Reproducible LLM Workflows in Economic History: Degener's Wer Ist's?

Forthcoming in Jahrbuch für Wirtschaftsgeschichte

This project develops a transparent and reproducible workflow for using large language models to turn dense historical print sources into structured research data. The paper demonstrates the workflow by digitizing Hermann A. L. Degener’s Wer Ist’s? (1911) and discusses how historians can combine OCR, preprocessing, and LLM-assisted extraction without losing reproducibility.

Abstract

This paper develops a transparent and reproducible LLM-based workflow for turning dense historical print sources into structured data, and demonstrates it by digitizing Hermann A. L. Degener’s Wer Ist’s? (1911). The source contains roughly 16,000 biographies spread across about 1,900 pages and is challenging due to dense typography, non-standard abbreviations, and frequent minor errors. The workflow has modular steps that are divided into small tasks, which also involve humans for verification. These modular steps include image pre-processing and OCR, biography assembly with sanity checks, variable identification, field extraction, validation, and normalization that expands abbreviations while flagging uncertainty. The workflow is also transparent; prompts, logs, model responses, and outputs link each biography back to the original source.

Code

Kiel Working Paper No. 2324, June 2026 with Sebastian Braun

Using parish-level data for Baden and Württemberg, this paper studies how Catholic-Protestant cleavages shaped electoral fragmentation from Weimar Germany into the postwar period. It shows how Nazi and later CDU mobilization changed the political effects of the religious divide, while Catholic areas remained less susceptible to far-right voting.

Abstract

How do social cleavages affect party-system fragmentation, and what changes when parties organize across rather than along those divides? We study the Catholic-Protestant cleavage in Germany from the Weimar Republic through the postwar period using the confessional mosaic of Baden and Württemberg. Based on more than 3,000 parishes, we document a hump-shaped relationship between the Catholic population share and electoral fragmentation in Weimar Germany: fragmentation was greatest in mixed parishes, where a cohesive Catholic bloc faced a divided Protestant center-right. The Nazi Party subsequently consolidated much of that fragmented Protestant vote during the Great Depression. After 1945, the cross-confessional CDU weakened the relationship between religion and fragmentation. Catholic areas nevertheless remained less receptive to the far right, a persistence that survey evidence links to church influence rather than differences in policy preferences. The results show that the political effects of social cleavages depend on how parties and organizations mobilize them.

Website Paper Dataset

Protestantism, Industrialization, and the Fertility Decline: Evidence from Religious Enclaves in Württemberg, 1871-1933

Working Paper with Sebastian Braun and Richard Franke

This project studies the relationship between Protestantism and fertility decline in the Kingdom of Württemberg between 1871 and 1933. Using data covering the universe of 1,764 localities, the paper provides large-scale descriptive evidence on the timing of fertility decline at the parish level in Germany and links religious enclaves to broader processes of demographic change and industrialization.

Abstract

The fertility decline in industrializing Europe marks a turning point in human history, and there is ongoing debate about its causes and their relative importance. This paper examines the impact of Protestantism on fertility decline in the southwestern German Kingdom of Württemberg, 1871-1930. Using data on the universe of 1,764 localities, we provide the first large-scale descriptive evidence on the timing of fertility decline at the parish level for Germany. We find that there is as much heterogeneity in the timing of fertility decline within Württemberg as there is across districts in Europe. To provide causal evidence on the effect of religion on the timing of the fertility decline, we exploit that religion was determined by decisions of local rulers in the 16th century, decisions that are arguably exogenous to fertility decisions of people in the 19th and 20th centuries. Furthermore, we show that neighboring parishes do not differ in economic, geographic, or institutional characteristics other than religion. Our results indicate that Protestant parishes experienced the onset of fertility decline significantly earlier than Catholic parishes, by about ten years. We hypothesize that an interaction between religious norms, secularization, and differential responses to changes in infant mortality drives the results and provide preliminary evidence for this notion.

Dataset