Empirically assessing corporate adaptation and resilience disclosure using AI

The extent to which firms are adapting and building resilience to environmental change is crucial information for financial institutions, regulators and governments but it has so far received considerably less attention than their mitigation actions. One of the principal barriers to adaptation and to the scaling up of adaptation financing is the dearth of reliable information. While the exposure of companies’ assets to physical climate risk due to environmental change can be calculated using models, additional information is needed to evaluate their vulnerability to physical climate change, how well they are adapting, and their broader alignment with societal adaptation and resilience (A&R) goals.

The paper builds on established sustainability disclosure frameworks to develop an A&R disclosure framework that the authors combine with the latest advances in large language models (LLM) to assess the sustainability reports of S&P 500 companies. The development of a common framework is expected to enhance companies’ A&R reporting while making public information more comparable and usable for investors and regulators.

Key points for decision-makers

• The study provides the first empirical analysis of the extent to which company disclosures include information on A&R.
• The authors develop a framework of 91 binary indicators and use it to assess the latest sustainability reports of the S&P 500 companies using an LLM to generate 42,030 datapoints for analysis.
• By using the disclosures of S&P 500 companies, the paper provides a snapshot of the disclosures of firms with a combined valuation of US$43 trillion regarding their adaptation to climate change and resilience building.
• On average, S&P 500 companies report against just 20% of the 91 indicators in the framework, with the least-reporting company only providing information for one of the 91 indicators and the most-reporting company providing information for 50 indicators.
• The most disclosed indicators include areas such as board oversight, social and environmental safeguards, and engagement with value chains. The least disclosed indicators relate to companies’ metrics and targets, nature-related financial risks, capital expenditure linked to adaptation or natural capital, and the precise mechanics of their risk assessments.
• Overall, corporate A&R disclosure is found to be deficient, particularly around risks, metrics and targets, underlining the need to consider other data sources when assessing firm-level risks and contributions to societal A&R goals. The methodology developed can be applied to other forms of corporate disclosure (e.g. US filings or other financial reporting), which may be deemed more trustworthy than companies’ sustainability reports.