Regularization Methods on Almost Ideal Demand System

Abstract

Almost Ideal Demand System (AIDS) has a flexible functional form that can first order approximate any sets of demand functions derived from utility maximizing behavior. This flexibility, however, comes with a cost of estimating numerously many parameters, hindering practical use of this model. In this paper, I suggest to use regularization methods on AIDS model to circumvent this curse of dimensionality. Monte Carlo simulation experiments, following Buse (1994), show that elastic net performs reasonably well under multicollinearity and sparsity environments.