Christian Vedel

Did religious missions influence firm-level productivity? This study examines the impact of the Inner Mission movement on productivity in early twentieth-century Denmark, a predominantly Protestant and homogeneous society undergoing rapid industrialization. Using data on 964 creameries and an instrumental-variables approach, we provide evidence that Inner Mission intensity reduced productivity in butter production, measured by the milk-to-butter ratio. The main mechanisms were Sunday closures, which disrupted economies of scale, and the fragmentation of creameries due to doctrinal disagreements among farmers. The results show how religiosity shaped patterns of modernization absent the usual colonial or missionary confounders, speaking to debates in development economics about how informal institutions and moral regulation can constrain firm behavior and productivity.

First-nature geography shapes the location of prosperity. I provide evidence by investigating the effects when it suddenly changes. In 1825 a storm breached the Agger Isthmus, connecting Denmark’s west Limfjord region to the North Sea. I demonstrate that trade followed, and prosperity relocated with it: population rose 27 percent within a generation — an elasticity of 1.6 relative to market access — with occupational shifts toward fishing and manufacturing. Fertility, not migration, drove the expansion. A mirror experiment, the waterway’s closure circa 1086–1208, caused symmetric declines in medieval coin and building finds.

OccCANINE is an open-source tool that maps occupational descriptions to HISCO codes, replacing weeks of manual coding with results in minutes. We fine-tune CANINE on 15.8 million description–code pairs from 29 sources in 13 languages, achieving 96 percent accuracy, precision, and recall. The approach generalizes to three further systems — OCC1950, OCCICEM, and ISCO-68 — all released open source. By breaking the “HISCO barrier,” OccCANINE democratizes access to high-quality occupational coding, enabling broader research in economics, economic history, and related disciplines.

Machine learning (ML) has rapidly transformed economic history, lowering costs of digitization, data linkage, and imputation, and making information in historical text usable at scale. This paper offers a practical guide to using these tools well. However, ML tools have also created new problems. Prediction errors are often systematically correlated with covariates of interest, so even highly accurate models can distort and sometimes reverse coefficients, and standard validation cannot detect this. Given that ML tools often perform worse for historical data, this problem is especially severe for the field of economic history. We also identify a solution to this problem. We show that recent debiasing methods can correct such bias for a wide class of applications, using a small, randomly sampled set of expert-coded labels while retaining the efficiency of large-scale prediction. We organize the field with a taxonomy of three ML tasks, survey the literature along it, and indicate where debiasing applies and where validation against proxies remains the only recourse. We close with best-practice guidance on digitization, model choice, and reproducibility.

History is rich in information about people’s occupations but far less forthcoming on outcomes such as income, wealth, or skills. A widely used solution is to proxy these outcomes by averages within occupational categories—so-called occupational scores—of which IPUMS OCCSCORE is the most prevalent. But existing scores rely on fixed benchmarks such as 1950 U.S. wages, discard ambiguity in historical titles, and are difficult to adapt across time and space; constructing a new score by hand for every project is infeasible. This paper introduces CHAOS, a replicable, automatic, and interpretable framework for converting historical occupational descriptions into outcome estimates given any source. The key insight is that every occupational score is a weighted average of observed outcomes, where the weights reflect the relevance of each piece of source information. These relevance weights can be estimated automatically using a classification algorithm such as OccCANINE—and the classifier’s own uncertainty, which would otherwise bias the resulting scores, can be removed with a transparent correction. Alongside the estimator we provide diagnostics that report, for any score, how much source evidence stands behind it. We demonstrate the utility of CHAOS through an application to historical wage reports covering 88,000 occupation–income pairs across U.S. states and other countries spanning the 19th century. Following the occupations of Campbell’s 1747 London Tradesman through the 18th and 19th centuries, we use these estimates to trace the evolution of skill premia across the Industrial Revolution.