high caste individuals but not among low caste individuals. Efficiency is highest (lowest) in homogenous groups with high (low) caste subjects. Disaggregation of the results into the behavior of first movers and second movers reveals that when the second mover is from the low caste, high caste first movers cooperate the least in the public good game but in the common pool game the least cooperative come from the low caste first movers. Uncooperative behavior towards in-group members is observed when low caste second movers withdraw a significant amount of contributions to the public good made by low caste first movers whereas high caste second movers exhibit the most (in-group) cooperative behavior in both games.