Vernon Smith Hall (formerly Metropolitan Building), #5043
May 29, 2015, 10:00 AM to 06:30 AM
This dissertation combines game theory with controlled laboratory experimentation to better understand the performance of two-sided matching markets. Two main questions are addressed. First, do market participants strategically misrepresent their preferences in these environments? Second, in markets with multiple equilibria, which equilibrium is more likely to be implemented?
Chapter 1: Information Effects in the Dynamic Gale-Shapley Mechanism: An Experimental Study
We investigate strategic behavior in a decentralized two-sided matching market in which we vary the level of information that is available about other agents' preferences. We find that this manipulation has implications for equilibrium selection, but does not affect the stability of final outcomes. This result is due almost entirely to a change in the behavior of one side of the market. In complete information environments, "proposers" often fail to play their dominant strategy and instead skip down their preference lists when making offers. This skipping behavior is sophisticated in the sense that proposers take into account how they are ranked by the other side of the market, forgoing offers to preferred match partners who are unlikely to accept. Our findings highlight the important role that behavioral biases can play in these environments.
Chapter 2: Truncation Strategies in Two-Sided Matching Markets: Theory and Experiment
We investigate strategic behavior in a centralized matching clearinghouse based on the Gale-Shapley deferred acceptance algorithm. To do so, we conduct a laboratory experiment to test whether agents strategically misrepresent their preferences by submitting a "truncation" of their true preferences. Our experimental design uses a restricted environment in which subjects always have a best response in truncation strategies. We find that subjects do not truncate their preferences more often when truncation is profitable. They do, however, truncate less often when truncation is dangerous - that is, when there is a risk of "over-truncating" and remaining unmatched. Our findings suggest that eliminating profitable opportunities for strategic behavior may not be sufficient to induce participants to report their true preferences.
Chapter 3: Equilibrium Selection in Two-Sided Matching Markets
We investigate equilibrium selection in the context of stable matching mechanisms. By focusing on symmetric equilibria, we illustrate the tension between the familiar concepts of payoff-dominance and risk-dominance in this strategic environment. In particular, for a large class of markets, symmetric equilibria in weakly dominated strategies will risk-dominate the symmetric equilibrium that produces the optimal outcome for one side of the market. However, this construction is sensitive to institutional details such as the assumptions imposed on agents' cardinal preferences. We also present a simple behavioral model that provides useful comparative statics predictions for a more restricted strategic environment. Even under conditions that are conducive to the emergence of the symmetric payoff-dominant equilibrium, the model argues for the selection of asymmetric equilibria in which a subset of agents free-rides on others' optimal behavior. Taken together, our results show the limitations of equilibrium selection theories in this context.