Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Confessions after failures are socially desirable. However, confessions also bear the risk of punishment. In a laboratory experiment I examine how confessions work. I analyze whether the willingness to punish harmful failures depends on how the harmed party has learned about the outcome. The harmed party can learn about the outcome via random detection or self-report by the performer. There are two major findings: first, confessions are a powerful instrument: punishment for confessed failures is less likely than for randomly detected failures. Second, confessions are much more likely to occur if there is no punishment.