Experimental designs for identifying causal mechanisms.pdf

J. R. Statist. Soc. A (2013) 176, Part 1, pp. 5–51 Experimental designs for identifying causal mechanisms Kosuke Imai, Princeton University, USA Dustin Tingley Harvard University, Cambridge, USA and Teppei Yamamoto Massachusetts Institute of Technology, Cambridge, USA [Read before The Royal Statistical Society on Wednesday, March 14th, 2012, the President , Professor V. S. Isham, in the Chair ] Summary. Experimentation is a powerful methodology that enables scientists to establish causal claims empirically. However, one important criticism is that experiments merely pro- vide a black box view of causality and fail to identify causal mechanisms. Specifically, critics argue that, although experiments can identify average causal effects, they cannot explain the process through which such effects come about. If true, this represents a serious limitation of experimentation, especially for social and medical science research that strives to identify causal mechanisms. We consider several experimental designs that help to identify average natural indirect effects. Some of these designs require the perfect manipulation of an interme- diate variable, whereas others can be used even when only imperfect manipulation is possible. We use recent social science experiments to illustrate the key ideas that underlie each of the designs proposed. Keywords: Causal inference; Direct and indirect effects; Identification; Instrumental variables; Mediation 1. Introduction Over the last century and across numerous disciplines, experimentation has been a powerful methodology to test scientific theories. As Neyman demonstrated in 1923 (see Neyman (1990)), the key advantage of randomized experiments is their ability to estimate causal effects without bias. However, one important c