This analysis is motivated by recognition that anti-poverty interventions often affect both the level and composition of assets held by beneficiaries. To assess the conventional view that assets uniformly improve childhood development through wealth effects, we use three waves of panel data from Tanzania and test whether different types of assets have differential effects on children's educational outcomes. Our results indicate that household durables and housing quality have positive effects, but agricultural assets have adverse effects on children's highest grade completed and exam performances. We use a Hausman-Taylor instrumental variable (HTIV) panel data estimator to identify the effects of both time-varying and time-invariant endogenous variables. We find that the negative effect of agricultural assets is driven by large agricultural equipment and livestock ownership and the negative effect is more pronounced among rural children, poor children, and children from farming households, presumably due to the higher opportunity cost of schooling.