Helene Winch, Head of Responsible Investing, lifts the bonnet on the AI bandwagon and finds that ESG research is one area of financial markets where AI is already being used.
For information purposes only. The views and opinions expressed here are those of the author at the time of writing and can change; they may not represent the views of Premier Miton and should not be taken as statements of fact, nor should they be relied upon for making investment decisions.
Data trawling
AI is well established and embedded into ESG data production and delivery to investors. Many of the ESG data organisations use AI to trawl the internet for negative and positive global news stories about companies ESG activities that could have a financial impact on companies’ stock prices.
The ESG data sector is growing rapidly with new entrants to this valuable market aiming to undercut the current dominant providers by relying on AI. With thousands of pages of company reports to analyse, AI can be used to evaluate ESG performance and ultimately produce company ESG ratings for investors to review.
If it gets measured, it gets done
Using AI to collect information to evidence sustainability claims and populate a vast number of company level metrics is an attractive proposition, especially one that can cover a vast universe of companies and at a reduced cost. In this respect AI has the potential to help investors comply with ESG reporting requirements such as SDFR (Sustainable Finance Disclosures Regulation).
But this is where the proverbial rubber hits the road with the AI bandwagon. An ESG assessment requires a certain amount of qualitative assessment and the final view is quite unique for each investor – two reasons perhaps that investors complain that third party ESG ratings often contradict on whether a specific company is a good, sustainable company or an evil polluter.
Joe Average?
ESG ratings that use a prescriptive check box approach can favour larger companies who may have a significant reporting and marketing resource to support the availability of ESG data. The AI processes that make estimates and assumptions to fill data gaps can only be based on approximate or average performances across a sector or region. This may be a suitable proxy when applied to a passive index or for reporting across a portfolio that may represent the “average” group of companies, but which company would market themselves as an average performer?
For example, the carbon emissions for a company that does not report or is of smaller market captalisation and is not required to report, is estimated using a local peer. A real time example could be the UK Property REIT sector, as many companies in this sector are outside of the scope of mandatory reporting a number of companies in the sector have the same estimated carbon metrics with no differentiation on good or bad – not helpful when trying to make an active investment decision based on their ESG credentials.
More than machine learning
It will be interesting to see how AI data capture develops. Meanwhile at Premier Miton we like to analyse a wide variety of sources to inform us on ESG management, including meeting companies, company site visits, desk-based research, alongside financial analysis. We feel that this multi-faceted approach brings us a broader view compared to an AI driven ESG assessment.
We use the information that we gather to drive our investment decisions and build our active portfolios. We very much hope that our active portfolios do not represent the market ‘average’ and hence the AI-driven ESG data which is sometimes based on this ‘average’ company performance may not always represent what is held in our active portfolios.
