Modifying areal interpolation techniques for analysis of data on food assistance benefits

Analyses aimed at identifying food deserts-defined as areas with limited access to healthy food-have garnered much recent attention from the news media, policy makers, and non-profit groups. Much of this research relies on the proximity of large grocery stores as a measure of food access. These studies have been limited by poor data quality, boundary effects, and scale dependence. Drawing on data from the Supplemental Nutrition Assistance Program (SNAP, formerly known as food stamps), we suggest an alternative approach that incorporates the distribution and redemption of food assistance benefits in low-income neighborhoods. This data is publically available, but at the zip code level, limiting its usefulness for neighborhood analysis. We use a three-class areal interpolation method to develop three disaggregation techniques that increase the usability of this data. These utilize several external data sources to weight the distribution of this data, including the U.S. Census, U.S. Geological Survey satellite imagery, and existing cadastral data. Our analysis, focused on the Twin Cities metropolitan region for federal fiscal year 2010, thus allows for a more accurate depiction of how residents actually access the food system.