An Experimental Analysis of the Impact of Reputations

Texas A&M Research Foundation, College Station TX

Investigators

Abstract

Reputation plays an important role in long term relationships. Long lived economic agents (such as shops, banks, government etc.) care about their reputations concerning whether they are honest, kind, ruthless or unforgiving for many possible reasons. Since reputations have the ability to profoundly effect market behavior, economists have studied reputations and their possible roles in a variety of ways. In the now classic approach to study reputation, pioneered by Kreps and Wilson (1982) and Milgrom and Roberts (1982), a sequence of short run players interact with a long run player whose preferences or type is unknown to them. The short run players observe the previous choices of the long run player. The availability of this information allows the long run player to build a reputation about her type. This literature suggests that the long run player's ability to build a reputation may allow the long run player to obtain a higher payoff than what would have been obtained by her had her previous actions not been observable (or her type been known). Therefore, reputation is thought to be good. Recent work by Ely and Valimaki (2003) suggests that reputations need not be beneficial for the long run player. That is, if the previous actions of the long run player are observed by successive short run players then the long run player may do worse than if such information was not available to the short run players. Therefore, reputation is considered bad. This project experimentally investigates whether the possibility of having reputations impacts behavior in the manner suggested by theory. Investigating this experimentally is important for at least three reasons: (1) theory gives mixed answers about the effect of reputation, (2) it is not easy to isolate these effects by looking at real-world economic data, and (3) real world data might not exist. A reputation framework which allows for both the good and the bad reputation result to emerge is designed. In this framework the good reputation result emerges in a specialization of the environment required for the bad reputation result. Experimentally observed behavior and outcomes are compared when short run players are informed about the previous choices of the long run players and when they do not have this information. This provides an explicit test of the impact of reputations. Intellectual Merit: The research demonstrates the significance of having reputations. Theoretical work suggests that in some environments it may benefit the long run agent and in others it may hurt her. This research clarifies the behavioral relevance of each of those predictions. The unified reputation framework that is developed provides the first experimental test of the bad reputation prediction. Broader Impact: Knowing whether reputation is good or bad allows researchers to more efficiently price services that provide reputations (if reputation is good). It may also point toward the need to design suitable mechanisms and institutions which help overcome the effect of reputation (if reputation is bad). The broader impact of the proposed research also includes guidance regarding when reputation may be hurting the market. It further provides support for graduate students doing research on related topics.