Hierarchical Bayesian Methods for Combining Surveys

• Activity Number: 83
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Yang Cheng*+ and Adrijo Chakraborty and Gauri Datta
• Companies: U.S. Census Bureau and University of Georgia and University of Georgia
• Keywords: Current Population Survey ; Gibbs Sampling ; Noninformative Priors ; Time Series

Abstract

In order to estimate the number of occupied households, US Census Bureau conducts many surveys. As a result, we get different estimates of the number of occupied households from these surveys. While each survey is useful, differences among the estimates they produce are sometimes very large. To resolve these differences, we propose in this study a hierarchical Bayesian method to obtain a more reliable estimate of the number occupied households by combining estimates from these surveys. Exploiting the repetitive nature of the surveys, we propose a time series model. We apply our method to the estimates from Current Population Survey (CPS)/Annual Social and Economic Supplement, CPS/Housing Vacancy Survey, American Community Survey, and American Housing Survey between 2002 and 2011 to produce a more reliable estimate of the number of occupied households. We implement our objective Bayesian method by Gibbs sampling.