The Potential for Big Data to Improve Neighborhood-Level Census Data

Abstract

The promise of “big data” for those who study cities is that it offers new ways of understanding urban environments and processes. Big data exists within broader national data economies, these data economies have changed in ways that are both poorly understood by the average data consumer and of significant consequence for the application of data to urban problems. For example, high resolution demographic and economic data from the United States Census Bureau since 2010 has declined by some key measures of data quality. For some policy-relevant variables, like the number of children under 5 in poverty, the estimates are almost unusable. Of the 56,204 census tracts for which a childhood poverty estimate was available 40,941 had a margin of error greater than the estimate in the 2007–2011 American Community Survey (ACS) (72.8 % of tracts). For example, the ACS indicates that Census Tract 196 in Brooklyn, NY has 169 children under 5 in poverty ±174 children, suggesting somewhere between 0 and 343 children in the area live in poverty. While big data is exciting and novel, basic questions about American Cities are all but unanswerable in the current data economy. Here we highlight the potential for data fusion strategies, leveraging novel forms of big data and traditional federal surveys, to develop useable data that allows effective understanding of intra urban demographic and economic patterns. This paper outlines the methods used to construct neighborhood-level census data and suggests key points of technical intervention where “big” data might be used to improve the quality of neighborhood-level statistics.