Can Agents Be Trusted? Experimental Evidence From The Monitoring Game

Abstract

Most principals can monitor their agents, but monitoring is usually costly and imperfect. This paper reports the experimental results of a Monitoring Game. In this sequential game, each principal decides whether to monitor an agent that he is randomly paired with and then the agent decides whether to cheat or be honest. Monitoring is costly, but it increases the probability that the agent will get caught. The experiment shows that cheating is commonplace, although most participants do not cheat if they are monitored. Additionally, cheating is more common when neutral terms are used in the instructions.