A Machine Learning Approach to Understanding the Determinants of Female Labour Force Participation Rate in India

Female labour force participation (FLFP) in India remains low despite improvements in education and economic growth, forming what is often referred to as the “FLFP puzzle.” This study examines which factors most strongly predict women’s participation in the labour market, and whether these patterns reflect individual characteristics or deeper constraints.

Using microdata from the Periodic Labour Force Survey (PLFS), the analysis applies logistic regression, Random Forest, and XGBoost models to evaluate the relative importance of demographic, socio-economic, and household-level variables. Across models, marital status, sector of residence, and life-cycle effects (captured through age) consistently emerge as the strongest predictors of participation, while education, religion, and social group play a comparatively limited role.

Comparative evaluation using multiple performance metrics shows that non-linear models outperform logistic regression, indicating the presence of complex relationships not fully captured by linear specifications. However, recall remains below 0.5 across all models, suggesting that fewer than half of actual participants are correctly identified. This highlights the limits of observable characteristics in fully capturing participation behaviour.

These findings indicate that FLFP is more strongly associated with household context, labour market conditions, and lifecycle dynamics than with individual attributes alone. At the same time, the results demonstrate that predictive models, while useful in identifying consistent patterns, cannot fully explain participation outcomes.

Overall, the study shows how machine learning methods can complement traditional approaches by uncovering stable predictive relationships, while also highlighting the need for richer data and integrated predictive–causal frameworks to better understand the drivers of female labour force participation in India.