Nonparametric analysis of returns to scale in the US hospital industry

This paper presents new estimates of scale economies for US hospitals. We show that the common translog specification of hospital costs is a misspecification, and employ nonparametric, local linear estimation with both continuous and discrete covariates. A bootstrap method is used to provide inferences regarding ray scale economies and expansion path scale economies for a large sample of hospitals covering 1984-1996. We find evidence of changes in the structure of hospital costs over the sample period, as well as evidence of locally optimal hospital sizes.