Using Qualitative Comparative Analysis to Study Causal Complexity

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

As an alternative to searching for necessary-but-not-sufficient causal conditions (i.e., studying causal commonalities shared by instances of an outcome), researchers should work forward from causal conditions, especially combinations of conditions, and assess their sufficiency. To assess the sufficiency of a cause or causal combination, the researcher must determine whether or not the cause or combination of causes in question always, or virtually always, produces the outcome in question. Evidence that there are instances in which the cause or causal combination is not followed by the outcome challenges the researcher's claim that the cause or causal combination is sufficient. The assessment of sufficiency, therefore, involves searching for cases that are similar with respect to the relevant causes and then assessing whether or not these cases display the same outcome.

For illustration, suppose two combinations of conditions exist that produce "defection-related turnover" in HMOs: (1) a change in management combined with a speed-up in the patient flow and (2) extensive use of outside medical specialists combined with management appropriation of the power to veto referrals to specialists. Because there are two sufficient combinations of conditions and no single condition is shared by the two combinations, the analysis of necessity would show that no single causal condition (and no causal combination) is necessary. Likewise, the analysis of sufficiency would show that no single causal condition is sufficient. Thus, no single cause is either necessary or sufficient. However, the analysis of the sufficiency of combinations of conditions would show that the two listed combinations are sufficient.

The relevant pattern of results is presented in Table 1. In the analysis of either of the two sufficient causal combinations, there would be cases in cell 1 because there is plural causation--more than one way to generate the outcome. However, both of the listed combinations would pass the test of sufficiency. In each test, cell 4 would be void or virtually void of cases, while cell 2 would contain instances of defection-related turnover explained by the causal combination in question. As long as there are cases in cell 2 and no cases (or virtually no cases) in cell 4, then the researcher may argue that the causal combination is sufficient, assuming that theoretical and substantive knowledge support this interpretation. Cells 1 and 3 are not directly relevant to the assessment of sufficiency.

More generally, Table 1 demonstrates that it is possible to assess the sufficiency of causal combinations one at a time, in isolation from one another. This conclusion is important because of its implications for the study of causal complexity. If, as I have argued, we live in a world of great causal complexity, then a common pattern will be for outcomes to result from different combinations of causal conditions. Although it might seem that causation this complex would befuddle analytic social science, it is clear from the simple example just presented that the analysis of the sufficiency of causal combinations can proceed in a straightforward manner.