Corporate board composition, protocols, and voting behavior: experimental evidence

We model experimentally the governance of an institution. The optimal management of this institution depends on the information possessed by insiders. However, insiders, whose interests are not aligned with the interests of the institution, may choose to use their information to further personal rather than institutional ends. Researchers (e.g., Palfrey 1990) and the business press have both argued that multiagent mechanisms, which inject trustworthy but uninformed "watchdog" agents into the governance process and impose penalties for conflicting recommendations, can implement institutionally preferred outcomes. Our laboratory experiments strongly support this conclusion. In the experimental treatments in which watchdog agents were included, the intuitionally preferred allocation was implemented in the vast majority of cases. Surprisingly, implementation occurred even in the absence of penalties for conflicting recommendations.