Assessing Disclosure Protection for A SOI Public Use File

• Activity Number: 306
• Type: Topic Contributed
• Date/Time: Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics*
• Author(s): Marianne Winglee*+ and Peter Sailer and Michael Weber and Richard Valliant and Jay Clark and Yunhee Lim
• Affiliation(s): Westat, Inc. and Internal Revenue Service and Internal Revenue Service and Westat, Inc. and Westat, Inc. and Westat, Inc.
• Keywords: subsampling ; Microdata ; disclosure risk ; information loss ; microaggregation clusters ; record linkage

Abstract

The Statistics of Income (SOI) program of the Internal Revenue Service (IRS) is mandated to provide data to support analyses of the tax system in the United States. For this purpose, SOI releases a data file each year that contains microdata from a representative sample of individual tax returns. A public-use version of this tax model file is released after disclosure control procedures to avoid potential recognition of individual taxpayers. Disclosure control techniques applied to the tax model file for public-use include suppression, global and local recoding, top coding, rounding, subsampling, and microaggregation (replacing reported data by the average value of a sorted cluster of three values). To address the recent concerns of information explosion, SOI has sponsored several evaluation studies to determine the availability of external data and the risk of data linkages. This paper discusses methods to measure the likelihood of correct matches for rare income returns and examines options to refine the subsampling method and the microaggregation clusters to enhance proper disclosure protection and to preserve the analytic value of the data.