Anne-Katrin Roesler
Assistant Professor
University of Toronto, Department of Economics
Research interests: Mechanism Design, especially Information Design, Information Economics, Game Theory.
Published, accepted, and forthcoming papers:
Multi-Dimensional Screening: Buyer-Optimal Learning and Informational Robustness (joint with Rahul Deb) [pdf]
(accepted at Review of Economic Studies)
A monopolist seller of multiple goods screens a buyer whose type is initially unknown to both but drawn from a commonly known distribution. The seller chooses a mechanism to maximize her worst-case profits against all possible signals from which the buyer can learn about his values for the goods. We show that it is robustly optimal for the seller to bundle goods with identical demands (these are goods that can be permuted without changing the buyer's prior type distribution). Consequently, pure bundling is robustly optimal for exchangeable prior distributions. For exchangeable priors, pure bundling is also optimal for the seller in the information environment (with the reverse timing) where an information designer, with the objective of maximizing consumer surplus, first selects a signal for the buyer, and then the seller chooses an optimal mechanism in response. We derive a formal relationship between the seller's problem in both information environments.
Learning Before Trading: On The Inefficiency of Ignoring Free Information (joint with Doron Ravid and Balázs Szentes) [pdf]
Journal of Political Economy, 2022, 130:2, 346-387
This paper analyzes a bilateral trade model where the buyer's valuation for the object is uncertain and she can privately purchase any signal about her valuation. The seller makes a take-it-or-leave-it offer to the buyer. The cost of a signal is smooth and increasing in informativeness. We characterize the set of equilibria when learning is free and show that they are strongly Pareto ranked. Our main result is that, when learning is costly but the cost of information goes to zero, equilibria converge to the worst free-learning equilibrium.
Buyer-Optimal Learning and Monopoly Pricing (joint with Balázs Szentes) [pdf] - (Supplemental Material, pdf)
American Economic Review, 2017, 107(7), 2072-2080
This paper analyzes a bilateral trade model where the buyer's valuation for the object is uncertain and she observes only a signal about her valuation. The seller gives a take-it-or-leave-it offer to the buyer. Our goal is to characterize those signal structures which maximize the buyer's expected payoff. We identify a buyer-optimal signal structure which generates (i) efficient trade and (ii) a unit-elastic demand. Furthermore, we show that every other buyer-optimal signal structure yields the same outcome as the one we identify, in particular the same price.
Working Papers:
Committee Composition: The Impact of Diversity and Partisanship on Voting Behavior [pdf] (updated)
We propose and analyze a committee voting model in which players with interdependent values take binary decisions. Each player holds two-dimensional private information: a private signal about the payoff state, and a private preference type that captures their level of partisanship. We show that, in equilibrium, committee members adopt cutoff strategies that are monotone in preference types. We identify a committee's composition, captured by the distribution of private preference types, impacts its decisions. As committee members are drawn from increasingly partisan populations, equilibrium cutoffs move away from the sincere voting threshold. Conversely, when committee members originate from more diverse populations, equilibrium cutoffs move closer to the sincere voting threshold. We discuss implications for the set of alternatives that different committees accept.
Private Learning and Exit Decisions in Collaboration (joint with Yingni Guo) [pdf]
We study a collaboration model in continuous time, with a positive arrival rate of a success in both the good and the bad state. If the project is bad, players may privately learn about it. At any time, players can choose whether to exit and secure the positive payoff of an outside option, or to stay with the project and exert costly effort. A player's effort not only increases the probability of success, but also serves as an investment in private learning.
We identify an equilibrium with three phases. In all phases, uninformed players exert positive effort. Players who become informed and learn that the project is bad never exert effort. Because players benefit from the effort of the others, informed players may not exit immediately. In the first, "no-exit" phase, informed players do not exit. In the subsequent, "gradual-exit" phase, they exit with a finite rate. In the final, "immediate-exit" phase, informed players exit immediately. We find that effort levels may increase in the no-exit phase, if the markup of effort in the bad state is positive. Surprisingly, increasing the payoff of the outside option encourages collaboration.
Information Disclosure in Markets: Auctions, Contests and Matching Markets [pdf]
We study the impact of information disclosure on equilibrium properties in a general model of a two-sided matching market that incorporates a large class of market design environments. In this model, each agent first privately observes an informative, but potentially noisy, signal about his private type. The agents then enter a matching stage in which they choose signaling investments to compete for match partners. In order to study the impact of information disclosure, we introduce a criterion that orders signals in terms of their informativeness. We show that information disclosure increases the expected total match output but may also increase wasteful signaling investments due to amplified competition within groups. The second effect may dominate, leading to a decrease in expected welfare. Disclosure effects on equilibrium properties depend on whether information is disclosed to agents on the short or the long side of the market. Implications of our findings to auctions, contests and matching markets are discussed.
Mechanism Design with Endogenous Information [pdf]
In mechanism design problems with endogenous information, regularity properties of the distribution of posterior estimates (types) are essential for tractability. Important properties are a monotone hazard rate, increasing virtual valuations or costs. Difficulties arise since these properties are not preserved under mixtures, and regularity of the prior distribution may not translate to the distribution of posterior types. In this note, we identify sufficient conditions on the primitives of an information structure, which guarantee that the distribution of posterior types has a monotone hazard rate, increasing virtual valuations or costs. These characterization results make it possible to study mechanism design problems with endogenous information, without imposing regularity conditions on the interim stage or restricting attention to specific information structures. Applications to information acquisition and disclosure in optimal auctions, and to allocation problems without money are discussed.
Other Contributions:
Advances in Mechanism Design: Information Management and Information Design [pdf]
The Bonn Journal of Economics, Volume V, No. 1, July 2016, pp. 13 - 24.