Content area
Full Text
Counterfactuals and Causal Inference: Methods and Principles for Social Research, by Stephen L. Morgan and Christopher Winship. Cambridge, UK: Cambridge University Press, 2007. 328pp. $27.99 paper. ISBN: 9780521671934.
Given the burgeoning popularity of counterfactual models, this monograph will certainly become a necessary addition to any serious researcher's arsenal of literature. It is clearly written and will help scholars utilize and evaluate both counterfactual models as well as the use of propensity scores. To the authors' credit, the monograph does not attempt to provide a cookbook approach to estimating these models, but rather emphasizes the need for a precise theoretical model of the social process under study when utilizing these techniques. And, even though by "theory" we are not necessarily referring to a need for a particularly large set of interdependent relationships specified from axiomatic principles nor any sweeping versions of grand social theory, it is nonetheless the case that the need to specify the precise causal model for any given relationship may seem quite daunting to one wishing to employ these techniques. Nevertheless, specifying the causal model is necessary if one hopes to claim estimating causal effects.
This major theme of needing theoretical clarity for guiding analyses is by no means unique to this particular modeling technique, but is nonetheless welcome. As is the case for structural equation modeling, counterfactual models require a theoretical rigor not always present in social science analyses. One positive consequence of this recurrent theme is that it will hopefully dissuade those who may be tempted to use propensity scores as a shortcut to avoiding such theoretical rigor. For these reasons, this monograph's constant focus on specifying the reasons for the relationship under study or possible confounding variables-and the consequent necessity of conceptualizing the counterfactual of interest-is necessary reading for any researcher wishing to utilize or evaluate studies using these techniques. Arguably, the risk of poorly conceptualized models is even greater for this counterfactual approach because of its focus on causal models. That is, given that this technique...