Designing Educational Policies: Evidence from Secondary Education in Paris – DesignEdu

This research develops econometric methods for estimating student preferences from their rank-order lists that are submitted in school choice systems. Empirical approaches used in the literature are usually based on the strong assumption of truthful revelation of preferences. However, most of the mechanisms used in practice restrict the number of schools being included in the rank-order list, which can lead students to adopt strategic behavior. We use a revealed preference method, assuming that the observed choices made by families reflect their preferences in some way. We propose a new method to estimate student preferences based on observed student choices, without assuming that such choices are literally true preferences. Instead, it only assumes that students do not make significant mistakes so that they do not miss the schools at which they have good chances of admission. We theoretically show that this is plausible even when each student knows little about others’ preferences. Using the 2013 school choice data in Paris, we show that the new method explains the student behavior better than the alternative economic approaches, which rely on truth-telling assumptions.

Our scientific approaches to analyzing school choice data can be used in many other settings, including centralized university admissions. We use them to analyze school choice data from Paris and empirically answer questions such as how school choice procedures and admission criteria determine student sorting across schools. Moreover, our methodology is currently being used by researchers from Columbia, MIT, Princeton, and University of Melbourne to study university admissions in Australia and Chile, in which the centralized DA mechanism is also used.

Our methodology is currently being used by researchers from Columbia, MIT, Princeton, and University of Melbourne to study university admissions in Australia and Chile, in which the centralized DA mechanism is also used.

Our main results are summarized in the working paper “Beyond Truth-telling: preference estimation with centralized School choice” by Gabrielle Fack, Julien Grenet, and Yinghua He. It is currently under revision for publication at the American Economic Review, one of the top academic journal in economics. The paper summarizes our analytical tools and some empirical results from the secondary education in Paris.

Project DesignEdu is a fundamental research project coordinated by Yinghua He (Toulouse School of Economics). It also associates Gabrielle Fack (Paris School of Economics) and Julien Grenet (Paris School of Economics). The project began in October 2014 and lasted 22 months. It received ANR assistance of € 56 686,24 for an overall cost of € 77 422,24.

This proposed research aims to understand better and thus improve the market design for allocating students to public schools. In the absence of monetary transfers, this allocation is usually implemented with centralized mechanisms. The theoretical literature has shown that the Gale-Shapley deferred-acceptance (DA) mechanism has desirable properties, such as stability and strategy-proofness, which justifies its adoption in many places (e.g. Boston, New York City, Paris, and Taiwan). The empirical analyses of the mechanism and its contribution to the efficiency and equity of education however remain limited. This research attempts to fill this gap by exploring the allocation mechanisms maximizing social welfare.

More specifically, we plan to answer three questions:
1) How to estimate student preferences from their submitted rank-order lists in school choice?
2) What is the optimal mechanism for allocating students to public schools?
3) What is the optimal affirmative action policy to reduce educational inequality?

The first project of this research aims to make a contribution by developing econometric methods for estimating student preferences from their submitted rank-order lists in school choice systems. Empirical approaches commonly used in the literature are based on the strong assumption of truthful revelation of preferences. However, most of the mechanisms used in practice restrict the number of schools being included in the rank-order list, which can lead students to adopt strategic behavior. Our goal is to provide methods that require less restrictive assumptions and are based on a theoretical foundation of student behavior. The results will be illustrated by Monte Carlo simulations and an empirical study with a data set from Paris.

The estimation of student preferences is the prerequisite for the second project, which aims to identify the optimal mechanism in terms of student welfare. The relative efficiency and distributive properties of alternative allocation mechanisms will be evaluated using simulations that are build on the estimates from the Parisian data. We will in particular compare the mechanism currently used in Paris with the unconstrained DA, which allows students to rank as many schools as they wish. Moreover, we will use the relaxation of the constraint in Paris to validate our approach – in 2013, Parisian students were allowed to rank 8 schools, increased from 6 schools, while in total there have been always 9 to 15 schools accessible to every student.

The third project in this proposal focuses on the impact of Affirmative Action policies on educational inequality. As widely documented in the fields of economics, education, and sociology, the persistence of educational inequalities still call for new approaches to address them. The policies in Paris for school choice are a recent innovation: Since 2008, students from low socioeconomic background receive preferential treatment under which they are essentially allowed to enroll in any school of their choice. We plan to analyze the impact of this policy on educational outcomes of both disadvantaged and other students, and thus to evaluate the effectiveness of this form of positive discrimination.

In summary, the results of this research will not only advance our scientific knowledge on the allocation mechanisms, but also contribute to the improvement of educational policies.