Poster Paper: Using Geographic Information Systems for an Alternative Definition of Neighborhood

Saturday, November 4, 2017
Regency Ballroom (Hyatt Regency Chicago)

*Names in bold indicate Presenter

A. Rupa Datta1, Kanru Xia2 and Joshua Borton2, (1)National Opinion Research Center, (2)NORC at the University of Chicago


Interest in estimating neighborhood effects or developing community-level measures spans a broad range of policy areas, including social inequality, child care access, labor markets and employment, access to healthy food, exposure to crime, and proximity to environmental hazards. Data availability constrains how ‘neighborhoods’ or communities are defined; as a result, empirical analyses of these concepts in the U.S. use a variety of proxy measures, such as ZIP codes, census tracts, municipalities, school districts, counties, and commuting areas or Census PUMAs. An ideal definition would be: feasible to construct, link well to other available data sources and benchmarks, have borders that are sensible given the research context, and be equally appropriate as a measure for a variety of relevant subgroups and settings. Each of the example proxy measures suffers disadvantages on some of these criteria. For example, ZIP codes have boundaries that are relevant for postal delivery but do not otherwise map cleanly to other data sources or meaningful policy units. On the other hand, Census PUMAs are feasible to construct, but are often much larger in land area and include many more people than would be ideal for notions of community or neighborhood.

The 2012 National Survey of Early Care and Education (NSECE) used GIS technology to develop a census tract cluster as a proxy for child care search areas. A child care search area, the geographic area in which parents might search, is typically quite compact -- within 3 miles of a household – but much larger in areas with sparse populations or geographic obstacles to travel such as rivers, highways, or mountains. Defined around an anchor tract where a household is located, the cluster includes any tract that overlaps at all with a circle of radius two miles around the population centroid of the anchor tract. The resultant cluster dimensions mimics child care usage in that densely populated areas have clusters that may be 3 miles across, while sparsely populated rural areas may have clusters that are 30 miles across. Clusters contain between one and 60 census tracts, depending on the local population density and geographic features. Using American Community Survey data, one can construct community characteristics by aggregating across all of the tracts within the cluster.

This paper investigates the use of the tract cluster as a measure of neighborhood by examining differences in estimates of key measures using three definitions of neighborhoods: census tracts, the tract cluster, and the county. We investigate the poverty density of neighborhoods in which households of different income levels reside, within-neighborhood vs between-neighborhood components of variation in black-white residential integration, and presence of a publicly-funded child-care facility within the neighborhood. We conclude with advantages and challenges of using the tract cluster as an analytic construct, possibilities for using the definition to such policy areas as food deserts/nutritional access, spatial mis-match in employment, and exposure to crime.