Operationalizing Intersectionality in Quantitative Research - What Do Classification Trees Reveal about Gendered Divisions of Housework?

Jenny Chanfreau, University of Sussex
Wendy Sigle, London School of Economics and Political Science (LSE)

Intersectionality emphasizes need to attend to multiple interlocking axes of marginalization and within-group diversity. With linear regression models, such diversity can be captured to some extent by introducing interaction terms. Although recent reviews of quantitative intersectionality found regression to be the most common method, previous scholars have outlined some limitations of this approach. Models with complex and high-level interactions can be difficult to present and interpret. Moreover, the small size of some sub-groups, particularly in survey data, can seriously limit which variables and how many variables can be interacted in regression models. In this paper, we consider the ways decision tree methods can be used to identify patterns of “intra-categorical” complexity. Focusing on the gendered divisions of housework in different-sex couples with small children, and using data from the first sweep of the UK Millennium Cohort Study, we use Classification and Regression Trees (CART) to identify profiles of families most likely to share housework equally. In a context where few couples actually share housework, CART can be used to identify small but potentially theoretically interesting vanguard groups (e.g. male breadwinning couples with 2+ children, highly-educated fathers and a large age gap between parents) that might otherwise evade researchers’ attention.

Keywords: Gender Dynamics, Families, Unions and Households, Inequality, Disadvantage and Discrimination, Data and Methods

See extended abstract.