Bayesian Persuasion by an Informed Mechanism Designer

We investigate how an informed designer maximizes her objective when facinga player whose payoff depends on both the designer's private information andon an unknown state within the classical quasilinear environment. Thedesigner can disclose arbitrary information about the state via Bayesianpersuasion and adopt arbitrary mechanisms. We characterize theRothschild-Stiglitz-Wilson (RSW) mechanism and identify three channels forachieving separation. While disclosing inefficient information is essential,providing a bonus for participation and randomization are supplementary. Theequilibrium is unique and robust under weak assumptions. Our results canprovide rationales for many phenomenons in practice