Random Signals and Strategic Confusion in Punishment Decisions∗

Understanding how effort and luck map to observed success or failure can be very difficult. However, people often have to make important judgments about punishments and/or rewards from that vantage point. In an experimental context where this judgment should be very easy to make, we study how effectively principals can punish the low effort of agents while ignoring random information. The innovation of this paper is to ask whether principals and agents are sophisticated about this “outcome bias.” Do principals ignore information that could lead them to excuse shirking or punish hard work? Do agents hide bad luck and advertise good fortune? We find substantial outcome bias, replicating previous work. Agents can lower their punishment through control of the principal’s information set. Principals, in contrast, show little or no demand for commitment against outcome bias. JEL Classification: C92, D63