Research

Abstract: Iran is the only country in the world with a legal, monetary market for kidneys.

In this paper, we study the Tehran market for kidneys, the largest organ market in Iran. With more than five years of transplant and demographic data, we show there is a substantial supply of living paid donors with a relatively high price elasticity. The average waiting time to receive a deceased or live kidney is less than one year. Family donations are extremely rare. To estimate welfare effects of the market, we build a simple dynamic model, where patients di er in their wealth levels. We argue that establishing a monetary market makes all patients better off, including those who do not purchase live kidneys, because paid donors reduce the congestion of the deceased kidneys queue. Our calibration exercise suggests that, if the U.S. legalizes a kidney market with outcomes similar to the Tehran market, the average U.S. kidney patient's quality adjusted lifetime will increase by about 7 years. In addition, this can save more than $5.8 billion in annual healthcare costs for the U.S. We discuss several ethical concerns of this market and examine them with our data.

Abstract: We study the consequences of unequal access to information among students in the Deferred Acceptance (DA) algorithm commonly used in school choice. In our model, there is a large population of students and two schools with many seats each. Students care about the school's overall quality (common value) and their idiosyncratic preferences (private value). Students are heterogeneous with respect to their idiosyncratic preferences and the information they have about the common component. A fraction of the students are informed, and the rest are uninformed (or more generally, partially informed). Unlike the informed students, the uninformed students optimally base their rankings not only on the expected utilities of the schools but also on the likelihood of being admitted to the two schools. For example, they are more likely to rank first a large school than a small school. This behavior is not present in a pure private values model. In our model, it is caused by the “admission curse”, namely that a student is more likely to get admitted to a school if it has a lower quality and hence is under-demanded by the informed students. That effect is stronger for smaller schools. We also study the consequences of providing a subset of students with priority in admissions. We show that while such priorities do not change the optimal ranking of informed students, they change the optimal rankings of the uninformed students since they reduce the admission curse. We show that priorities can backfire: offering a group of students priority in some schools, particularly in lower-quality schools, can reduce the equilibrium welfare of those students. 

Abstract:  I study an allocation problem where buyers’ values consist of a common and a private component and buyers are inherently heterogeneous in their access to common value information. This problem covers a wide range of allocation problems including auctions for financial assets and art and collectable items. Milgrom and Weber (1982) show how winner’s curse distorts bidders’ incentives to bid in these environments. I show that when agents suffer at different level from winner’s curse due to information inequality, equilibrium of the second price auction is ex-ante inefficient. I use a model where buyers might have either a high or a low private value and their information structure is in form of connected partitions around the common value. Buyers are then heterogeneous in precision of their partitions. I show that informationally inferior buyers with high private value lose to low value buyers with better information in an equilibrium in undominated strategies. I then assess that introducing rationing by putting a cap on bids in a second price auction can improve both ex-ante efficiency and expected revenue. A price cap with rationing mitigates winner’s curse for some of the high value buyers and raises their equilibrium bid. My analyses provide a justification for use of Or-Best-Offer auctions in the real world.

Abstract:  A less-understood feature of financial markets is their fragmentation, which is the same security being traded in different venues and markets and at significantly different prices and speeds of execution. Earlier studies mostly attribute such fragmentation to the aspects directly related to the design of the markets, such as their degree of transparency or the mechanisms used for price discovery. However, here, we show that if competitive dealers offer different prices for different levels of trade immediacy, two markets endogenously emerge: An illiquid market with zero bid-ask spread (match-making) and a perfectly liquid market with a positive spread (market-making). The model attributes the recent drop in the bid-ask spread after the enaction of regulations that increased the dealers' cost of liquidity provision to a shift from market-making to match-making.

Project Summary: Advertising positions are sold in real time auctions upon each search. Online platforms such as Microsoft Bing search engine use the information on user's demographics and history to offer targeted bidding to the advertisers. Targetted advertising improves efficiency and value of advertising. However, if the user market is highly segregated among competitors, targetting weakens competition in auctions. We theoretically found the relationship between effect of targetting on revenue and correlation advertisers' values across user segments. 

We used the data from advertising  auctions in the Microsoft Bing search engine. Given the high-frequency and high dimensional nature of the data, we benefitted from implementaion of an unsupervised learning algorithm designed for the advertising auction data to find clusters of competing advertisers. We found a high variation of market segregation among competition advertisers across different keywords and assessed that reducing targeting helps boosting revenue in auctions where users are highly segregated based on their value for the advertiser.