Stochastic Frontier Analysis

From MaRDI portal





This book is concerned with the development of a modified econometric approach to the estimation of production, cost and profit frontiers, and the estimation of the technical and economic efficiency with which producers approach these frontiers. Traditional textbook presentations of production economics, from Samuelson's ``Foundations of Economic Analysis to the present day, treat producers as successful optimizers. They produce maximum outputs allowable by the technology, they minimize the cost of producing, they maximize profit. Econometric practice, beginning with the pioneering work of \textit{C. Cobb} and \textit{P.H. Douglas} [A theory of Production. Am. Econ. Rev., Suppl. 18, 139-165 (1928)], has followed this theoretical paradigm. The least squares based regression techniques were used to estimate the parameters of production, cost, and profit functions. In these techniques the error terms were assumed to be symmetrically distributed with zero means. The only source of deviation from the estimated function was assumed to be statistical noise. However, not all producers are always so successful in solving their optimization problems. Consequently not all producers succeed in minimizing the expenditure required to produce the outputs. For this reason it is desirable to move the analysis of production, cost, and profit from the traditional functions toward frontiers. Thus a production frontier characterizes the minimum input bundles required to produce various outputs, or the maximum output with various input bundles. Producers operating on their production frontier are technically efficient, and producers operating beneath their production frontier are technically inefficient. The econometric implication of the proposed reformulation from functions to frontiers is that symmetrically distributed error terms with zero means are not appropriate when analyzing producer behavior. Consequently, error terms associated with frontiers are composed of a traditional symmetric random-noise component and a new one-sided inefficiency component. These composed error terms cannot be symmetric and they cannot have zero means. They must be skewed (negatively in the case of production, revenue, and profit frontiers and positively in the case of cost frontiers), and they must have nonzero means (negative in the case of production, revenue, and profit frontiers and positive in the case of cost frontiers). The authors description of the Chapters of the book is the following: Chapter 2 is devoted to the analytical foundations of producer theory and efficiency measurement. In Section 2.2 we characterize production technology with production frontiers in the single-output case and with distance functions in the multiple-output case. We also characterize technology with dual cost, revenue, and profit frontiers, which provide increasingly exacting standards against which to measure producer performance. In Section 2.3 we define producer performance in terms of technical efficiency, and we measure technical efficiency with distance functions. In Section 2.4 we define producer performance in terms of economic (cost, revenue, and profit) efficiency, and we measure economic efficiency relative to cost, revenue, and profit frontiers. We also show how to decompose each type of economic efficiency into technical and allocative components. Chapter 3 is concerned with the estimation of technical efficiency. In Section 3.2 we develop and show how to estimate cross-sectional production frontier models, both deterministic and stochastic, although most of our effort is directed toward stochastic production frontiers. In Section 3.3 we develop and show how to estimate panel data production frontier models, where technical efficiency is initially time invariant and then is allowed to vary through time. In Section 3.4 we discuss the problem of heteroskedasticity in stochastic production frontier models. Chapter 4 is concerned with the estimation and decomposition of cost efficiency. In Section 4.2 we develop and show how to estimate cross- sectional stochastic cost frontier models, in both single-equation and simultaneous-equation settings. In Section 4.3 we develop and show how to estimate panel data stochastic cost frontier models, again in single-equation and simultaneous-equation settings. In Section 4.4 we discuss a pair of novel approaches to the estimation of cost efficiency. Chapter 5 is concerned with the estimation and decomposition of profit efficiency. In Section 5.2 we develop and show how to estimate single-output stochastic profit frontier models, using both primal and dual approaches. In Section 5.3 we develop and show how to estimate multiple-output stochastic profit frontier models, again using both primal and dual approaches. We pay little attention to the distinction between cross-sectional and panel data models, since the various estimation techniques developed in Chapters 3 and 4 apply equally well to the estimation of stochastic profit frontiers. In Chapters 3-5 inefficiency is modeled by introducing additional error components and assigning distributions to them. Inefficiencies are then estimated as functions of the parameters of these distributions. In Chapter 6 we take a very different approach, in which both technical and allocative inefficiencies are modeled parametrically, on the assumption that producers optimize with respect to shadow prices, which are parametrically related to actual prices. In Section 6.2 we develop and show how to estimate and decompose both cost and profit efficiency in a cross-sectional setting. In Section 6.3 we do the same thing in a panel data setting. This parametric approach has some advantages, and some disadvantages, relative to the error component approach developed in Chapters 3-5. Estimation of efficiency is the first of two tasks. The second task is to explain variation in estimated efficiency. In Chapter 7 we discuss alternative approaches to the explanation of variation in efficiency. In Section 7.2 we discuss some early approaches to explanation, and we find these approaches wanting. In Section 7.3 we discuss a variety of recent approaches to explanation, which we find superior to the early approaches. Essentially these recent approaches achieve explanation by making the one-sided inefficiency error component a function of the explanatory variables. If efficiency varies, either across producers or through time, its variation constitutes a source of producer performance variation. One measure of performance is productivity change, and so if efficiency changes through time, it makes a contribution to productivity change. Chapter 8 concludes the book by incorporating efficiency change into models of productivity change, which heretofore have tended to neglect the contribution of efficiency change. In Section 8.2 we develop a primal approach, based on a stochastic production frontier, to the estimation and decomposition of productivity change. In Sections 8.3 and 8.4 we develop a pair of dual approaches, based on stochastic cost and profit frontiers, to the estimation and decomposition of productivity change. At least four topics are missing from the book. First, we do not discuss the estimation and decomposition of revenue efficiency relative to stochastic revenue frontiers. Second, we do not explore the efficiency with which producers pursue nonconventional objectives. The third and fourth omissions are perhaps more serious. We do not discuss the Bayesian approach to stochastic frontier analysis, and we do not discuss semiparametric approaches to stochastic frontier analysis. Our reason for omitting these two topics is that the two literatures are small and not yet influential.




Cited in
(only showing first 100 items - show all)








This page was built for publication: Stochastic Frontier Analysis

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4507503)