Comparing Open-Source Record Linkage Software to Existing Approaches: A Case Study using Census Data

The U.S. Census Bureau has used entity resolution, also referred to as record linkage or
deduplication, for the decennial censuses and other programs such as post-enumeration surveys for
several decades. In particular, entity resolution was used on names and common characteristics in
the 2020 Censusto deduplicate censusresponses and to link people rostered in the Post-Enumeration
Survey (PES) to people enumerated in the Census. In preparation for the 2030 Census, the Decennial
Census research agenda includes assessing different record linkage methodology and software. This
paper discusses and compares BigMatch, which is a matching software developed at the Census
Bureau, against two recent options of open-source software, fastLink and Splink. This paper
provides results such as metrics of accuracy measured against 2020 PES clerically reviewed data,
for each software.