Reducing the Risk of Data Disclosure through Area Masking: Limiting Biases in Variance Estimation

• Activity Number: 462
• Type: Topic Contributed
• Date/Time: Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Inho Park*+ and Sylvia Dohrmann and Jill Montaquila and Leyla Mohadjer and Lester R. Curtin
• Companies: Westat and Westat and Westat and Westat and Centers for Disease Control and Prevention
• Keywords: disclosure control ; segment swapping ; stratified multistage sampling ; weighted mean ; design effect ; national health and nutrition examination survey

Abstract

When preparing public files, variables used to compute variances are sometimes masked to limit the risk of data disclosure. For example, in area surveys with a limited number of primary stage units (PSUs), the original PSUs are split and recombined to construct pseudo-PSUs with swapped second-stage units (segments). Masking PSU identifiers is an effective way to reduce data disclosure risks. However, such masking results in biased variance estimates, as observed in previous work (Dohrmann et al. 2002, 2004). This paper considers alternative masking strategies, with a focus on reducing the effects of segment swapping on the variance estimates, and the resulting bias. Data from the National Health and Nutrition Examination Survey is used to illustrate the impact of the new masking strategies on variance estimates computed using the masked PSUs.