Duration Dependence in Longitudinal Consumer Panel Data: A Continuous-Time, Stochastic Modelling Approach

Abstract

<p>Due to growing competition for market share, the retail environment is becoming increasingly specialized. Consequently, current retailing practices are concerned with targeting both spatially and aspatially defined population segments which, in turn, has led to the need to acquire detailed information on individual purchasing patterns. The collection of longitudinal data associated with the development of 'scanner panels' will result in a consumer behaviour data explosion, generating extensive, spatially disaggregate and detailed purchasing histories recorded at the individual level. A crucial element in the use of 'scanner panels' for retailing practices resides in the development of appropriate methodologies for analyzing this data and methods which exploit fully the information contained in these detailed records. This thesis considers the application of an event-history modelling approach to consumer shopping behaviour using longitudinal data from the Cardiff Consumer Panel, 1982. The focus of this research is duration dependencies involved in store switching behaviour. It is reasoned that the probability of choosing a particular type of retail outlet depends on both the type of store visited in the preceding trip and the time elapsed since that trip. Duration dependence between store switches can then be expressed in the distributional form of the hazard rate. Results indicate that an event-history approach provides a valuable methodology for examining consumer shopping behaviour. Duration dependence is found to be an important influence on store switching behaviour and the distributional form of duration dependence is seen to vary with different types of stores. Furthermore, measured sources of heterogeneity contained in such models indicate how different individual characteristics may be important in explaining the store choice behavioural process.</p>