Using a large language model (LLM) method, researchers extracted approximately 2.9 million ESG observations from 9,000 corporate reports. The results show that reporting has become significantly more comprehensive over the past ten years, but the trends in actual ESG indicators vary widely. Many social indicators have largely stagnated, although progress is evident, particularly in the area of gender equality. In the environmental sector, significant improvements are visible for some indicators, while for others there has been little or no progress. The trend in Scope 3 emissions is particularly illustrative: The sharp increase in reported total emissions does not necessarily mean that performance has deteriorated accordingly. It is largely attributable to the fact that companies now track and disclose more emission categories. The study thus also demonstrates how strongly perceived sustainability performance depends on the availability of the underlying data.
For companies, the framework provides a basis for systematically analyzing their own reporting and sustainability performance over time and in comparison with industry competitors. This makes it easier to identify both transparency gaps and actual areas where action is needed. For policymakers, the study provides a comprehensive starting point for monitoring which areas have available corporate data, where gaps still exist, and whether new reporting requirements, such as the ESRS, will actually lead to a better information base in the future. The methodology can also be regularly updated and applied to additional companies, countries, and regulatory requirements. For society, the project makes information accessible that was previously scattered across thousands of extensive PDF reports or available only through commercial data providers for a fee. The open dataset and freely available code enable, for example, researchers, the media, NGOs, investors, and employees to systematically compare corporate data and track progress toward self-imposed goals. In this sense, the study creates a public good that improves the verifiability of corporate sustainability activities.
With the study’s findings and the open dataset, the team around WiSo Professor Maximilian Müller is making an important contribution to making sustainability information more transparent, comparable, and accessible to the general public.