Education Market Design in the Presence of Peer Effects: Theory and Evidence From South Korea

(with Sam Hwang)

Schools can also differentiate by peer composition among other aspects. Different rules elite or private schools are subject to compared to public schools have been heavily debated because of their effects on student distribution. In many countries former group admit students earlier than the latter, which we call Sequential Admissions (SA). Furthermore, the former group can use academic criteria in admissions unlike the latter. We study the effect of these admission aspects on welfare and distribution. Our analysis has important distinguishing features. First, it includes private schools as well as public schools, whereas previous studies on admission rules focused on centralized public school allocation. Second, students can have preferences for peer composition at schools. This is an important factor that can determine sorting behavior. First we theoretically study equilibria of the admission games complicated by the peer effects. Then we estimate a structural model using detailed high school applications data from Seoul to run counterfactual simulations. We show that SA can approximate centralized admission schemes well. This is important since complete centralization is known to increase welfare but hard to implement in many cases. Regarding admission criteria, we show that use of academic criteria in a subset of schools increases the desirability of these schools. The reason is that high performing students want to coordinate to study together and academic screening provides this. This suggest that, school choice is also a coordination game, not just an object allocation problem.

Education Reform and Behavioral Response: Evidence From South Korea

(with Sejin Ahn and Sam Hwang) (under revision)


We study an education reform resulting in delayed ability tracking for South Korean students during the 1960s-70s. The reform ended a practice of sorting students into elite and non-elite middle schools via admission exams, postponing ability tracking until the high school level. A discontinuity in the probability of students’ facing admission exams based on a birth-date cutoff enabled us to identify the causal effect of the reform on short- and long-run outcomes. We find that the reform increased both the incidence of private tutoring as well as hourly wages amongst students from wealthy households. A causal mediation analysis shows that private tutoring is an important pathway for the effect of the reform on university graduation and hourly wage. Our findings suggest that education reforms can interact with household behavior to yield unintended policy outcomes, especially in developing countries with well-established private tutoring markets.

The Role of Outside Options in Boston Mechanism


This paper examines ex-ante welfare from centralized public school allocation for students who cannot go to private schools when there are others who have such options. I show that when a private school is preferred to only the last option, students prefer Boston Mechanism (BM) to Deferred Acceptance (DA). Analyzing the case of desirable private schools in a model with three public schools, I demonstrate that students who are marginal in the decision of which school to report as top choice are better off under DA compared to BM; whereas inframarginal students are better off under BM. I show that a distribution of preferences with full support guarantees the existence of students who are better off under BM compared to DA. Assuming uniform distribution of preferences allows one to find the fraction of students who are better off under BM compared to DA. Analysis of the effect of private school entry on the welfare of students who cannot access private schools demonstrates that under mild conditions either there are students who are strictly better off after the entry or welfare of none of the students change.

Work in progress:

How does heterogeneity in beliefs affect students in the Boston mechanism? (with Sam Hwang) (revised version coming soon)