BACKGROUND: Preterm birth (PTB) and small-for-gestational-age (SGA) disproportionately affect women who are Black or Asian. Structural racism produces health inequalities. Identifying latent socioeconomic classes may help to understand the role socioeconomic position (SEP) plays in this inequality. METHODS: We included women from the baseline survey of the UK-based Millennium Cohort Study who had a live singleton pregnancy and gave birth from 1 September 2000 to 11 January 2002. Relative risks (RR) with 95% confidence intervals (CI) for PTB and SGA were estimated for ethnic groups compared with women who were white, with adjustment for SEP. Latent SEP classes were then identified using diverse prospective socioeconomic data. Mediation of health inequality via SEP and latent SEP class was tested. RESULTS: Among 17 701 included women, 6.7% (95% CI 6.2%, 7.1%) experienced PTB and 7.0% (6.5%, 7.5%) SGA. We found evidence that the association between ethnic groups and PTB was mediated by latent SEP class for women who were Bangladeshi, Black African, Black Caribbean and Pakistani, with indirect 'effects' of RR 1.08 (1.01, 1.16), 1.07 (1.01, 1.14), 1.06 (1.00, 1.12) and 1.06 (1.00, 1.13), respectively, relative to White. When using the simple measures of maternal education, household income and marital status, we found no evidence of mediation except for a potential protective effect among Indian women, relative to White. We found similar evidence for SGA, with indirect effects through latent SEP class of RR 1.35 (1.19, 1.52), 1.32 (1.17, 1.48), 1.26 (1.12, 1.41), 1.27 (1.13, 1.42), respectively. When using the simple measures, we found evidence of mediation only among Black African and Black Caribbean women, with RR 1.16 (1.04, 1.30) and 1.12 (1.00, 1.26), respectively, relative to White. CONCLUSION: The determinants of inequality appeared to differ by ethnicity. We demonstrated the mediating role of individual-level SEP and a role for latent class analysis to interpret complex combinations of socioeconomic data.