MET Seminar - Josue Ortega (Queen’s, Belfast)
| Dates: | 11 October 2023 |
|---|
| Times: | 17:00 - 18:00 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | School of Social Sciences |
|---|
| Who is it for: | University staff, Current University students |
|---|
|
Title: The Limits of School Choice with consent (joint with Gabriel Ziegler)
Abstract: The deferred acceptance (DA) algorithm is used all around the world to assign pupils to schools due to its stability and strategy-proofness. However, DA can generate an assignment that is Pareto inefficient for students and, in some cases, it may assign all students to their worst or second worst school. The inefficiency of DA has been also documented in real life. Kesten (2010, QJE) proposed an influential modification of DA, known as efficiency-adjusted deferred acceptance (EADA), which fixes the Pareto inefficiency of DA. A vast literature has now shown that EADA is not only Pareto efficient, but also satisfies a plethora of desirable properties both theoretically and in the lab. For these reasons, EADA is to be implemented in Flanders. Despite these good properties, we show that EADA can generate allocations that are significantly rank-inefficient and unequal. In particular, we prove that for any number of students, there exists a school choice problem where half of the students are allocated to a school in the bottom half of their preferences according to EADA, even though a different allocation assigns all but one student to their top school, and no student to a school worse than their second choice. Furthermore, we analyze random, iid one-to-one matching problems. In our main Theorem, we approximate the asymptotic expected average rank in EADA, which is in the order of $\log(n!)/n$. This result has an interesting corollary: the quality of student placement in {EADA is asymptotically the same as in DA}, which implies that the size of the efficiency gains of EADA over DA vanish when the number of students is large. This result explains the small efficiency gains of EADA versus DA documented empirically (Ortega and Klein, 2023 GEB).
Manchester Economic Theory (MET) schedule: http://daviddelacretaz.net/seminars/
Speaker website: https://www.josueortega.com/
Travel and Contact Information
Find event
Boardroom 2.016
Arthur Lewis Building
Manchester
