NCRN Virtual Seminar - Center for Disclosure Avoidance Research

Speakers: Marlow Lemons (U.S. Census Bureau) and Paul Massell (U.S. Census Bureau)

Title: A Method to Improve Data Swapping at the U.S. Census Bureau (M. Lemons)

Abstract: Data swapping is one of several disclosure avoidance methods that the Census Bureau implements to uphold confidentiality mandated by law.  The Center for Disclosure Avoidance Research (CDAR) is currently studying the use of n-cycle swapping as a means to protect respondent identity in large-scale data.  N-cycle swapping, a variant of data swapping, uses permutations of size 'n' to swap data records rather than swapping them in pairs.  In this talk, we will discuss the processes surrounding n-cycle swapping, CDAR's current studies and challenges, and future projects and data products involving this disclosure avoidance technique. (archived presentation)

Title: Cell Suppression as used for Protecting Magnitude Data Tables (P. Massell)

Abstract: The most common data products released by the Economic Directorate of the Census Bureau are magnitude data tables.  Common magnitude variables in these tables are ‘sales’ (aka ‘receipts’), and ‘number of employees’. In this method, an agency uses the p% rule for determining which cells reveal too much information about particular establishment or company value contributions to the cell. Such a cell is declared sensitive and is suppressed. However, since Census tables are typically additive, additional cells, called ‘secondary’ suppressions, must also be suppressed in additive to make it impossible for a table user to recovery the value of any sensitive cell.  Using techniques from operations research, Census Bureau researchers developed methods for finding these secondary suppressions in a way that minimizes information loss from the table. Good software had been developed about 1990 for implementing cell suppression. We will discuss improvements to the method that have been implemented in the current version, such as better protection at the ‘company level’, handling of negative values, and improved processing of linked tables. (archived presentation)


Apr 01, 2015, 3:00pm to 4:30pm EDT
4600 Silver Hill Rd.
Suitland, MD
United States