We examine the educational production function and efficiency of public school districts in Illinois. Using nonparametric kernel methods, we find that most traditional schooling inputs are irrelevant in determining test scores (even in a very general setting). Property tax caps are the only relevant factor that is related to districts' financial constraints and have predominantly negative associations with test scores. Therefore, insufficient resources may be partially responsible for the lack of growth in test scores. For most other relevant inputs, we find substantial heterogeneity in the returns, which helps reconcile some of the puzzling results in the literature. We further find that there exist inefficiencies in school districts. Moreover, the level of test scores, commonly used as a measure of school effectiveness, (while related) differs substantially from our efficiency scores, and standard parametric approaches drastically underestimate school efficiency. We discuss the policy implications of our results.