A0857
Title: Testing sign agreement
Authors: Deborah Kim - University of Warwick (United Kingdom) [presenting]
Abstract: The focus is on the problem of testing sign agreement among a finite number of means. This problem naturally arises in numerous empirical contexts, including detecting sign-opposing average treatment effects across subgroups in randomised controlled trials, testing the homogeneity of local average treatment effects (LATE) across subpopulations, and examining the testable implications of the assumptions underlying LATE. For the null hypothesis that the means are all non-negative or all non-positive, two novel bootstrap tests are proposed: The least favourable and the hybrid tests. Unlike existing procedures, both tests accommodate arbitrary dependence among estimators for any finite number of means. It is shown that both tests control their asymptotic sizes uniformly over a large class of nonparametric distributions. Results from simulation studies in finite samples indicate that the rejection probabilities of both tests attain the nominal level under the null. The Hybrid test exhibits higher power than the least favorable test against a broad class of alternative distributions, though the opposite holds when comparing only two means. An empirical application to disapproval ratings of the Trump administration demonstrates the practical utility of both tests.
