Periodic Reporting for period 4 - Frontiers in Design (Frontiers in Mechanism Design: Methodology and Applications)
Periodo di rendicontazione: 2021-03-01 al 2022-08-31
The standard mechanism design paradigm relies on three key fundamental assumptions: 1. The designer of the institution—the `principal'—does not have any privileged information: is uninformed. 2. The designer chooses the mechanism and commits to it once and for all. 3. Misreporting information is costless. These assumptions often fail in today’s “big data” world: Firms (online retailers, insurance companies, banks) do have privileged—and often certifiable —information that may affect the contractual terms that they propose. Also, they interact repeatedly with the same agents, and, as they learn about them, they attempt to change the terms by making personalized offers. Finally, oftentimes, regulators (for instance environmental agencies) seek to design mechanisms in the presence of costly fraud.
This project developed tools and methodologies for designing optimal mechanisms for those highly relevant situations. It applied the methods to the design of online marketplaces when consumer privacy is at stake and to the design of tests (such as emission tests) when firms may have incentives to manipulate test results.
The central novel methodological contribution is a revelation principle for settings in which the designer can only commit to today's mechanisms but to future ones. We identified a canonical class of mechanisms, which we coined Direct Blackwell Mechanisms these are direct revelation mechanisms coupled with an information disclosure policy, modelled as a Blackwell experiment. Direct Blackwell mechanisms encode the rules that determine the allocation and the information the designer obtains from the interaction with the agent. Our result simplifies the search for the designer-optimal outcome by reducing the agent’s behavior to a series of participation, truthtelling, and Bayes’ plausibility constraints the mechanisms must satisfy. This fundamental tool will allow researchers to address practical questions when designers have less than full commitment to future contractual arrangements: These include the design of social insurance programs, monetary policy and contracts that govern international trade.
The generality of the settings under which this and the other methodologies were developed opens the door to an array of policy-relevant applications in active and policy-relevant areas of economics (design of transparency regulation in financial markets, taxation policy); political science (design of term length) and in computer science (design of algorithms in the presence of manipulations).
1. The revelation principle under limited commitment
2. The new concept of interim optimal mechanisms that allows for the tractable characterization of credible disclosure rules in general informed information design settings.
3. The tools to study optimal mechanism design in the presence of fraud.
Within economics, removing the assumption of commitment to long-term contracts allows to revisit with more realistic models important and policy-relevant issues including the design of compensation schemes, debt and mortgage contracts, monetary policy, fiscal policy, and social insurance. The techniques introduced in "Mechanism Design with Limited Commitment" in Doval and Skreta 2022 will allow researchers in public finance and macroeconomics to revisit models of optimal policy under more realistic assumptions.
The project also has interdisciplinary impacts to computer science and political science. On the one hand, mechanism design looms large in computer science, with applications to the design of online platforms, ad auctions, and the dynamic allocation of computational and communication resources. Here, too, the assumption of commitment is often unrealistic: platforms interact with consumers repeatedly and re-optimize. On the other hand, problems of limited commitment are ubiquitous in political science and a methodology that speaks to the design self self-enforcing institutions will advance the literature on political mechanism design.
Similarly, the notion of interim optimality provides the class of credible disclosure mechanisms that can be use to analyze optimal information design problems in economics, political science, accounting, finance and beyond.
Finally, the tools of test design with falsification have direct applications in computer science theory when one seeks to design reliable algorithms in the presence of manipulations.
The project has produced three published papers and another three papers that are already revised and resubmit in journals:
Part 1: Mechanism Design by an Informed Principal
“Selling with Evidence,” (with Frederic Koessler) published in Theoretical Economics.
“Informed Information Design,” (with Frederic Koessler) revise and resubmit at Journal of Political Economy.
Part 2: Mechanism Design with Limited Commitment
“Mechanism Design with Limited Commitment” (with Laura Doval); published in Econometrica in July 2022.
“Optimal Mechanism for the sale of a durable good” (with Laura Doval); second round revise and resubmit at Theoretical Economics.
“Purchase history and product personalization” (with Laura Doval) revise and resubmit at Rand Journal of Economics.
“Constrained Information Design: Toolkit ” (with Laura Doval) was revised and resubmit at Mathematics of Operations Research.
Part 3: Mechanism Design in the presence of costly fraud.
“Information Design under Falsification” (with Eduardo Perez-Richet); published in Econometrica in May 2022.
The PI has given numerous talks at the most prestigious institutions around the globe and organised a successful conference in 2019. In 2022, the PI co-organized a week of ESSET around topics central to this grant.
Research findings of the program have been presented in prestigious schools such as the Jerusalem Institute of advanced studies (2 lectures on mechanism design by and informed principal); the Simons institute of theoretical computation at UC Berkeley (a lecture on mechanism design under limited commitment by co-author Laura Doval). Papers produced by this grant are already taught at Stanford, MIT, Columbia, Northwestern, Caltech and UCL among other universities.