Borrowing Strength from Census Data to Assess Survey Disclosure Risk

• Activity Number: 71
• Type: Topic Contributed
• Date/Time: Monday, August 9, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics
• Author(s): Christopher Duddek*+
• Companies: Statistics Canada
• Keywords: disclosure risk assessment ; public-use microdata files

Abstract

In practice, disclosure risk is often measured using survey data. Due to small sample sizes, however, this information can lead the statistician to alter data more than necessary in the production of public microdata files. In other words, when measuring disclosure risk, it is impossible to know if unique combinations in the sample are unique in the population since population uniqueness is typically unknown. This paper explores the potential of using a publicly available census microdata file to identify the risk of population uniqueness for an unrelated survey. Though the concept of "borrowing strength" is associated with small-area estimation, it is applicable in this context because we can use the great mass of Canadian census data to evaluate disclosure risk for a small-scale survey. Our quarterly survey, the Canadian Changes in Employment survey, is particularly appropriate for this analysis due to the fact that we have a survey frame which allows us, in a limited way, to identify population uniqueness. We compare a census-derived disclosure risk measure to a multiplicity-table measure.