NSF-BSF: Mechanism Design for All

Harvard University, Cambridge MA

Investigators

Abstract

School choice programs that match students to schools are one example of what is called in economics a “matching market”. Many such matching markets use a centralized mechanism, or algorithm, to match applicants with institutions. Increasingly wider adoption of such mechanisms (not only in school choice programs but also in physician residency programs and others) results in a profile of participants in economic mechanisms that is more diverse than ever before. For this reason, the economic literature is slowly shifting from “idealized participants”, who are perfectly rational and never make mistakes, to studying these mechanisms both theoretically and empirically, taking into account that the participants are real and all too human. Yet, quite often, empirical and experimental studies are conducted on some of the best students in the country, either because they are run in labs at leading US universities, or because even studies analyzing real-world matching markets focus on those markets that match recent graduates with employers (for instance, resident doctors are assigned to hospitals). These students and graduates exhibit various background traits not necessarily found in individuals participating in real-life mechanisms: graduates are fairly well educated, have impressive reading comprehension skills in English, are trained in analytical reasoning, and have the ability to focus and follow complex instructions. Moreover, college-educated participants are much more likely to have trust in institutions; and they have a more uniform cultural background than the general population. It is unclear to what extent experiments that focus on introducing and explaining new mechanisms that are run on college-educated subjects (who are well trained and have the ability to follow new, complex instructions) would properly inform the design of mechanisms that are ultimately meant to be applied to everybody, including subjects with different backgrounds, abilities, and skills. This research studies (and aims to improve) the design of mechanisms for all people from all populations. Unfortunately, those agents who are less adept at strategizing in situations where the mechanism is not strategy-proof, are less likely to recognize a mechanism that is strategy-proof (and for which strategizing is not beneficial). These agents may belong to less advantaged groups, have less trust in institutions, or have fewer people in their social circles whom they can lean on to verify scientific claims. To reach an equitable outcome, it therefore does not suffice for the mechanism to be strategy-proof; its strategy-proofness must also be understood by all participants, from all populations. For example, the deferred acceptance mechanism, which is known to have desirable properties and yield desirable outcomes in theory, is anything but easy to understand for the people it is supposed to serve. A double-edged challenge here is therefore to (a) develop the theory of easy-to-understand mechanisms; and (b) empirically and repeatedly test their performance on a large enough set of subjects (e.g., broad online samples) to make it possible to study heterogeneity of outcomes. This project has three aims. First, to improve our empirical understanding of the way real people, from real populations, understand mechanisms (strategically and otherwise) and interact with them. Second, to improve our theoretical understanding of behavioral issues traditionally neglected in mechanism design. Finally, to theoretically derive, and empirically test, more efficient and inclusive mechanisms (and their descriptions) in such behavioral models. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.