International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Using Choice Modelling for Inferring Latent Spatial Demand: Estimation of a Market Demand Model for the Mobile Consumer
Ari Pramono, Harmen Oppewal

Last modified: 11 May 2015


The decision where to locate a store is one of the most important decisions in retail marketing (Ghosh and Craig 1983). In order to find the best location for their outlets, retailers need to assess where their current or potential customers are located, in other words, they need to assess the spatial distribution of demand. A main challenge in inferring the spatial distribution of demand from observed sales is that sales are observed at a specific point in space, at the retail outlet (Duan and Mela 2009) , while the location of the actual demand  itself (e.g . at the household level) remains latent.  This issue exacerbates when the demand occurs in the context of a trip to some main destination. This happens for example when motorists have to refuel their car. This paper aims to address this issue by proposing a combination of disaggregate choice modelling approach and an aggregate spatial interaction model to derive the spatial demand for fuel.

Many approaches have been presented to provide estimates of spatial demand, ranging from the early retail models, such as the Retail Gravity Model  (Huff 1964) and Multiplicative Competitive Interaction Model  (Nakanishi and Cooper 1974; Nakanishi and Cooper 1982) to approaches using Spatial Interaction Modelling  (e.g. (González-Benito 2005) and  (Roy and Thill 2004)). There are also several studies that apply choice modelling methods and model predict retail demand by modelling consumers’ choice of retail location (e.g., Arentze, Oppewal, and Timmermans 2005; Borgers and Timmermans 1987; Dellaert et al. 1998; Fotheringham 1988; Timmermans, Borgers, and van der Waerden 1992).

Despite their differences, all of the above approaches share three basic assumptions. Firstly, that decision makers are fully rational and make a trade-off between the attraction of a given good and the effort required and cost involved in obtaining the good (cf. Dion and Cliquet 2006). Secondly they assume that shopping activities are located at the consumers’ main destinations or at least at a destination that is part of their chain of destinations.  Thirdly, it is assumed that the cost of accessing the store (distance, travel time or cost)  is a monotonously increasing function of the distance between the consumer’s origin location and the shopping  (destination) location .

Implications of these basic assumptions in regards to the spatial distribution of demand are, firstly, that the demand has a “cone-shaped” spatial distribution around the point of sales, secondly that Tobler (1970)’s first law of geography applies that “everything is related to everything else, and near things are more related to distant things”, in spatial competition among firms and lastly that Hotelling (1929)’s minimum spatial differentiation  principle applies (positive agglomeration effect).

However, while they hold true for many retailing applications, these assumptions do not apply for decision situations where consumers choose and consume products (e.g. fuel retailing and convenience shopping) while they are on their way to somewhere else.  Indeed, studies in fuel retailing reveal that the agglomeration effect is found to be negative (Netz and Taylor 2002), and show there is little evidence of local competition

Conference registration is required in order to view papers.