Confidentiality and Disclosure Limitation Methodology: Challenges for National Statistics and Statistical Research

Stephen E. Fienberg


For many years, pledges of confidentiality to respondents in censuses and surveys were interpreted in an absolute fashion and agency statisticians created conservative rules for disclosure avoidance which they believed would prevent disclosure of confidential information. During the past twenty-five years, the field of disclosure protection has undergone a ``statistical transformation'' and begun to utilize the advances that have occurred within the field of statistics itself. This paper provides an overview of the statistical issues that are related to the evolving area of statistical disclosure limitation methodology and illustrates the ideas in the context of two examples of current interest to U. S. government statistical agencies and others.

Keywords: Decennial census releases; Inferential disclosure; Intruder; Longitudinal datafiles; Matrix masking; Perturbation methods; Simulated data.

Here is the full postscript text for this technical report. It is 3059967 bytes long.