Market Structure Analysis Using Generalized IRT Models and Hierarchical Bayes Estimation Procedures

• Activity Number: 325
• Type: Topic Contributed
• Date/Time: Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and Marketing
• Author(s): Lynd D. Bacon*+ and Douglas Rivers and Jeff Hunter
• Companies: Lynd Bacon & Associates, Ltd. and Stanford University and General Mills, Inc.
• Keywords: Market Structure Analysis ; IRT ; hierarchical Bayes ; MCMC ; product strategy ; consumer insights

Abstract

There is a long tradition of quantitative analysis procedures for doing demand-side market structure analysis. These methods include internal methods in which the dimensions and nature of competition between brands or products is inferred from preference/choice data, and external methods in which the dimensions are identified a priori. Our paper adds to the work in this area by extending a technique described by Clinton, Jackman, and Rivers (2003) for analyzing political role call data, and applying it to the analysis of market structure. The method we describe produces maps that represent consumers and brands/products in a single, low-dimensional space by using a hierarchical Bayes specification of what is effectively a generalized multidimensional Item Response Theory (IRT) model. Our general specification is in the spirit of IRT extensions described by Junker (1997), and by Patz and Junker (1997). The key advantages of our method include that it provides a natural way to integrate perceptual and preference data of multiple measurement types and from multiple sources. We apply our model to consumer brand usage data spanning product categories.