Estimation of a Doubly Heteroscedastic Stochastic Frontier Cost Function

Abstract This article extends the results obtained by Caudill, Ford, and Cropper in estimating and testing stochastic frontier cost functions in the presence of heteroscedastic inefficiency using cross-section data. It suggests the estimation of the model under the assumption of heteroscedasticity in both random terms. This avoids any possible misspecification bias and the use of incorrect tests. The same data as used by Caudill et al. supports the new specification. Moreover, we find that firm-specific inefficiency measures are extremely sensitive to the proposed correction for heteroscedasticity. KEY WORDS: Heteroscedastic mixture of normalsLikelihood ratio testStochastic frontier