Ping An uses artificial intelligence to detect greenwashing
AI-driven transparency indicators complement ESG ratings and analytic tools in market
This article has been reprinted with permission from Ping An.
(Hong Kong, Shanghai, 8 December 2020)
Natural Language Processing (NLP) technology, used to analyze language from company disclosures, can help to detect potential “greenwashing” by high emission companies, according to the latest report from the Ping An Digital Economic Research Center (PADERC), a member of Ping An Insurance (Group) Company of China, Ltd. (HKEx:2318; SSE:601318), and the Brevan Howard Centre for Financial Analysis at Imperial College London, the world’s leading climate finance research center.
The report, “Climate Disclosures and Financial Performance”, also found that artificial intelligence (AI)-based climate disclosure indicators perform better than some existing ESG ratings at differentiating green companies from other high emission companies – making these tools a valuable complement to ESG ratings for investment analysis.
The study found that firms with better disclosure of financial impact metrics tend to have higher valuations, lower leverage and lower cost of capital, after controlling for carbon emissions and other firm characteristics. Large cap firms that follow the TCFD recommendations tend to have higher valuations, but small and medium cap firms engaging in climate disclosures may still offer considerable opportunities for appreciation. A range of existing ESG ratings, however, appear to have a weak or inconclusive bearing on valuations.
“AI-based indicators offer a valuable addition to the asset manager’s toolkit to enhance and refine their investment screening process, with more objective information on the climate risk exposure of firms,” said Chenxi Yu, Deputy Director of Ping An Digital Economic Research Center. “They can also detect potential greenwashing that may be at play in particular companies.”
The researchers developed a series of AI-based indicators related to climate risks and financial impacts, drawing from the guidelines of Task Force on Climate-related Financial Disclosures (TCFD) for relevant words and expressions. The NLP techniques automatically assessed the coverage of the indicators in the climate risk disclosure reports of US and Chinese firms in the S&P 500 and the CSI 300.
Under-reporting of climate metrics by high emission firms may be erroneously rewarded
The study found that AI-driven indicators perform better than some traditional ESG ratings to detect corporate “greenwashing” – providing misleading or incomplete information to give the impression that a company is more environmentally responsible than it actually is.
The indicators were better than some ESG ratings in differentiating between so-called “brown” firms – those in high emission industries, such as mining, transportation, and infrastructure -- and lower emission firms.
The indicators found patterns in the disclosure among brown firms, including:
- Under-reporting of the capital and financial impacts of climate risks, such as the impact of stranded assets and liabilities for oil and coal companies
- Limited disclosure of Scope 3 emissions (all indirect emissions, except for emissions generated by purchased energy, that occur in the value chain of the company, including upstream and downstream emissions)
Furthermore, the study found that some ESG ratings penalize companies for climate risk disclosure, which may be encouraging selective non-disclosure instead of transparency from other companies.
Providing companies, asset managers and investors a meaningful tool
The study shows the potential of AI-driven climate risk disclosure indicators as an effective tool to analyze the impact of climate risk on business value, such as:
- Helping asset managers structure meaningful decarbonization strategies – differentiating companies that may play a crucial role in transitioning to a low-carbon economy
- Helping investors inform and support portfolio tilts – allowing investors to articulate their view on the pace at which information on climate risk exposures will be dynamically incorporated in market valuations
- Helping investors better understand climate risk premiums beyond emissions – although a carbon risk premium has been documented for high emission firms, the picture is more refined once climate risk disclosure indicators are taken into account
- Helping investors capitalize on the increase of climate awareness – AI-driven indicators can help to inform investment policies aimed at making the most of forward-looking metrics of climate change during the transition to a low-carbon economy
- Helping investors better articulate their view on climate risk-return tradeoffs – investors could use AIbased indicators to specify competing constraints in their portfolio optimization engines, to identify portfolios achieving their desired risk-return tradeoffs
- Helping companies understand how climate risk disclosures can add shareholder value – climate disclosures can reduce the cost of capital
Three streams of research
This report, the second on climate risk and financial innovation, combines Ping An’s expertise in financial technology with the Brevan Howard Centre’s academic research on investment risk. The research partnership focuses on developing methodologies using artificial intelligence (AI) and big data to assess the risks for investment assets from climate change and other ESG-related factors.
PADERC is a professional institution specializing in macroeconomics and policy research, using big data and artificial intelligence to provide insights on macroeconomic trends, including developments in ESG disclosure. The Brevan Howard Centre connects the financial economics expertise of Imperial College London Business School with other disciplines, including engineering and computational finance.
Its three streams of research include financial stability and financial regulation, comparative financial systems and designing new financial structures, and financing development, environment protection and medicines.
Article published in Risk Consulting 2/2021