Experimental Identication of Causal Mechanisms

Abstract Experimentation is a powerful methodology that enables scientists to empirically establish causal claims. Even in social sciences where traditionally observational studies were domi- nant, experiments are becoming increasingly common. However, one general and important criticism of experimental methodology is that experiments merely provide a black-box view of causality { although experiments can identify an average causal eect, they cannot explain how such an eect,comes about. If true, this represents a serious limitation of experimenta- tion, especially for social science research whose main goal is to identify causal mechanisms. In this paper, we rst establish that two existing approaches commonly employed by social scientists to address this critique suer,from serious methodological problems. To overcome this limitation, we consider alternative experimental designs that can be used to help identify causal mechanisms. Finally, we discuss the situations in which these experimental designs are applicable and illustrate their use by applying them to a decision neuroscience experiment. Key Words: causal chain, causal inference, causal mediation, decision neuroscience, poten- tial outcomes, randomized experiments We thank Luke Keele for our ongoing collaboration on related topics, and Adam Glynn and Walter Mebane