Attribution Bias, Blame, and Strategic Confusion in Punishment Decisions∗

Disentangling the roles of intentional effort and luck is critical when evaluating decisions made by others under uncertainty. Previous research shows that this can be a very difficult task. In a principalagent experiment where an agent’s effort is always perfectly observable, we find an undue influence of luck on punishment decisions made by principals. Our experiment identifies attribution bias as a mechanism behind this behavior—lucky agents are perceived to exert more effort than unlucky ones. We test the sophistication of both the principals and agents about this bias. Agents show positive willingness to pay for control over the principal’s access to information about luck but principals show no demand to eliminate news about the agent’s luck from their information set. We replicate these findings with third parties tasked with punishing agents, and consider policy implications for what information should be available to principals. JEL Classification: C92, D63