Explaining Preference Heterogeneity with Mixed Membership Modeling

Understanding the drivers of preferences requires models and covariates that explain their cross-sectional variation in models of demand. In this paper, we develop an integrated model of choice and covariates where parameters of the covariate model serve as regressors to explain preference heterogeneity in the choice model. We investigate alternative models to deal with many potential covariates and find that a grade of membership model provides empirical and conceptual advantages in explaining preferences. Our proposed model is illustrated with two conjoint data sets.