ESG Investing: Harness the Power of AI
Updated: May 14
GaiaLens, Miles Clayton (Agility PR)
The ESG investment industry’s “one-size fits-all ESG” is unlikely to achieve asset owners’ specific ESG objectives while it could hinder their returns. ESG analysis should not be left alone to the humans when powerful forms of Artificial Intelligence (AI) now exist to better address asset owners objectives. This is a crucial issue as ESG objectives have shot up the agenda for asset owners, posing incredibly difficult questions for the entire ESG food chain that finds itself mainly unprepared to manage these newly prioritized fiduciary requirements.
AI, and machine learning specifically, can tap the vast data resources of news flow, CSR reporting, and financial datasets and use this to scale bespoke ESG risk analysis for an asset owner, and direct humans to engage with companies on the most important topics to achieve a plan’s objectives. None of these things are possible at scale with traditional set ups.
Up until now, investment managers have answered the demand for ESG strategies with cost sensitive bolt-togethers, or homogenous definitions of ESG which now present a dangerous disconnect between what an investment board wants and what it will get. This could present one of the biggest risks to institutional asset owners over the next decade, as their ESG and return objectives remain homogenized for the convenience of the ESG investment industry.
Up until now, investment managers have answered the demand for ESG strategies with cost sensitive bolt-togethers, or homogenous definitions of ESG which now present a dangerous disconnect between what an investment board wants and what it will get.
ESG Ground Truths?
One of the problems with setting up an ESG index or rating system today is, in machine learning parlance, the universal ‘ground truths' are not yet set firm. Institutional asset owners, pension consultants, asset managers and pension trustees often have different views of what socially or environmentally-responsible investing is and therefore what an ESG index should measure. Only segregated accounts, of the sort reserved for large institutions, allow that sort of purity of definition. However, artificial intelligence (AI) can help us establish an objective truth faster and more consistently once those ground truths are agreed.
Investors are already discovering that the components of ESG may be in conflict with each other and not the holy trinity originally envisaged. Take Cambodia's recent appeal for foreign investment to help develop its new oil discovery. My ‘E' antenna says, ‘avoid this black carbon-heavy fossil fuel'. But my ‘S' conscience has its heartstrings tugged by the plans of one of the world's poorest countries to use oil revenues to develop health and education systems. As for ‘G', well, Cambodia has come a long way since the despotic years of rule by Pol Pot and the Khmer Rouge.
What does AI bring to the world of ESG rating?
How then can AI algorithms do a better job in spotting and objectively analyzing ESG risk? First, given the poor availability of ESG data (and the danger of misreporting), AI offers a powerful approach to data acquisition and validation. It's capable of scouring thought leaders' comments on social media, influencers such as Amnesty International's red-flags, as well as news flow on events and sentiments. You can therefore eliminate a reliance on either what a company says about itself or what its management tells a human analyst.
Second, ratings are often criticized for inconsistencies over time, sector, country and more. Machines are inherently objective and can be scaled across vast investment universes. This makes AI-led ESG risk assessment a compelling solution in principle.
Algorithms will, in the future, be capable of being run over all the available structured and unstructured information linked to tens of thousands of listed businesses, tirelessly uncovering developments worthy of analyzing and scoring, all in a consistent and objective manner. While there are huge challenges to developing AI in this area, the current alternative (human analysts) are expensive, inconsistent, and necessarily subjective.
While there are huge challenges to developing AI in this area, the current alternative (human analysts) are expensive, inconsistent, and necessarily subjective.
ESG Risk or Responsibility?
Many of the largest pension funds in the US are beginning to affirm that a specified proportion of their investments must be net-zero within a certain timeframe. They have established a roadmap for their investment managers whereby that percentage must increase steadily year on year.
So not only must the investor's investment managers meet that percentage target, but they would also be wise to increase weightings in companies that have set fairly aggressive net zero emissions targets. However, a focus on Green House Gas (GHG) emissions may not extend to covering other legitimate ESG targets such as over-extraction of water in manufacturing or microplastics pollution.
However, if as an investment manager you go beyond these highly-focused remits into other ESG-related targets that are not a priority for the asset owner and that are perceived to limit investment returns, you risk losing your mandate. It's a balancing act for asset managers - you cannot afford to make too many changes too quickly.
Asset managers must move at the pace of the investors, working with their changing ESG priorities. However, it's critical to be able to tailor ESG measurement systems to investors' bespoke demands and focus areas.
However, it's critical to be able to tailor ESG measurement systems to investors' bespoke demands and focus areas.
What this means is that, over the next decade or so, there will be less and less capital moving to companies that don't change their behavior in line with the Paris Accord and other ESG targets. The key will be to ascertain which risky areas to divest from this year, and which need to be tackled next year and so on. A divestment road maps might need to be built by investment managers covering the next 10 years or so, and this must be signed off by the investors.
Saints or Sinners: All relative?
ESG risk analysis will need to consider also whether it's about the level of GHG emissions that a company is putting out today or the degree of reduction intended or achieved. One also has to consider how material emissions are to the business a company operates in.
For some companies like BP or Shell, the scale of the task is clearly massive. So, arguably they should be rewarded for meeting or exceeding their emission reduction targets and diverting resource into clean energy projects successfully. These are arguably the companies whose efforts to go green will yield the most positive results in terms of global GHG emission reductions because their impacts are the most material. But are GHG emissions an absolute sin or a relative one?
Investors will need to decide therefore if they are puritans or pragmatists. In other words, do you decarbonize your portfolio of investments aggressively, come what may, potentially leaving some of the worst offenders with no direct incentive to complete their emission reduction programs?
Or, as a pragmatist do you instead accept that some businesses are materially worse than others and engage with them to support their transition for the good of all? Or, thornier still, is whether you levy the same punishment on an emerging market (EMs) energy company as a European one, given that tighter regulations will likely take longer to be enforced in EMs.
AI: The perfect solution for the ESG Conundrum?
AI, a catch all for machine learning and associated techniques, is an inevitable future for ESG risk and investment until global standards of reporting can be agreed. If it were highly challenging for a team of human investment analysts to objectively digest the financial reporting and salient news flow on thousands of listed equities, adding ESG’s terabytes of NGO data sets, CSR reports and more would make it relatively impossible. The main shortcoming of almost all current ESG approaches, however, is the dangerous disconnect between asset-owner objectives and the ubiquitous one-size-fits-all ESG.
None the less, the future of ESG investing is likely to be heavily reliant on AI given the expansive data resources now available and that appropriately applied machine learning can help solve the problems afflicting current ESG rating agencies: subjectivity, inconsistency and opacity.
... the future of ESG investing is likely to be heavily reliant on AI given the expansive data resources now available and that appropriately applied machine learning can help solve the problems afflicting current ESG rating agencies: subjectivity, inconsistency and opacity.