A New Approach to Protect Confidentiality for Census Microdata with Missing Values

• Activity Number: 158
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Yajuan Si* and Jerome P. Reiter+
• Companies: Duke University and Duke University
• Keywords: synthesis ; missing data ; census ; confidentiality ; multiple imputation

Abstract

Protection of the sensitive values at high risk of disclosure or values of key identifiers is a crucial problem. Meanwhile, there usually exists missing data in the census. Multiple imputation has been a useful tool to protect data confidentiality and handle missing data. We propose a new approach to census microdata dissemination: sampling simultaneously with synthesis and missing data, which generates multiple imputed datasets that simultaneously handle missing data and disclosure limitation. The basic idea is to fill in the missing data first in the census to generate multiple complete populations, then replace the identifying or sensitive values in each population with multiple imputed values, and finally release samples from these multiple imputed populations. This article develops methods to obtain valid inferences from the new three-stage imputation process.