NCJ Number
47186
Date Published
1977
Length
31 pages
Annotation
PROCEDURES FOR SAFEGUARDING THE CONFIDENTIALITY OF DATA DURING STATISTICAL ANALYSIS BY OUTSIDERS AND IN INTERFILE LINKAGES ARE CONSIDERED.
Abstract
INTRAFILE ANALYSIS BY OUTSIDERS INVOLVES PROVIDING ACCESS TO A SINGLE DATA FILE FOR PURPOSES OF STATISTICAL ANALYSIS. IN SUCH SITUATIONS, DELETION OF DIRECT PERSONAL IDENTIFIERS (NAME, SOCIAL SECURITY NUMBER) MAY NOT BE ENOUGH TO PRESERVE CONFIDENTIALITY. RESTRICTIONS ON THE RELEASE OF DATA THAT ARE PUBLICLY LISTED ELSEWHERE OR 'ERROR INNOCULATION' OF PARTICULARLY SENSITIVE VARIABLES MAY BE REQUIRED. MICROAGGREGATED RELEASE (CREATING 'SYNTHETIC AVERAGE ACTUAL INDIVIDUALS) IS SAFE BUT RESULTS IN A LOSS OF STATISTICAL EFFICIENCY. IN-FILE CAPACITY TO RUN OUTSIDERS' ANALYSES, WITH RANDOMIZED ROUNDING OF FREQUENCY TALLIES (A PROCEDURE FOR COUNTERING ANONYMITY-BREAKING POSSIBILITIES), IS THE BEST MEANS OF ASSURING CONFIDENTIALITY. INTERFILE LINKAGE REFERS TO A SEARCH FOR STATISTICAL RELATIONS BETWEEN THE DATA CONTAINED IN TWO CONFIDENTIAL FILES. INTERFILE LINKAGE HAS GREAT POTENTIAL VALUE FOR PROGRAM EVALUATION AND CAN BE ACHIEVED WITHOUT THE RELEASE OF INDIVIDUALLY IDENTIFIED DATA FROM EITHER FILE BY 'MUTUALLY INSULATED FILE LINKAGE' PROCEDURE FOR LINKING FILES WITHOUT MERGING THEM. LINK FILE BROKERAGE IS UNACCEPTABLE ON CONFIDENTIALITY GROUNDS, AND MICROAGGREGATION AND SYNTHETIC LINKING BY MATCHING ARE ACCEPTABLE ON STATISTICAL GROUNDS. FOR BOTH INTERFILE AND INTRAFILE DATA USE, IT WOULD BE BENEFICIAL FOR THE GOVERNMENT TO FUND INTERNAL STATISTICAL ANALYSIS CAPABILITIES IN IMPORTANT ADMINISTRATIVE ARCHIVES IN THE PUBLIC SECTOR (CENSUS BUREAU, SOCIAL SECURITY, INTERNAL REVENUE, CRIME, WELFARE) AND THE PRIVATE SECTOR (MEDICAL AND AUTOMOBILE INSURANCE). REFERENCES ARE PROVIDED. (AUTHOR ABSTRACT MODIFIED--LKM)