Using data from BRIGHT, an integrated program that aims to improve school participation in rural communities in Burkina Faso, we investigate the impact of school subsidies and increased access to education on child work. Regression discontinuity estimates demonstrate that, while BRIGHT substantially improved school participation, it did not reduce – in fact may have increased - children’s participation in economic activities and household chores. This combination of increased school participation and work can be explained by the introduction of a simple non convexity in the standard model of altruistic utility maximizing households. If education programmes are implemented to achieve a combination of increased school participation and a reduction in child work they may either have to be combined with different interventions that effectively reduce child work or they may have to be tuned more carefully to the incentives and constraints the child labourer faces.