Improving the Annual Survey of Local Government Finances' Sample Design and Estimation

• Activity Number: 422
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Government Statistics Section
• Author(s): Joseph Barth* and Elizabeth Love
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: Estimation ; Empirical Bayes ; Cut-off Sampling ; Sample Design ; , Public Sector

Abstract

The sample for the Annual Survey of Local Government Finances (ASLGF) is drawn every five years based on data from the most recent Census of Governments. In 2009 we introduced a modified cut-off sampling procedure which reduced the sample size while maintaining our quality standards. In 2014 we improved the design by examining multiple subsampling rates for the cut-off portion of the sample and considering their impact on mean squared error. We also considered three different estimators in conjunction with this sample design: the Horvitz-Thompson, calibration, and empirical Bayes estimators. All combinations of estimator and subsampling rate were evaluated through a Monte Carlo simulation to determine subsampling rates which maintain a three percent coefficient of variation threshold for all states. In this paper we show how we performed the modified cut-off sampling, describe the estimators used, and present the evaluations used to validate our final methodology.