Navigating ESG data sets and ‘scores’

By Gary Vaughan-Smith, chief investment officer, SilverStreet Capital

As politicians continue to focus on creating regulation that meets net-zero pledges, the number of ESG data service providers entering the market swells. The proliferation of service providers, with divergent ESG topics, measurement criteria and ESG ‘scores’ creates several challenges within this sustainable universe. For pension schemes looking to assess sustainable investment opportunities through ESG data analysis the picture is increasingly complex.

Without a method of standardisation ESG data sets and subsequent ‘scores’ risk being misleading. Positive ESG scores tend to sit with those companies with a higher market cap able to funnel resources into reporting. ESG scores may become more biased as companies get ‘better’ at reporting the ESG data.

Most ESG analytics is focused on factors that lead to improved stock market performance. This may sound a logical approach but, to take environmental factors as an example, the focus should be on what that company is doing to the environment and not which factors drive better financial performance.

Some of the services providing ESG scores and weightings have based weightings and criteria on historic stock market relative performance. Providers have been quick to show a portfolio tilted towards companies with high ESG scores will produce outperformance. However, this approach has generally led to the usual back-testing bias issues that many quant fund managers grapple with day-to-day: a model works well in a back-test but stops doing well on live data. This seems to have happened with ESG scores: weightings and factors are carefully chosen so that the score would have, in retrospect, helped you out-perform. The marketing data shows out-performance (high ESG score companies out-perform low ESG score companies) leading up to launch of the new service but not thereafter.

As a result of these difficulties, the data provided by ESG vendors is inconsistent with each other and this is a material issue that is not easily resolvable, probably impossible.

How can ESG management be implemented and monitored?

An approach that leads to informed and independent assessments on ESG risks, where companies can show improvements over time is the preferred course. Our approach in summary is threefold; select ESG performance standards (the equivalent of IFRS accounting standards), conduct independent annual ESG audits and ensure full transparency, with reports published on-line with action points for management.

How will machine learning and AI change how data is analysed?

Machine learning techniques are used to support ESG scores by analysing unstructured text (articles) from a large variety of sources in different languages and typically lead to daily scores in trend, volume, and performance relative to 26 ESG factors. The advantage of this approach is that ESG factors become closer to ‘real-time’. One example cited has been Facebook where the AI ESG score started deteriorating some time before the Cambridge Analytica scandal came out in 2018.

AI can be applied in a much broader way than these text-based approaches. One example is the reading of satellite data to search for fires, deforestation, or industrial emissions. These can be linked to precise locations and thus to companies. Emissions could be measured real-time and more accurately. Similarly, analysis of satellite data should be able to identify (for example) CO₂/methane emissions or fertiliser pollution in rivers, tracing back to source. River pollution is often periodic (bursts of chemicals released over time) so real-time monitoring should be able to keep an accurate track of some key ESG issues such as this and make companies more accountable.

Is the regulation supporting the use of data robust and futureproofed?

The current situation is one of evolving regulation and diverse approaches to ESG data. Inevitably, things will change over time, and we believe that the end-game is an approach outlined above with a common set of standards equivalent to IFRS and independent ESG audits. There will be less interest in “scoring” and more on “reporting”.

How does it need to change with the times/technology?

ESG data management and reporting needs to evolve so that ESG standards improve, rather than being some sort of exercise to game a score. This is best done through a commonly accepted set of standards which need to be flexible enough to allow ESG management and audits to be customised to that company and the sector that it operates within.

In respect of data provision and fund management, a frenzy has emerged which is ultimately based on a flawed approach. It is difficult to produce an ESG score which a third party might independently produce. Indeed, that third party might have a substantially contrary view to that company’s ESG score. Data providers are salivating at the fee income as are fund managers who can ‘manufacture’ ESG products in multiple forms using these scores. The correct approach, and the long-term future, will be company specific reporting against standards and this will require old fashioned analysis by investment managers rather than simplified scoring approaches.