BACKGROUND:
Ethnic inequalities in maternal and infant health persist across low‑ and middle‑income countries (LMICs). Whilst we know socioeconomic status (SES) matters, the contribution of inequalities in health services to ethnic inequalities in infant mortality is underexplored.

METHODS:
We analysed 266,222 births from 35 LMICs (2009–2022; Demographic and Health Surveys). Within each country we quantified ethnic inequalities using weighted mean absolute deviation (WMAD) and its relative form (RWMAD). We applied Oaxaca–Blinder decomposition to ANC gaps between ethnic groups to partition gaps into SES‑explained and unexplained components. Using pooled logistic regression with SES covariates and country fixed effects to predict infant mortality (IMR) under two counterfactuals: ANC equalised (i) to the country mean and (ii) to the best‑performing ethnic group.

RESULTS:
Within‑country ethnic inequalities were substantial. Country‑level ANC coverage and content WMAD averaged 0.35 points (RWMAD 0.07), while IMR WMAD averaged 5.5 per 1,000 (RWMAD 0.21). SES only explained a small proportion of the inequalities in ANC proportion (0.214 (95% CI 0.201–0.227)). Improving ANC in underserved ethnic groups could substantially reduce ethnic inequalities within countries – upto 30% reduction in inequalities. Countries with larger inequalities in ANC had larger inequalities in IMR, and these were the countries with the largest potential reductions from equalised increasing ANC coverage.

CONCLUSIONS:
Equalising ANC content across ethnic groups is associated with meaningful reductions in ethnic IMR inequalities within countries, however most gaps in ANC coverage and content were not captured by SES. This is consistent with an interpretation of structural/systemic determinants in driving inequalities. Policy should pair poverty and education efforts with targeted, intercultural and quality‑improvement actions for underserved groups and facilities.