Cooperation Among Strangers with Limited Information About Reputation

Abstract
The amount of institutional intervention necessary to secure efficiency-enhancing cooperation in markets and organizations, in circumstances where interactions take place among essentially strangers, depends critically on the amount of information informal reputation mechanisms need transmit. Models based on full backward induction find that the necessary information is very large and recursive in nature, implausibly so outside of circumstances where there is a formal institution to process it. Models that relax backward induction find that the information demands may be quite modest. The experiment we present indicates that subjects condition their cooperation on information about their partners' immediate past action, while additional, recursive information about the partners' previous partners' reputation triggers and additional conditional response. Taken together, our results suggest that reputation is a more robust enforcement mechanism for cooperation than standard models suggest; informal stictures that punish cheaters and those who fail to punish cheaters is sufficient to generate substantial cooperation even when the benefits from cooperation are relatively modest.