RESEARCH

Endogenous Market Segmentation for Lemons

Job Market Paper, Nov. 2008

I explore the possibility of endogenous segmentation in a decentralized market where asymmetric information about the quality of goods may result in only low-quality goods trading (the lemons problem). Endogenous segmentation alleviates information asymmetry, potentially improving market efficiency, without employing costly signalling or screening devices. The incentives for sellers to sort themselves are endogenously generated by buyers' behavior, and vice versa. I show that endogenous segmentation is supported by a monotone market arrangement: if a market is segmented into multiple submarkets, higher-quality submarkets entail more quality uncertainty and attract relatively fewer buyers. The result has implications for applied work, including the structure of multiple marketplaces or platforms, the informativeness of costless advertisement for experience goods, and the role of non-binding list prices in decentralized markets.
JEL Classification Numbers: C72, D82, D83.
Keywords : Lemons Problem; Market Segmentation.

The Coase Conjecture with Incomplete Information on the Monopolist's Commitment

Forthcoming in Theoretical Economics, first draft: Apr. 2007

A key to the Coase conjecture is the monopolist's inability to commit to price, which leads consumers to believe that a high current price will be followed by low future prices. This paper studies the robustness of the Coase conjecture with respect to these beliefs of consumers. In particular, there is uncertainty over whether the monopolist is committed to a price (i.e. she may be a commitment type). Consequently, consumers are no longer certain that the price will change over time. I consider two kinds of commitment types. A behavioral commitment type charges an exogenously given price, while the rational commitment type optimally chooses a price. I show that the Coase conjecture is robust with regard to uncertainty over the monopolist's commitment. When the probability of behavioral types is sufficiently small, as in the original Coase conjecture, the monopolist earns the competitive profit. When the probability of behavioral types is positive, unlike in the original Coase conjecture, there will be positive delay. But the delay disappears as the probability approaches zero. When the commitment type is rational, unless the probability of the commitment type is sufficiently high, both the normal and committed monopolists charge the competitive price, and thus there is no delay.
JEL Classification Numbers: C72, C78, D82.
Keywords : Coase conjecture, reputational bargaining, rational commitment

Communication with a Possibly Honest Expert

Joint with Jonathan Pogach, this version: Nov. 2008 (first draft: Mar. 2008)

This paper studies a variant of the standard cheap talk game in which the sender is possibly honest. The honest sender always recommends the ex-post optimal policy to the receiver. The presence of the honest type introduces intrinsic value for messages, necessitating a different characterization than Crawford and Sobel (1982). We characterize two classes of equilibria that satisfy two natural properties. In addition, we compare this model to a model in which the sender may be an advocate. The advocate may manipulate information, but only for the sake of the principal (i.e. the sender with no bias). We show that the receiver is often better off with a possibly honest sender than with a sender who may be an advocate.
JEL Classification Numbers: C72, D82, D83.
Keywords : Cheap talk, Honesty, Advocacy

Perfect and Imperfect Learning in Bargaining

Work in progress (preliminary and incomplete draft available)

I study a bilateral bargaining problem with one-sided incomplete information where the uninformed player makes all the offers and learns the opponent's private information over time. I consider two kinds of learning, perfect and imperfect. Learning is perfect if private information may be revealed completely, while it is imperfect if the uninformed player receives only a noisy signal. I show that in the limit as bargaining friction becomes negligible, (1) there is delay with perfect learning if and only if learning is fast, whereas (2) delay disappears with imperfect learning, no matter how fast and informative learning is. I also show that the result regarding imperfect learning does not imply that a noisy signal is ignored in bargaining. It amplifies the effect of perfect learning on the bargaining outcomes, if they are present together.
JEL Classification Numbers: C72, C78, D82, D83.
Keywords : Delay in bargaining, Learning in bargaining

Ambiguity, Informational Size and Efficient Auctions

Dec. 2006 (preliminary and incomplete draft available)

This paper studies an interdependent value auction under ambiguity, where ambiguity is represented by a commonly known set of priors. I provide conditions under which an augmented Vickrey auction with conditional values yields an ex post efficient allocation. In addition, I suggest the definition of informational size under ambiguity. I show that the necessary payment to bidders for efficiency is arbitrarily small if agents are sufficiently informationally small.