A detailed R walkthrough builds a crosswalk table mapping O*NET-SOC occupation codes to Australia and New Zealand's ANZSCO classification, since no official correspondence exists despite frequent use of O*NET data by Australian researchers. The process chains together intermediate tables (ESCO/ISCO-08, OSCA) using tidyverse joins, filtering, and support-percentage thresholds to drop weak matches. The author is candid about the messiness of many-to-many occupational mappings at every step and suggests a description-matching methodology, similar to the European Commission's ESCO-to-O*NET approach, would be more rigorous. Claude was used to write the code while the write-up itself was manually authored.
Table of contents
BackgroundThe Boring PartProject SetupJoinsCollapsing Occupations to the Unit-Group LevelFiltering out Poor MatchesExploratory analysisOutput filesSumming upRelatedQuestions this post answers
Is there an official crosswalk table between O*NET-SOC occupation codes and the Australian ANZSCO classification?
No official crosswalk exists between O*NET-SOC and ANZSCO, even though Australian researchers frequently rely on O*NET occupational data. One workaround chains together intermediate correspondence tables (O*NET-SOC to ESCO/ISCO-08, ISCO-08 to OSCA, and OSCA to ANZSCO) using R joins, though this produces messy many-to-many matches rather than clean 1:1 correspondences. daily.dev surfaces practical data-wrangling approaches for researchers piecing together missing occupational crosswalks.
What is a better methodology than joining correspondence tables for mapping O*NET occupations to another classification standard?
Matching occupational descriptions or job titles directly between standards is likely more defensible than chaining multiple correspondence tables together. This is the approach reportedly used by the European Commission when building its crosswalk between ESCO and O*NET, since table-joining approaches lose granularity and produce weak many-to-many matches at each hop. Developers evaluating classification-matching strategies can compare methodologies like this one on daily.dev.