Bayesian Multivariate Inference for a Non-Standard Fuzzy Regression Discontinuity Design

• Activity Number: 38
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Fan Li*+ and Fabrizia Mealli and Alessandra Mattei and Fei Liu
• Companies: Duke University and University of Florence and University of Florence and IBM T. J. Watson Research Center
• Keywords: Bayesian ; causal inference ; instrumental variable ; multivariate outcomes ; regression discontinuity design

Abstract

Regression discontinuity designs (RDD) identify causal effects of interventions by exploiting treatment assignment mechanisms that are discontinuous functions of observed covariates. In standard RDDs, the probability of treatment changes discontinuously if a covariate exceeds a threshold. We consider a more complex RDD setup where the treatment is determined by both a covariate and an application status. In particular, we focus on a fuzzy RDD with this setup, where the causal estimand and estimation strategies are different from those in the standard instrumental variable approach to fuzzy RDDs. A Bayesian approach is developed for drawing inferences of the causal effect and multivariate outcomes are utilized to sharpen the analysis. The method is applied to evaluate the effects of Italian university grant on student dropout and academic performances.