The Gosnell Prize

Fu and Slough re-examine one of the most common empirical practices in political science---estimating heterogeneous treatment effects (HTE) as a test of causal mechanisms. Despite the popularity of HTE analyses, how HTEs connect to causal mechanisms is often only discussed with intuitions without a formal framework. By explicitly connecting the mediation framework and HTE analyses, Fu and Slough develop a framework to understand when the existence of HTEs can support inferences about the activation of a mechanism. They demonstrate that interpreting HTEs as evidence for mechanisms requires what they call the exclusion assumption---the moderator of interest introduces heterogeneity only with respect to indirect effects specific to the mechanism of interest and does not introduce heterogeneity with respect to other mechanisms. As the authors articulate, the logic is straightforward, but it remains implicit in most uses of heterogeneous treatment effects to detect mechanisms. They also clarify that there is no logical ordering between this exclusion assumption and the sequential ignorability assumption necessary for mediation analysis, implying that HTE analyses are neither more nor less agnostic tests of mechanisms than mediation analysis. Fu and Slough's paper will help a wide range of applied scholars understand how to properly interpret HTEs as a mechanism test, and we also expect that their paper will inspire additional methodological developments about HTE analyses, mechanisms, and causal inference. This paper is also a great example of recent developments in theoretical implications of empirical models (TIEM). We congratulate the authors on this excellent contribution to political methodology.