Modeling Interdependent Consumer Preferences

An individual's preference for an offering can be influenced by the preferences of others in many ways, ranging from the influence of social identification and inclusion, to the benefits of network externalities. In this paper, we introduce a Bayesian autoregressive discrete choice model to study the preference interdependence among individual consumers. The autoregressive specification can reflect patterns of heterogeneity where influence propagates within and across networks. These patterns cannot be modeled with standard random-effect specifications, and can be difficult to capture with covariates in a linear model. Our model of interdependent preferences is illustrated with data on automobile purchases, where preferences for Japanese made cars are shown to be related to geographic and demographically defined networks.