Green Finance Ratings: With Artificial Intelligence, instead of a guessing game

Anyone who invests money these days is looking not just to increase their wealth, but also to save the world a little bit – or at least help make climate reversal easier. “Sustainability is a megatrend in the financial industry”, says Jan-Alexander Posth of the ZHAW Institute of Wealth & Asset Management. The growing range of products in the sector is evidence of this. Financial data provider Morningstar counts 7,486 sustainability funds; these often have the identifier ESG in their name. The abbreviation stands for taking into account the impact on the environment (Environment) and people (Social), and a corporate management ranking (Governance). According to Morningstar, in the first nine months of 2021 investments in these funds doubled to $3,893 billion. This sum doesn’t include money that is invested directly in stocks, bonds or other financial instruments according to ESG criteria. 

Institutional investors such as pension funds are increasingly bringing their investments into line with ESG criteria. According to the Swiss pension fund index maintained by Credit Suisse (CS), these pension institutions have invested 32.1 per cent of their money in a “very sustainable” manner. This estimate is based substantially on the ESG ratings of the financial products used, as scored by financial services provider MSCI.

“Our mission is to provide more reliable benchmarks for selecting a green, sustainable investment.”

Jan-Alexander Posth,  Institute of Wealth & Asset Management

But there’s a catch. With ESG ratings, one rating agency may give a company a score of ‘good’, but another agency arrives at a conflicting conclusion, says Jan-Alexander Posth, whose main areas of research include sustainable finance, and digitisation & artificial intelligence. That means anyone who wants to invest sustainably is groping about in the dark. It’s the same for companies that want to fight against global warming. It’s still unclear to them what it is specifically that they need to change.

This is where the Institute of Wealth & Asset Management comes in. “Our mission is to provide more reliable benchmarks for selecting a green, sustainable investment”, explains Posth, who holds a doctorate in physics and has worked in the finance sector for a long time. Artificial intelligence is proving its worth here. One project, for instance, aims to use satellite data to show how specific mining companies are impacting on biodiversity. It’s not just about using ordinary satellite images to detect deforestation in remote primeval jungles. Multispectral data, ranging from the infrared spectrum to radar systems and allowing for very detailed statements about plant species, geographical features and much more, will also be used. While this scheme, led by Tomasz Orpiszewski, is in its initial stages, a second project is in the process of implementation. This second project looks at how sustainable a company’s technologies are. The project uses AI to evaluate a company’s patent portfolio – identifying firms that are particularly active in driving innovations in the area of CO₂ reduction, for example.

In a first stage, the database is a set of 2.5 million patents from 1990 to 2021, in English. The patents come from a number of countries and cover the European Patent Office’s patent class Y02, with climate-relevant technologies (Climate Mitigation Technologies). The project, financed by Innosuisse, is run jointly with the University of Basel and project partner EconSight GmbH, a firm specialising in analysis of the competitive environment in technologies of the future. “The patents a company submits today will usually be in use on the market in three to four years”, explains Jochen Spuck, Chief Technology Officer of EconSight. He conducts in-depth patent analyses for companies and identifies those that are likely to perform well given the realities of climate change. The results are sometimes unexpected. For instance, at first glance the oil companies Shell, Exxon and Total are making equal efforts towards sustainability, but Spuck believes that, thanks to a variety of relevant ‘green’ patents – in areas such as photovoltaics, battery technology and automation – the French firm Total is set to significantly outperform its two rivals.

“The machine recognises, for instance, whether ‘bank’ means the side of a river or a bank where I can open an account”

David Jaggi, Institute of Wealth & Asset Management

EconSight has already classified thousands of patents. These patents are now on a deep learning workstation as data learning sets, explains David Jaggi, a member of Posth’s project team. A deep learning workstation is a very powerful computer that is optimised for use with artificial intelligence. It sifts through the patent texts with a deep learning language module that understands not only individual words, but also their context. “The machine recognises, for instance, whether ‘bank’ means the side of a river or a bank where I can open an account”, says David Jaggi. Using the learning set and some of the 2.5 million patents, it is training computers. It sets out each patent text in numeric form and automatically compares patterns and correlations. Then the work gets serious: a first pass in which a previously unused set of patents is used. “This is where it becomes apparent how well the algorithm can cope with data that it has never seen”, says Jaggi.

The success of the algorithm can be measured against the patents and companies that EconSight has already assessed. Does the computer come up with a similar result? If not, corrective actions are called for: make adjustments, fine-tune the algorithm, go through new rounds of training until the goal is achieved: a reliable classification indicating how well a company is positioned in terms of the climate-relevant areas of technology. Up to now, almost every ESG rating provider has used a different method for its calculations, says project team member Linus Grob, “so it’s difficult to compare the results of the various rating providers”. An established index is still lacking and much of the information is based on self-declarations by the companies in question, which are difficult to verify.

A transparent, authoritative rating is of prime importance, as project leader Posth makes clear. “To bring about the change to a sustainable society, we need to redirect financial flows in such a way that innovative approaches and sustainable technologies are favoured and promoted, while modes of production that damage the climate are made more expensive.” Redirecting without guard rails and signposts is an impossibility: “In the absence of reliable data and guidelines for decision-making, the change will be unsuccessful in the end.”