On the Suitability of Econometric Demand Models in Design for Market Systems

Abstract

A goal of design for market systems research is to predict demand for differentiated products so that counterfactual experiments can be performed based on design changes. We review conventional methods and propose an additional method to evaluate the suitability of econometric demand models estimated from revealed preference data for use in product design studies. We evaluate one demand model form from literature and two newly constructed forms for new vehicle demand along existing metrics of fit and predictive validity as well as a newly developed metric of proportional substitution sensitivity. We show that a model that includes horizontally differentiated preferences for size performs better under metrics of fit and predictive validity but that no model relaxes the IIA property satisfactorily to avoid exploitation by design optimization. We conduct design studies separately, applying each demand model form assuming the automotive market is in Bertrand–Nash price equilibrium. Results illustrate that the influence of the demand model form on the optimum in terms of design variables and expected firm profit is significant.