Determinants of Maternal Wages across the Wage Distribution: Evidence from Quantile Regression

Understanding the factors that shape mothers' earnings is central to debates over wage inequality, family policy, and the labor market consequences of childbearing. While a substantial literature has documented the average determinants of maternal wages, conventional mean-based Ordinary Least Square (OLS) regressions can obscure important heterogeneity in how individual and family characteristics operate across the entire wage distribution. Returns to education, age, and family structure may differ substantially between mothers at the lower and upper ends of the wage distribution, with implications for how we interpret wage gaps and design targeted interventions.

This study draws on linked administrative data from the Oregon Health Authority, the Oregon Department of Human Services (ODHS), and the Oregon Employment Department (OED), which were probabilistically matched using non-public unique identifiers. The population data consisted of first-time mothers in Oregon with singleton births during 2016–2017, followed quarterly across a 48-month window centered on childbirth. Maternal quarterly earnings were derived from administrative wage records reported by employers to the OED, providing a population-level, longitudinal measure of income that avoids the underreporting problem which is common in survey-based wage data. The final estimation sample consisted of 289,281 mother-quarter observations.

We used simultaneous quantile regression to estimate the conditional effects of mother's race, education, age, and marital status on maternal quarterly wages at the 10th, 25th, 50th, 75th, and 90th percentiles of the conditional wage distribution. Bootstrap standard errors based on 200 replications were used for inference. This approach allowed the marginal effect of each covariate to vary across the distribution rather than being constrained to a single average effect.

Returns to maternal education and age were highly skewed toward the upper end of the distribution. The marginal effect of an additional level of education at 90th percentile was approximately eight times larger than at the 10th percentile, with a similar gradient for maternal age. Model fit improved across quantiles, with pseudo R² rising from 0.05 at the 10th percentile to 0.30 at the 90th, indicating that observable characteristics explain considerably more wage variation among higher-earning mothers. Secondly, the wage gap between white and nonwhite mothers reversed sign across the distribution: white mothers earn more at the bottom but less at the top, conditional on education, age, and marital status, consistent with strong positive selection among nonwhite mothers who reach the upper tail. Potentially because nonwhite mothers face structural barriers to high-wage employment disproportionately, those who reached the upper tail are a highly selected group with unusually strong unobserved characteristics (e.g. ability). Further, we found that marriage is associated with a wage premium that grows monotonically across the distribution, suggesting that benefits of marriage for maternal earnings are concentrated among higher earners.

Taken together, these results demonstrate that mean-based estimates substantially understate the heterogeneity in the determinants of first-time mothers' wages. Future extensions will incorporate mother fixed effects, add regional dimension and examine race-by-education interactions. Improved understanding of maternal wages across the distribution highlights areas where targeted cash assistance policies may have the greatest impact in offsetting economic insecurity.