Publications

A selection rather than a complete list. It is arranged to show the arc the unit works along: first the systems that make handwritten archives readable, then the research those data enable — which is the only real test of whether the infrastructure was worth building. Complete and current lists are on the SDU research portal.

Methods and infrastructure

• Dahl, C.M., Johansen, T.S.D., Sørensen, E.N., Westermann, C.E., & Wittrock, S.F. (2026). DARE: A large-scale handwritten date recognition system. International Journal on Document Analysis and Recognition. doi:10.1007/s10032-026-00587-5 

_{Dates are the primary key in historical record linkage. This is the system that reads them.}

• Vedel, C., Dahl, C.M., & Johansen, T.S.D. (2024). Breaking the HISCO barrier: Automatic occupational standardization with OccCANINE. EHES Working Paper 255. arXiv:2402.13604 

_{96% accuracy across 13 languages, trained on 14 million description–code pairs; released openly as a working tool.}

• Dahl, C.M., Johansen, T.S.D., Sørensen, E.N., Westermann, C.E., & Wittrock, S.F. (2023). Applications of machine learning in tabular document digitisation. Historical Methods, 56(1), 34–48. doi:10.1080/01615440.2023.2164879 

_{The end-to-end pipeline — layout analysis, transcription, structuring — for tabular archival sources.}

• Dahl, C.M., Johansen, T.S.D., Sørensen, E.N., & Wittrock, S. (2023). HANA: A handwritten name database for offline handwritten text recognition. Explorations in Economic History, 87, 101473. doi:10.1016/j.eeh.2022.101473 

_{An openly released benchmark of 1.1 million handwritten name images.}

• Johansen, T.S.D., Dahl, C.M., Hwang, S.I.M., & Squires, M. Improving historical census transcriptions: a machine learning approach. Working paper.

_{Retranscribing the historical US censuses raises linkage rates most for the groups hardest to link — changing not just how much data there is, but whom it is about. See the project page.}

• Johansen, T.S.D., Koschnick, J., & Vedel, C. (2026). How to deal with machine learning bias in economic history. Working paper.

_{What systematic transcription error does to downstream estimates, and what to do about it.}

Health, human capital and the life course

• Baker, J.L., Bjerregaard, L.G., Dahl, C.M., Johansen, T.S.D., Sørensen, E.N., & Wüst, M. (2026). Universal investments in toddler health: learning from a large government trial. The Economic Journal, ueag057. doi:10.1093/ej/ueag057

_{A large-scale evaluation of an early-life state intervention, on a dataset transcribed and curated entirely by machine from archival records.}

• Bjerregaard, L.G., Johansen, T.S.D., Dahl, C.M., & Baker, J.L. (2024). Duration and intensity of being breastfed and educational attainment, income and labour force participation. Public Health, 237, 37–43. doi:10.1016/j.puhe.2024.08.026

_{Machine-transcribed early-life records linked to adult outcomes — the life-course design in its simplest form.}

• Bjerregaard, L.G., Wüst, M., Johansen, T.S.D., Sørensen, T.I.A., Dahl, C.M., & Baker, J.L. (2023). Cohort profile: the Copenhagen Infant Health Nurse Records (CIHNR) cohort. International Journal of Epidemiology, 52(6), e340–e346. doi:10.1093/ije/dyad096

_{How an archive of handwritten records becomes a documented, governed cohort others can apply to use.}

• Bruun-Jensen, M., Winther, S., Schmidt, S.E., & Dahl, C.M. (2024). DARE-ISC model for prediction of 1-year ischaemic stroke risk. BMJ Open, 14(5), e076640. doi:10.1136/bmjopen-2023-076640

_{Modelling on the full Danish administrative registers — the register environment historical records must ultimately link into.}

Economic and social history

• Dahl, C.M., Hansen, C.W., Jensen, P.S., Karlsson, M., & Kühnle, D. (2023). School closures, mortality, and human capital: evidence from the universe of closures during the 1918 pandemic in Sweden. IZA DP 16592 / CEPR DP 18399; revise and resubmit, American Economic Journal: Economic Policy. Working paper. 

_{500,000 archival death certificates and closure records for 2,100 school districts, transcribed by machine.}

• Dahl, C.M., Hansen, C.W., & Jensen, P.S. (2022). The 1918 epidemic and a V-shaped recession: evidence from historical tax records. The Scandinavian Journal of Economics, 124(1), 139–163. doi:10.1111/sjoe.12456 

_{Archival tax records used to identify the economic consequences of an epidemic.}

• Borowiecki, K.J., & Dahl, C.M. (2021). What makes an artist? The evolution and clustering of creative activity in the US since 1850. Regional Science and Urban Economics, 86, 103614. doi:10.1016/j.regsciurbeco.2020.103614 

_{A cultural- and economic-history study built entirely from digitised census records — the same infrastructure serving a humanities question.}

• Bentzen, J.S., Boberg-Fazlić, N., Sharp, P., Skovsgaard, C.V., & Vedel, C. (2026). Holy cows and spilled milk: the impact of religious missions on firm-level productivity. Journal of Development Economics. 

_{Firm-level historical data at scale, from a database of 1,419 Danish cooperative creameries.}