Direct and indirect causal effects:
a helpful distinction?

Donald Rubin
Harvard University, USA

Although the terminology of direct and indirect causal effects is  relatively common, I believe that it is generally scientifically unhelpful  without further explication.  This assessment is based on repeated  experience with the confusion it creates in important examples in the  social and biomedical sciences.  After reviewing my perspective on causal  inference based on the concepts of potential outcomes and assignment  mechanisms, this presentation will discuss two distinct ways to formalize  the issues that arise in circumstances where this terminology is used.

The first is based on the concept of principal stratification, and the  second is based on the concept of a compound assignment mechanism,  which is just a special case of general assignment mechanisms.  Simple  artificial examples will be used to illustrate the differences between the  two conceptualizations, and to show that it is easy to become confused  when using the direct and indirect jargon to describe causal effects.

Even the great R.A. Fisher was a victim of this confusion, as will be  documented.