207 – Applications of Calibration and Empirical Bayes Estimation Methods
Evaluating Calibration Estimators for the Annual Survey of Local Government Finance
Elizabeth L. Love
U.S. Census Bureau
Joseph J. Barth
U.S. Census Bureau
Bac Tran
U.S. Census Bureau
The Governments Division of the U.S. Census Bureau conducts the Annual Survey of Local Government Finances (ALFIN). The ALFIN provides statistics about the financial activities of state and local governments across the country. We currently use calibration to estimate these finance statistics. Calibration methods adjust sampling weights so that the adjusted weight totals agree with reliable known totals, e.g., census totals (or census counts) obtained from the Census of Governments. In previous cycles of the ALFIN, survey analysts used decision-based estimation, a technique that performs hypothesis tests that allow combining strata when possible to reduce the variance and improve the accuracy of survey estimates. In this evaluation, we develop a design-based Monte Carlo simulation experiment in which we draw repeated samples from the 2007 Census of Governments data using the ALFIN sample design. We compute the decision-based, calibration, and Horvitz-Thompson estimates that use the generated sample and the 2002 Census of Governments data as auxiliary information. We then compare mean squared errors of these estimators.