Evaluating the Quality of a National Voter Registration File Using Administrative Records

• In Event: Measuring What Matters: Improving Data Quality and Decision-Making under Uncertainty

Abstract

This report presents a comprehensive, nationwide assessment of the data quality of 2020 commercial voter registration files. Each voter registration record is processed through the Census Bureau’s Person Identification Validation System (PVS), which uses probabilistic matching across multiple administrative data sources to assign anonymized person identifiers (PIKs). These PIKs serve as a powerful linking key, enabling systematic, large-scale evaluation of data accuracy, quality, and consistency across state voter registration files and supporting a rigorous assessment of overall list maintenance quality. Using this linkage infrastructure, the study produces the most detailed national estimates to date of duplicate voter registrations—both within states and across states. It also measures the number of deceased individuals who remain on voter registration lists by linking the PIKs to the Census Numident, which contains death information maintained by the Social Security Administration. Together, these measures provide a clearer understanding of how well voter registration systems reflect the current eligible voting population and where gaps remain. In addition to these error measures, the study evaluates the completeness and reliability of demographic and address information within voter registration files. The relationship between data quality and successful PIK assignment is examined to identify the types of records most vulnerable to non-linkage and to understand how missing or inconsistent information affects the ability to integrate voter files with other datasets. This analysis offers additional insight into how specific data limitations can shape the quality and usability of voter registration files for policy, research, and election administration purposes. The principal contribution of this paper is its methodological novelty: to my knowledge, it is the first to apply nationwide administrative record linkage to a complete commercial voter registration dataset. While existing literature has documented concerns about list maintenance through case studies, audits, or state-level analyses, this research provides a unified, empirically grounded national picture of duplication, outdated records, and data reliability. The findings offer new insights into longstanding discrepancies between survey-based and administrative estimates of voter registration rates and provide actionable guidance for improving voter list maintenance and strengthening the validity of research that relies on these data.

Author

• Andres Felipe Mira, U.S. Census Bureau
  Presenting Author