Store patronage : the utility of a multi-method, multi-nomial logistic regression model for predicting store choice

Faclor, multinomiallogistic regression and cluster analyses are used in combination to provide a predictive model of store patronage behaviour for consumers in Cardiff, Wales. A subset of variables and factors that are important for consumers when choosing a supermarket were used to provide a picture of each store's clientele. Multinomial logistic regression allowed an overall model of supermarket choice to be developed and also enabled comparisons to be made of individual supermarkets within the sample. A detailed picture of store patronage is presented along with predictions about store choice for a number of "consumer clusters". The results demonstrate the utility of the predictive multinomial models when used in conjunction with other analytical techniques and reinforces a number of studies that have investigated patronage behaviour.