Non-Compliance Can Bias Estimates of Heterogeneous Treatment Effects

• In Event: Advances in Experimental Methodology

Brief Overview

We study how common methods for studying conditional treatment effects in experiments may be biased by noncompliance. We identify this bias using multiple experiments and conclude with guidance for addressing noncompliance in survey experiments.

Authors

• Scott Clifford, Texas A&M University

• Geoffrey David Sheagley, University of Georgia