Demand for data-screening services that can help customers find investments that meet environmental, social and governance (ESG) standards has soared in recent years.
More than 20 per cent of global fund assets under management were invested using at least one restriction screen (which enables investors to avoid certain sectors or companies) by the end of June 2023 — 10 times more than three years earlier, according to a report by Morgan Stanley.
As the Morgan Stanley researchers point out, demand for more “exclusionary” investments has risen not only in response to investors’ changing values, but also as a result of rapid changes in regulatory requirements — such as the EU’s Sustainable Finance Disclosure Regulation, which sets out mandatory ESG disclosures to be made by asset managers.
The facility to screen investments means investors are able to not only steer clear of, for example, weapons makers or thermal coal or tobacco companies — which are the most common exclusions — but also to invest in ways that emphasise broader beneficial outcomes, such as cleaner energy or gender equality.
Morgan Stanley estimates about 8 per cent of global fund assets are now invested sustainably.
How is the screening carried out?
Index and data providers agree that they would not be able to meet the explosion in demand for ESG screening and rankings had there not been a parallel jump in the capabilities of the technologies to provide them.
Morningstar Sustainalytics, one of the best known ESG data and index providers, says it uses information retrieval and extraction technologies, coupled with different types of artificial intelligence — such as machine learning, natural language processing, and symbolic AI (which tries to mimic human abilities to deal with concepts and behaviour rules) — to screen investments against hundreds of ESG criteria.
This, it says, means it monitors hundreds of thousands of publicly available sources, starting with disclosures published on websites, media publications, regulators’ and non-governmental organisations’ publications, and providers of non-standard data, such as independent research on supply chain risk.
“In all cases, there is an element of curation, where our analysts check the output of the automated process to ensure adherence to the highest quality standards,” says Arik Brutian, senior vice-president of digital innovation at Morningstar Sustainalytics.
Vinit Srivastava, chief executive and co-founder of MerQube, a specialist index provider that has moved into the ESG sector, points out that data has “proliferated” in volume since interest in investing according to sustainable principles gained ground.
This proliferation and the resulting confusion and contradictions that arise from trying to make sense of different data sources, means there is a greater need than ever for what he refers to as “prompt engineering”, which is used in generative AI models. “It’s very key in all of this,” he says. “The questions you ask generate the response that you get.”
Srivastava notes that any information that is difficult to interpret, however, represents an opportunity for any investor looking for a competitive edge. One of the niche areas that MerQube works with uses natural language processing to analyse the language used by senior executives in earnings calls, which as he says are forward-looking, as opposed to earnings reports, which deliver a historical picture.
“Technology has been an enabler,” says Srivastava.
What are the technology challenges in ESG data screening?
Morningstar’s Brutian says one of the main challenges for those screening investments on ESG criteria relates to the quality of original data. “The quality of publicly available data does not always meet our expectations, nor is ESG disclosure yet standardised,” he says, adding that Morningstar Sustainalytics has created technology-supported safeguards to try to ensure the reliability of data and the trustworthiness of sources.
Other challenges relate to the speed with which regulations are changing. A sustainable finance implementation timeline provided by the European Securities and Markets Authority, indicates more than 20 deadlines relating to requirements for new disclosures between early 2021 and 2028.
Each time a law changes, so must the data-screening database. In some cases, too, public agencies can impose sanctions or restrictions — for example, on Chinese suppliers judged to have connections to military establishments or to forced labour in the country’s Xinjiang region.
For this kind of screening, there is no alternative to the kind of in-depth, shoe-leather research conducted by traditional risk consultancies, even if it is supplemented by natural language processing and machine learning. “Without technology, we would not be able to do half of what we do,” says one specialist supply-chain screening expert. “But there’s no substitute for smart humans.”
Data-screening professionals have also had to develop ways of comparing data over time — for example, a company’s promises on climate targets. Morningstar Sustainalytics says it uses processes that allow for “triangulating” targets reported by companies and comparing them to those companies’ emissions data.
Leonardo Bonanni, founder and chief executive of Sourcemap, a specialist in supply chain data, says its systems can check for fraud. For example, if a company claims it is using recycled materials, Sourcemap can verify that there are transaction records to back that up.
This work has been helped massively, Bonanni says, by what he calls “robust AI” technologies, such as natural language processing, rather than the generative AI technologies such as ChatGPT that have recently made headlines. Srivastava agrees. “AI techniques are not new; what has changed is the processing power,” he says.
