To estimate the determinants of stunting using rich data from a birth cohort study from urban South Africa and to examine the various mechanisms, both proximate and distal, through which maternal education affects stunting.

Multivariate regression analysis using birth cohort data, where the outcome variable was stunting at age 2 years, and multiple mediator analysis to identify pathways from maternal education to stunting.

South Africa’s largest metropolitan area, Soweto-Johannesburg.

Participants of Birth to Twenty Plus, a longitudinal cohort study of children born in 1990 (n 691).

In multivariate analysis, the birth weight Z-score (−0·084; P<0·001; 95 % CI −0·11, −0·06), the mother’s openness towards modern health care, captured by a vaccination score (−0·05; P=0·04; 95 % CI −0·10, −0·00), and a better-quality care environment (−0·015; P=0·04; 95 % CI −0·03, −0·00) were found to be negatively associated with stunting. Having experienced symptoms of illness related to ears and eyes increased the risk of stunting (0·038; P=0·01; 95 % CI 0·01, 0·07). Results of the mediation analysis showed that maternal education had an indirect effect on stunting largely through socio-economic status and the antenatal environment (measured by the birth weight Z-score).

Overall, many of the factors that were protective against stunting in the final analysis, whether they operated through maternal education or not, were related to the mother’s contribution to the child’s life. This reinforces the idea that to minimise stunting, enhanced antenatal and postnatal services to better support and empower mothers may be important.