Abstract:
ZIP Codes have traditionally been represented cartographically as polygon features. Polygon-based representations of ZIP Codes are derived from point features employing interpolation techniques. There has been an increasing understanding in the literature that these interpolations can introduce error into the results of spatial analyses that employ ZIP Code polygon representations. This research uses multiple ZIP Code datasets which have been collected from eleven different sources. First the study seeks to determine if the ZIP Code representations are identical as they are purported to be. Comparisons are made based on several spatial characteristics; specifically area, level of overlap, and centroid location. It has been determined that frequently there are considerable, in some cases statistically significant, differences in these measurements across polygon representations. The consequences of these differences for the results of spatial analyses that employ ZIP Code polygons are explored through typical applications. These applications include a test for spatial autocorrelation, an examination of network distances, a ranking of ZIP Codes by school quality and availability, an examination of segregation level, and a description of accessibility to emergency rooms. Finally, a statistical comparison of ZIP Code polygon datasets is made, demonstrating that the representations are not identical and in many cases are not even statistically similar. Conclusions regarding best practices for ZIP Code polygon use and suggestions for future research are provided.
