|
Abstract:
|
This paper describes a two step methodology for measuring person duplication in a census. Firstly, an exact match is conducted between census records and an auxiliary file of administrative records covering most of the Census target population to find easily identified duplicates, based on name and demographic variables. A name frequency file is used to assign first and last names. Secondly, an estimate of less easily identified duplicates is derived using clerical verification of a sample of potential duplicates obtained by probabilistic record linkage, based on name and demographic variables. Geographical proximity of addresses is also considered when calculating the strength of the probabilistic linkage. The methodology, its advantages over previous methods used at Statistics Canada, innovative features, and initial quantitative evaluations of the exact match methodology are outlined.
|
