Production frontiers and efficiency measures : concepts and applications

Abstract

Two major criticisms have been levelled against the statistical approach to measuring production efficiencies. First, the sampling distributional assumptions artificially imposed on the one sided-error term used to characterize inefficiency are somewhat restrictive. Moreover, alternative distributional assumptions can lead to substantially different results for the estimated technical efficiencies; making it difficult to provide an economic and practical justification of the choice of a particular distribution. Within the spectrum of inefficiency sampling distributions proposed in the literature, the half- or truncated normal has received a relatively wider applications than others such as gamma and exponential. Quite often the choice of the distributions is based on ease of empirical estimation. Second, the specification of the stochastic frontier production function in the statistical approach assumes that the effects of technical inefficiency on input productivity (or elasticity) are the same for each input with the resultant neutral shift of the frontier production function from the ‘average’ and firm-specific realized production functions. In other words, the frontier and the other production functions have identical slope coefficients (input elasticities) but different intercepts so that they merely represent neutral shifts from one another. While some attempts have been made recently in response to the first criticism, the second one appears to have so far attracted very little attention in the production frontier and efficiency literature. Thus the primary objective of this thesis is to develop an alternative conceptual framework to production efficiency measurement that aims at obviating both of these criticisms. Empirical illustrations based on survey agricultural data sets from Sri Lanka, China and India are provided to show the workability of the proposed procedures. The thesis format is as follows. Chapter 1 gives an overview of production efficiency analysis with more emphasis on technical efficiency measurement. Chapter 2 establishes for subsequent applications, the modelling, estimation and testing procedures of linear models with heterogeneity in both intercepts and slopes. Chapter 3 discusses and empirically demonstrates a method of measuring firm- and input-specific technical efficiency within a varying coefficients production function framework. Chapter 4 extends this method to a panel data context and discusses the measurement of temporal firm-specific technical efficiency and shifts over time of the frontier production functions (that is, technological progress). Chapter 5 focusses on total factor productivity growth over time. It explains a method to decompose the sources of total factor productivity growth into technological progress and changes in technical efficiency within the framework of the varying coefficients frontier production function approach discussed in the preceding chapters. In chapter 6, a primal method based on a varying coefficients production function is developed for estimating allocative efficiency. An empirical illustration is provided. The concluding chapter highlights some of the issues not explicitly addressed in the concluding final section of each chapter and also points out some of the problems, mainly of empirical nature, that may be encountered. Some directions for further investigations are briefly suggested.