Using Qualitative Comparative Analysis to Study Causal Complexity

Health Services Research, Dec, 1999 by Charles C. Ragin

When they face causation this complex, case-oriented researchers are likely to typologize their cases. That is, rather than seeing their cases as instances of the "same thing," they differentiate types of outcomes, for example, types of "defection-related turnover." The analysis of similarities and differences among cases will play a key part in the process of developing types. The investigator will try to identify causal commonalities within each type and causal differences between types. In effect, the search for commonalities will narrow as different combinations of causal conditions are linked to specific types of the outcome. In the end, a one-to-one correspondence appears between the different types of the outcome and the sets of conditions linked to each type.

There is nothing wrong with resorting to typologies. Often, it is a very creative process that leads to important theoretical insights and advances. However, typologizing may not be the best way, and it is certainly not the only way, to address causal heterogeneity. The danger is that types can proliferate to the point where the number of cases conforming to each type may be quite small. This potential proliferation of types stems in part from an overemphasis on the search for commonalities, understood as necessary conditions; but it is also an effect of the rigidity of the notion underlying it: that each outcome must result from a single causal condition or a single combination of conditions. Once it is acknowledged that the same outcome may result from several different combinations of conditions, then the decision to typologize seems less straightforward.

How should case-oriented researchers address causal heterogeneity? What should they do when a given outcome results from several different combinations of conditions? QCA is designed specifically for the study of this type of complexity. At its core, it is nonstatistical, but probabilistic criteria can be incorporated at various points in the procedure (see Ragin 1999, 2000). QCA is intended to work hand-in-hand with case-oriented inquiry. It is based on case-oriented knowledge and at the same time provides the means for structuring case-oriented inquiry and for systematizing the results of this type of inquiry. Before describing QCA, I first address the issue of causal complexity, sketching the core of the problem. Even though causal complexity does undermine the examination of single causes (and the use of additive models more generally), I show that causal complexity is not resistant to analysis, per se.

NECESSITY, SUFFICIENCY, AND CAUSAL COMPLEXITY

In many fields of social science, the assessment of necessary conditions is very important. Necessary conditions have very powerful policy implications. For example, in some situations it is possible to prevent a bad outcome by removing or blocking one of its necessary conditions. However, it is important to recognize that designs that work backward from multiple instances of an outcome to identify shared antecedent conditions (1) cannot address sufficiency and thus may fail to identify decisive causes, and (2) make restrictive assumptions about causation, namely, that theoretically relevant, nontrivial necessary conditions exist. I contend that the search for necessary conditions, while often useful, sometimes constricts case-oriented research in unproductive ways. It is much more fruitful to allow for the possibility that a given outcome may follow from a variety of different combinations of theoretically relevant causal conditions.