Verfassungs-, Sozial- und Wirtschaftsgeschichte

Journal of Data-centric Machine Learning Research (2025) abstract: The correct detection of dense article layout and the recognition of characters in historical newspaper pages remains a challenging requirement for Natural Language Processing (NLP) and machine learning applications in the field of digital history. Digital newspaper portals for historic Germany typically provide Optical Character Recognition (OCR) text, albeit of varying quality. Unfortunately, layout information is often missing, limiting this rich source’s scope. Our dataset is designed to enable the training of layout and OCR models for historic German-language newspapers. The Chronicling Germany dataset contains 801 annotated historical newspaper pages from the time period between 1617 and 1933. The paper presents a processing pipeline and establishes baseline results on in- and out-of-domain test data using this pipeline. Both our dataset and the corresponding baseline code are freely available online. This work creates a starting point for future research in the field of digital history and historic German language newspaper processing. Furthermore, it provides the opportunity to study a low-resource task in computer vision. Keywords: Digital History, Machine Learning, Optical Character Recognition (OCR)

Abstract. Based on new aggregated data on initial and seasoned equity offerings (IPOs and SEOs) on the Berlin Stock Exchange before the First World War on a monthly basis and in combination with several other available datasets, we test the hypothesis presented by Hanselaar, René Stulz, and Mathijs (2019. Do Firms Issue More Equity when Markets Become More Liquid?” Journal of Financial Economics 133 (1): 64–82) that market liquidity significantly influences issuance activity on the stock market. For the first time, we can substantiate this hypothesis using historical data. Indeed, the number and value of SEOs and IPOs can be explained quite well with the help of past market liquidity. Keywords: financial history; Germany pre-1913; stock issues; market liquidity JEL Classification: N23; G12

Abstract. In the German Empire, corporations almost always paid a dividend to their shareholders. Dividends have been cut or increased in line with the development of profits. We demonstrate that the target dividend and the average dividend tended to be nearly the same. If the dividend paid deviated from the target, we measure an extraordinarily fast return towards the target. Our analysis of the change in dividends, the payout ratio, and the dynamics of the dividend level provides evidence in favour of the agency theory of dividend policy. Companies with good investment opportunities paid comparatively high dividends. An improvement in shareholder protection weakened this effect. Best practice voting rights at the company level have systematically influenced dividend levels, the relevance of investment opportunities for the dividend policy and the speed of adjustment after a deviation from the target.

Artificial intelligence (AI) methods have become indispensable in science. However, a "one-size-fits-all" approach isn't enough for researchers: For each project, AI programs and models must be adapted and trained, if not programmed. The University of Bonn imparts these skills to its researchers from all disciplines in the "AI for All" seminar series.