Viability of demographic and behavioural independent variables in quantile regression models in predicting retail patronage

Abstract

The purpose of this paper is to model and predict consumer patronage behaviour in terms of relative shopping frequencies in two shopping destinations – in the city centre and in an edge-of-town retail park. Two different quantile regression models for both destinations, the demographic and the behavioural, were constructed based on existing literature and data collected in 2001. In the demographic model the independent variables were the distance separating respondents and respective shopping destinations, and demographic variables describing the households. The behavioural model consisted of statements concerning the importance of various store choice criteria. The revealed preference approach was adopted, i.e. information revealed by past behaviour was used to extract the most powerful predictors of relative visit frequencies. The study is based on longitudinal data collected over a period of 11 years (2001, 2003, 2006, and 2011),which enabled the monitoring of the long-term effects of retail change on consumer behaviour. Access to longitudinal data enabled us to test the viability of the models using both in-sample (2006) and out-of-sample (2011) predictions. The results strongly support the existing literature emphasising the importance of distance and accessibility in store choice.