We develop a theoretical model where child labour results from a household's trade-off between sending a child to school or to work. Education is considered as a risky investment, since the survival of the child is not certain. We explore the effects of public expenditure on education and health on child labour. On the one hand, we establish that health expenditure reduces child labour more when child mortality rate is high. On the other hand, a moderate aversion to risk is a necessary condition for expenditure on education to have a positive effect on child labour. Our theoretical results are tested empirically on data from 66 developing countries between 1985 and 2000.