Willy Cheng
Site Selection Methodology for Automated Banking Machine Network: A Case Study of the Royal Bank of Canada © 2002
This paper demonstrates the use of spatial techniques that can be used to assist site selection for an Automated Banking Machine (ABM) network. Regression analysis and spatial interaction modelling are used to predict the ABM demand and to allocate the demand to facilities, respectively. The empirical information utilized in the calibration relates to Edmonton, Alberta in 2001. Regression analysis is used to predict ABM demand, in terms of number of transactions, at the Enumeration Area (EA) level. It is hypothesized that the ABM demand is related to market size, income, socio-demographic characteristics and employment level. It is shown that income is not statistically significant. The independent variables in the final regression model "explain" 60.4% of the variation in the dependent variable (ABM demand). The specific independent variables include: number of households, population with post-secondary and university education as their highest level of education, population aged 65 and above, participation rate and unemployment rate. Only population age 65 and above, and unemployment rate show a negative relationship with the dependent variable. In addition, a brief analysis also concludes that 71.17% of the ABM demand is satisfied by ABMs located at banks, while 14.01%, 6.25% and 9.41% are satisfied by ABMs located at convenience, institutional and retail locations, respectively. Spatial interaction models are calibrated for bank, convenience and retail location categories. Institutional location is excluded as their transactions are generated by capture, not by attractiveness. Also, it is concluded that a combined model including all of the three location categories is inappropriate, since it cannot provide a good representation on the difference between the three categories. All three spatial interaction models show that distance between origin and destination, number of ABMs available, level of competition, retail activity and accessibility are all statistically significant factors. However, distance between origin and destination is the most important predictive variable among all three models.