Heterogeneous Conjoint Choice Designs
Published 5/1/2005
Author: Zsolt Sándor and Michel Wedel
View this content
Executive Summary
Because conjoint choice experiments have become a preferred tool for the collection of information on consumers’ preference structures, the question of how to better design such experiments is important. Conjoint choice experiments have become important tools for collecting data that underlie product design, pricing, and branding decisions. This article proposes heterogeneous designs and demonstrates their improvements compared with homogeneous designs. Heterogeneous designs contain several subdesigns that are offered to different respondents, but the designs that have been considered thus far in the marketing literature are homogeneous insofar as each respondent is given the same choice sets. The proposed approach improves such designs by exploring the potential information gain from the use of a larger number of choice sets employed across rather than within respondents. Although the principle of the heterogeneous design is applicable to any choice model, this article follows previous work in the field and develops a design construction procedure for standard Bayesian and mixed logit designs.
This study demonstrates that the construction of several choice designs and the random allocation of respondents to these designs yield substantially higher efficiency. It is important to note that the use of the heterogeneous design comes at no additional cost or response burden to respondents and is simple to implement. The results show that if there is uncertainty about the parameter values, heterogeneous designs produce substantial improvements in efficiency compared with previously proposed (Bayesian and mixed logit) designs. This is important because in most cases, when conducting a conjoint study, the purpose is to estimate parameters about which there is little or no prior knowledge. If the true parameters are far from those that are assumed in the design construction, the heterogeneous designs are much better than are the homogeneous designs because they produce a higher spread of the design points, which enables more precise estimation of the parameters.
Biography
Zsolt Sándor has an M.Sc. in mathematics from the Babes-Bolyai University in Cluj, Romania (1993); he completed postgraduate studies in economics at CERGE in Prague, Czech Republic (1996) and has a doctoral degree in economics from the University of Groningen (2001). Currently, he is a postdoctorate researcher at the Tinbergen Institute, Erasmus University Rotterdam. His main research interests are various statistical issues related to discrete choice models. The title of his docoral dissertation is “Computation, Efficiency, and Endogeneity in Discrete Choice Models.” His work has appeared in Journal of Marketing Research, Marketing Science, Journal of Econometrics, and Transportation Research B.
Michel Wedel has an M.Sc. in biomathematics from the University of Leiden (1981), an M.Sc. in statistics from the Netherlands Statistical Society (1986), and a doctoral degree in marketing from Wageningen University (1990). He was head of a statistics and computing group at the TNO Nutrition Research Institute in Zeist (the Netherlands, 1982–1991) and Professor of Marketing Research at the Department Economics, University of Groningen (the Netherlands, 1991–2002). Currently, he is the Dwight F. Benton Professor of Marketing at the University of Michigan Business School. He has published more than 100 articles in peer-reviewed academic journals, which have been cited more than 1000 times. He has written two books and has edited special issues of Journal of Econometrics and the International Journal of Research in Marketing. He serves on the editorial boards of the Journal of Marketing Research, Marketing Science, Journal of Marketing, Journal of Classification, and the International Journal of Research in Marketing. He has consulted for more that 25 different companies in the nonprofit and profit sectors.
J Marketing Research, Volume 42, Number 2, May 2005
View Table of Contents.