Modeling Interdependent Consumer Preferences.pdfVIP

SHA YANG and GREG M. ALLENBY^ A consumers preference for an offering can be influenced by the pref- erences of others in many ways, ranging from social identification and inclusion to the benefits of network externalities. In this article, the authors introduce a Bayesian spatial autoregressive discrete-choice model to study the preference interdependence among individual con- sumers. The autoregressive specification can reflect patterns of hetero- geneity in which 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. The authors illustrate their model of interdependent preferences with data on automobile purchases and show that preferences for Japanese-made cars are related to geographically and demographically defined networks. Modeling Interdependent Consumer Preferences Preferences and choice behavior are influenced by a con- sumers own tastes and the tastes of others. People who identify with a particular group often adopt the preferences of that group, which results in interdependent choices. Examples include the preference for particular brands (e.g.. Honda Odyssey) and even entire product categories (e.g., minivans). Interdependence may be driven by social con- cerns, endorsements from respected people who increase a brands credibility, or learning the preference of others who may have information not available to the decision maker. Moreover, because people engage in multiple activities with their families., coworkers. neighbors, and friends, interde- pendent preferences can propagate across and through mul- tiple networks. Quantitative models of consumer purchase behavior often do not recognize that preferences and choices are interde- pendent. Economic models of choice typically assume that a consumers latent utility is a function of brand and attrib- ute preferences, not the preferences of others. Preferences are assum