Online Program

Methods for Assessment of Incremental Disclosure Risks Incurred Through the Release of Special Tabulations
*John Eltinge, Bureau of Labor Statistics 

Keywords: Confidentiality, data quality, disclosure limitation, loglinear model, remote access, sequential query system, stakeholder utility functions,subpopulations

Disclosure limitation methods for tabular data have tended to focus on cases that involve (1) a well-defined set of prospective tables that are specified a priori; or (2) tables produced through multiple requests submitted through a remote-query system. Each of class (1) and (2) can potentially involve a tabular analyses of a single dataset, or analyses of a sequence of datasets associated with distinct points in time. Some statistical organizations also encounter a third class of tabular analyses that are based on special requests from one or more stakeholders. This third class of analyses may have several features that are not well defined a priori, including the subpopulations and tabular classification structure of interest; the level of estimator precision needed by the primary stakeholders; and the degree of sensitivity related to cell-level and unit level disclosure risks. This paper outlines the abovementioned features; adapts some standard disclosure analysis tools to evaluate disclosure risks incurred within this third class of tabular analyses; and explores the prospective use of standard tabular masking methods to ameliorate the abovementioned risks. Special attention is directed toward characterization of risks incurred through specific sequences of special-tabulation requests, and decision rules related to these requests.