Heard about the chatbot ChatGPT? Artificial intelligence is advancing rapidly, says Arjun Brara – and could soon be used to refine ESG ratings and expose greenwashing
The significance of environmental, social and governance (ESG) is growing for actuaries. And as its effects in the stock markets become more pronounced, even the procrastinators are finding it harder to ignore.
In 2020, a Sunday Times investigation into Boohoo hit the headlines, having found that the online fashion giant was exploiting its factory workers. Despite receiving largely positive ESG ratings from multiple sources, the company was found to be paying factory workers £3.50 per hour and failing to adopt covid social distancing guidelines. After the scandal broke, its share price almost halved.
This is just one example of many, showing how important ESG considerations are to investors and the wider markets. It is also becoming clear that ESG investments tend to outperform the rest of the market – for example, Morgan Stanley’s 2020 Sustainable Reality report found that sustainable funds in the US outperformed traditional funds by 4.3% in that year. This is further increasing investor confidence and interest in ESG.
Are ESG ratings trustworthy?
However, there are concerns over ESG ratings and the way they are derived. The ratings are currently unregulated, so there is little scrutiny over how they are calculated. A study in 2020 by the University of Zurich found that there is, on average, a correlation of 0.54 between the ESG ratings given to the same companies by different providers. This relatively low correlation shows that the ratings in themselves cannot be relied upon in the same way we rely on credit ratings, which have a correlation of 0.99.
One reason for the low correlation is that providers weight various ESG elements differently. An example of this is Tesla, which scores higher among providers that weight ‘environmental’ parts of ESG more highly, and lower among those that weight ‘social’ parts more highly. ESG strategies such as ‘best-in-class’ or inclusionary/exclusionary investing may be less affected by this issue than thematic or impact investing, as they are relative measures. However, it still raises questions over the reliability and comparability of the ESG information that drives the ratings, especially given the lack of regulation.
Other considerations might be bias – ESG ratings are still gaining traction, and if an agency upgrades a well-known company’s ESG rating, the company may promote the agency by advertising the fact that it has been certified by it. There have been similar issues around greenwashing, such as companies releasing plans relying on technology that is yet to be invented and may not come to fruition.
The data challenge
The availability of relevant and credible structured data is cited by asset managers as one of the most significant challenges in ESG investing, according to EY’s 2020 Global Institutional Investor survey. While there is a shedload of unstructured data out there, the structured data that we typically use for analysis is not available in large quantities, or over a longer period. This is improving as time goes on and companies become eager to disclose their ESG credentials, but there is a long way to go before we have the kind of data quality and quantity that we have with, say, historical mortality data. Confidence in this data is not helped by stories of companies trying to ‘greenwash’ their products.
The rate of data production is increasing exponentially – 90% of the data currently in existence was created in the past two years. Every minute, almost half a million tweets are sent on Twitter, and more than 800,000 comments and status updates are posted on Facebook. Every day, 300m new Facebook photos are uploaded and 100 million people use Instagram Stories.
A lot of this data is public, but too unstructured and impractical for any human or classical computing algorithm to work through. Artificial intelligence (AI) may be able to help us digest it and start making sense of the noise using ‘sentiment analysis’. This can be done in a way that allows us to compare how people feel about the ethics of different companies. Such data mining is increasingly being explored in the general investments space, including by large institutions such as Goldman Sachs, and is starting to be picked up more widely.
The role of AI
More and more investment companies are incorporating AI into their decision-making, as it helps guard against a major financial risk: the risk that they have invested in a company that engages in heavy greenwashing, which then becomes public and causes a significant hit to the share price.
Tokyo’s Government Pension Investments Fund – the world’s largest pension fund – is already basing a lot of ESG investment decisions on AI-powered analysis, with AI models constantly watching the markets and unstructured online data. This means the fund can get updates daily, as opposed to quarterly or annually, and catch in real time any evolving situations that might have a significant impact on a company’s ESG profile.
Some companies are even developing specific AI tools that look only for greenwashing. While AI is effective at pulling out significant information, this will likely be used to highlight key pieces of information so that an expert can review them and decide how they could affect market prices.
Towards the start of this year, the news was awash with ChatGPT3, a generative model that is capable of simulating human speech. It can summarise large volumes of text data and pick out the key points that are relevant to a specific area of concern, and is evolving further. This is a good example of how far AI has progressed in terms of understanding the unstructured language data around us, and the speed at which it is progressing.
The problem of limited structured data could be tackled by using AI techniques to leverage the large volume of unstructured data, and structuring it so it is more interpretable for humans. AI is evolving, and quickly – so while these are some of the things making inroads today in ESG, this could change or develop rapidly, and both actuaries and data scientists should be ready to evolve alongside, to meet the challenge.
Arjun Brara is a senior actuarial consultant at EY, specialising in sustainable finance and technology innovation
