Categorical Variables Without Categorical Thinking? A Relational Reading of the Sri Lankan Demographic and Health Survey

Abstract

Connell’s theory of gender relations is among the most influential and comprehensive frameworks for analyzing gender. Its influence across methodological approaches has been uneven, though, and Connell herself has argued that relational theory is incompatible with statistical analyses, which, by relying on categorical dummy variables to indicate the sex or gender respondents, essentialize and fundamentally misrepresent gender. We argue that categorical variables do not require “categorical thinking.” We outline three necessary steps for linking statistical analyses and relational theory: contextualizing statistical rates and means; highlighting within-group variation and the process through which it is produced; and contextualizing the data collection and research process more broadly. To illustrate our approach we critically examine data from the 2006–2007 Sri Lanka Demographic and Health Survey. We show that, far from being incompatible, relational theory is often vital for understanding the meaning of statistical data, as well as for critiquing and evaluating any resulting claims. When interpreted within a relational framework, statistical data can also clarify how gender structures the lives and experiences of people of all genders.