Strausz, Roland (Humboldt University Berlin)
In mechanism design with (partially) verifiable information, the revelation principle holds if allocations are modelled as the Cartesian product of outcomes and verifiable information, giving rise to evidence-contingent mechanisms. Consequently, incentive constraints characterize the implementable set. The revelation principle does not hold when an allocation is modelled as only an outcome so that mechanisms are non-contingent. Yet, any outcome implementable by an evidence-contingent mechanism is implementable by a non-contingent mechanism, provided it can both extend and restrict reporting information. A type-independent bad outcome implies the latter property.
revelation principle; mechanism design; verifiable information