Understanding how companies are managing the risks and opportunities of climate change is critical for investors, financial institutions and analysts. Corporate earnings calls are a valuable source of information and fill gaps in climate change data. We use transcripts of these calls in Europe and the US over the past two decades to assess how companies are afected by four climate hazards: storms, cold weather, heat waves and wildfires. Our approach involves several steps. First, we develop a classification system (taxonomy) for each climate hazard by reviewing scientific reports. This taxonomy is then expanded by identifying semantically similar words using a Word2Vec model. We then identify sentences in the transcripts that contain these climate-related keywords. Using generative AI techniques, specifically GPT 3.5, we analyse these sentences to gain insights into how individual companies are exposed to climate change risks. We distinguish between negative impacts (risks) and potential benefits (opportunities) for their business activities. We also identified three main channels through which climate risks afect diferent companies: 1) disruptions to the company's supply chain, 2) impacts on the company's demand, and 3) direct damage to the company's assets and operations. Our findings show that exposure to physical climate risk varies widely across sectors in terms of the types of events and the channels through which they afect firms. This innovative dataset has the potential to provide investors and analysts with accurate information on past climate risk exposures, thereby enhancing their understanding of how climate change may impact economic activity and corporate decision-making.
amount of data that needs to be properly analyzed. This task in particular, when applied to the diversified naIn an era of increasing environmental uncertainty, un- ture of large corporations spanning various geographical derstanding the exposure of firms to physical climaterelated financial risks has become imperative. Indeed, the increasing likelihood of acute hazards such as floods, storms, wildfires, heat waves, and cold waves presents a formidable challenge to corporations worldwide. These risks pose big challenges for companies, not just in terms of how they might afect the value of buildings and infrastructure or generate business interruption, but also because they can disrupt supply chains and afect the demand for products. Capturing the full spectrum of physical climate-related financial risks demands a large regions, increases the complexity of assessing their vulnerability to climate-related hazards.
Our objective is to utilize NLP techniques to extract information from earnings call transcripts, thereby discerning firm-level indicators of physical risk exposure.
In particular, the indicators that we extract ofer information on firms’ exposure in 3 diferent dimensions. Firstly, the exposure metrics are hazard-specific; the hazards considered in this paper are the following: wildfires, floods, hurricanes, cold waves, heat waves, and droughts. Secondly, we aim to diferentiate between the adverse efects of these hazards on business activity, identifying whether they pose risks or present opportunities for firms. Thirdly, we seek to delineate the channels through which these hazards exert their influence, distinguishing between direct impacts, such as the destruction of firms’ physical infrastructure, and indirect efects, such as disruptions to supply chains or reductions in consumer demand.
Our paper mainly contributes to the strands of finan
3. Extracting Physical Risk Metrics
3.1. Taxonomy definition
licly listed companies catalogued in the Refinitiv Eikon
database. Our dataset includes companies from 17
sectors, and with headquarters in 34 countries, although we
report a concentration of 68% in North America and 23%
in European countries.
text analysis to measure firm-level exposure to climate
risk. From early applications to more recent
advancements, studies utilizing text analysis have provided
insights into market perceptions, sector vulnerabilities, and
regulatory developments related to climate change. By
leveraging textual data from diverse sources including
news articles ([
Our findings highlight that the efects of climate hazards on firms are not uniform, with variations observed across sectors, time frames, and types of hazards. Moreover, we find a significant portion of firms that exhibit positive exposure to these hazards and a majority of firms indirectly impacted, often through shocks in their end markets or disruptions in their supply chains.
70% with the NHM Boolean model. This strategy enables the exclusion of overly ambiguous keywords that could introduce excessive noise into our exposure metric. Additionally, we drop keywords that do not occur within our documents. This refinement process ends with a taxonomy of 43 keywords, from our initial sample exceeding 80.
Finally, we enrich our taxonomy through the
implementation of a Skip-Gram Word2Vec model [
To extract the aforementioned indicators, we craft unique prompts for each task. For our study, we devise three zero-shot prompts, each with explicit instructions detailing: a) the tasks to be executed, including descriptions of the indicators to be generated, b) the format of the desired output, which is a JSON document with labelled variables, c) the analytical context, encompassing a brief definition of an earnings conference call, the paragraphs subject to analysis, the keyword identifying the hazard type, accompanied by a succinct hazard group description and the NACE 1 prevalent economic sector of the company.
Class: Indirect
To evaluate the model’s precision, we manually annotate a subset of paragraphs for each task and compare human and machine classification. Given that the sample distribution in task 1 predominantly favours True Positives (the Boolean model is applied solely to keywords with 90% and 80% Precision, as delineated in section 3.1), we augment our validated database with out-of-sample Negative paragraphs (without any mention of physical hazards) to balance our dataset, sourcing from the ambiguous text snippets dropped by the Boolean models.
The model’s performance results are in Table 2 for each task. Balanced accuracy approximates 83% for the initial task and 80-76% for the subsequent tasks. F1 scores for the first task equals 77%, 71% for the second task and 70% for task three.
Subsequently, we apply the model to the entire corpus of identified paragraphs, with each paragraph being categorized in terms of exposure, risk versus opportunity, and exposure channel. Only paragraphs verified by GPT in Task 1 were deemed valid from the initial set pinpointed by the Boolean model. These paragraph indicators were then aggregated at both the transcript and firm levels. other hazards, we source events from Wikipedia. The outcomes of this analysis are depicted in Figure 2. The data reveals an increase in the number of exposed firms to hurricanes following the years in which the considered hurricanes struck the US. A similar trend is observed for the other risks. These findings provide preliminary evidence of a consistent correlation between exposed firms and the incidence of major physical events. 4. Analysis An analysis of the distribution of firms exposed to physical hazards reveals distinct patterns. Initially, an asIn this section, we aim to provide an analysis of indicators sessment of sector-specific vulnerability (refer to Figure derived at the firm level. Our initial focus is on evaluating 3) indicates a pronounced disparity in hazard exposure whether the total number of exposed firms in our sample among sectors. Flood-related events constitute the prito a specific hazard tends to be higher in years coinciding mary risk across all sectors, followed by cold waves as with significant events. To accomplish this, we conduct a the secondary hazard. Heatwaves and wildfires present preliminary assessment as follows. After determining the a more varied pattern of exposure. Specifically, the agriindividual exposure of each firm, we track the number cultural sector (NACE A) and water-related industries of exposed firms across diferent hazards over time. Sub- (NACE E) are comparatively more susceptible to heatsequently, we compile a list of major global events. For waves and droughts. In opposition, the construction hurricanes, we reference NOAA data2, selecting Katrina industry faces a heightened risk from wildfires. and Sandy as significant events with a category exceed- Furthermore, it is significant to note that a considing 5. Additionally, we include hurricanes Harvey, Irma, erable fraction of firms are subject to multiple hazards and Maria, which occurred in the same year (2017). For (illustrated in Figure 6), with nearly 30% of all firms categorized under this multi-hazard exposure. 2https://www.nhc.noaa.gov/news/UpdatedCostliest.pdf
I_AARG II_BNNGM ._FANUCM c._LEECDold_wav_TEAERWe flo._TFSRNCOod_hur_TEADRGicNaAneCs__TASPNRHEwSindectIL_TEHOorheat_IJ_TCwave_dL_ERrought _FPROM wildfiI_ADNNMre _ANUHQM ,_TSARR AN T UN0.50 O C
The second analysis we conduct aims to examine the risk and opportunity patterns associated with each haz- is also a proportion of firms that are indirectly afected, ard across various sectors. As depicted in Table 3, a especially through cold waves and floods. The results significant majority of firms present negative exposure of the analysis at the sectoral level are shown in Figure to hazards. This phenomenon is especially marked in the 6. The mining sector stands out because it is directly case of floods and wildfires. Nonetheless, there exists a exposed to all risks, a result somehow expected. The agrisubstantial proportion of firms that demonstrate positive cultural sector tends more often to be exposed directly exposure, especially after cold waves and heat waves. rather than indirectly, while among the most indirectly Looking at the diversification by economic sector could exposed sectors, we find the trade and accommodation provide some intuitions in explaining this tendency in services. our data. Figure 5 portrays the distribution of risk and opportunity across sectors for diferent hazards. It be- 5. Conclusion comes apparent that certain sectors face relatively lower exposure to specific hazards, whereas others might even Our study reveals a multifaceted landscape of climate reap advantages. For example, firms engaged in electric risk exposure among firms. The application of NLP techpower generation (NACE D) are intuitively positioned niques to earnings call transcripts allows the identifito gain from cold waves, as demand for utilities tends to cation of firm-specific, hazard-related exposure metrics rapidly increase during such events. that vary significantly across diferent dimensions. Our
Next, we investigated which sectors are impacted di- findings indicate that while physical climate hazards genrectly through damage to assets and operations, and erally pose risks to business activities, there are instances which indirectly, disruptions to the supply chain or to where they can present opportunities, particularly for the demand. It is clear from Table 3 that firms are more ifrms that benefit from the aftermath of such events, and likely to be indirectly rather than directly exposed to most notably, the highest share of physical risk exposures physical hazards. While there is more than half of the for firms come from an indirect source, being either a ifrms which is indirectly exposed to the hazard, there shock in demand caused by hazard impacts in end mar
cold_wave flood_hurricanes_wind heat_wave_drought wildfire
kets, or disruption of value and supply chains in the firms’ network.
The taxonomy developed has proven efective in discerning the nuanced impacts of climate hazards. This has facilitated the deployment of generative language models, and the consequent granular understanding of the direct and indirect channels through which these hazards afect firms, highlighting the complex interdependencies within supply chains and consumer markets. Our research underscores the importance of considering these multifaceted efects when assessing climate-related ifnancial risks.
Looking forward, the insights garnered from this study could serve as a valuable resource for navigating the evolving landscape of climate risks. Investors and financial institutions could leverage this research to enhance their understanding of how climate change may impact economic activity and corporate decision-making. Ultimately, our work contributes to the broader discourse on sustainable finance, emphasizing the need for innovative approaches to understand and mitigate the financial repercussions of climate change. Additional work is still required to leverage our newly created database and investigate more deeply the intricate impacts of physical climate risk on corporations.
We would like to thank Jakob Jones and Michele Tursi for their support in the construction of the Earnings call database. We would also thank Michele Filannino for his guidance on GPT deployments and perceptive feedback, and Professor Giovanna Bua and Professor Daniel Kapp for their availability to share data and provide valuable insights. We also thank participants at the 2023 International Fintech Research Conference in Naples.