This paper develops and tests a simple model of parental allocations which stresses the e¢ ciency-equity tradeo§ they face when investing in children of varying ability. The empirical component of the paper focuses on inequalities in schooling and child labour. Using siblingbased Öxed e§ect models, I show that large di§erences in IQ test scores between siblings do not translate into large di§erences in schooling. The evidence also suggests that richer households compensate more than poorer ones. On child labour participation, the data show that there is a great deal of variation amongst brothers along this margin. If parents take compensating action they may be able to attenuate any harmful e§ects arising from these di§erences. I show that in fact there are substantial adverse e§ects: participation in paid employment during elementary school leads to rise of 8 percentage points in the probability of being a§ected by one of seven acute morbidity conditions. These results suggest that even if parents act to compensate along some dimensions, in the end they may be forced to pick and choose amongst their children for other allocations, leading to large inequalities.