Firm-level climate change exposure

The report develops a method to quantify firm-level exposure to climate change by analysing earnings call transcripts from over 10,000 firms across 34 countries between 2002 and 2020. It captures market participants’ attention to climate-related opportunities, physical risks, and regulatory shocks using climate-relevant bigrams. The measures reflect “soft” information exchanged between analysts and managers and complement existing indicators based on emissions or physical climate events. The method identifies material variation in exposure across and within industries and over time.

Data

The analysis uses English-language quarterly earnings call transcripts obtained from Refinitiv Eikon. The final sample contains 86,152 firm-year observations after excluding countries with limited coverage. Additional data sources include S&P Global Trucost emissions data, OptionMetrics for option-implied variables, Compustat for firm financials, and external datasets on climate news coverage. Green job creation data are taken from Burning Glass, and green patenting data rely on patent authority records, using priority dates to determine timing.

Quantifying firm-level exposure to climate change

The authors adapt a machine-learning keyword discovery algorithm to identify climate-related bigrams. A short initial list of unambiguous climate bigrams seeds the search, and the algorithm expands this set by identifying statistically associated bigrams within relevant sentences. Four exposure measures are constructed: overall climate exposure, opportunities, physical shocks, and regulatory shocks. Exposure is defined as the share of the transcript containing these bigrams, scaled by transcript size.

Validating exposure measures

Several validation tests confirm the method’s effectiveness. A structured human audit of over 2,000 transcript fragments shows higher predicted accuracy in identifying climate-related content for firms with higher exposure scores. Keyword robustness checks indicate that removing individual initial bigrams does not materially change results. Industry patterns are plausible: utilities show the highest exposure, reflecting both regulatory scrutiny and opportunities in renewable energy; heavy construction, transportation equipment, and electric equipment also rank highly.

Within-industry variation is sizeable. For instance, TotalEnergies exhibits more than seven times the opportunity exposure of ExxonMobil despite comparable regulatory exposure, reflecting market perceptions of their differing strategies towards renewable energy. The exposure measures correlate positively with emissions and public climate attention indices, particularly through opportunity and regulatory dimensions.

Variance decomposition of exposure measures

Between 70% and 96% of the variation in exposure occurs at the firm level, indicating substantial heterogeneity among firms within the same sector. Only about half of this variation appears persistent, suggesting that exposure evolves over time. Measurement error accounts for 5–10% of variation, increasing slightly at the firm level but remaining modest overall.

Applications of exposure measures

The measures predict real economic behaviour. A one-standard-deviation increase in climate exposure is associated with a 109% rise in green job postings in disruptive technologies in the following year. Similarly, green patenting increases by 72% under the same change. These effects are driven by opportunity and regulatory exposure; high-exposure firms reduce non-green hiring and innovation.

Climate exposure also relates to financial market outcomes. Physical exposure is linked to lower leverage after the Paris Agreement. In credit markets, overall exposure is negatively associated with CDS spreads. In equity markets, exposure is priced, and a climate exposure factor earns a positive risk premium in conditional asset pricing tests. Option-implied risk measures also vary with exposure, indicating that markets incorporate climate considerations into risk assessment.