Research

Abstract: In a principal-agent interaction I characterize when the agent can benefit from the ability to covertly acquire hard evidence. The same condition determines whether the principal's payoff is monotone or U-shaped in the cost of evidence acquisition. I derive implications for aggregate welfare.

Online appendix. 

Abstract: I consider a model of monopoly insurance contracting where the consumer has access to endogenous, costly evidence of his risk type (such as a test result). I characterize when the consumer is worse off if the insurer is allowed to condition contracts on evidence and when the ability to contract on evidence leads to a Pareto improvement. I also show that allowing contracting on evidence has the potential to either increase or decrease aggregate welfare. I fully characterize the optimal mechanism in the monopoly case, which features 'low powered’ contracts. I compare the results to an analogous setting with perfect competition: Under perfect competition, when evidence acquisition costs are low, the ability to contract on evidence is always Pareto improving. For intermediate costs, I uncover a new source of unraveling. The results are relevant to policy debates over the use of genetic information in health and life insurance.

Abstract: I consider a broad class of economic environments where a principal contracts with an agent under adverse selection and the agent can credibly disclose information to the principal. I show that there is an equilibrium that interim Pareto dominates the equilibrium without evidence if and only if the optimal mechanism without evidence assigns the outside option to a set of types satisfying a ‘gains from trade’ property. The results apply to a range of economic environments including insurance markets, financial markets and goods markets with quality-based price discrimination. 

Supplemental material

Abstract:  I define Weak Implementation under incomplete information. A social choice set is weakly implementable if the set of equilibrium outcomes of some mechanism is a nonempty subset of the social choice set. Weak implementation is a more natural objective than either full or partial implementation in many cases. I show that there are social choice sets where every subset can be weakly implemented yet the set cannot be fully implemented. I give a complete characterization of the weakly implementable social choice sets under a weak restriction on preferences. As a corollary, I show that in independent private values environments the set of interim efficient social choice functions is weakly implementable whenever it is partially implementable. Using similar techniques I also extend existing characterizations of full implementation.

Abstract: A target equilibrium in a game of complete information is called robust to incomplete information when all nearby games of incomplete information have equilibria which generate similar ex-ante distributions over actions as the distribution generated by the target equilibrium. Robustness to canonical elaborations considers only nearby games with a special structure. I show that robustness to incomplete information and robustness to canonical elaborations are equivalent when the equilibrium concept in the nearby incomplete information games is agent normal form correlated equilibrium. 

Online Appendix

Abstract: Many settings exhibit matching between two sides of a market and also disclosure of some payoff-relevant information gathered at a cost. An important application is the recent adoption of optional SAT in college admissions. To capture the salient features of this and other applications, we develop a model in which heterogeneous students can invest at a cost in learning an attribute that is payoff-relevant for colleges, then decide whether to disclose it, and finally match with colleges that differ in their quality. We show that equilibria entail more investment when disclosure is voluntary than when it is mandatory. Our main result is a complete description of who benefits from voluntary investment and disclosure. In particular, we show that an interval of low types of students strictly prefers mandatory to voluntary disclosure, while the opposite is true for high types if colleges can be divided into top and non-top schools. We discuss several implications for the adoption of optional SAT in college admissions. Since our framework subsumes many applications, we also analyze the matching between entrepreneurs and investors, where there can be both investment and disclosure externalities.

Abstract: Startups are a key driver of innovation but are frequently the target of acquisition by incumbent firms. How does this affect the types of products that are developed? We study a model in which an entrant decides whether to develop either a product that is a partial substitute for an incumbent’s product or an unrelated product. After the product is developed, the incumbent may choose either to buy out the entrant or to engage in Bertrand competition. We show that when buyouts are possible, the entrant is more likely to develop the substitute product. Moreover, this effect reduces aggregate welfare compared to a counterfactual where buyouts are not possible.

Abstract: I explore the welfare consequences of costly evidence acquisition in a broad class of contracting environments. An initially uninformed agent contracts with a principal. Before choosing whether to participate in a mechanism, the agent can observe, at a cost, a payoff-relevant signal which can be credibly disclosed to the principal. The principal may commit to a mechanism in which allocations are contingent on disclosure of a signal realization. I find that the principal's expected payoff is either non-increasing or U-shaped in the cost of evidence, and derive a condition that precisely distinguishes the two cases. In contrast, the agent's payoff is maximized at intermediate costs of evidence. Applications include insurance and labor markets, and public procurement. 

Abstract: I consider environments in which an agent with private information can acquire arbitrary hard evidence about his type before interacting with a principal. In a broad class of screening models, I show that there is always an evidence structure which interim Pareto-improves over the no-evidence benchmark whenever some types of the agent take an outside option in the benchmark case, and additional weak conditions, including either a single-crossing condition or state-independence of the principal's payoffs, are satisfied. I show that the sufficient conditions are tight and broadly applicable. Addressing concerns about multiple equilibria, I show how a planner can restrict the available evidence to ensure that an equilibrium which interim Pareto-improves over the benchmark case is obtained. Furthermore, I show that Pareto-improving evidence can arise endogenously when agents choose what evidence to acquire (and disclose).