Evidence on the deleterious impact of sustained use of polynomial regression on causal inference

Abstract

It is standard in regression discontinuity analysis to control for thirdor fifth-degree polynomials of the assignment variable. Such models can overfit, leading to causal inferences that are substantively implausible in which variation is somewhat arbitrarily attributed to high-degree polynomial or the discontinuity. We discuss in the context of two… (More)

4 Figures and Tables

Topics

  • Presentations referencing similar topics