Choosing an Empirical Production Function: Theory, Nonnested Hypotheses, Costs of Specifications

This paper considers the specification of a production function, using crop-yield response to agricultural lime as an example. The paper first reviews neoclassical production theory, observing that this theory specifies neither the form nor the arguments of a production function. Several functions that are consistent with the theory are then evaluated by goodness-of-fit tests, by nonnested-hypothesis tests, and by "costs of misspecification." Neither goodness-of-fit tests nor nonnested-hypothesis tests provide a clear choice among the candidate functions; costs of misspecification provide some choice.