The causes and consequences of child labour are examined theoretically and empirically within a household decision framework, with endogenous fertility and mortality. The data come from a nationally representative survey of Indian rural households. The complex interactions uncovered by the analysis suggest that mere prohibition of child labour, or the imposition of school attendance, could make things worse, and would be difficult to enforce. Beneficially reducing child labour requires changing the economic environment to which the work of the children constitutes, in the great majority of the cases, the rational response. Suitable policies include reductions in the cost of attending school, and public health improvements. The effects of these policies go far beyond direct impacts. Health policies have favourable indirect repercussions on the school attendance, demand for educational material, and labour participation of children. Educational policies have favourable indirect repercussions on the nutritional status of children. Both types of policies discourage fertility. Income re-distribution may be helpful, but land re-distribution could be counterproductive.