Much of the valuable information that companies communicate to their shareholders about their ESG performance and the social and environmental risks facing the business lie not in the tagged financials nor even in the structured tables and graphs embedded in annual reports, but rather in the paragraphs that flow around the numbers and figures.
For instance, corporate disclosure of climate-related risk, as defined by the US Securities and Exchange Commission’s (SEC) interpretive guidance issued in Feb 2010, can be included in one or more of a number of sections of the 10-K filing — an electronic document that can span more than 100 pages. Climate change poses direct risks to businesses (and their investors) from changes in weather patterns, precipitation and resource availability; extreme weather events; rising sea levels; and changes in disease patterns and ecosystems. Risks derive indirectly from laws, rules and agreements aimed at setting a price on GHG emissions or imposing fines and penalties for emission violations. Compliance with disclosure obligations imposes costs of monitoring and reporting. Potential legal and reputational sanctions may be imposed on companies identified as egregious contributors to climate-induced calamities via their role in the accumulation of GHGs in the atmosphere. New technologies that reduce emissions or provide alternatives to fossil fuel-powered products, for instance, may pose both challenges and opportunities to businesses.
However, without much digging and cross-referencing, it is extremely difficult, time consuming and fundamentally expensive to get the full picture of climate-related risks for just one company, much less an informed survey of climate reporting across companies and reporting periods.
The Climate Risk Disclosure project was developed as an automated system, relying on a set of rules-based algorithms, for finding and describing climate-related disclosures. It digests large amount of narrative disclosure drawn from annual reports filed with the SEC by both domestic and foreign companies. It distils and broadly categorises the text relevant to climate change. As such, it has created, and continually updates, a corpus of corporate statements about climate risk made since 2009 by over 6000 companies.
The question remains: How can we understand, compare and use narrative corporate climate-related disclosures — without too much reading?
Making sense of the wealth of information contained in text is a rapidly evolving science. New digital methods of data visualisation allow us to mine text for relationships, trends and other insights on that even the most informed reader might overlook.
In a recent collaboration with Science-Po’s MediaLab[i], a subset of the CookESG Climate Risk Disclosure database was subjected to a digital mapping exercise staged as the culmination of a high-profile seminar for academics, researchers, developers and designers in Paris, France.
The project looked at climate disclosures contained in 6 years of 10-Ks filed by around 600 Russell 3000 companies in selected industry groups (apparel and textiles, electrical utilities and coal producers, food and agriculture, insurance and oil and gas).
The methodology revealed unique insights at a number of levels: across the entire corpus of disclosures, within individual industry groups and in comparing company to company within an industry.
A combination of natural language processing and network analysis tools exposed that the content of climate disclosure can be usefully grouped into eight clusters related to the physical impacts of climate change, the climate impacts on the markets for oil and gas and electricity, the regulatory environment and also to emissions measurement and disclosure.
Superimposing heat maps on this cluster mapping — i.e. using colour as an indication of where each industry sector is situated within the overall map — shows that industries exhibit somewhat distinct patterns of disclosure. Apparel and textile companies showed sparse disclosure and tended to focus on risk management and economic conditions — in particular, supply chain impacts. Food and agriculture and insurance companies focused squarely on adverse physical climate impacts. Electrical utilities provided the greatest volume of disclosure and focused on the impacts of climate on the markets for gas and heating as well as on the EPA regulations relating to GHG emissions monitoring and reporting, GHG emission permits and renewable portfolio standards. Oil and gas companies are primarily concerned with regulatory risk, EPA regulations and the market drivers of oil and gas prices.
Understanding disclosure in aggregate provides a context for evaluating individual disclosures. So when a company goes further than its peers in recognising certain kinds of risk, this potentially points to an under-reported area of risk and may signal that this company has developed a more sophisticated approach to climate risk analysis. For instance, electrical utilities that place greater weight on physical climate risk stand out. What they say may merit a closer read in understanding whether others in the industry are overlooking something.
The radial depictions of individual company disclosures provide a potentially invaluable heuristic. American Electric Power (AEP), for instance, places relatively greater emphasis on physical climate risk than others in the electric utility industry, stating unequivocally “climate change creates physical and financial risk”. The 10-K discusses the impact of weather on distribution and transmission systems, the risks of extreme weather, and the effect on demand for energy of altered seasonal weather. Alliant Energy’s (LNT) 2014 10-K contains substantial disclosure related to gas and heating market, discussing the impact of weather on residential and commercial sales and the impact of weather on generation, purchase and distribution of electric energy. By contrast, the 2014 10-K of Calpine Corp (CPN) while containing a sizeable volume of disclosure addressing regulatory impacts, contains only a very general reference to the physical impact of climate change.
Conversely, this methodology helps us identify companies that are far less willing than their peers to discuss climate change – potentially indicating less willingness to address the related risks to the company. Consider the volume of disclosure provided in their 2014 10-K reports by Occidental (OXY) and Exxon (XOM), two of the largest oil and gas companies in the world, to CVR Energy (CVI), a much smaller oil and gas company. While CVR covers in considerable detail regulatory developments relevant to the oil and gas industry as well as a range of physical climate risks, Exxon and Occidental devote just a few paragraphs and fail to demonstrate how issues covered are relevant to their businesses.
There are a number of ways in which the radial depictions of individual company disclosures can be further refined. For instance, in representing where the disclosures are situated in the 10-K report: in general disclosures made in Management Discussion and Analysis (MD&A) tend to be more company-specific and informative than disclosures made in the Risk Factors section. Consider the disclosures of insurers Alleghany Corp (Y), Loews Corp (L) and Travellers Companies (TRV): much of their climate-related disclosure is made in the MD&A section. By contrast, Universal Insurance Holdings (UVE) makes very little disclosure made under MD&A of its 2014 10-K, yet according to the radial diagram this filing to contains more disclosure.
The results of this exercise are presented in a series of refined graphics that serve as an example of how data-visualisation strategies can be used by investors and other advocates to navigate and critically analyse corporate climate risk disclosures and, potentially, other domains of sustainability reporting. Combining text visualisation with more sophisticated machine learning algorithms could help to scope out systemic risks in corporate risk management approaches to climate change and provide investors with more tangible methods for incorporating climate risk into their investment decisions.
[i] The Médialab is a research laboratory within Sciences-Po, the prestigious Paris-based university of social sciences. It develops and leverages digital methods to better understand complex issues in the social sciences including, amongst others, how humans are addressing and preparing for the impacts of climate change.