Comparing Cell Perturbation to Cell Suppression for Statistical Disclosure Control of Tables

• Activity Number: 438
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics
• Author(s): Paul B. Massell*+
• Companies: U.S. Census Bureau
• Keywords: cell suppression ; cell perturbation ; statistical disclosure control ; confidentiality

Abstract

Statistical tables often contain cells whose values are too sensitive for public release. The method of cell suppression has been used for decades at statistical agencies to protect such cells. For additive tables, it involves suppressing not only the sensitive cells but additional cells called secondaries. These latter cells are needed to ensure that the additive relations do not allow full recovery of the sensitive values. The goal of disclosure control is to ensure that the best estimates of the sensitive cells have a sufficiently wide range, i.e., are sufficiently uncertain. It is known that one can achieve this degree of uncertainty in ways that are less extreme than suppression. With such methods, e.g., cell perturbation, an approximate value is released for every cell and table additivity is preserved, but a larger number of cells may need to be modified than in cell suppression. We comment on some recent work along these lines, e.g., work of Juan-Jose Salazar Gonzalez and joint work of Ramesh Dandekar and Lawrence Cox. We discuss some theoretical and implementation aspects of these methods.