Bridging the Data-Theory Divide Using Regionalization and Uncertainty for Neighborhood Identification

Graduate Thesis Or Dissertation

Citations

Citeable URL: https://scholar.colorado.edu/concern/graduate_thesis_or_dissertations/x059c765v

Abstract

Census tracts are often equated with neighborhoods to study small-scale social phenomena. The association of tracts and neighborhoods suffers from two major problems: data quality and theoretical implications. Tract-level data in the American Community Survey (ACS) contains high uncertainty measured as sampling error. Neighborhoods are inherently vague, formed from a mix of complex social, political, economic, historical, and environmental factors. This analysis attempts to reconcile statistical and geographic uncertainty accompanying the association of tracts and neighborhoods by applying Spielman & Folch’s data-driven regionalization algorithm to identify neighborhoods in the Denver metro area using ACS data. The results of the algorithm are unstable, undermining efforts to compare data-driven regions to neighborhoods as defined in theory or practice. The challenge of reconciling data and theory as demonstrated in this experiment suggests that rather than equating tracts and neighborhoods, tracts should be interpreted as zones that can be aggregated in flexible combinations.

Creator