QUALITATIVE COMPARATIVE ANALYSIS VIS-A-VIS REGRESSION
2
For all three of these categories, QCA is found to require more restrictive assumptions
that regression analysis. Therefore, with respect to these foundational issues, QCA does
not represent an improvement over regression.
Introduction
This paper proceeds from the belief that Qualitative Comparative Analysis has made
a major contribution to methodology in the social sciences, and, in particular, that Charles
C. Ragin’s (1987, 2000) work on this technique has been invaluable. For these reasons,
Qualitative Comparative Analysis deserves close scrutiny.
Qualitative Comparative Analysis (hereafter QCA) commands wide attention in part
because some of its features fit quite closely with standard qualitative intuitions about causal
inference. In particular, this approach more directly captures the idea of placing central
analytic emphasis on comparing different types of cases, while deemphasizing comparisons of
abstract ”variables” in isolation from the cases that they describe. Writings in this tradition
have likewise pushed many scholars to pay more attention to the potential importance of
interaction terms in causation. Research in the QCA tradition has therefore had a markedly
positive effect on the general intellectual climate of social science methodology.
In addition to raising these important themes, Ragin has presented QCA as a pow-
erful tool for causal inference. One of the most frequently discussed and potentially most
important bases for this claim is that this approach is said to rely on fewer restrictive as-
sumptions about the causal processes under study. For example, Ragin argues that QCA
avoids the ”homogenizing” or ”simplifying assumptions” behind regression analysis and
other statistical techniques (e.g., Ragin, 1987 x, xii, 32, 61-64, 103, 105, 166; Ragin, 2000).
Likewise, the approach is said to make ”no assumptions about the empirical scope or power
of the causes examined in social research” (Ragin, 2000).
Such assertions are compelling because they promise an alternative, in nonexperimen-
tal contexts, to mainstream quantitative analysis that avoids the major weakness of that
approach: i.e., untested analytic assumptions (Collier, Brady, & Seawright, 2004). Indeed,
Convention