Exploring Socio-Economic and Built-Environment Determinants of Alcohol Expenditure for the City of Toronto: A Spatial Analysis Approach

Abstract

Alcohol expenditures can provide a different perspective on alcohol use by providing an understanding of the consumer demand for alcohol, the effect of alcohol sales on the economy and the impact on the household budget. Previous studies have focused on alcohol expenditure at the individual-level and have not considered population level factors or the influences of geographical variation. The goal of this study was to examine the socio-economic and built-environment characteristics associated with alcohol expenditures at the small-area level in the City of Toronto.

Alcohol expenditure data consisting of purchases in licensed premises and purchases in stores for the year 2010 were retrieved from the Survey of Household Spending (SHS) at the Dissemination Area (DA) level. Socio-economic and built-environment variables were retrieved from the 2006 Census of Canada and DMTI Enhanced Points of Interest (EPOI) data, respectively. Multivariate spatial regression models were used to analyze the associations between alcohol expenditure and both socio-economic and built-environment variables (i.e. alcohol outlet density, restaurant density and subways).Global Moran’s I identified geographic variation for both types of alcohol expenditures. Local Moran’s I identified three hot-spots and three cold-spots for licensed premises expenditures whereas four hot-spots and two cold-spots were identified for purchases in stores expenditures. The spatial Durbin error model was identified as the best spatial model for both types of alcohol expenditures. For licensed premises expenditures, positive associations were found with seven socio-economic variables and three built-environment variables. For purchased in stores expenditures, positive associations were found with three socio-economic variables and subway intercepts.

This study was the first to identify socio-economic and built-environment characteristics associated with alcohol expenditures at the DA level for the City of Toronto. Additionally, these findings highlight the importance of examining associations between built-environment variables and health behaviours, as well as the importance of considering the actual geography of the study region. Finally, the spatial regression models, visualizations and Geographical Information System (GIS) methods employed in this study can be collectively applied as a toolkit to study other health behaviours that present with geographical variation.

Future studies should examine additional types of built-environments for associations with alcohol expenditures. To create an even more complete understanding of what drives alcohol expenditures, the findings from this study should be combined with individual-level characteristics in spatially-explicit multilevel models.