Economists have long argued that markets play a key role in aggregating information; what individuals know affects the prices they will pay and charge. How this process works varies from market to market. For example, by analyzing auction markets economists have been able to develop a strategic foundation for information aggregation.
In many markets, however, trading is not conducted through centralized auctions. Instead, prices are often determined through local interactions between a small number of agents, for example, in decentralized markets such as those for labor and over-the-counter securities. This award funds research in economic theory with a goal of understanding how decentralized markets aggregate information that is dispersed among market participants.
Decentralized markets are modeled with a search framework. Pairs of traders bargain over prices and search for other traders if they do not reach an agreement. The analysis shows that information aggregation in decentralized markets differs fundamentally from information aggregation in auctions. The PI is building and analyzing two different models of this process.
In the first, there is a decentralized setting with a single buyer. The buyer seeks to procure a service, repair or cure, and contacts sellers for price offers. The cost of the service depends on private characteristics of the buyer, such as the complexity of the repair or the health status. The number of sellers contacted is endogenous and depends on the private information of the buyer. Moreover, being contacted reveals information to the seller. Sellers also receive noisy information about the buyer's characteristics through inspection or diagnosis. While each individual seller's information is imperfect, all sellers jointly posses sufficient information to determine the buyer's characteristics. Whether and how this dispersed information is aggregated is the question examined. Preliminary results show that information aggregation in the decentralized market is much harder (in a sense made precise later) than in standard auctions. This continues to be true even when the standard auction setting is extended to permit the buyer to solicit bidders.
The second model analyzes the flow of information in decentralized settings with many buyers and sellers who trade a single type of good. Aggregate demand for the good (scarcity) is unknown. Buyers and sellers learn individually about the true market condition by interacting with other traders. As in the first sub-project, the focus of the analysis is on the transmission and aggregation of dispersed information. If information is aggregated, the price reflects the true scarcity of the good, that is, the price reflects the good's true economic value.
The broader impacts include graduate student training and a better understanding of key features of important markets like the job market.
Overview: Decentralized Markets with Uncertainty This research project studied the ability of decentralized markets to function well in the presence of uncertainty by aggregating information that is dispersed among market participants. Examples for decentralized markets are the markets for labor, housing, and over-the-counter financial assets. In these markets, prices are not usually determined by a central market clearing institution. Instead, prices---and the terms of trade more generally---are negotiated between individual traders. Moreover, these markets can be quite intransparent, meaning that the market participants do not observe much beyond what happens in their immediate interactions. Therefore, uncertainty – for example, uncertainty about the state of the market or the value of the asset that is being traded – may affect decentralized markets much more than centralized markets and the presence of uncertainty may in particular reduce the efficiency of the resulting allocation. Why Some Markets are Intransparent In this project, I developed new tools for the analysis of decentralized markets with uncertainty that I then applied to gain insights into when decentralized markets are nevertheless working well --- and when they are not. For example, I studied whether sellers of assets have incentives to choose transparent or intransparent trading institutions in the project "Auctions in Markets: Common Outside Options and the Continuation Value Effect" (with Gabor Virag) For that paper, I developed a new model of auctions embedded in larger market contexts and I demonstrated that a well-known principle of auction design theory may fail. The "linkage-principle" stipulates that sellers would voluntarily choose to make as much information available to the buyers as possible. However, as demonstrated in the paper, this is not true when auctions are part of a larger market place. Then, sellers may have incentives to withhold information from buyers that could improve the value of their outside option. Thus, intransparent marketplaces may arise endogenously through the choice of specific traders who have incentives to impede the spread of information, thereby making it harder for the market to aggregate information and exacerbating the negative effects of uncertainty. Information Aggregation is more Difficult in Decentralized Markets Similar insights emerge in other papers: Intuition developed for centralized markets or for bargaining of traders in isolation needs to be qualified when applied to decentralized markets. As another example, I studied adverse selection in decentralized markets, that is, the effect of uncertainty about the value of the object that is being traded (e.g., the value of a used car or the productivity of a worker). An important part of decentralized markets is the contact process or the search for trading partners. With adverse selection, being contacted by a trader may contain information about the value of that trader’s good. In two papers, "Search with Adverse Selection" and "Auctions with Bidder Solicitation" (with Asher Wolinsky), I have shown that this may worsen the adverse selection problem and lead to significant inefficiencies in markets in which traders choose to contact others. Thus, decentralized markets may not be able to deal as effectively with problems of adverse selection as more centralized marketplaces. In particular, under conditions in which a more centralized market place (i.e., an auction without the contacting problem) is aggregating information well, a decentralized market place (i.e., search) may fail to aggregate any information at all.
