INTRODUCTION

In Chapter 3 technical inefficiency was modeled as an error component within a stochastic frontier framework. In Chapter 4 and parts of Chapter 5 cost inefficiency and profit inefficiency were also modeled as error components within a stochastic frontier framework. Thus the basic strategy in previous chapters has been to construct a composed error stochastic frontier model, and to extract estimates of inefficiency from the parameters describing the structure of the two error components. This procedure is straightforward in single-equation models, in whih the sole objective is to estimate technical or economic inefficiency. However this procedure is much less straightforward in simultaneous-equation models, in which the dual objectives are to estimate and decompose economic inefficiency.

In this chapter we change our strategy. We do not estimate stochastic frontiers. Instead, we model all types of inefficiency parametrically, through the introduction of additional parameters to be estimated, rather than through an error component. The error structure of the estimating equation (or system of equations) is conventional, the same as that employed in the estimation of cost, revenue, or profit functions in the nonfrontier literature. Thus the error term is distributed normally in a single-equation model, and the error terms are distributed as multivariate normal in a system of equations model.

Technical inefficiency is introduced in two alternative ways. Output-oriented technical inefficiency is introduced by allowing the production function intercept to vary across producers; this generates a cost function in which producers' output vectors are scaled differentially, and a profit function in which producers' output price vectors are scaled differentially.