Spatial data uncertainty for location modeling: Ghost blocks and their implications

• Published in: Applied geography (Sevenoaks) 2024-05, Vol.166, p.103266, Article 103266
• Main Authors: Grubesic, Tony H., Wei, Ran, Helderop, Edward
• Format: Article
• Language: English
• Subjects: California; case studies; Census; geography; Housing units; issues and policy; Location modeling; Master address file; Spatial analysis; spatial data; Uncertainty
• ISSN: 0143-6228; 1873-7730
• DOI: 10.1016/j.apgeog.2024.103266

Census blocks are administrative units that serve as statistical areas for the decennial Census in the United States. Visible and nonvisible features bound blocks, including roads, railroads, streams, property lines, and city boundaries. The Census Bureau builds blocks using the Master Address File (MAF), which includes field-verified geographic information about the location of housing unit addresses. Unfortunately, there are substantial errors in the counts of housing units at the block level, even with the purported quality checks by the Census Bureau. This paper aims to detail a method of identifying problematic blocks (i.e., ghost blocks) that report the presence of housing units, but no such units exist. Further, we identify the implications of using ghost blocks in location models using the maximal covering location problem (MCLP) in a case study for sensor locations in Los Angeles, California. We discuss policy implications and strategies to address these errors for developing higher-fidelity location models. •Details a method of identifying problematic U.S. Census blocks that report the presence of housing units, but no such units exist.•Identifies the implications of using ghost blocks in location models.•Discusses policy implications of using ghost blocks.•Provides strategies to address data uncertainty and to develop higher-fidelity location models.