Obtaining costly unverifiable valuations from a single agent

摘要

A principal needs to elicit the true value of an object she owns from an agent who has a unique ability to compute this information. The principal cannot verify the correctness of the information, so she must incentivize the agent to report truthfully. Previous works coped with this unverifiability by employing two or more information agents and awarding them according to the correlation between their reports. We show that, in a common value setting, the principal can elicit the true information even from a single information agent, and even when computing the value is costly for the agent. Moreover, the principal’s expense is only slightly higher than the cost of computing the value. For this purpose we provide three alternative mechanisms, all providing the same above guarantee, highlighting the advantages and disadvantages in each. Extensions of the basic mechanism include adaptations for cases such as when the principal and the agent value the object differently, when the object is divisible and when the agent’s cost of computation is unknown. Finally, we deal with the case where delivering the information to the principal incurs a cost. Here we show that substantial savings can be obtained in a multi-object setting.