Estimation of a Doubly Heteroscedastic Stochastic Frontier Cost Function

This article extends the results obtained by S. B. Caudill, J. M. Ford, and D. M. Gropper (1995) in estimating and testing stochastic frontier cost functions in the presence of heteroskedastic inefficiency using cross-section data. It suggests the estimation of the model under the assumption of heteroskedasticity in both random terms. This avoids any possible misspecification bias and the use of incorrect tests. The same data as used by Caudill, Ford, and Gropper supports the new specification. Moreover, the author finds that firm-specific inefficiency measures are extremely sensitive to the proposed correction for heteroskedasticity.