An Evaluation of Different Small-Area Estimators and Benchmarking for the Annual Survey of Public Employment and Payroll

• Activity Number: 78
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 4:00 PM to 5:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Bac Tran*
• Companies: U.S. Census Bureau
• Keywords: Government Units ; Empirical and Hierarchical Bayes ; Composite Estimator ; Horvitz-Thompson Estimator ; SPREE ; Monte Carlo Simulation

Abstract

The Annual Survey of Public Employment and Payroll (ASPEP), conducted by the Government Division of the U.S. Census Bureau, provides statistics on the number of federal, state, and local government civilian employees and their gross payrolls. Different small area estimators can be produced using the ASPEP data and auxiliary information from the preceding Census of Governments. We develop a design-based Monte Carlo simulation experiment in which we draw repeated samples from the 2007 Census of Governments data using the ASPEP sampling design and compute a wide range of estimates that use the generated sample and the 2002 Census of Government data. We then compare simulated design-based biases, variances, mean squared errors of these estimators. We repeat the experiment using the 2012 Census of Government data in order to understand if these properties change over years. The estimators covered under our simulation study includes: Horvitz-Thompson, SPREE, traditional composite, and empirical Bayes and hierarchical Bayes methods. Lastly, we present different benchmarking approaches and compare their performances.