ATLASS

Automated transcription and linkage of Swedish archival school data

Sweden's school archives contain something unusual: systematic, individual-level records of children — attendance, grades, health examinations, teacher assessments — kept consistently across the country for generations, and preserved. They are also almost entirely handwritten and, until now, effectively unusable at scale.

ATLASS transcribes them automatically and links them into SwedPop, the Swedish national research infrastructure that already holds harmonised population data from parish registers, censuses and later administrative sources. The point of linking rather than simply transcribing is that it closes a gap in the life course: SwedPop can already follow a person from birth to death, but the school years — where much of the human-capital story actually happens — have been missing.

What it makes possible

Once school records are linked to the rest of a life, a set of long-standing questions become tractable with real data rather than proxies. How much of adult inequality is already visible in childhood? What did schooling reforms actually change for the children exposed to them, as against the cohorts just before? How do early health, family circumstance and school performance interact in producing adult outcomes — and how much of that transmits to the next generation?

Because SwedPop is an infrastructure rather than a project dataset, the answer is not a single paper. The transcribed records become available to the wider research community on SwedPop's own access terms, which is the durable form of the contribution.

BDAD's contribution

The unit supplies the transcription and linkage layer: handwritten text recognition tuned to Swedish school records, structured extraction of the fields that matter, occupational and categorical standardisation, probabilistic linkage into existing SwedPop identifiers, and quantified accuracy reporting so that users of the data know what they are working with.

ATLASS is the second national infrastructure — after Geolinking in the UK — to adopt the unit's methods, and the Swedish sources are a natural fit: the underlying transcription models were originally developed and validated on Scandinavian institutional registers.

The team

The project's composition explains its ambition. Mikael Lindahl works on education, human capital and the transmission of advantage between generations; Gustav Kjellsson and Martin Karlsson are health economists, Karlsson also an economic historian and a long-standing collaborator of this unit — the study of school closures during the 1918 pandemic in Sweden is joint work. BDAD supplies the transcription and linkage.

Between them that covers the whole chain the project depends on: getting the records out of the archive, joining them to the rest of a life, and then asking what schooling did to human capital and to health. ATLASS is in that sense a continuation of an existing collaboration rather than a new one — the same people, with the school records finally available.