Access to education is usually found to be highly correlated with household income and wealth. This correlation often instigates an expectation that increasing income of the poor households will lead to greater human capital accumulation. This paper exploits randomized roll-out of a large-scale livelihood development programme for the ultra-poor in Bangladesh to measure the effect of asset transfer and livelihood supports on children’s schooling. We find limited impact on enrolment although this programme has been extremely successful in transforming the economic lives of the ultra-poor and causing substantial increases in their income and productive assets. The beneficiary households are also found to have increased their expenditures on education. This increase in educational investment, however, has not affected educational attainment during the evaluation period. We also find that the programme increased the extent of child labour immediately after asset transfers. The level of this impact on children’s work declines two years after the interventions ended. The increases in child labour are concentrated in activities related to livestock rearing, which is the primary type of asset transferred in this programme. However, we do not find evidence indicating a trade-off between children’s enrolment and work. The evidence suggests that asset transfer programmes can be more effective by including additional components focusing on improvement in educational outcomes instead of relying primarily on spillover effects through income gain.