Maximum Likelihood Estimation of Input Demand Models with Fixed Costs of Adjustment

Abstract

Traditional models of input demand rely upon convex and symmetric adjustment costs. However, the fortune of this highly restrictive approach is due more to analytical convenience than to empirical relevance. In this note we examine the model under more realistic hypothesis of fixed costs, show that it can be cast in the form of a Double Censored Random Effect Tobit Model, derive its likelihood function, and finally evaluate the performance of the ML estimators through a Monte Carlo experiment. The performances, although strongly dependent on the degree of censoring, appear to be promising.