This paper studies optimal education subsidies when parental transfers are unequally distributed across students and cannot be publicly observed. After documenting substantial inequality in parental transfers among US college students with similar family resources, I examine its implications for how the education subsidy should vary with schooling level and family resources to minimize inefficiencies generated by borrowing constraints. Unobservable heterogeneity in parental transfers creates a force to heavily subsidize low schooling levels chosen by borrowing-constrained students with low parental transfers. This force is stronger for rich families, but it is weakened if heterogeneity in returns to schooling also leads to different schooling choices. These mechanisms are quantified using a calibrated model. Quantitative analysis suggests a reform that reallocates public spending toward the first two years of college. The reform also reduces the gap in subsidy amounts by parental income during early years of college.