In this article, researchers from Spain’s central bank apply text mining techniques to analyze the TCFD (Task Force on Climate-related Financial Disclosures) recommendations on climate-related disclosures of the 12 significant Spanish financial institutions using publicly available corporate reports from 2014 until 2019. To analyze the reports, they used NLP (natural language processing) techniques using NER (named entity recognition) for information extraction.
In the analysis, applying their domain knowledge, first they create a taxonomy of concepts present in disclosures associated with each of the four areas described in the TCFD recommendations. This taxonomy is then linked together by a set of rules in query form of selected concepts. The queries are crafted so that they identify the excerpts most likely to relate to each of the TCFD’s 11 recommended disclosures.
By applying these rules researchers estimate a TCFD compliance index for each of the four main areas for the period 2014-2019 using corporate reports in Spanish. They also describe some challenges in analyzing climate-related disclosures. The index gives an overview of the evolution of the level of climate-related financial disclosures present in the corporate reports of the Spanish banking sector.
The results indicate that the quantity of climate-related disclosures reported by the banking sector is growing each year. The study also suggests that some disclosures are only present in reports other than annual and ESG reports, such as Pillar 3 reports or reports on remuneration of directors.