Draft Multivariate Visual Diffusion for Social Groups

We are interested in the role that visual (social) identity plays in product adoption. How does the type of people who own and are seen to use a particular product/service influence others' likelihood of purchase? For example, might men be less likely to buy a sporty coupe automobile if they observe lots of women are driving it? Might these cross-group effects be more likely in certain car categories as opposed to others? To address these questions, we have acquired data on over 26 million automobiles sold throughout the U.S. from January 1999 till January 2008. We use Bayesian probability models to identify whether the adoption of a given make and model by one social group can be explained " visually " by previous adoption by another social group. Furthermore, much extant research suggests that social contagion is a local phenomenon (that is, an individual's purchase decision for a given product depends in part on whom else he actually sees using that product) — " you can not be affected by what you can not see " — hence, we incorporate local visual spatial effects in our model. In addition, we are able to determine how these patterns vary across both geographic areas as well as across makes, models, and categories of cars. Finally, since our main hypotheses surround visual diffusion of automobiles, we relate the degree of local visual diffusion to various local correlates (e.g. climate, topology, income, commuting variables).