Observability of partners' past play and cooperation: Experimental evidence

The observability of partners' past play is known to theoretically improve cooperation in an infinitely repeated prisoner's dilemma game under random matching. This paper presents evidence from an incentivized experiment that reputational information per se may not improve cooperation. A structural estimation suggests that a certain percentage of players act according to the "Always Defect" strategy, whether or not the reputational information is available. The remaining players adopt available cooperative strategies: specifically, the tit-for-tat strategy when reputational information is not available, and a strategy that conditions on the matched partner's past play when reputational information is available.