Second-order beliefs in reputation systems with endogenous evaluations - an experimental study

We investigate a repeated public good game with group size two and stranger matching. Contributions are public information and each participant evaluates her partner's contribution. At the beginning of each period, participants receive information regarding the evaluation of the previous period. Evaluations are subjective judgments, hence our reputation system allows for some degree of noise. There are two information treatments: Each participant receives information either about her partner's evaluation, or about her own and her partner's evaluation. The results show that although participants condition their contributions on their partners' evaluations, this information alone is insufficient to raise contributions. Only if participants also know their own evaluations, we find an increase in contributions.