Alcohol retail density and demographic predictors of health disparities: a geographic analysis

Abstract

Objectives: We examined whether the geographic density of alcohol retailers was greater in geographic areas with higher levels of demographic characteristics that predict health disparities.

Methods: We obtained the locations of all alcohol retailers in the continental United States and created a map depicting alcohol retail outlet density at the US Census tract level. US Census data provided tract-level measures of poverty, education, crowding, and race/ethnicity. We used multiple linear regression to assess relationships between these variables and retail alcohol density.

Results: In urban areas, retail alcohol density had significant nonlinear relationships with Black race, Latino ethnicity, poverty, and education, with slopes increasing substantially throughout the highest quartile for each predictor. In high-proportion Latino communities, retail alcohol density was twice as high as the median density. Retail alcohol density had little or no relationship with the demographic factors of interest in suburban, large town, or rural census tracts.

Conclusions: Greater density of alcohol retailers was associated with higher levels of poverty and with higher proportions of Blacks and Latinos in urban census tracts. These disparities could contribute to higher morbidity in these geographic areas.

FIGURE 1

Relationship between percentage of Black population and number of alcohol retailers per 1000 population in (a) urban (n = 44 883), (b) suburban (n = 6385), (c) large town (n = 6536), and (d) small town/rural (n = 7537) areas: US Census tracts, continental United States, 2000. Note. Census tracts are categorized according to the 4-tiered Rural Urban Commuting Area (RUCA) classification system. Each dotted line indicates the regression line for that graph's RUCA category. The background points and solid lines represent the national sample and are repeated for all graphs, for purposes of visual comparison. Percentage of Black population is log scaled to account for outliers in the model. A 1-point increase in the x-axis represents an increase of 10 percentage points.

FIGURE 2

Independent associations between retail alcohol density per 1000 population and (a) percentage of Black population, (b) percentage of Latino population, (c) percentage of families in poverty, (d) percentage of women with less than a high school education, (e) median number of rooms in one's housing unit, and (f) average number of people in the household: US Census urban tracts, continental United States, 2000. Note. Graphs demonstrate where median, interquartile range, and 90th percentile would lie on the regression line. For example, for percentage of Black population, the 90th percentile is 62%, and the third quartile for percentage families in poverty is 15%. The curves are based on a multivariate regression with residual standard error: 0.986 on 42 842 degrees of freedom, adjusted R² = 0.2668, and F statistic = 2599 on 6 and 42 842 degrees of freedom, with 2034 observations deleted because of missing values. Variables were scaled to make relationships linear. A 1-point increase in the x-axis represents an increase of 10 percentage points.