The Differences Between Analyst-Created and AI-Derived ESG Scores

The two types of ESG research play different roles and have different advantages


By Simon Turner



As the importance of integrating ESG data into investment decision-making processes becomes clearer, there’s also a growing awareness of the two distinct types of ESG research available: analyst-created and AI-derived. As the names suggest, the former is subjective research by ESG analysts, while the latter is objective analysis derived from Artificial Intelligence algorithms which don’t depend on human opinion. The outputs of these two ESG approaches are markedly different from one another. Understanding the differences, and how the two types of ESG data can be used to your advantage, is the key to maximizing their value.


Analyst-created ESG research: the mainstream approach


Most ESG research, historically, has been created by ESG analysts, based mainly on corporate sustainability reports and management conversations. This type of ESG research tends to be mostly subjective. As the mainstream source of ESG information, it has had a leading role to play in the growth of ESG-focused assets in recent years. For good reason: ESG analysts are for the most part highly qualified professionals with extensive specialist knowledge which can be used to create value for their corporate clients and/or ESG-focused investors.


However, there’s a noteworthy challenge with analyst-created ESG research: there’s a wide divergence of analyst opinions. As shown below, the analyst-created ESG scores for the same companies are vastly different. Even though these are some of the world’s most analyzed companies, the divergence of analyst ESG scores for Apple, Amazon, Facebook and others can vary considerably.

Source: https://cyanreef.com/too-many-cooks-in-the-kitchen-why-the-esg-scores-differ-so-much/


As a result, the correlation of ESG ratings is a remarkably low 0.4, far lower than the 0.9 correlation for credit ratings. Many observers are surprised by this variance, and rightly wonder why.


There are two main reasons.


First, each ESG rating agency has its own internally-designed methodology for assessing performance, and there’s been little progress towards a common standard. In fact, each ESG rating agency closely guards its methodology as intellectual property, so the prospect of an industry standard emerging or transparency into its calculation is unlikely.


And second, ESG analysts each have their own subjective opinions on ESG performance, often defined by qualitative rather than quantitative judgments. Humans tend to agree about numbers more than words, so the range of outcomes is higher when qualitative assessment enters the fray.


Take ESG materiality (the likely impact of each ESG factor on financial outcomes) for example. Each ESG rating agency has its own method for assessing materiality, an infamously fluid concept which changes over time. Sustainalytics, one of the larger ESG rating agencies, has a customized approach to ESG materiality for each of the 138 sub-groups that it assesses, and regularly reviews its ratings manually to ensure that all current risks are reflected. The complexity and fluidity of their approach highlights what a moving target ESG analysis proves to be.


The subjectivity challenge in analyst-created ESG research


Beyond the vast range of ESG scores, there are some fundamental challenges which arise from the emphasis that ESG analysts place on sustainability reports.


First, the publication of sustainability reports is an annual event which includes ESG information from the preceding year, so all analysis based upon sustainability reports is at best backward-looking by a year or two.


Second, companies choose their own goals, updates, and stories as well as which ESG data to report. So management teams are setting their own goal-posts and then presenting their own perception of how they are performing against those goals. With this backdrop, there’s the obvious risk that some companies use their sustainability reports to “greenwash.”


Case in point: there’s a bias in the ESG ratings world towards awarding higher scores to larger companies which leverage their larger budgets to produce longer and glossier ESG reports. This risk of rewarding ESG communication skill and resources rather than underlying ESG factors is especially prevalent as investors become aware that stronger ESG performance tends to produce better investment outcomes.


So subjectivity is a big deal in the ESG world, with research depending upon subjective inputs affected by these limitations.


AI-derived ESG research and its challenges


AI-derived ESG research effectively addresses the timing and subjectivity challenges of analyst-created ESG research. This research is compiled by trawling through vast amounts of unstructured data in ways beyond human capability. It’s a markedly different approach to traditional ESG analysis in that it is an automated process which doesn’t rely on subjective human opinion.


The reasons for utilizing AI-derived ESG data are compelling. AI-derived ESG data deliver an investment edge beyond what is available in the mainstream ESG data market. It allows investors to use up-to-date ESG data, track ESG sentiment over time, identify stock-specific ESG risks, and augment analysis with the most meaningful ESG inputs. So AI-derived data is unbiased and current, enabling savvy investors to use it to gain an advantage over competitors who rely exclusively on traditional ESG data sources.


The benefits of using analyst and AI-derived research in combination


So the differences are apparent. Are both types of ESG research useful?


The short answer is yes.


Analyst-created ESG research creates value through the unique ESG rating agency methodology employed, and the subjective expertise of the ESG analysts conducting the research.


AI-derived ESG research is emerging as the ESG data sector’s gap-filler. It provides more current data which are objective and incorporate a more extensive range of inputs than analyst research. It provides a useful context and benchmark to assess the validity of analyst-based research.


For these reasons, there’s a strong argument to use AI-derived ESG research in combination with analyst-created research as they complement and supplement each other.


Why not leverage the best of both worlds?


With more and more investors connecting the dots between ESG ratings and company performance, access to high quality ESG data is becoming a source of competitive advantage. Analyst-created and AI-derived ESG research each have their roles to play in the transition toward a more sustainable world. These two types of ESG research offer compelling benefits in combination beyond what each approach alone can achieve.


At Global Imprint, we’re here to help you make sense of the ever-evolving ESG landscape. Reach out to our team to set up an introductory call or complete our ESG Readiness Survey to get your ESG baseline.