Heterogeneous Conjoint Choice Designs

Abstract

Previous conjoint choice design construction procedures have produced a single homogeneous design that is administered to all study participants. In contrast, this article proposes to construct a limited set of different designs. The principle of heterogeneous designs is applicable to a variety of types of models. This article illustrates this principle for Bayesian designs, taking into account prior uncertainty about the parameter values, and for mixed logit designs that accommodate respondent heterogeneity. The authors develop and investigate a computational procedure that enables quick and easy implementation. Although the number of different designs in the optimal set is small, the authors use a Monte Carlo study to demonstrate that their heterogeneous design achieves substantial gains in efficiency compared with homogeneous designs.