The Consequences of Person-Based Record Linking Methods on Empirical Results

• In Event: Causal Inference and Algorithmic Discovery for Criminology and Public Policy

Abstract

Person-based administrative record linkage is crucial in empirical social science research, enabling researchers to combine individual-level information from multiple sources. However, the choice of linking method can significantly impact the accuracy and reliability of the linked data, affecting empirical results and conclusions. This is particularly important when analyzing subgroups defined by demographic attributes, as the performance of a linking algorithm often varies by subgroup. While prior studies have highlighted the impact of errors in person-based data linking on empirical findings, there is a lack of research on how the choice of linking method itself can influence results, especially for subgroup estimates. In this paper, we compare several commonly used person-based data linking strategies and their associated software implementations to demonstrate how their performance affects coefficient estimates from regressions, with a focus on subgroup-specific estimates. Our goal is to emphasize the importance of person-based data linkage as a methodological choice and provide guidance for researchers in selecting an appropriate linking method. We discuss implications for transparency and reproducibility, offering recommendations for best practices in documenting and justifying data linking choices.

Authors

• Zubin Jelveh, University of Maryland

• Robert Stewart, University of Maryland

• Sarah Tahamont, University of Maryland