Extrapolation in multiple-cutoff regression discontinuity designs
Date
2019-08
Authors
Vázquez Baré, Gonzalo
relationships.isContributorOfPublication
Cattaneo, Matías D.
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad de San Andrés. Departamento de Economía
Abstract
Si bien la mayor parte de la literatura econométrica sobre diseños de regresión discontinua se enfoca en el caso de un solo cutoff (valor de corte), frecuentemente se analizan programas o políticas en los que distintas observaciones enfrentan distintos umbrales. El presente trabajo introduce la idea de selección a cutoffs y estudia identificaci on de efectos de tratamiento cuando unidades enfrentando distintos cutoffs pueden diferir en variables observables y no observables. El resultado principal provee condiciones bajo las cuales es posible identificar y estimar un efecto de tratamiento fuera del cutoff. Los resultados son ilustrados usando datos de Oportunidades, un programa anti-pobreza de gran escala en México en el cual elegibilidad a nivel de hogar fue basada en un índice de pobreza con distintos cutoffs según la región geográfica.
Although the bulk of the econometric literature on Regression Discontinuity Designs (RDDs) focuses on the case of a single cutoff, researchers frequently study programs or policies in which different observations face different thresholds. This paper introduces the idea of selection into cutffs and studies identification of treatment e ects when units facing different cutoffs are allowed to di er in both observable and unobservable characteristics. The main result provides conditions under which it is possible to identify and estimate a treatment e ect away from the cutoff. The findings are illustrated using data from Oportunidades, a large-scale anti-poverty program in Mexico in which household eligibility was based on a poverty index with cuto s varying across geographical regions.
Although the bulk of the econometric literature on Regression Discontinuity Designs (RDDs) focuses on the case of a single cutoff, researchers frequently study programs or policies in which different observations face different thresholds. This paper introduces the idea of selection into cutffs and studies identification of treatment e ects when units facing different cutoffs are allowed to di er in both observable and unobservable characteristics. The main result provides conditions under which it is possible to identify and estimate a treatment e ect away from the cutoff. The findings are illustrated using data from Oportunidades, a large-scale anti-poverty program in Mexico in which household eligibility was based on a poverty index with cuto s varying across geographical regions.
Description
Fil: Vázquez Baré, Gonzalo. Universidad de San Andrés. Departamento de Economía; Argentina.