Online shopping is increasing steadily, and could lead to substantial impacts on trip generation, destinations, and timing. Understanding the potential transportation impacts requires us to better understand the adoption of this new shopping alternative (or “channel”). Using data collected from a web-based survey of two university towns in Northern California (N=967) in 2006, we developed models of the intention to purchase one of two different product types: book/CD/DVD/videotape or clothing/shoes. We were especially interested in exploring the nature of taste heterogeneity (differences across people in the importance placed on factors affecting the decision), including the best way to identify and model it. The model types we used include logistic regression (LR), latent class models (LCM) and LR with interaction terms.
Preliminary results showed that product type matters; variables such as trustingness and store brand independence are only significant for book purchases, while others such as efficiency/inertia and being female are only significant for clothing purchases. Accordingly, later models are product-type-specific.
The main findings are as follows. With respect to the e-shopping application context, we found, first, that product type, and general and channel-specific shopping attitudes, in addition to previously-identified effects (such as sociodemographics) clearly contribute to the purchase intention. Second, channel-specific perceptions substantially differ, on average, by product type. Therefore, it is dangerous to elicit general channel perceptions (or, comparative judgments not distinguished by individual channel) without regard to product type. With respect to the methodological approach, we found that empirically, LCM is not always superior to a conventional LR model with interaction terms. Instead, it can function as a useful diagnostic tool for dealing with taste heterogeneity, by leading us to more intelligently specify a conventional model with interaction terms. The latter often yields a more parsimonious and better-fitting model.
This study constitutes an early application of taste heterogeneity analysis to an e-shopping context. The models developed here improve our understanding of people’s shopping behavior, which will ultimately improve our ability to predict its impacts on transportation demand. Accordingly, our methodological approach and our specific results are of value to both marketing researchers and transportation planners.
Key words: internet/online shopping, store shopping, product type, logistic regression model, market segmentation, latent class model, taste heterogeneity, attitudinal factors