As our society balances community needs and individual rights, an energetic data dissemination program by government agencies in support of public policy research must have parity with an effective disclosure limitation program. The substantial growth which has occured in the number of publicly collected microdata files in recent years has made it increasingly difficult, however, to reconcile the public's interest in scientific inquiry with its equally legitimate interest in protecting individual privacy. This project seeks to promote the usefulness of federal statistical data through the development of methodologies that will allow important types of social data to be accessible while assuring an appropriate level of confidentiality. Specifically, the aim is both to develop and evaluate disclosure-limiting techniques that will help guide microdata dissemination. The disclosure-limiting data disseminating approach advanced by Duncan and Lambert in 1985 will be extended to both complete census data and sample data. Examination will be made in the DL context of a variety of proposed disclosure avoidance techniques, for both reidentification and data base query situations. There are four kinds of resolutions between the tension between dissemination and disclosure: ethical guidelines to encourage responsible use of data; administrative rules to control access to data; legal specifications that delineate when data may or may not be used; and statistical techniques to mask data on individuals without compromising information about the group. The approach embraced in this proposal has potential applicability to all four kinds of disclosure limitations. To date, only statistical controls have had any exploration in this framework. The proposed research would effect a significant extension to ethical guidelines, administrative rules, and legal specifications.
