Appendix: Qualitative Causal Models and the Potential-Outcomes Framework

Summary

The argument in Chapter 2, that a suitably broad reading of the potential-outcomes framework captures most of the ideas and intuitions about causation that have been important in both qualitative and quantitative methodological work in the social sciences, loses all traction if in practice adopting the potential-outcomes framework as a shared causal language would serve to rule out important qualitative causal hypotheses and models. Hence, it is important to demonstrate that at least many such models fit well with the potential outcomes setup. The discussion below argues not only that the potential-outcomes framework can successfully represent a set of key qualitative causal ideas, but also that the process of translation illuminates key aspects of those causal ideas that may otherwise be more difficult to spot. This argument will be developed through a discussion of necessary and/or sufficient causes, INUS and SUIN causes, path-dependent and critical-juncture causal models, and ideas about causal mechanisms and pathways. Brief concluding reflections will be offered regarding some other causal categories sometimes seen as related to a qualitative-quantitative divide in causal conceptualization.

Necessary and/or Sufficient Causes

Qualitative methodologists have drawn a great deal of attention to causal models involving necessary and/or sufficient causes (Dion 1998; Braumoeller and Goertz 2000; Ragin 2000; Goertz and Starr 2002; Seawright 2002; Goertz and Levy 2007). Applied qualitative researchers have of course made heavy use of such concepts for many years, as has the occasional piece of quantitative research. In brief and in deterministic form, necessary causes are those causal factors whose absence makes an outcome impossible; however, their presence need provide no special information about the likely score of the dependent variable. Sufficient causes are those factors whose presence makes the outcome inevitable, although their absence may tell us little about the probable outcome.

Some scholars extend the concept of necessary and/or sufficient causes from this deterministic starting point toward a probabilistic formulation that allows for some exceptions (see especially Ragin 2000). In probabilistic necessary/sufficient causal formulations, some predefined proportion of exceptions to the defining rule is allowed. For example, a cause may be 85% necessary if at least 85% of cases without the cause also fail to have the outcome.