On the model selection criteria for demand system: Theil’s minimum entropy measure and its modification

Several models are proposed for analyzing consumer demand systems. Empirical researchers usually, adopt one or more models and test some constraints based on demand theory, and then evaluate the plausibility of the models. However, they rarely adopt a statistical model selection procedure in their evaluation. The main reason is that the dependent variables are different in each model. Theil’s minimum entropy measure is a candidate for a practical model selection procedure, but its performance has not been studied carefully and sufficiently. In this paper, we conduct a Monte Carlo simulation to ascertain its performance. Simulation results show that the criteria based upon entropy do not perform well. We propose a two-step selection procedure, utilizing the criteria. This procedure performs better.