Abstract
New products and ideas often exhibit heterogeneity in spatio-temporal diffusion, including spatial clustering at multiple scales. For example, adoption of the Toyota Prius hybrid electric vehicle is far higher on the US West and East coasts than in the Midwest; adoption clusters are also observed at the city scale. Does heterogeneity in adoption arise from differences in local conditions and adopter preferences, such as gasoline prices, affluence or political affiliations, or from endogenous social influence processes that build consumer familiarity with new products? The answer conditions policies to promote the adoption of technologies, for example, to mitigate greenhouse gas emissions. Resolving this question is challenging because information conditioning individual adoption decisions is communicated through multiple channels, including interpersonal interactions, direct experiences with the product, media attention, and advertising. We develop a novel model in which spatio-temporal diffusion arises from the interaction of product utility and social exposure across multiple geographic scales. We test the model on the diffusion of the Prius in 4 US cities, selected to capture variation in local conditions and the actual extent of Prius adoption observed, using a dataset capturing all new Prius registrations at the ZIP code level during the first decade of its introduction. Variation in Prius adoption is primarily explained by social influence through local interactions within each ZIP code, amplifying underlying heterogeneities in the utility consumers derive from the Prius. Our analysis highlights the importance of jointly representing local conditions and social influence at different scales to explain adoption clustering.