Using AI to Manage ESG & Investment Risk
Investment risk is taking on a new meaning and how AI can help us understand it
The definition of investment risk has evolved significantly this decade, with the growing inclusion of environmental, social and governance (ESG) practices in investment decision making. This is a concept known as materiality.
When investing into a basket of securities on behalf of its client, an asset manager should seek to keep its exposure to materiality to a minimum, by avoiding products linked to issuers (and possible investees) heavily weighted with ESG risk.
For example, one may look to reduce the physical climate risk in their portfolio by steering clear of assets which would be imperilled by rising sea levels or temperature. Climate transition risk can be mitigated through the avoidance of assets which may be disfavoured by anticipated policy changes, such as coal producers. Investors may also keep clear of companies which are prone to controversy or operate in controversial sectors, to avoid controversy risk where a company is disfavoured by its customers for ethical reasons, affecting its financial performance, or legal risk, the cost of legal settlements against a company stemming from ESG misdeeds.
A metric used to gauge the ESG credentials of a specific firm is an ESG ‘score.’ The industry standard for calculating a company’s ESG score is the MSCI ESG score. MSCI scores are calculated using a rules-based methodology that rates companies on a scale of AAA to CCC, by analysing thousands of data points across roughly 33 ESG issues, typically using artificial intelligence. These activities are typically carried out by ESG ratings providers and are utilised by asset managers. Some asset managers develop their own proprietary ESG ratings.
However, an over reliance on simplistic ESG scores can be a perilous strategy, especially when using them to build investment portfolios. ESG scores have tendencies to focus on how well companies manage their internal processes rather than the real-world impact of their products and services. There are also disparities and inconsistencies between different ESG ratings providers, making it hard to gauge the true ESG credentials of a particular company.
Currently, there are two solutions to this problem that, in GreySpark’s view, capital market firms should be aware of.
Rather than focusing on the ESG score of a company, and being subject to the challenges outlined above, an investor may be better off quantifying ESG risks through calculating the financial impact of ESG on the risk/return profile of a company, or ‘ESG in Dollars and Cents.’ Typically, such a figure is derived using proprietary AI, natural language processing (NLP) and algorithms.
One such example of a company providing ESG-based Dollar risk/return profiles is Yieldrive. The fintech company has 500 billion data points, using AI to scrape information across regulatory filings, annual and ESG reports and credible media sources to calculate the ESG financial impact of a specified firm.
Another alternative is using sentiment analysis to drive ESG decision-making.
Sentiment analysis is a much bigger driver of investment decision making in financial markets compared to 20 years ago, with markets becoming more sensitive to news developments. This is otherwise known as information inelasticity, where it takes a smaller amount of Dollar-valued information (i.e., a piece of breaking news) to move a market by a given amount (in terms of price).
At the same time, a structural shift is taking place, with institutional asset owners shifting from active investing to passive investing strategies.
Around 30 years ago, 95 per cent of institutional assets were actively managed, meaning that a fund manager was picking and choosing investments on behalf of an institutional investor. Today, this figure is less than 45 per cent, meaning that 55 per cent of institutional assets are sidelined and not participating in the price discovery mechanism seen in active management, with passive investors now grabbing a larger slice of the pie.
In the main, this growth in passive investment has been fuelled by retail markets, which is markedly more affordable to the average retail investor. This helps to explain the greater sensitivity to market news outlined above, with retail investors tending to harbour more emotion and reactiveness when it comes to making investment decisions in the wake of market developments.
As such, with markets becoming more retail-oriented, conducting sentiment analysis across novel news and social media platforms can provide credible forms of alpha in 2024.
However, to be conducted effectively, a more dynamic type of sentiment analysis, in comparison to traditional methods, is required.
Rule-based sentiment analysis is the traditional method of sentiment analysis. Around since the 2000s, this method typically uses NLP to identify whether a piece of text is negative, positive, or neutral. In contrast, automated sentiment analysis relies on machine learning (ML) techniques. In this case, a ML algorithm is trained to classify sentiment based on both the words and their order. The success of this approach depends on the quality of the training data set and the algorithm.
Specifically, automated sentiment analysis can gauge the interest across news and social media platforms, by ingesting and categorising data inputs and providing investment signals or actions based off of findings. Current portfolios can then be benchmarked against the findings.
In an ESG sense, this kind of sentiment analysis can gauge public opinion around ESG issues across social media channels. This can help financial institutions identify emerging concerns, track the impact of their ESG initiatives, and adjust their strategies and communications accordingly.
As is the case with several other variations of AI, a key concern remains model biases. Sentiment analysis algorithms are typically trained on existing data, which may contain inherent biases and reflect social misconceptions. In addition, gen AI models recognise patterns from the past and applies them to the future even though past performance is not always an indicator of future trends. As a result, human oversight is still required to a large degree.
As such, when seeking to really understand the ESG credentials of a firm, investors may look beyond conventional ratings, and utilise AI technology to quantify risk/return and sentiment. Given the increasing influence of retail investors in capital markets in 2024, it may be more prudent to look beyond traditional ESG ratings and towards AI-based frameworks to generate alpha and gauge the true ESG credibility of a firm.
For further information, please do not hesitate to contact us at london@greyspark.com with any questions or comments you may have. We are always happy to elaborate on the wider implications of these headlines from our unique capital markets consultative perspective.