Optimal selling time with evolving private information

Abstract

We study the problem of when to sell one indivisible object to a buyer whose value from immediate allocation follows a privately observed Markov process. Incomplete information transforms what would otherwise be a standard Markovian optimal stopping problem into a dynamic mechanism design problem. We characterize dynamic incentive compatibility by the envelope formula and integral monotonicity. We first solve the relaxed problem that maximizes the seller's revenue over feasible stopping rules while ignoring integral monotonicity, and then give conditions under which the optimal rule of the relaxed problem is dynamically implementable. Under dynamic single crossing and stochastic monotonicity, the optimal rule of the relaxed problem takes a threshold form, and incomplete information weakly delays sale relative to complete information along every realization. We also give a condition for the optimality of a one-step look-ahead rule. Examples illustrate how the informational state depends on the type process.