Should a typical developing country invest more in agriculture or education? At what stage of development is it optimal to invest more in each of these sectors? Every developing country government grapples with these questions annually when designing a national budget. In this paper, I provide estimates of agricultural returns to schooling in Malawi- evidence of such returns implies a more complex non-separable decision process to answer the first question. While a large development economics literature has documented the effects of schooling on agricultural incomes, such estimates are potentially biased because of unobserved heterogeneity and selection bias. In this paper, I use 2010-2013 two period nationally representative panel survey data in Malawi and rely on the exogenous education policy changes and spatial variation in access to schooling to identify effects of schooling on agricultural incomes. In addition, I use recent econometric methods to correct for selection into income activities within a panel data and instrumental variables estimation framework. I find annual agricultural returns to one additional year of schooling in Malawi that range from 3% to 7%.