Research

Trade Delay in Bargaining over Lemons

Abstract: A seller holds a good that may be of either high or low quality, and a buyer receives an exogenous imperfect private message before bargaining begins. In a benchmark with an uninformed seller, the Coase conjecture emerges and is driven by two key properties: the market-clearing price is invariant to the seller’s deviation, and the seller becomes increasingly pessimistic after repeated rejections. By contrast, when the seller privately observes quality, prices serve not only as offers but also as signals that affect the buyer’s expected valuation. When no message can reveal the seller to be of the low type, deviations can induce sufficiently pessimistic off-path beliefs, so the off-path market-clearing price falls below the on-path one. This breaks the invariance of the market-clearing price and makes early market clearing suboptimal. When some message reveals the seller to be of the low type, the high-type seller knows with certainty that the buyer is optimistic, which breaks the seller-pessimism property. Both cases lead to a breakdown of the Coase conjecture and to trade delay in the limit. The paper thus provides a unified account of how lemons-market valuation and second-order uncertainty break the Coase conjecture through distinct channels.

Accuracy Discrimination in Personalized Product Recommendation

Abstract: I examine how a platform designs personalized product recommendations with discriminatory accuracy based on consumers’ preference intensities to maximize the probability of trade. The platform can increase recommendation accuracy for a consumer at the cost of the consumer’s privacy and higher endogenous prices set by third-party firms. When consumers’ types are observable, the platform’s optimal mechanism is perfectly misaligned with consumer preferences: it offers accurate recommendations only to “picky” consumers who benefit the least, thereby minimizing ex post consumer surplus. In contrast, when the platform must screen consumers, it behaves as if maximizing ex post consumer surplus by allowing consumers to self-select their most preferred accuracy levels. The results highlight the importance of requiring platforms to obtain consumers’ consent before providing personalized services, even when the platform’s interests are partly aligned with those of consumers.