Measuring Exposure Across Spatial Scales: Association Between Tobacco Retail and Poverty in Washington State

BACKGROUND: It has been asserted that the distribution of retailers of unhealthy products, such as tobacco, disproportionately affects socially disadvantaged neighborhoods. Attempts to examine relationships between tobacco retail and poverty have struggled to find an appropriate measure of retail exposure. Measures of proximity may work well at a local scale on a relatively uniform landscape. However, at a statewide or national scale, proximity can merely reflect population density - in dense urban areas, all distances are shorter than in sparse rural areas.  We present a dimensionless measure of relative exposure to tobacco retail that allows for meaningful comparison across spatial scales.

METHODS: We geocoded addresses of tobacco retail outlets in Washington State, obtained from state license information for October 2015. We computed population weighted distance (PWD) for tobacco retailers by census tract following Zhang et. al. 2011. We then computed PWD by census tract for all retail based on 2010 tax parcel land use. The ratio of PWD for tobacco to PWD for retail parcels provided a dimensionless measure of relative tobacco exposure. We used spatial lag regression to examine the association between relative tobacco exposure and poverty by census tract in Washington State.

RESULTS: Relative tobacco exposure by census tract was uncorrelated with population density (r=.028), street network density (r=.021) and urbanicity (r=-.044). The distribution of relative tobacco exposure by census tract was slightly skewed and heavy-tailed. There was significant spatial autocorrelation in relative tobacco exposure by census tract (Moran’s I = .3539, p< .001), indicating that spatial regression was called for. In spatial lag regression, relative tobacco exposure was positively associated with log-transformed percent poverty by census tract (p=.005). This relationship persisted after controlling for other spatial variables including population density, urbanicity and street network density.

CONCLUSIONS: Association between tobacco retail and poverty can be demonstrated locally, but at the state level, meaningful comparison across spatial scale can be problematic. By using a dimensionless metric of relative exposure, we demonstrate a positive association between tobacco retail and poverty across Washington State.