Applying Alternative Variance Estimation Methods for Totals Under Raking in SOI's Corporate Sample

• Activity Number: 568
• Type: Topic Contributed
• Date/Time: Wednesday, August 4, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Valerie Testa*+ and Kimberly Henry and Richard Valliant
• Companies: IRS and IRS and University of Maryland
• Keywords: administrative data ; survey sampling ; raking ; Taylor series approximation

Abstract

SOI's Tax Year 2006 Corporate sample is a stratified Bernoulli sample of approximately 110,000 corporate income tax returns. Raking adjustments are performed using the sample's design strata, related to the return's size of assets and income, and the primary industry, based on collapsed categories of the six-digit NAICS code. We apply several alternatives to estimate the variance of national- and domain-level totals of several key variables of interest: ignoring raking, post-stratification, a Taylor series approximation, and the delete-a-group jackknife replication estimator (with 100 and 200 groups). Results demonstrate that the poststratified total had the highest variance estimates, while the linearization and Jackknife when implemented incorrectly produced variance estimates that were too small, despite large sample sizes.