GIS mapping of retail food access for populations of different poverty status

This research utilized GIS technology to compare two databases and investigate how community poverty level relates to food access in Philadelphia, PA. Retail food listings were purchased from Dun and Bradstreet (D&B) (N=4,643) and determined using inspection records from the Philadelphia Health Department (PHD) (N=8,317). Facilities of interest were identified by NAICS code and by keyword, respectively. Addresses were geocoded to census tracts in Philadelphia, PA. Census tracts were divided into quintiles using Census Bureau poverty data. GIS overlay analysis was used to group locations within tracts. Chi-square statistic was utilized to test interaction between poverty and store type. Using either database, a significant (p<0.05) interaction was found between poverty and the distribution of food markets, indicating that rates of all grocery stores (including corner markets) were highest in high poverty areas. Further analysis of D&B revealed 11.8% of markets in low poverty areas to be chain markets compared to only 1.5% of markets in high poverty areas. Using PHD, supermarkets alone made up 6.9% of markets in lowest poverty areas and only 4.4% in the highest poverty areas. There was a significant interaction between poverty and the distribution of fast-food and full service restaurants using PHD (p<0.05) but not D&B (p = 0.09). Fast-food comprised 49.7% and 41.7% of restaurants in high and low poverty areas, respectively. This study demonstrates not only differential access to food for different populations but also the need to investigate different sources of data for food access research.