erc/metu
INTERNATIONAL CONFERENCE IN ECONOMICS  IV
September 13-16, 2000, Ankara

Estimating Technical Efficiencies from Panel Data: A Semi-Parametric Frontier Approach

Morteza Haghiri (University of Sakatchewan,Canada)
James F. Nolan (University of Sakatchewan,Canada)
Kien C. Tran (University of Sakatchewan,Canada)

Abstract

The historical discussion of measuring productivity and efficiency dates back to forty-nine years ago when Debreu (1951) and Koopmans (1951) addressed implicitly and Farrell (1957) proposed explicitly these issues to the economic literature. The problem of measuring efficiency of production units is important to both economic theorists and economic policy makers. Before Farrell’s (1957) proposition, a number of studies and attempts addressed the issue, which failed to combine these measurements into any satisfactory measure of efficiency because of neglecting the theoretical side of the problem. Efficiency means a comparison between observed and optimal values of a firm’s output and input, which can be viewed from both output (output-orientated measures), and input (input-orientated measures) sides. It can be measured as technical, allocative, and overall that is a product of the other two. To date, most efforts in measuring efficiency have been made by the estimation of frontier functions. The basic idea is to project the optimal (frontier) production function from a set of observations of particular production units and then calculate the distance of geometric locus of other production units from that optimal. Frontier functions can be estimated by three different approaches: (a) a parametric model that is a particular analytical function with a priori fixed number of parameters; (b) a non-parametric model, which is robust with respect to the particular form of the frontier and to the distributional assumptions; and (c) a semi-parametric model that is a combination of both parametric and non-parametric aspects. Semi-parametric models are often estimated in two steps: first, the effects of inefficiency are eliminated by a non-parametric method. Second, the parametric part of the model is estimated by standard parametric methods based on the filtered efficient data. This paper has four main sections: the first section explores the theoretical aspect of efficiency. A brief discussion of all methods of estimating technical efficiency is presented in section two. Empirical analysis of estimating technical efficiency from semi-parametric frontier models with panel data is presented in the third section. Finally, section four concludes some remarkable points.

Economic Research Center
Middle East Technical University
06531 Ankara Turkey
Phone: + 90 312 210 3044, 210 2003
Fax: +90 312 210 1244
e-mail: metuerc@metu.edu.tr