Decomposing the gender gap in subjective well-being: evidence from a machine learning-based switching framework

Human well-being has received increasing attention as an indicator of individual and social health and as a consideration in public policy. Gender differences in subjective well-being (SWB), a widely accepted indicator of human well-being, have long been documented. However, explanations of these gender differences remain varied and inconclusive. This study utilizes Gallup World Poll data comprising 1,911,212 observations from 2009 to 2022. Employing advanced machine-learning techniques within an Exogenous Switching Treatment Effect Model (ESTEM) framework, we decompose the gender gap in SWB into differences in characteristics and differences in regime-specific returns. Conditional on observed characteristics and under a common predictive regime, the model predicts higher SWB for men. In contrast, the regime-specific returns component is associated with lower predicted well-being under the male regime. This component also increases over time, widening the estimated gender gap in SWB, while its magnitude varies across countries. The results suggest the potential relevance of institutional flexibility and reduced gender-role pressures for addressing gender disparities in human well-being.