Measuring Factor Substitution with Neoclassical Models: An Experimental Evaluation

The purpose of this paper is to evaluate the performance of neoclassical cost functions in estimating input substitution. In contrast to earlier studies, this analysis assumes that the "true" technology is represented by each of three large-scale, engineering, process analysis models. Experimental methods are used to evaluate the effect of the definition of factor inputs, the characteristics of the technologies, and the presence of residual discharge constraints on measures of input association. The findings support the use of approximate significance tests as an integral part of the interpretation of the neoclassical estimates. They suggest that input aggregation reduces the ability of neoclassical methods to isolate input associations. Finally, they indicate that residual discharge constraints can, if ignored in modeling, distort measured substitution relations.