Design and Implementation of a Historical German Firm-level Financial Database

Abstract

Broad, long-term financial, and economic datasets are scarce resources, particularly in the European context. In this paper, we present an approach for an extensible data model that is adaptable to future changes in technologies and sources. This model may constitute a basis for digitized and structured long-term historical datasets for different jurisdictions and periods. The data model covers the specific peculiarities of historical financial and economic data and is flexible enough to reach out for data of different types (quantitative as well as qualitative) from different historical sources, hence, achieving extensibility. Furthermore, we outline a relational implementation of this approach based on historical German firm and stock market data from 1920 to 1932.

Entity Matching with Similarity Encoding: A Supervised Learning Recommendation Framework for Linking (Big) Data

“…This use case refers to the matching of entities from repeated annual cross-sections of firm data extracted via OCR from historical German yearbooks. Gram et al (2022) describe the data extraction process, detail the underlying relational data model, and provide summary statistics for a subset of the fields extracted for the period 1920 -1932. The database is implemented according to the FAIR principles and will be made fully available the the public in the upcoming years.…”

“…Finally, we present an application of our matching framework in a domain with dirty firm-level financial data that we extracted from historical archives by using Optical Character Recognition (OCR) software (Kamlah et al, 2022). The data represent German firms operating in the period from 1910 to 1919 with non-harmonized and non-standardized attributes extracted from the "Handbuch der deutschen Aktiengesellschaften" (see also Gram et al, 2022). In a 5-fold cross-validation with 30% train and 70% test random sample splits, our framework achieves an average 99.36 F-score in the test sub-sample.…”