The value of information disclosure under local learning. The case of fixed types.

A large population plays a two-period sequential common agency game. Agents are long lived, while principals are short lived. Preferences and technology are additively separable in time and time independent. At the onset, agents are matched in pairs under private information of individual types. At the end of the first period, in each pair the principal can disclose membersÕ reports, in which case members remain together in the second period, or conceal information, in which case members are randomly rematched and in the second period their type remains private information. We show that an equilibrium exists in which information disclosure is efficiency enhancing. Remarkably, information disclosure would have zero value if reassembling agent pairs was not an option, as in the standard one agency literature.