Qualitative Comparative Analysis vis-a-vis Regression
Jason Seawright
University of California, Berkeley
Paper Presented at the 2004 Meetings of the American Political Science Association
Abstract
Comparisons of Qualitative Comparative Analysis versus regression analysis have not
adequately considered the assumptions about causation behind each of these tools. Yet it is
important to ask the generic question: how many untestable, or hard-to-test, assumptions
must be met for us to believe the results of any particular analytic technique?
This paper attempts to address this question by considering three of the most im-
portant families of assumptions employed in regression analysis: assumptions about correct
functional form, missing variables, and inferring causation from association. For each as-
sumption, the role of corresponding assumptions in QCA will be explored and illustrated
through an analysis of left party electoral fortunes in Latin America. Regarding the func-
tional form of causal relationships, QCA builds strict assumptions into measurement proce-
dures. Regarding missing variables, while earlier versions of QCA require a strong assump-
tion of no causally relevant missing variables, more recent procedures allow some kinds of
missing variables, but build in mutually contradictory statistical assumptions about those
variables. Correcting these contradictions essentially converts QCA into an application of
regression analysis. Regarding inferring causation from association, QCA makes causal in-
ference on the basis of patterns of association by pure assumption. That is, association is
assumed to have a one-to-one relationship with causation.
Page 1 of 29 
Convention