A1152
Title: Modelling consumer heterogeneity in ranking conjoint data via structural equation modeling
Authors: Chung-Ping Cheng - National Cheng Kung University (Taiwan) [presenting]
Abstract: Conjoint analysis decomposes product choice into part-worth utilities but usually assumes all consumers share the same utilities. Using structural equation modeling (SEM), heterogeneity can be addressed by treating each respondent's part-worth vector as multivariate-normal random effects, jointly estimating mean utilities and their covariance structure. This SEM framework is extended to ranking data, converting rankings into pairwise comparisons that yield binary indicators for every product pair. Using data from 87 consumers, eight low-calorie chocolate profiles (28 comparisons per person) are analyzed. The SEM shows an acceptable fit. Four of seven random-effect variances are significant, revealing pronounced heterogeneity, especially for flavor and packaging, while fixed effects confirm strong preferences for walnut flavor, deluxe packaging, and moderate price sensitivity. This ranking-based SEM recovers individual differences lost in fixed-effect models, enables richer market segmentation and what-if simulations, and demonstrates how mixed-effects conjoint analysis can accommodate rating, ranking, or binary data within a single unified framework.
