Shengwu Li, Obviously

My other student on the job market, Shengwu Li, is a powerful theorist (with a strong interest in market design) as well as a behavioral & experimental economist.

His JM paper, "Obviously Strategy-Proof Mechanisms" considers the classical question on what makes it easier for people to understand whether a mechanism is strategy-proof. For example, while people in general know not to overbid in an ascending auction, in a strategically equivalent second price sealed bid auction participants in experiments bid both above and below their private value.

Shengwu says that a strategy obviously dominates another if, basically, at the earliest time in the game tree when the the two strategies deviate: the best outcome of the deviation can be no better than the worst outcome under the original strategy. So, to understand whether a strategy obviously dominates another, agents have to only compare the best in a set of possible outcomes from deviating to the worst outcome under the original strategy. This is, of course, intuitively, much simpler than computing expectations over outcomes.

The beauty of Shengwu's paper is that he then finds two equivalences for mechanisms that are obviously strategy-proof (OSP), that is that implement the dominant strategy equilibrium in obviously dominant strategies. The first equivalence concerns a specific model of bounded rationality: Consider an agent who only knows at each moment what the set of possible outcomes can be, but not how those outcomes depend on actions of other players. Shengwu shows that a strategy is obviously dominant in a game if and only if it is dominant in every game the boundedly rational agent cannot distinguish from the current game. His second equivalence concerns a classical mechanism design problem: Consider a mechanism designer who lacks commitment power, but has partial commitment. Specifically, the planner can bilaterally communicate and commit vis-a-vis an agent to actions observable to that agent. Shengwu then shows that an allocation rule can be implemented by an OSP mechanism if and only if there is a bilateral commitment game that implements it and in which each agent has a dominant strategy. The paper therefore makes a nice link between a model of cognitive limitations and one in which the mechanism designer has only partial commitment.

Shengwu's paper is very complete, as it has two more parts. The first is an experiment where he confirms for the standard second price versus the english auction but also for a novel auction format that implementations via OSP generate "better" behavior than the implementation in only dominant strategies. Shengwu then shows that this is also the case when comparing an OSP implementation of a serial dictatorship to a strategy-proof implementation. Finally, Shengwu provides for a wide class of situations a characterization of the set of OSP-implementable allocation rules.