Abstract
Contrary to many other types of spatial decisions, shopping destination choice behavior is highly repetitive. For the practitioner looking for good predictors of store patronage, for reliable marginal utility estimates and reliable market share predictions, a central concern is with the type of data best suited to the research question, given the existing logistic and financial constraints. Different approaches can be recognized in the literature in which conventional discrete choice models are applied to shopping destination choice problems. In this paper, two of the most common practices are assessed and compared. First, the choice model is estimated with all choices of a relevant destination observed during a certain period of time (pooled cross-sectional data). The alternative approach consists in an estimation with the choice of the destination where the majority of purchases takes place (cross-sectional data). In the particular data set employed here, no evidence is found to support the idea that a multinomial logit model estimated with cross-sectional data does not perform as well as a model estimated with pooled cross-sectional data. Both models are found to be similar in their ability to identity the main predictors of store choice. Models developed on either data sets have marginal utility estimates that exhibit no statistically significant differences. Finally, market share predictions derived from both models are not statistically different. It appears, therefore, that there is no need to collect repeated patronage data over an extended period of time. The practitioner who wishes to use a conventional discrete choice model may avoid spending much time and money by gathering limited data on regular patronage patterns. In addition to this practical implication, the conclusions suggest that regular shopping destinations are chosen in accordance with the same behavioral motives as ancillary destinations are.
Get full access to this article
View all access options for this article.