Interpolating Population Distributions using Public-use Data with Application to the American Community Survey

• Activity Number: 310 - Modern Approaches to Small Area Estimation with Spatial Modeling and Machine Learning
• Type: Topic Contributed
• Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 11:50 AM
• Sponsor: Survey Research Methods Section
• Author(s): Matthew Simpson* and Scott H. Holan and Christopher Wikle and Jonathan Bradley
• Companies: SAS Institute and University of Missouri and University of Missouri and Florida State University
• Keywords: Bayesian statistics; density estimation; official statistics; income distribution

Abstract

Statistical agencies publish aggregate estimates of various features of the distributions of several socio-demographic quantities of interest based on data obtained from a survey. Often these area-level estimates are tabulated at small geographies, but detailed distributional information is not necessarily available at such a fine scale geography due to data quality and/or disclosure limitations. We propose a model-based method to interpolate the disseminated estimates for a given variable of interest that improves on previous approaches by simultaneously allowing for the use of more types of estimates, incorporating the standard error of the estimates into the estimation process, and by providing uncertainty quantification so that, for example, interval estimates can be obtained for quantities of interest. Our motivating example uses the disseminated tabulations from the American Community Survey to estimate U.S. Census tract-level income distributions in order to construct income segregation indices. Then using the indices we study the correlates of income segregation.