Latent Class Conjoint Choice Models: A Guide for Model Selection, Estimation, Validation, and Interpretation of Results

Abstract

The consideration of preference heterogeneity in consumer choice behavior has become state of the art. In addition, the identification of consumer segments remains essential for marketing managers. For disaggregate consumer choice data representing the basis of segmentation, the latent class multinomial logit (MNL) model is currently the most popular approach for estimating segment-specific preferences. After addressing the theoretical background of the latent class MNL model, we use an empirical choice-based conjoint data set to illustrate model estimation and validation, as well as how the estimation results should be interpreted. A particular focus lies on the model selection process, i.e. the determination of an appropriate number of segments. We further work out interpretation pitfalls when the existing preference heterogeneity of consumers is ignored. This will ultimately provide a guide for applying the latent class MNL model regarding model selection, estimation, validation, and interpretation of results both from a statistical and managerial perspective.