When AI and ESG investing collide in emerging markets
It is crucial for investors to educate themselves on ESG investments. In order to make better-informed decisions about the risks and opportunities presented by sustainability issues (such as the industry transition to a net zero economy), sustainable investors rely on ESG data but are faced with multiple challenges. After all, data is the lifeblood of well functioning financial markets. However, we cannot deny the fact that there are existing gaps, availability problems and inaccuracies within ESG data. Meanwhile, ESG data is proliferating. Indeed, more companies report data and alternative sources of data are becoming more and more available through forums, social media platforms and conference calls, for example. This article will explore how investors can use AI to navigate this landscape and make better informed decisions.
Today, there is no longer a debate regarding the integration of ESG in investments strategies is generating significative returns. Depending on the market chosen and the way investors are integrating ESG, the demand in ESG strategies is evolving. That being said, it doesn’t mean that the ESG data question is an issue solved by itself, because when you are trying to integrate the best you can ESG to your investments, it is still challenging for multiple reasons such as coverage of companies and data providers. This is the main reason why we have seen ESG denominating products in emerging markets represent relatively small market shares of the total of this asset class. However, this is a great point in time for this to change drastically. Indeed, the relative share of ESG products in equity and bond markets is predicted to flourish in emerging markets in the upcoming years. Hence, the strong will of investors to integrate ESG data of better quality within the emerging asset classes because this is where the maximum impact is believed to be generated. Therefore, the true need at the moment is for data to be less and less backward looking, and rather more forward-looking-focused, which is crucial when a decision is needed for a future investment in a future emerging market. The last element that is going to affect massively the type of data will be the impact of the regulations. Now the main question is “How can we create ESG data that is much more forward-looking and that will be easily accessible?” As ESG data is the fuel of ESG strategies investments, it is predicted that ESG data will become a public common good. To create this type of offer, public entities are needed to help create it and address new ways to gather the data and put it at disposal of investors, which brings the necessity for AI technologies. Before covering AI applications on treating ESG data, it is important to understand the different terminology that will be employed in the following sections of the article:
Figure 1: Artificial Intelligence Terminology Review
AI in Action: Using Natural Language Processing to Unlock ESG Data
Issues with data quality and disclosure of sustainability data are yet to be solved, but most importantly, it’s the unstructured sources of ESG data that are underused because of its interpretation complexity. Machine learning algorithms are the key to these issues. It can help analyze information at scale and provide insights. In particular, the use of “Natural Language Processing” and the “Computer Aided Analysis of Unstructured Text” are being leveraged to train models and store data in order to read technical texts while keeping in mind the identification of the environmental, social and governance risks and conduct sentiment analysis. So, it tells if a risk term appears in text in a positive, neutral or negative way. Today, the models are able to analyze all these metrics with 87% accuracy, which is a good performance compared to the industry’s standards. Great inputs to take from those models are:
- Issues with the greatest negative sentiment
- Positive momentum when time to focus on transitional opportunities
Now, why does all this matter? For example, take the word “fatality.” It could automatically be associated with a negative context and analysts could jump quickly to conclusions. However, analysts could miss out on the point where it was a negative sentence and there were no fatalities involved. At the same pace, Natural Language Processing won’t make the same mistake a human could when being faced to a massive amount of information.
Figure 2: Natural Language Processing applied to ESG data
Decision-making can be quite hard to understand here. If these models are meant to make impact, it is essential that they can be trusted and are being efficient. Building an explanatory feature is important, as well as tracking, managing and transparency about data bias and model performance. Today, it is challenging to predict the potential of impact of sharing data and to find ways to use AI to extend analytics, so it is important to look for ways to leverage data science in order to support investing.
Why do emerging markets particularly suffer from more data gaps than developed markets?
Mainly, we are faced with an issue of coverage when trying to leverage sustainability data from emerging markets. Indeed, when not covered by larger data provider institutions, it is an additional challenge for analysts to pull insights and predict their next move with or without AI models. However, there is a potential with AI to gather different sources of information to cover that gap. Also, this gap highly reflects the structure of the industry that is extremely American/European centric. So far, there are few to no asset managers from emerging countries, so most of the CAPEX regarding ESG developments in the past have been made in the core markets of where the industry is. The good thing is that this trend is likely going to change as we see more and more major financial players in emerging countries developing and deploying their own AI strategies to address ESG questions on their own markets.
How can these AI techniques can leverage ESG data?
Today, it is important to develop models that will interpret the ESG data currently available, but also models that will create the data needed to fill the gap and fix this quality issue we have been discussing throughout this article. The purpose is to gather all unconventional sources of information and treat good quality ESG insights.
At this moment in time, the landscape for data today, more particularly for climate data, is unrecognizable from the point we were 2 years ago, so we can predict that 2 years from now, it will have completely changed once again. AI and quantitative investment techniques in general have a large role of helping investors handle the nature of that inflexion point. Now, a broad range of sustainable data sets are available that are ranged from a highly structured company reported data and surveys that are conducted by professionals to very unstructured data sets based on satellite imaging and various text sources. But even within these categories (structured vs. unstructured), the raw inputs disagree, tell different stories like, for example, how much carbon does company X emit. And so, AI techniques help articulate and navigate across this incoherence. The purpose is to learn about which providers of information are more appropriate to focus on and which providers can be combined and improve insights while being of good quality.
On the other hand, NLP techniques can help identify new ESG insights that were once considered as plain data. It is important to understand the way companies are talking about sustainability and intersecting it with non-text data on what the companies are actually doing (speeches, social media, etc.). The main purpose is to look for opportunities where there are agreements and disagreements between these two types of information through Natural Language Processing.
Examples on how companies use AI to better understand ESG issues
Providing ESG analytics and investment solutions for creating more sustainable portfolios with greater impact, Impact Cubed use technologies to source, harvest and clean data as well as create analytics or large data sets already harvested. Technology enables them to provide on-demand coverage of every listed equity globally. One particular area that has been exploited by advance technologies are issues about sovereign debt in emerging markets. Looking at their ESG scores, there is a clear wealth bias like, for example, Norway and Sweden at the top of the list. Surely, there are other emerging countries that are investing, but maybe started on a lower level and have a long way to do in order to catch up to their sovereign debt. So, in order to find the investments and evaluate their true ESG score, Impact Cubed have developed a quantum that harvested massive amount of data filtering 20 years, 190 countries and 29 ESG factors to create an applied series of algorithms to transform data in pathways describing the expected rate of change for any level of sovereign debt for each country. So now, fixed income investors have a much more differentiated and nuanced way to look at ESG in emerging markets, and a bigger basket of emerging market countries to consider in their portfolio. For example, some countries in the Middle East are moving at much faster pace on these pathways on gender equality issues than what a conventional ESG score would show.
Amundi focuses on specific needs that are not necessarily covered yet by publicized solutions. Using NLP, they are interested in collecting information such as the company’s commitments and plans to phase out coal, for example, so that Amundi can gather a large pool of companies to engage with them, challenge them and make change happen. It is quite complicated to collect this type of information. Even though it can seem straightforward, there is nothing really well-structured at the moment from providers so Amundi develop techniques to save their analysts’ time so that rather than be focused about collecting the data, they would be more oriented towards the analysis process and the engagements with the company afterwards. In order to reach the commitments promised about decarbonization of portfolios, it is critical to dedicate the time to educate companies and solve together a way to make their data tell another story.
In conclusion, ESG investing is growing more and more in popularity among investors and engages a strong call for financial ESG data to be available to assess risks and improve their investment decisions. Up until today, there has been a need for better and faster ESG data delivery to help investors form robust-investment strategies. Artificial intelligence (AI) technologies, such as machine learning and NLP, are emerging in the financial practice, and it is important for investors to understand how to approach such technologies and interpret ESG data for better use.