The estimation of education production functions has provided direct evidence about the effectiveness of various educational policies. Most specifically, the existing analyses that explain differences in student outcomes by the influences of families, peers, and schools strongly indicate that the current provision of schooling is very inefficient. Commonly purchased inputs to schools—class size, teacher experience, and teacher education—bear little systematic relationship to student outcomes, implying that many conventional input policies are unlikely to improve achievement. At the same time, differences in teacher quality, defined in terms of effects on student performance, have been shown to be very important, even if not closely related to salaries or readily identified attributes of teachers. In policy terms, this work has led to a general conclusion that how resources are used is often more important than how much resources are used.