Couples’ Household Labor: New Insights with Observed-Synthetic Data Using Supervised Machine Learning and Actor-Partner Interdependence Model

• In Event: 30802 - Machine Learning and Classical Statistical Approaches: Trade-offs, Integration, and Debate (Co-sponsored by Section on Mathematical Sociology)

Abstract

Gender inequality in the division of household labor remains persistent to a certain extent despite decades of progress in both the work and family spheres. Paradoxical patterns were found in existing studies, where couples with higher social status, resources, or egalitarian views to reduce gender inequality tend to exhibit more traditional gendered constraints. This study argues that gender gaps in household labor and the gendered impacts of key social status indicators are interweaved in the theoretical discussions and obscure the picture. Further, the empirical analyses of couples’ divisions of household labor must consider dyadic observations of couples’ household labor and couple-based perspectives of social status and within-couple power relations—educational assortative mating, household specialization, and occupational structure. While the American Time Use Survey (ATUS) provides rich socio-demographic information on households and the high-quality time diary of one sampled individual per household, its capacity for couple analysis is constrained by the missing-by-design of the partners’ household labor. To overcome the empirical limitation, this paper combines machine learning and multiple imputations to construct observed-synthetic dyads for couples’ household labor time. This approach enables couple-level analysis with seemingly unrelated regression in the Actor-Partner Interdependence Model framework, allowing within-couple interdependence in model residuals by unobservable factors for unbiased estimates. Using ATUS 2014-2019 with 24,694 couples, this paper finds narrowed gender gaps by asymmetric and gendered effects of higher social status. The effects of education were exhibited mainly through the college-educated women’s partners’ input compared to their counterparts’ partners. Occupational structure influences couples through the professional women’s significant reduction in household labor time, although not relevant to men’s. However, while women also benefit from full-time employment to have narrowed gender gaps across various forms of household specialization, women still performed significantly more household labor even when their partners were part-time or non-working.

Author

• Xingyun Wu, Johns Hopkins University