Using Weighting to Improve Cell Suppression Pattern in Annual Survey of Manufacturers

• Activity Number: 462
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Government Statistics Section
• Author(s): Bei Wang*+
• Companies: U.S. Census Bureau
• Keywords: cell suppression ; LP ; ASM ; weighting ; cost

Abstract

The Economic Census and other related products use cell suppression to protect sensitive data. Annual Survey of Manufacturers (ASM) provides sample estimates of statistics for manufacturing establishments between the Economic Censuses. Beginning with 2012 Economic Census, we are switching to a Linear Programming (LP) instead of network methodology, which has been used in the cell suppression program since the 1992 Economic Census. The new methodology avoids undersuppression and reduces oversuppression. A cell suppression LP problem consists of number of constraints and one objective function. The number of constraints is determined by the size of data and the number of relationships in each dimension. For example, for one of the ASM table group with 21,781 cells, the problem has 12,721 constraints and more than 43,572 variables. In the objective function, a cost (weight) is assigned to each cell, usually its value. However, each product has its own characteristics and different requirements, we feel it is possible to fine tune the weighting of the program to reach an optimal for that products. In general , this means fewer cells and value suppressed. While the constraint