Mechanism Design with Ambiguous Transfers

This paper introduces ambiguous transfers to study the (partial) implementation problem. We show that under all profiles of utility functions, any efficient allocation rule is implementable via an individually rational and budget-balanced mechanism with ambiguous transfers, if and only if the Beliefs Determine Preferences (BDP) property is satisfied by all agents. This property holds generically when there are at least two agents. It is weaker than the necessary and sufficient conditions for implementation via a Bayesian mechanism. Therefore, ambiguous transfers may provide a solution for situations where Bayesian mechanism design is impossible. In particular, efficiency becomes achievable generically in two-agent problems, which contrasts the impossibility results in the literature