Endowment heterogeneity, incomplete information & institutional choice in public good experiments

We study centralized and decentralized enforcement in social dilemmas with income inequality and incomplete information. Subjects are randomly assigned different endowments and, across treatments, these endowments can either be observed or not. After gaining experience with peer punishment and a simple central authority, groups voted on their preferred enforcement institution. Under complete information (endowments observed), most groups voted for peer punishment. Under incomplete information (endowments unobserved), most groups voted for central authority, and results suggest this preference was largely driven by subjects with lower incomes. Since free-riding could not be targeted when incomes were not observed, subjects with larger incomes tended to under-contribute, encouraging groups to self-impose central authority.