Developing Adoptable Disclosure Protection Techniques: Lessons Learned from a U.S. Experience

摘要

The development of new disclosure protection techniques is useful only insofar as those techniques are adopted by statistical agencies. In order for technical experts in disclosure limitation to be successful, they are likely to need to interact with the appropriate statistical offices. This paper discusses just such a successful interaction in the United States. It describes the foundation that three major U.S. agencies - the Census Bureau, the Social Security Administration, and the Internal Revenue Service - laid in order to develop more useful statistical products. These included a proposed synthetic data public-use file based on the confidential microdata from all three agencies. Since then other governmental organizations, such as the U.S. Congressional Budget Office, have become involved with this inter-agency effort, which seeks to provide researchers and other users in the broader statistical community with a data utility often possible previously only with access to the confidential microdata. The confidentiality implications for all three agencies - and the potential for more - of a successful conclusion to this work would be enormously beneficial to data users, data producers, and data respondents. This paper describes the importance of developing the necessary framework, which includes an understanding between statistical office decision makers and the technical experts, before beginning such an endeavor. It provides a description of how this effort even became possible, and uses the history of events and related lessons to describe essentials that might be useful for other national statistical offices facing similar constraints and goals.