Previous work has found that in social dilemmas, the selfish always free-ride, while others will cooperate if they expect their peers to do so as well. Outcomes may thus depend on conditional cooperators� beliefs about the number of selfish types. An early round of the game may be played anonymously, so that contributions cannot be traced back to particular individuals. By protecting low contributors from potential sanctions, this encourages selfish types to reveal their true preferences in their play. We offer a simple model illustrating when revelation of types can increase contributions, and when only an anonymous game can separate types. As a proof of concept, we run a laboratory experiment involving a two-stage public goods game with an exclusion decision between stages. An anonymous first stage led to significantly higher stage-two cooperation than a revealed first stage, a slower decline across the 15 repetitions, unusually high final-stage contributions relative to previous work, and greater profits. Statistical analysis shows that the anonymous first stage reduced uncertainty about types, and this preserved cooperation and led to greater efficiency. Our results suggest that customs such as anonymous church donations may play an important role in building social trust.
