Unobserved Preference Changes in Conjoint Analysis

During the completion of a conjoint task, fatigue, boredom, or task simplification strategies and learning effects may cause unobserved changes in individual consumer preferences. We propose a hierarchical Bayesian dynamic switching model that we call Change Point Heterogeneity model (CPHM) to take both preference heterogeneity and heterogeneity with respect to unobserved non-stationarity (i.e. change) into account. We discuss full conditional Gibbs sampling of the unknown model parameters and apply the model to data from a metric conjoint study. The substantive results indicate that approximately half of the consumers in the sample changed their preferences during the course of the conjoint task in two different ways. The majority of consumers in the sample appreciates the design variables from the start and approximately half of them resorts to some simplification of the evaluation task towards the end. A smaller group of consumers struggles with the evaluation task in the beginning. Approximately half of those learn to appreciate the design variables at least to some extent later during the task. To the extent that answering behavior prior to task simplification / after a learning phase better reflects 'true' preferences, conventional conjoint models would underestimate the effect of the design variables on the preferences of these consumers