Modelling store choice: a segmented approach using stated preference analysis

Recent structural changes in the UK grocery retailing sector have led to a shift in patterns of consumer shopping behaviour that increasingly can only be properly understood and modelled through the division of consumers into two groups: those who shop at superstores and those who do not. A disaggregated approach to the analysis of store choice is therefore necessary. In a survey undertaken in Cardiff, South Glamorgan, stated preference data were collected, which involved the observation of respondents' most preferred alternatives in a number of hypothetical choice scenarios. These repeated observations of choice allow the calibration of small sample and even individual-specific utility functions. This is achieved through the use of the 'exploded logit model', an extension of the multinomial logit that exploits the properties of ordinal preference data. Canonical correlation analysis is used to relate variations in individuals' preferences to their socio-economic and demographic characteristics. Market segments of like-minded individuals are then defined in terms of those characteristics found to be significant. Mobility, affluence and time-constraints, as measured by car ownership and the employment status of the principal shopper, are found to be significant segmentation criteria. Validation of the models is successful, with the segmented models performing better than the aggregate model