Economic foundations of conjoint analysis

Abstract Observational data on either individual or aggregate demand is often not sufficient to identify consumer preferences due to lack of variation in prices or product features, or the desire to study product features not currently available. Choice-based conjoint analysis offers a solution to this problem by creating hypothetical product choices via experimental design and collecting demand data using survey methods. Choice-based conjoint designs can apply to a pure discrete choice model of demand in which consumers only purchase one unit from a set of products or to more general settings (termed “volumetric” conjoint analysis) in which consumers have the opportunity to purchase more than one product and to consume continuously variable quantities. This chapter provides the economic foundations for choice-based conjoint analysis along with efficient statistical methods for estimation and prediction of demand. We review the requirements that formal economic and statistical analyses impose on the design of conjoint surveys as well as the analysis of the conjoint survey data. A number of variations on conjoint analysis which appear in the literature do not conform to these requirements and, therefore, may provide erroneous inferences and predictions. We provide an example where both demand data and conjoint survey data are available for the same consumers and provide some conclusions regarding what aspects of preference estimation are consistent between actual marketplace data and conjoint survey data. All models and procedures in this chapter are available as R packages.