American Economic Association

Journal of Business and Economic Statistics Invited Session

Paper Session

Friday, Jan. 4, 2019 2:30 PM - 4:30 PM

Atlanta Marriott Marquis, L508

Hosted By: Econometric Society

Chair: Todd Clark, Federal Reserve Bank of Cleveland

Breaking Ties: Regression Discontinuity Design Meets Market Design

Joshua Angrist

Massachusetts Institute of Technology

Abstract

Centralized school assignment algorithms uses stochastic tie-breaking to discriminate between otherwise similar applicants. School matches may employ non-lottery tie-breakers like test scores, randomly assigned lottery numbers, or both. The New York City public school district, for example, uses test scores, grades, and interviews to rank applicants to screened schools, combined with lottery tie-breaking at other schools. We develop methods to identify causal effects of school attendance in such settings. Our approach generalizes the standard regression discontinuity design to allow for multiple treatments and multiple running variables, some of which are randomly assigned. We show that lotteries generate assignment risk at non-lottery programs for applicants with tie-breakers away from non- lottery cutoffs, while non-lottery variation randomizes applicants near cutoffs regardless of lottery risk. These methods allow us to assess the predictive value of New York City’s school report cards. Grade A schools boost SAT math scores and the likelihood of graduating with a Regents diploma. These results are sharpened by estimation strategies that exploit the combination of lottery and screened school risk. Both sources of risk generate similar results. Grade A effects are also similar for screened and lottery schools. The lecture draws on joint work with Atila Abdulkadiroğlu, Yusuke Narita, and Parag A. Pathak.

Discussant(s)

Justine Hastings

Brown University

John Friedman

Brown University

Ariel Pakes

Harvard University

JEL Classifications