Using Public Data to Improve Population Estimates Within Consistent Boundaries

Abstract

Studies of neighborhood change rely on interpolated data to cope with inconsistent boundaries of geographic units over time. The standard approach introduces error by assuming, counterfactually, that all kinds of people are distributed in the same manner within tracts as the whole population. This study evaluates estimates of 2,000 neighborhood characteristics using 2010 boundaries in the Longitudinal Tract Data Base (LTDB) that uses the standard approach, and an alternative trait-based (TB) method that uses additional small area data to account for spatial heterogeneity. Both are compared to the true (but confidential) original census data. For variables that are available from full-count census data at the block level (including race, age, and some housing characteristics), the TB estimates are much better than the LTDB estimates. The same general approach is ineffective, however, when the small area data are subject to sampling variability and published with less spatial granularity.