Frameworks for Deriving Canonical Person Names in Integrated Administrative Data

Administrative and survey data integrations frequently produce multiple name variants for the same individual, complicating record linkage and downstream statistical use. We examine a selection of methodological frameworks for deriving a single canonical name for each person in a large-scale, multi-source environment. The frameworks span hierarchical choice, source weighting, recency, component completeness, cross-source frequency, random selection, as well as hybrid methodologies that utilize multiple methods. We illustrate the performance of these methods using synthetic data designed to mimic common inconsistencies observed in real-world systems. This small-scale simulation delivers a starting point to compare methodologies in terms of accuracy and computational expenditure while providing deeper insight on how data structure and source quality influence outcomes. The approach is broadly applicable to organizations seeking to aggregate and canonicalize data across varied sources while maintaining confidentiality protections