Heterogeneous income gradients in early motherhood in Mexico: a population-based multi-method study, 2000-2020

Early motherhood remains a major public health and equity challenge in low- and middle-income countries, especially in Latin America. While poverty, education, gender norms, and barriers to sexual and reproductive health are well-established determinants, how household income translates into (or fails to translate into) delayed childbearing remains insufficiently understood-particularly in highly unequal settings marked by informality and fragmented social protection.

To examine the association between household income and early motherhood in Mexico, and to disentangle direct and indirect socioeconomic pathways linking income to early childbearing.

We conducted a population-based, multi-method analysis using harmonised municipal microdata from the 2000, 2010, and 2020 Mexican Population and Housing Censuses (n = 3.05 million women aged 12-24 years, representing 30.6 million individuals). Total household income was imputed using Quantile Random Forests trained on nationally representative income-expenditure surveys. Our analytical strategy combined: (1) machine learning models to identify key correlates and non-linear individual-level patterns of early motherhood; (2) Bayesian spatial models (INLA) to quantify subnational clustering and contextual heterogeneity; and (3) dynamic structural equation modelling to assess direct and lagged relationships among income, schooling-related indicators, union formation, and early motherhood outcomes.

Early motherhood prevalence declined modestly from 21.46% (2000) to 19.31% (2020) and remained concentrated in socioeconomically disadvantaged households. In individual-level models, marital status, age, household structure, and dependency ratio were the dominant correlates; income ranked eighth in feature importance. Partial dependence plots indicated a non-linear inverse L-shaped association: predicted risk increased up to approximately Int$ 1,760 (2018 PPP) per adult equivalent, then plateaued and declined slightly. Parity-specific models showed that income's predictive relevance weakened with higher parity, with U-shaped patterns at the lowest income levels for second and ≥third births. Municipal analyses revealed persistent spatial clustering after adjustment for key predictors, and residual "place" effects consistent with unmeasured contextual constraints shaping early motherhood. In dynamic models, the direct income-early motherhood path was not statistically significant. Overall patterns were consistent with predominantly indirect linkages operating through schooling- and union-related indicators and reinforcing disadvantage dynamics at the municipal level.

In Mexico, early motherhood remains strongly shaped by intersecting social and territorial inequalities. Income gradients are modest and appear to operate mainly through schooling lag and union-formation pathways rather than as an isolated driver. Equity-oriented strategies are likely to be most effective when they integrate school retention, prevention of early unions, and youth-friendly sexual and reproductive health services, prioritising municipalities with persistently elevated risk.