IFB researchers, Professor Ben Groom and Dr Chunying Guo of the University of Exeter, will be presenting IFB outputs at the World Biodiversity Forum in June. The title of their talk is When stakeholders ask: Measuring biodiversity, climate, and natural disaster risks and opportunities from earnings conference calls.
Abstract
Accurate, forward-looking measurement of biodiversity-related risks and opportunities at the firm level is essential if investors, regulators and standard setters (e.g. TNFD/ISSB) are to integrate nature into financial decision-making. Yet existing measures—typically based on news or 10-K keyword counts—are noisy and often overstate how many firms actually discuss biodiversity risk, as it is hard to distinguish genuine biodiversity discussions from generic, or metaphorical, uses of terms.
We therefore contract earnings conference call (ECC)–based indicators using an expert-labelled sample of ECC excerpts combined with ChatGPT-assisted annotation and fine-tune BERT models, and examines whether these disclosures are financially material. We assemble a global panel of ECC transcripts for listed firms from 2001–2024 and split each call into short excerpts, structured around the prepared remarks, analyst Q&A exchanges, and closing statements. Environmental and biodiversity specialists manually label a high-quality sample of excerpts into several categories (biodiversity, climate risk and opportunity, and natural disasters), explicitly flagging cases where dictionary methods misclassify language (e.g. “technology ecosystem”) as biodiversity-related. We then use this labelled set with ChatGPT-assisted annotation and fine-tune transformer models (BERT) to estimate the independent probabilities of belonging to each category. Aggregating these probabilities yields firm-year, LLM-based indicators of nature-related disclosure. In parallel, we construct a keyword-based baseline for biodiversity and climate from ECCs, which we currently use for the main empirical tests and to benchmark the LLM measures.
Using the keyword indicators, we document a strong upward trend in biodiversity disclosure in ECCs from 2001 to 2024, starting from very low levels and with a sharp jump around 2020–2021. Comparing with climate change, we find that climate disclosure starts from a much higher and more stable baseline and also rises around the Covid period, so that biodiversity remains at lower absolute levels but grows more rapidly over time. Also, linking biodiversity disclosure to firm characteristics and market outcomes, we find that higher biodiversity exposure is positively associated with social ESG scores and negatively associated with stakeholder-engagement scores, while largely unrelated to current profitability. Firms that mention biodiversity exhibit slightly lower subsequent excess returns and higher market beta, suggesting that markets currently interpret biodiversity primarily as a source of systematic risk rather than a short-term opportunity.
Taken together, the results demonstrate that ECC-based biodiversity indicators—starting from a keyword baseline and moving towards LLM-based measures—offer a promising, scalable tool for bringing nature into financial analysis.
The full presentation will be available after the Forum.