School Choice Under Imperfect Information

46 Pages Posted: 28 Jan 2020

Date Written: January 13, 2020

Abstract

As in many school districts around the world, prospective high-school students in Ghana are assigned to schools through a centralized system. Using administrative data on applications, we report that virtually all students adopt a weakly dominated strategy, and matching outcomes show that approximately 15% of students end up unassigned, while almost half of schools have at least 1 vacancy. In order to rationalize choices in this setting, we build and estimate a model, where students engage in a costly search process to acquire information over school characteristics. The key insight of the model is that schooling decisions are exerted without the full examination of all available options, which may lead to sub-optimal choices. Our empirical application documents a substantial welfare loss: distance traveled to schools could be divided by 4. Counterfactual simulations show that if a planner were to restrict choices and assign the highest test score student to the most selective school, welfare would increase by 72%. We propose a mechanism design reform, and show that collecting preferences over a limited number of school attributes would recover most of the lost welfare.

Keywords: school choice, uncertainty, consideration set, search

JEL Classification: C53, D61, I20

Suggested Citation

Ajayi, Kehinde and Sidibe, Modibo, School Choice Under Imperfect Information (January 13, 2020). Economic Research Initiatives at Duke (ERID) Working Paper No. 294, January 2020, Available at SSRN: https://ssrn.com/abstract=3524535 or http://dx.doi.org/10.2139/ssrn.3524535

Kehinde Ajayi (Contact Author)

World Bank ( email )

1818 H Street, NW
Washington, DC 20433
United States

Modibo Sidibe

Duke University ( email )

100 Fuqua Drive
Durham, NC 27708-0204
United States

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
183
Abstract Views
1,042
Rank
297,507
PlumX Metrics