Unmasking Climate Change Impacts: Traversing Storms, Cold, Heat and Fire in Corporate Earnings Calls through a Hybrid Taxonomy and GPT-based Methodology

What do earnings calls reveal about how storms, heat and wildfires hit companies?

Michele Cimino, Annalisa Molino, Maria Paola Priola, Lorenzo Prosperi and Lea Zicchino extract firm-level physical risk signals from what managers tell analysts in their paper "Unmasking Climate Change Impacts: Traversing Storms, Cold, Heat and Fire in Corporate Earnings Calls through a Hybrid Taxonomy and GPT-based Methodology".

They analyse 101,069 English-language earnings call transcripts from 3,152 listed firms in 34 countries between 2002 and 2022, mostly headquartered in North America (68%) and Europe (23%).

A keyword taxonomy built from IPCC and NOAA glossaries flags hazard-related passages. GPT-3.5 then classifies each passage by hazard, risk or opportunity, and direct or indirect impact.

Their main conclusions include:

• The final 68-keyword taxonomy captures hazard discussions in 12,309 transcripts from 1,886 firms, and the number of exposed firms rises after major hurricane years.
• Hazards are mostly a threat, as 66% of exposed firms report negative effects, rising to 71% for storms and floods, which are the leading hazard in every sector, ahead of cold waves.
• 18% of exposed firms report gains overall, 27% for heatwaves and droughts and 24% for cold waves, with power generators benefiting as utility demand spikes during cold snaps.
• 51% of exposed firms are affected through disrupted supply chains or shifts in end-market demand, against 22% through direct damage to their own assets and operations.
• Mining is directly exposed to every hazard, agriculture and water utilities are most sensitive to heatwaves and droughts, construction to wildfires, while trade and accommodation are hit mostly indirectly.
• Nearly 30% of firms are exposed to several hazards at once, a reminder that climate risk rarely materialises through a single peril even within one single business.

The paper extends the earnings call measures of Sautner and co-authors, featured earlier in this series, from climate attention to transmission channels.

Earnings calls add evidence on how hazards actually reach revenues and supply chains, which matters as IFRS S2 and ESRS E1 require disclosing physical risks and their effects across the value chain.

Linking these indicators to returns, credit spreads or realised losses is however not obvious, and testing newer models would show whether the signals can inform pricing and investment decisions.