Specification tests and a reformulation of frontier models: An alternative approach to efficiency estimation

Following Farrell's (1957) definitions of firm's technical and allocative efficiencies and the formulation of composite error stochastic frontier model by Aigner, Lovell and Schmidt (1977), a substantial amount of research work has been performed measuring firm's technical inefficiency using this model with various distributional assumptions of technical inefficiency. Most stochastic frontier models are overly restrictive both in terms of the functional form and the distributional assumptions of the error terms. Unlike other econometric models, frontier models are rarely subjected to rigorous specification tests. The first part of this dissertation is devoted in developing several straightforward specification tests for these models. The tests are based on information matrix and moment tests principles.Composite error frontier models assume only technical inefficiency in their formulation and estimation. The parameters estimates are not robust against the distributional assumptions of the asymmetric error term. Further, little success has so far been achieved in estimating firm specific inefficiency level. Jondrow, Lovell, Materov and Schmidt (1982) and Greene (1990) have suggested measures of inefficiency based on the distribution of inefficiency, conditional on the entire composite error. Usually, a firm's input allocation is quite likely to deviate from the optimal inputs allocation locus, in which case the assumption of no allocative inefficiency does not adequately represent a firm's true performance. Under the assumption of Cobb-Douglas inputs transformation technology, and based on the information on observed output level, cost, input quantities and their prices, the second part of this dissertation finds expressions for both technical and allocative efficiencies for the cost and output models separately, in terms of observable characteristics. Since the level of these efficiencies affect observed cost and output and thereby deviate from their respective frontier value, a firm's observed cost and output models have been derived separately, adjusting the frontier functions for their respective efficiency levels. The frontier functions are the results on first-order conditions of either cost minimization or output maximization (see Nerlove (1965), p. 107). These alternative models, we derive, do not have an asymmetric error term and are dual to each other. Therefore, this approach is a major step in removing asymmetric error problems in frontier literature. These observed cost and output models now can be used to estimate the parameters of the production function and thereby to find the firm specific estimate of overall efficiency and its factors; namely, the allocative and technical efficiency levels.Finally, the theoretical developments of this dissertation are used to analyze the well-known data set from the U.S. steam electric generating plants previously used by Cowing (1970), Schmidt and Lovell (1979).U of I OnlyETDs are only available to UIUC Users without author permissio