This paper uses exogenous variation in rainfall across districts in Uganda to estimate the causal effects of household income shocks on children's enrollment and academic performance conditional on gender. I find negative deviations in rainfall from the long-term mean to have negative and highly significant effects on female enrollment in primary schools and the effect grows stronger for older girls. I find no effect of rainfall variation on the enrollment of boys and young girls. Moreover, I find that when schooling is free of charge and both marginal boys and girls are enrolled, a negative income shock has an adverse effect on the test scores of female students while boys are not affected. The results imply that households respond to income shocks by varying the amount of schooling and resources provided to girls while boys are to a large extent sheltered - a finding consistent with a model where parents' values of child labor differ across sexes.