Empirical Analysis of Intersectionality

Dubrow, Joshua Kjerulf.  2008.  “How Can We Account for Intersectionality in Quantitative Analysis of Survey Data?  Empirical Illustration of Central and Eastern Europe.”  ASK: Society, Research, Methods 17: 85-102.

 

 

 

Emerging from feminist scholarship over fifteen years ago, intersectionality theory has enjoyed a spectacular rise in popularity (Davis 2008).  In general, the common principles of intersectionality theory are the following: individuals belong to multiple demographic categories, so that the same individual has a specific gender, an ethnicity, and a social class position, among others; some categories provide advantages and some disadvantages, with each having roots in social stratification structure; not only each of these disadvantaged categories has its separate attitudinal and behavioral consequences but also jointly. (Steinbugler et al 2006; Warner 2008).  Despite sharp differences between variants of intersectionality theory, intersectionality has been called a research paradigm, forming “both a normative theoretical argument and an approach to conducting empirical research that emphasizes the interaction of categories of difference” (Hancock 2007: 63-4).  While most analyses of the influence of demographics on attitudes and behavior focus on the additive effect of categories, intersectionality theory focuses on the effect of categorical intersection above and beyond the effects of its components.  

 

How can we account for intersectionality in the quantitative analysis of survey data?  The few available guides speak in very broad terms and they do not provide empirical illustrations that would fit to the theoretical statements (Hancock 2007; McCall 2005; Weldon 2006; Bowden 2008; Warner 2008).  Reviewing the literature on the intersectionality theory it becomes obvious that there is a need of discussing and illustrating the basic methodological problems of how to test its main propositions. 

 

NOTE:  Accounting for intersectionality is rare in quantitative social science research for a number of reasons (Hancock 2007: 66; Simien 2007: 264).  First, it is difficult to capture the complexity intersectionality presents; the most popular places to publish quantitative social science research does not devote space enough for the length such an article would require (McCall 2005: 1787).  In addition, research that attempts to account for intersectionality rarely discuss the problems with testing the theoretical assumptions inherent in their models.  For example, one cannot assume that intersectionality can always be accounted for with interaction terms, as not all models of intersectionality agree that the constituent elements of the term are meaningful.

 

 

In building on past research, the aim of this article is to (a) examine the challenges of incorporating intersectionality into quantitative survey analysis; (b) compare and contrast the unitary, or additive approach – the most common way to analyze the influence of demographic variables on attitudes and behavior – with the core of the intersectionality theory, represented by the multiplicative approach; (c) propose the most appropriate statistical means for analyzing each, focusing mainly on multiplicative interaction terms; and (d) illustrate these approaches using the European Social Survey data for Central and East European countries. 

 

NOTE:  In discussing intersectionality, I do not make reference to the political bases of scientific research, e.g. that categorization leads to oppression.  While relevant to understanding intersectionality in its entirety, at this juncture reference to the dialectical relationship of politics and research is beyond the aims of this study.

 

 

In this paper political protest is the dependent variable. It is well established that the level of political protest in Central and East European countries is much lower than in Western Europe (Dubrow et al. 2008). However, the individual determinants of political protest within Central and East European countries have not been extensively examined.  Looking at gender, ethnicity, social class, and their intersection is intended to elucidate the mechanism through which individual attributes influence political behavior.  Substantively, I will test the basic propositions of intersectionality theory. 

 

Challenges of Incorporating Intersectionality into Quantitative Analysis of Survey Data

Applying intersectionality to quantitative analysis of survey data poses several difficult yet surmountable challenges.  For example, in the literature it is stressed that variable oriented analyses impose “within-case independence of categories” (Hancock 2007: 66; the history of the argument against variable analysis stretches back many decades: see Blumer 1956).  While the original survey data usually have separable demographic categories, combinations of them can be constructed in the form of interaction terms so that categories are not independent of each other. 

 

An additional challenge is that surveys are usually not designed with intersectionality in mind and demographic categories are represented in a limited number of cases with which to construct intersections (Hancock 2007: 66; see also Bowleg 2008: 314-317; McCall 2005: 1787).  Whether this challenge applies depends on particular surveys. Quantitative researchers interested in the intersectionality paradigm have enough cases for demographic items in the European Social Survey (ESS) to adequately construct and analyze intersections. 

 

NOTE: Survey questions such as “are you male or female?” are phrased in such a way as to divorce the categories from their institutional contexts, itself idea that is contrary to the intersectionality paradigm (Hancock 2007: 66).  See Bowleg (2008: 314-317) for how to construct survey data with intersectionality in mind.

 

 

Still another challenge is how to choose among demographic items.  A lesson from the democracy and descriptive representation literature is helpful here.  When constructing a representative governance body, some choices must be made as to which demographics should be represented.  Mansbridge (1999) and Dovi (2004) argued that each body requires a demographic composition appropriate to the task.  Thus, social contexts suggest which social cleavages and their intersections are the most salient for intersectional analysis.

 

Cross-national research has demonstrated that gender, ethnicity and class have profound consequences for a wide array of attitudes and behaviors and thus these variables should be tried first. 

 

Note, however, that as the intersectionality paradigm progresses, the focus on “master” categories – demographics that are theorized to encompass all aspects of a person’s identity, such as gender  — may give way to other, “emergent” – particularized and contextually contingent — categories of heretofore undiscovered but nonetheless salient social cleavages (Warner 2008: 457-9).

 

NOTE:  Hancock (2007) advocates Ragin’s fuzzy-set logic approach to guide collection of intersectionality appropriate data.  Warner (2008) suggests substituting master categories with constructs: the correlates of the master category can be identified and molded into a construct that has the properties of the master category that are strictly relevant to the attitude or behavior being studied (2008: 461).  For example, instead of a dichotomous gender variable, the correlates of being in either one category or another would be identified with existing questionnaire items; the “essential properties of being female” would be the explanatory variable, as opposed to the insensitive “gender” variable.

 

 

The intersectionality paradigm forces researchers to explore intersectionality in a somewhat inductive way, as existing theories that emerged from the unitary approach are unreliable guides. Weldon (2006) argues that because the intersectionality paradigm does not now provide general principles as to the attitudinal and behavioral consequences of intersecting categories within social systems — other than there are simultaneous advantages and disadvantages — research should be devoted to fleshing out these relationships (245).  Both of the intersectionality approaches “changes the relationship between the categories of investigation from one that is determined a priori to one of empirical investigation” (Hancock 2007: 67). 

 

Overall, despite concerns regarding the ability of current survey data and quantitative methods to address intersectionality, researchers equipped with these tools are up to the task.