Academic journal article Political Research Quarterly

An Introduction to Crisp Set QCA, with a Comparison to Binary Logistic Regression

Academic journal article Political Research Quarterly

An Introduction to Crisp Set QCA, with a Comparison to Binary Logistic Regression

Article excerpt

The authors focus on the dichotomous crisp set form of qualitative comparative analysis (QCA). The authors review basic set theoretic QCA methodology, including truth tables, solution formulas, and coverage and consistency measures and discuss how QCA (a) displays relations between variables, (b) highlights descriptive or complex causal accounts for specific (groups of) cases, and (c) expresses the degree of fit. To help readers determine when QCA's configurational approach might be appropriate, the authors compare and contrast QCA to mainstream statistical methodologies such as binary logistic regressions done on the same data set.

Keywords: comparative politics; political methodology; qualitative methods

Introduction

Qualitative comparative analysis (QCA)1 techniques are intended as methods for bridging the gap between qualitative (case study oriented) and quantitative (variable oriented) approaches in social scientific research. For simplicity of exposition here, we will limit ourselves to the dichotomous form of QCA, namely, to what is called crisp set QCA. The crisp set form of QCA allows for direct comparison to standard statistical techniques for handling variables treated as dichotomous and allows us to better compare and contrast the uses and theoretical objectives of QCA with those of more traditional methods so that readers may better judge for themselves when use of QCA is appropriate. To more clearly show the nature of differences between QCA and standard statistical approaches, we provide both a crisp set QCA and a binary logistic analyses of one particular data set, Charles Ragin's data on welfare states (see Ragin 2000, Table 10.6).

We begin with a discussion of four basic elements of QCA: (1) data tables, (2) truth tables, (3) solution formulas, and (4) parameters of fit. We then introduce three general aims for presentation of empirical analytic results that are not specific to QCA but to which QCA - due to its location at the intersection between case study and variable-oriented research - must pay particular attention. These aims consist of (a) displaying relations between variables, (b) indicating which descriptive or causal accounts apply to specific (groups of) cases, and (c) expressing the degree of fit of the proposed solution to the empirical data from which it was generated. For each of the standard elements of QCA we consider the degree to which it satisfies each of the three central goals of QCA data presentation identified previously.

Finally we compare QCA with logistic regression analyses of data on welfare states in sixteen countries.

Central Features of QCA

QCA is based on the twin ideas of necessity and sufficiency.2 Its motivations include a concern for unraveling causally complex structures in terms of equifinality, multifinality, and asymmetric causality (see discussion in the following) that tend to be omitted or slighted in most discussions of mainstream statistical methods.3 It is also explicitly configurational in approach (Rihoux and Ragin 2008). Moreover, unlike many statistical techniques, QCA does not require that at least some variables be measured at an interval or ratio level. In particular, for simple crisp set QCA, the data are in the form of (dichotomous) set membership scores in underlying concepts.

While QCA is sometimes thought to be strictly limited to small n, this is erroneous (see e.g., Ragin and Fiss 2008, and examples of large ? analyses such as Ragin and Bradshaw 1991 and Miethe and Drass 1999). However, even when the number of cases is large enough that we might apply standard statistical methods appropriate for dichotomous or ordinal data, the goals of QCA are different from those of other statistical techniques, and the results it produces are also different. As Ragin (2008b, especially 176-89) emphasized, while standard statistical techniques are good at distilling the net effect of single variables, QCA, by virtue of giving premium to causal complexity, seeks to detect different conjunctions of conditions (configurations) that all lead to the same outcome. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.