2
conclusion as to which causes apply.
Second, QCA has most frequently been applied within the field of comparative
method in which a limited number of extant cases reign but where analyses fail to culminate
in a summary measure. With few exceptions, QCA has not been applied to survey data in
which cases are individuals. By contrast, logistic regression is frequently applied to such data,
and analyses by default include probability and summary measures. The paucity of
individuals as cases in QCA has delayed the method from adopting probability and prevented
it from being effectively compared with other standard methods of survey research. Logistic
regression is therefore the case in point. It is expected that an application of QCA to survey
data will facilitate a comparison with logistic regression without loss of its inferential logic.
And it is also expected that logistic regression will be able to overcome limitations modelling
complexity so as to refute such criticism, as well as to make it comparable with QCA.
This paper has two general goals. First, to demonstrate how QCA can be applied to
survey data and large N, to suggest how QCA can incorporate probability, and to offer ways
in which logistic regression can respond to QCA analysts' claim of its inability to account for
causal complexity. Second and concurrently, to compare the degree to which QCA and
logistic regression converge, complement, or diverge from one another with views to causal
effects, model summaries, and complexity.
First, Braumoeller (1999) holds that the trend toward simplification in the social sciences
stems from a prevalence of complexity-deferring methods. If QCA demonstrates greater
complexity than logistic regression, an initial conclusion may be that the latter is a
complexity-deferring method. But second, method is also a logistical question of how data is
collected and analyzed, in contradistinction to design, which is the logical question of
ensuring that the evidence obtained enables the analyst to answer the research question (Vaus,
2001). If logistic regression demonstrates causal complexity equal or greater to QCA, then a
careful choice and calibration of method is as important to research as are theory and design.
1
The paper uses terms from different research traditions; antecedents, causes, conditions, and
independent variables will therefore be used interchangeably.
2
The Qualitative Comparative Method (QCA) used in this paper is based on the method as
advanced by Ragin (1987). Later developments alter some aspects of the method while also
introducing new ones (Ragin, 2000). These recent developments are not found pertinent to
this analysis. Studies that contrast or complement QCA with other methods include eg, Ragin,
Mayer, and Drass (1984), Ragin and Bradshaw (1991), Kangas (1994), Berg-Schlosser and
De Meur (1997), Foweraker and Landman (1997), Olli (1999), and Ragin, Shulman,
Weinberg, and Gran (2003).
Convention