A1506
Title: A nonparametric test for social spillovers and interference
Authors: Abhimanyu Gupta - Queen\'s University (Canada) [presenting]
Margherita Comola - Universite Paris-Saclay and PSE (France)
Camila Comunello - University of Essex (United Kingdom)
Abstract: A novel nonparametric test is proposed for cross-unit dependence occurring through peers' attributes (interference) or peers' outcomes (social spillovers). While the test is nonparametric, it is reminiscent of classical Lagrange multiplier-type diagnostic tests. Indeed, the test statistic requires only estimates under the null to be computed and therefore avoids nonparametric estimation altogether, and is shown to have a convenient asymptotic standard normal distribution. Empirical illustrations are presented to show that the test is effective at detecting cross-unit dependence arising in a nonparametric fashion that existing approaches might not detect effectively.
