A comparison between the geostatistical potential model and the discrete choice logit models in the estimation of spatial market potentials

This paper is devoted to the comparison of two main approaches for the estimation of the spatial market potential of a retail product over a retail trade area. The first approach is based on the Discrete Choice Logit Models while the second approach is based on the recently introduced Geostatistical Potential Model. In both the approaches, the spatial market potential is evaluated by considering the sales data of spatially distributed stores, the features of the stores and a set of spatial covariates taking into account the competition between nearby stores. It is shown that, when compared with the Discrete Choice Models, the Geostatistical Potential Model is able to handle all this information in a more natural and easier way and that it is richer in terms of model output. This allows to have a better insight into the spatial features of the market potential and to facilitate the decision making process when the spatial market potential is exploited in order to increase the sales volume. As a case study, the problem of estimating the spatial market potential of a daily newspaper and the problem of optimizing the network of newsstands are tackled.