This paper uses new data on school district capital stocks, stochastic frontier analysis, and a value-added measure of school quality to provide the fi rst direct evaluation of the relationship between school facilities and school district costs. We fi nd that the cost of education increases as the capital stock increases, suggesting either that school districts are grossly overcapitalized or that nicer facilities refl ect an important, unmeasured dimension of school quality. We also fi nd that cost function estimates for Texas are largely insensitive to the exclusion of the capital stock measures, suggesting that stochastic frontier cost function estimates without a capital stock measure are unlikely to be biased.