After the COVID-19 pandemic, stakeholders across all fields are paying more and more attention to environmental, social, and governance issues, collectively known as ESG. Recent years have witnessed major uptake in investments into this sector – more and more ESG financial products such as bonds, loans, and even mortgage products are being introduced by financial institutions. However, there are still underlying challenges when it comes to collecting data for ESG ratings, and AI technology is emerging as a promising solution.
What are ESG financial products?
Sustainable investing, as the World Economic Forum defines, is a range of activities that invest in green projects or companies that demonstrate social values [1]. Sustainable financial products or ESG financial products, in turn, are financial institutions' offerings, such as sustainable funds, green bonds and social bonds, that promote sustainable investing. Despite being relatively new compared to other traditional financial products, ESG-related financial offerings have enjoyed significant growth over recent years. In fact, over US$3.2 trillion has been invested in sustainable projects in 2021, doubling that of 2019 [2]. Moreover, sustainable lending activity grew to US$322 billion in 2021 from a mere US$6 billion in 2016 [3].
Given its rising popularity, sustainability has been high on financial institutions' agendas. Indeed, according to Kearney, 29 out of 30 largest financial service providers cite sustainability as a strategic focus [4]. Driving forces include:
- Investors' growing interest in ESG: the 2021 Natixis Global Survey of Individual Investors finds that globally, 21% of the respondents have invested in ESG, while twice as many investors (49%) show interest in such investment. 35% of the respondents require their advisors to exclude companies with conflicting values [5].
- Higher return from sustainable investing: A study by Kearney shows that companies with the highest ESG ratings outperform those with the lowest by 40% [4]. It is because of two reasons. Firstly, consumers are more attracted by companies who share their values – with 75% of American consumers citing it as a reason for choosing a brand. Secondly, ESG companies have lower costs as a result of reduced raw materials and energy consumption. McKinsey estimated ESG initiatives to have an impact on operating profit by as much as 60% [6].
- Lower risks: ESG companies enjoy lower risks than their peers as they have a lower chance of encountering regulation breaches, reputational damages, and potential lawsuits. Indeed, studies estimated that risks from sustainability issues, such as rising costs and supply chain disruption, can damage as much as 70% of earnings [4]. As a result, investors with lower risk thresholds prefer ESG investment. According to a study by the Jönköping International Business School in Sweden, ESG bonds appeal to private and institutional investors due to their lower risk, even with lower interest rates. [7]
ESG Data – still a gray area
Despite its soaring popularity, there are certain concerns involving ESG financial products. One of the most prominent challenges is the lack of reporting standards in some markets, making the process of collecting and analyzing ESG data confusing for financial institutions. For example, in the US, while companies may voluntarily report according to International Capital Market Association (ICMA) standards or offer third-party compliance reporting, there are no standard principles set out by the U.S. Securities and Exchange Commission. There are hundreds of third-party ESG rating agencies on the market; each has its own criteria for assessing ESG performance. Telsa is a good example of such disparity in ESG rating, as they have received an A rating from MSCI, a B- rating from S&P Global, and a high-risk rating from Sustainalytics.
Such data inconsistency presents a major setback for financial institutions and individual investors to adopt sustainable investing. Indeed, a BlackRock survey found that 53% of surveyed investors cited "poor quality or availability of ESG data and analytics," and 33% cited "poor quality of sustainability investment reporting" as the two biggest barriers to adopting sustainable investing [8]. Worse, the inconsistency in ESG ratings has even led to more serious problems, such as greenwashing – the act of providing misleading information to deceive the public into thinking one's products are environmentally friendly. A study by two professors from the University of Northern Iowa and the University of South Carolina found that managers are more likely to publicly emphasize ESG when they do not meet their earnings expectations set by analysts. On the contrary, they speak less of ESG when outperforming [9]. It means that fund managers focusing on publicly outspoken ESG companies may invest in less green and financially underperforming companies than they intended. Indeed, intentionally or not, the world's 20 largest ESG funds have been found to hold investments in fossil-fuel companies, each investing in 17 companies on average [10].
AI Technology to the rescue
The good news is that artificial intelligence (AI) tools are now available that can collect and analyze more data on ESG risks and opportunities than ever before. These tools enhance data quality, effectively analyze it, and generate new opportunities.
One AI application in ESG ratings is the use of Natural Processing Language (NLP) in ESG data identification, collection, and analysis. NLP is a branch of AI that can be trained to understand texts and spoken words in the same way as humans. This way, NPL can be utilized to analyze unstructured data such as ESG reports, news articles, or even corporate spokespersons' speeches and convey hidden messages based on their tones and expressions. Not only NPL offers a capability that previously could only be carried out by human inspectors, but it also provides a more cost-effective approach with higher accuracy. For instance, a study by IFC has tested NPL on 11,000 text documents and managed to identify more than 45 million ESG risk terms, with an accuracy rate of over 85% [11].
Another AI application in ESG ratings is carbon emission prediction. Although a company's scope 1 emission (direct emission) can be measured and monitored, scope 2 (indirect emission from the generation of energy consumed) and scope 3 (indirect emission from other activities in the company's value chain) are more challenging to capture. Since scope 2 and 3 influencing factors are beyond a company's control, data varies based on the consumption the company makes on these factors. As data varies, the estimation of a company’s total emission varies, which can be overestimated by as much as 200% [12]. AI opens a new opportunity for emission prediction by studying a company's multiple factors to provide a more accurate estimation. For instance, a study by Harvard Business School has used AI to study 15 widely available factors provided by public companies, such as total assets, sales, and profit margin to estimate a company's emission. The result shows that machine learning models applied in this study managed to improve prediction accuracy over the traditional regression models [13].
Interested in learning more about AI applications in BFSI? Check out this article
Sources
[1] World Economic Forum. What is sustainable finance and how it is changing the world
[2] PwC. Greening your financial products
[3] PRI. Sustainability-linked loans: A strong ESG commitment or a vehicle for greenwashing?
[4]. Kearney. European banks can—and should—do more to lead on ESG issues
[5] Natixis. Values alignment is only the tip of the iceberg for ESG
[6] McKinsey & Company. Five ways that ESG creates value
[7] Eurasian Econ Rev. Drivers of green bond issuance and new evidence on the “greenium”
[8] Deloitte. Globally Consistent ESG Reporting
[9] Harvard Business Review. An Inconvenient Truth About ESG Investing
[10] The Economist. Sustainable finance is rife with greenwash. Time for more disclosure
[12] Worldquant. Using AI to Tackle the ESG Data Challenge
[13] Harvard Business School. Machine Learning Models for Prediction of Scope 3 Carbon Emissions
