Modelling the behavior of fuel retail consumers

The study is motivated by the need for better insight into the nature of the demand for fuel stations, in particular how this demand is influenced by the fact that consumers have to make their decision where to refuel “on-the-fly”. There is only limited literature on this topic. Most of the existing studies in refuelling consist of aggregate analyses that have a lack of emphasis on understanding the consumers’ decision making process and approaches to choice modelling not taking into account the unique characteristics of the fuel retail market. Fuel retailing is distinct from most other retailing in both supply and demand side aspects. The supply side displays a high level of vertical integration and the product is commoditized and homogeneous. The demand side is characterized by the fact that the fuel consumer is moving across the market space and makes the decision where to refuel “on-the-fly”, during their trip to some final destination. These distinctions have significant implications for the modelling of the demand for fuel (or gas) stations. Firstly the final destination and the refuel location do not coincide. Secondly there is endogeneity in the choice set development process. The third complexity is that the impact of the spatial structure of alternatives on the consumer’s choice depends on the consumer’s direction of movement. The current research was conducted to enhance the capacity of existing theories to model “on-the-fly” choice behaviour, especially in the fuel retailing market. The main objectives of the research include the investigation of the consideration set generation process in on-the fly situations, the spatial modelling of fuel retail consumer’s choice behaviour, and the modelling of spatial competition among fuel stations. The study develops and tests a model that is based on the random utility model framework but with some modification in the specification of the spatial variables to take into account an individual’s relative path of movement. The model is estimated using data obtained from a field survey in the city of Bandung, Indonesia. For the survey, sampled consumers were intercepted during their refuelling activities and interviewed to gather information regarding their trip attributes and their alternative refuelling location options. The research then separated into three different modelling studies, firstly, a study of the role of accessibility as a determinant of an individual’s choice set, secondly, a study into the modelling of the refuelling activity as an on-the-fly choice, and finally, a study into the modelling of spatial competition in fuel retail market. The first study is a simulation study of refuelling choices of simulated individuals in a hypothetical grid-city. Two different simulations reflecting different spatio-temporal simulations are generated to answer the research question whether variation in the decision maker’s level of accessibility to the alternatives impacts the role of the consideration set in the spatial choice model. The simulated data are used to estimate both a single-stage simple Multinomial Logit model and a two-stage Constrained Multinomial Logit in order to test whether the inclusion of the consideration stage improves the modelling performance. The study findings confirm that the role of the consideration set depends on the level of accessibility. The more restrictions are imposed on access, the more important the consideration set becomes in setting the outcome of the choice process and the more superior the two-stage model becomes compared to the single stage model. The second study models actual refuelling behaviour as an “on-the-fly” choice. Two models are developed by using the field observation data from Bandung to empirically model the two different modes of the consumer’s refuelling behaviour, as defined in the literature: a planned choice mode and an ad-hoc search mode. The results from the study confirm our proposition that a consumer in planned choice mode emulates a typical stage-wise choice behaviour, signified by a rational compensatory trade-off between spatial and non-spatial effects with a significant role for the accessibility-driven consideration set, while ad-hoc search consists of sequential search behaviour, wherein spatio-temporal limitations diminish the role of the accessibility driven consideration set and signify the role of spatial variables over a non-spatial attractiveness variable. In comparison to other spatial models in the literature, the current model performs better in terms of both the model fit criteria and predictive capability. By integrating the choice model developed in the current study with the spatial interaction model framework, the spatial substitutability among all gas stations in the study area is modelled. The main objective of this exercise is to propose a costumer oriented market delineation, in which the spatial substitutability between spatially differentiated firms is measured through the perspective of consumer consideration rather than through the firms’ spatial locations. The study finds that the spatial competition between firms is heavily affected by the spatio- temporal context. First the substitutability between pairs of spatially competing firms varies based on their spatio-temporal situations. Secondly, the relationship between spatial competition and spatial structure (e.g. clustering and spatial differentiation) is dependent on the consumers’ behavioural modes, that is, whether they are in choice mode or in search mode. Finally, in the last section of the thesis, the managerial implications and the limitation of the study are discussed together with some proposals for future research avenues.