Lucas Barros

Research

Quasi-Ironing

Ironing was introduced in Myerson (1981) to handle monotonicity constraints natural to the maximization problems that arise in mechanism design. The method was then generalized in Toikka (2011). If there are two variation constraints, however, the value of the maximization problem is not invariant to ironing. This paper introduces a new method coined quasi-ironing. It refines ironing based on a quasiconvex, rather than a convex, envelope. This method works as well as ironing under a single variation constraint, but can accommodate multiple such constraints. The main application studied is that of screening buyers on the Hotelling line with unknown location and valuation for a homogeneous good.

Distortion-free Screening Problems

In a conditionally efficient mechanism, each type either receives its favorite outcome or nothing. This paper characterizes the screening problems that exhibit the following property: for every distribution over types, there exists a revenue-maximizing mechanism that is conditionally efficient. This characterization is then applied to two benchmark cases. (i) When all types share the same favorite outcome, the result reduces to and generalizes a known characterization in bundling. (ii) When each outcome is some type's favorite, the desired property identifies a familiar environment: location preferences for a homogenous good, with transportation costs satisfying a directed triangle inequality.

 The Limits of Information Acquisition in Cheap Talk

Can designing one’s own information substitute for commitment in communication? This paper studies a sender who commits to an information acquisition policy but communicates with a receiver through cheap talk. The main result shows that, for finite action spaces, acquisition design attains the Bayesian persuasion payoff for every prior if and only if classical cheap talk, with no acquisition design, already does. That is, whenever commitment to communication is essential for persuasion, designing the sender’s information alone is insufficient. The result clarifies the role of commitment in information design and identifies a limitation of endogenous information acquisition.