William Duan

University of British Columbia · Economics · 2026–27 job market

Fields

Applied Micro, Theory

Job market paper

Selling Mechanisms under Misspecified Learning

Buyers in a repeated selling mechanism hold a misspecified parametric model of their environment and learn by Bayesian updating from allocation outcomes alone. Because the mechanism generates the outcomes buyers learn from, its design shapes what they come to believe. This paper asks when a seller can choose a mechanism under which a given, possibly misspecified, belief is sustained by learning, and characterizes such mechanisms. With a single buyer, conditions on the buyer's model alone guarantee that a unique mechanism draws every initial belief to the target belief. With two buyers, local stability is established under conditions on the buyers' models and the value distribution, and a computational method identifies mechanisms that extend the basin of attraction to every belief below the sustained one. The main revenue implication is that value-dependent beliefs can allow the seller to exceed the Myerson benchmark, whereas value-independent beliefs cannot.