Richard Breen, Guanghui Pan
Abstract The so-called causal revolution that has spread through economics and into adjacent social sciences, including sociology, has been very much concerned with developing methods by which to arrive at credible causal estimates, especially in nonexperimental, observational settings. In the language of experiments, it has focussed on internal validity. But much less attention has been paid to external validity, that is, whether a causal relationship holds in situations other than the one in which it was found. Thinking about external validity obliges us to consider why we are trying to estimate causal relationships in the first place, and what we think they are for. In this paper we discuss in greater detail what we mean by internal and external validity and set out the assumptions required for causal estimates to have both. We consider some examples from sociological research and the challenges to external validity that they illustrate. We urge sociologists, especially those engaged in nonexperimental research, to pay more attention to external validity. But we also stress that this is important not only for causal research but also for other research, and we illustrate issues of external validity that may arise in a wide variety of noncausal studies. We conclude with some practical suggestions and remarks concerning some of the general issues that arise from our work.