Estimation of Technical and Allocative Inefficiencies in a Cost System: An Exact Maximum Likelihood Approach

Estimation and decomposition of overall (economic) efficiency into technical and allocative components goes back to Farrell (1957). However, in a cross-sectional framework joint econometric estimation of efficiency components has been mostly confined to restrictive production function models (such as the Cobb-Douglas). In this paper we implement a maximum likelihood (ML) procedure to estimate technical and allocative inefficiency using the dual cost system (cost function and the derivative conditions) in the presence of cross-sectional data. Specifically, the ML procedure is used to estimate simultaneously the translog cost system and cost increase due to both technical and allocative inefficiency. This solves the so-called ‘Greene problem’ in the efficiency literature. The proposed technique is applied to the Christensen and Greene (1976) data on U.S. electric utilities, and a cross-section of the Brynjolfsson and Hitt (2003) data on large U.S. firms.