Abstract
Empirical studies of fertility in traditional agriculture report either a positive or a negative estimate for landholding. Estimation with data from rural Bangladesh, in this study, shows that the uncontrolled effect of landholding on fertility is positive; however, when the impact of pooling land in extended families is controlled for, the effect turns negative; yet the addition of a quadratic term for landholding provides a significantly U‐shape relationship. The non‐monotonic relationship is analysed in terms of ‘prestige‐cost’ of child labour which is assumed to increase with landholding, but to generate differential child contributions to household income and potential old‐age support to parents.