Date of Award

Spring 5-15-2022

Author's School

Graduate School of Arts and Sciences

Author's Department


Degree Name

Doctor of Philosophy (PhD)

Degree Type



Market design is an area of study that focuses on number of issues encountered in a wide range of market places, utilizing game theory, mechanism design, algorithms, and laboratory research. We contribute to the literature in the two main areas of market design: matching markets and auction markets.

In many matching markets, agents meet and match over time under incomplete information. For example, in labor markets, workers and employers typically search for matches over time, holding interviews sequentially. This setting is substantially richer than a static matching environment, as the pool of participants evolves endogenously since matching decisions at any time must anticipate the distribution of options available at later dates, which are themselves determined in part by current decisions. However, the matching literature studies these dynamics in a limiting way by assuming that the matching environment is stationary and matching opportunities are ephemeral, evaporating at the next period if the match is not executed. These assumptions are not realistic and inconsistent with many dynamic matching markets. Chapter 1 and Chapter 2 are the first theoretical and experimental analysis, respectively, of dynamic matching markets without imposing the above-mentioned simplifying assumptions.

In Chapter 1, we study a two-period dynamic matching environment in which agents meet randomly and can decide whether to match early, or defer to the second period. Crucially, agents who defer retain the option to match with either partner in the second period. This novel form of “recall” captures situations where, e.g., a firm and worker can conduct additional interviews before agreeing to a contract. We find that recall has a profound impact on individual incentives and on aggregate outcomes. We show that the likelihood to match early is non-monotonic in type: while low types are not attractive, the highest types prefer to wait. Early matches thus occur between the good-but-not-best agents. The option value provided by the first-period partner provides a force against unraveling, and we show that an equilibrium with deferral exists even with participation costs.

In Chapter 2, we designed decentralized matching experiments to study agents’ early matching incentives in two sided one-to-one dynamic matching markets. The environment is designed to test our theoretical results, given in Chapter 1, as well as to capture the treatment effects of recall on agents’ behavior in dynamic matching environments. In our experimental sessions, participants engage in a sequence of decentralized dynamic matching markets with incomplete information varying agents’ ability to recall, which is either no recall or perfect recall and per period participation cost, from no cost to a small cost. We experimentally prove that recall has a significant and profound treatment effect on individual incentives and on aggregate outcomes. In the experiments, we verify our theoretical predictions showing that (i) similar and relatively high type pairs match early, (ii) the probability of matching early as a function of type is an inverse U-shaped function, and (iii) markets do not unravel even if each meeting is costly.

The last two Chapters are devoted to auction design in markets with intermediaries. In Chapter 3, we study intermediation chains in financial over-the-counter markets in which agents interact with counter-parties for lending and borrowing liquid assets with collateral under incomplete information. Given a financial network, liquid assets are allocated from initial lenders to final borrowers through a chain of intermediaries. In our model, agents with risky investment opportunities decide to invest the liquid asset or to be an intermediary and lend the liquid asset to other agents. Bilateral prices and lending decisions are determined by a sequence of security bid auctions. We show that the underlying network structure determines the level of competition, which in turn determines the split of expected surplus and the efficiency of asset allocation. We find that incomplete financial networks yield inefficient allocation, even though each lending in the equilibrium intermediation chain is efficient. Moreover, we show that equilibrium allocation from an initial lender to a final borrower does not necessarily proceed along the shortest path, which contrasts the general intuition that lenders and borrowers prefer to minimize the number of intermediaries.

Chapter 4 is a short discussion paper closely related to Chapter 3. In this last chapter, with simple examples, we show that efficiency is hard to achieve when intermediaries are not avoidable. Moreover, contrary to some results in the literature, we show that even if each seller sets an optimal trading protocol, efficiency is not guaranteed even under very simple network structures. Thus, we show that the results obtained in favor of efficiency crucially depend on some strong assumptions imposed on the underlying network structure.


English (en)

Chair and Committee

Brian Rogers

Committee Members

Ana Babus

Available for download on Wednesday, May 20, 2026

Included in

Economics Commons