Maxmin mechanism in a simple common value auction

This paper models the data generating process of common value auctions in a parameter-estimation way, known as the classical approach in statistical inference. Viewing the true value of the object as a parameter that nobody ever knows, we let our value function be the average of the individual estimations (signals) of all agents, a robust estimation of the parameter. Under this simple value function, we select almost uniquely the mechanism that gives the seller the largest portion of the true value in the worst situation among all the direct mechanisms that are feasible, ex-post implementable and individual rational. Our Maxmin mechanism, which randomly assigns the object to one agent, provides the seller (n−1)/n of the true value when expected revenue is concerned, where n is the number of the agents.