Temporal Harmonization of Industry Classifications with Swedish Register Data

This project addresses a challenge posed by the substantial revision of the Swedish Standard Industrial Classification (SNI) codes implemented in 2007 by Statistics Sweden. The absence of a well-defined mapping between the old and new codes has introduced significant obstacles for longitudinal industry analyses of Swedish labor market data. When aggregating longitudinal data for temporal machine learning (ML), unsystematic changes in categorical variables are difficult to deal with. Our objective is to develop a machine learning-based probabilistic model to harmonize SNI codes over time which contributes to making Swedish labor market data accessible for machine learning applications and other longitudinal studies.