Responsible investment has skyrocketed in recent years. Driven by demand from retail and institutional investors alike, some USD 30 trillion are currently invested in products with an ESG component. Unsurprisingly, the production of ESG data has intensified too in order to facilitate such investments by responding to transparency and performance requirements. Today, the market is literally flooded with data and there seems to be no way of turning off the tap. This information serves as an essential raw material when assessing the ability of a security, share or bond to generate value based on the issuer’s responsibility and sustainability policy. Given the sheer volume of information available, artificial intelligence may be the key to absorbing, managing, analysing and harmonising ESG data.
On the whole, companies are increasingly open about this topic. Some even expound on it at great length. In 2011, according to a study by KPMG, fewer than 20% of companies listed on the S&P 500 published information about their ESG commitments. Nearly a decade on, the picture has totally changed and most large caps go as far as publishing dedicated sustainability reports to highlight their efforts.
The complex maze of ESG data
In no time at all, trying to exploit the masses of ESG data out there has become a labyrinthine task. New entry points keep springing up and, in the absence of clearly established standards on modelling financial ratios, analysts and investors alike are struggling to get their bearings. In its defence, this multidimensional mass of data needs to fully cover the three criteria that make up the ESG triptych: environmental, social and governance. There are infinite parameters for each of these criteria.
Supplied by companies, public bodies or their proprietary research teams, new rating agencies have appeared over the last decade with the aim of collating and decrypting this maelstrom of information. Vigeo-Eiris, Sustainalytics, EthiFinance, ISS-Oekom and Robeco-Sam, to name but a few, have carved out a niche for themselves using methods that are varied to say the least. The data they exploit are different, and the ratings and categorisations they arrive at tend to be equally so. Each agency has its own way of doing things and establishes its own framework, and since they are not yet subject to regulatory standards, there is currently a certain lack of coherence to the ratings and categorisations that they publish.
As demonstrated in a recent study by the MIT Sloan School of Management, these differences are partly due to the use of different criteria when setting up the ratings. They may also be caused by different scales being applied to the same variables. These divergent approaches can throw up some real contradictions.
Artificial intelligence as a tool for enhancing the uniformity of ESG data
Ultimately, we can expect ESG data to become standardised and processed using a single system that makes them easier to compare. Consolidation in this sector, as demonstrated by Standard & Poor’s recent purchase of RobecoSam, is another pointer that things are headed in this direction.
The proposed advances in artificial intelligence, big data and machine learning will also make it easier to process and model these data. Indeed, the present ratings system is undermined by the obsolescence of traditional analysis tools and a lack of expertise in analysing and integrating ESG factors.
AI “for sustainability purposes*” is already used in many companies and is having a huge impact. Development managers are using it to design products and services that can stand out for their ability to generate sustainable value. As an industry with an insatiable appetite for information, asset management is a prime candidate for AI.
Aspects of AI including predictive analysis, statistical projections and sequencing lend themselves perfectly to the uniquely shape-shifting environment of ESG data. To start with, AI is able to help detect and correct bias and inconsistencies.
Its algorithms, architecture and resulting calculation power mean that data can be tagged and managed more rigorously – more fairly even. All this means that millions of pieces of information from thousands of sources can be synthesised and transcribed in much simpler and more structured formats than they are today.
In the more immediate future, asset managers with an ESG approach need to make the best of the data as they are right now, warts and all. Whatever the quality, relevance or granularity of the raw data gathered, they need to be reprocessed because of their heterogeneous nature, and that comes at a significant cost. At the moment, the priority is to adapt them to ensure they make a proper contribution to building the best possible ESG portfolio. This work involves sourcing, organising and using the data.
With this in mind, the ideal scenario would be to bring about a miniature value chain with tools and processes for facilitating data exploitation. This chain has several links, from selecting providers to automating reporting, and from creating internal methodologies to organising the data for investment purposes. It is crucial to reduce the number of external and internal data sources. It is equally important that complementary financial and non-financial data can be exploited together. Lastly, we must ensure that the chosen tools can adapt to all these functions.
This substantive work, carried out with a real focus on perspective, is likely to streamline the decision-making process and result in appropriate investment choices that will prove even more popular with investors. Although ESG investments are likely to experience strong growth in the decades ahead, their success will increasingly depend on asset managers’ ability to manage increasingly complex sets of data.
*AI "for sustainability purposes”: using artificial intelligence in products and markets to create sustainable value and exploit previously untapped sources of value.