Making the Cobb-Douglas Functional Form an Efficient Nonparametric Estimator Through Localization

This paper offers a new method for nonparametric estimation - based on the Cobb-Douglas form - that is simple, easy to interpret, efficient among the class of all linear nonparametric estimators, and nearly efficient among all nonparametric estimators. Thus empirical analysts who prefer the tractability of the Cobb-Douglas form may return to using a localized version of it with confidence that it is an efficient nonparametric estimator. This technique is applied to Nerlove's (1964) classic data set to estimate a cost function for U.S. electric utilities