Expected Number of Random Duplications Within or Between Lists

• Activity Number: 314
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): William E. Yancey*+
• Companies: U.S. Census Bureau
• Keywords: record linkage ; deduplication ; Stirling numbers

Abstract

The U.S. Census seeks to determine duplicately listed individuals by searching across lists to identify records with the same name and birth date. The question arises of how many of these agreeing records are random agreements, two different people with the same name and birth date. To formally answer this question, we consider first the familiar Birthday Problem and then the more complicated Collision Problem. For each of these problems we exhibit the explicit probability distributions from which we can compute means and variances for some parameter values. We apply this result to voter registration lists for Oregon and Washington to estimate the number of "false matches" occur across these lists.