Causal inference with observational data

Causality is central to our understanding of the world and central to scientific explanation. In recent years two approaches to causality have come to prominence and have had a major impact on the social sciences: these are the counterfactual or potential outcomes model of causality and the approach that understands causality in terms of a causal structure represented by a graph. I present both of these and explain how they can be used to identify causal relationships in situations when we do not have access to experimental data. I discuss the principles underlying the most widely used strategies for estimating causal effects in these situations. Finally, I discuss questions of external validity, and, in particular, the conditions under which sociologists' causal estimates can be of more than historical interest.