Marriage during childhood and adolescence adversely affects maternal and child health and well-being, making it a critical global health issue. Analysis of factors associated with women marrying ≥18 years has limited utility in societies where the norm is to marry substantially earlier. This paper investigated how much education Nepali women needed to delay marriage across the range of ages from 15 to ≥18 years. Data on 6,406 women aged 23-30 years were analysed from the Low Birth Weight South Asia Trial on the early-marrying and low-educated Maithili-speaking Madhesi population in Terai, Nepal. Multivariable logistic regression models assessed the associations of women's education with marrying aged ≥15, ≥16, ≥17 and ≥18 years. Cox proportional hazards regression models quantified the hazard of marrying. Models adjusted for caste affiliation. Women married at median age of 15 years and three-quarters were uneducated. Women's primary and lower-secondary education were weakly associated with delaying marriage, whether the cut-off to define early marriage was 15, 16, 17 or 18 years, with stronger associations for secondary education. Caste associations were weak. Overall, models explained relatively little of the variance in the likelihood of marriage at different ages. The joint effects of lower-secondary and higher caste affiliation and of secondary/higher education and mid and higher caste affiliation reduced the hazard of marrying. In early-marrying and low-educated societies, changing caste-based norms are unlikely to delay women's marriage. Research on broader risk factors and norms that are more relevant for delaying marriage in these contexts is needed. Gradual increases in women's median marriage age and increased secondary education may, over time, reduce child and adolescent marriage.