Motivating Standards for Disclosure Avoidance

• Activity Number: 403
• Type: Topic Contributed
• Date/Time: Thursday, August 12, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics
• Author(s): Philip M. Steel*+
• Companies: U.S. Census Bureau
• Keywords: confidentiality ; HIPAA ; record linkage

Abstract

Advances in computing and methods of access to data have made statistics an integral part of modern life. From marketing to evaluating public policy to clinical studies, solid data and good statistics are no longer the earmark of good, cutting-edge science, but rather an underpinning of everyday life. Many of these advances also raise privacy concerns and can create a trade-off between inadvertent disclosure of personal data and the accuracy of the inference from that data. Often data can be protected without harming the utility of the data and at very little cost. In some contexts protection and utility collide head on. Disclosure avoidance seeks to establish a best practices consistent with a statistical use of person level data and offer alternatives when utility cannot be maintained. This paper is intended as a general introduction to disclosure avoidance. We demonstrate some simple data attacks and techniques to block them. We give several examples of more complicated situations arising from the nature of the data and describe the outcome of those situations. We look at regulation in the federal data system, in HIPAA and in NIH funding requirements.