The ever-rising demand for ESG has not only been driving the need for standards and regulations, but also the need for consistency on ESG ratings.
The ever-rising demand for ESG has not only been driving the need for standards and regulations, but also the need for consistency on ESG ratings. Investors are especially concerned that rather than accurately measuring a company’s reputational risk implied by its business conduct, the inconsistent ESG ratings between ESG data providers lead to greenwashing – either because firms can select the best score offered by various providers, or because the company itself provides self-disclosures that mask risk.
RepRisk builds its daily ESG research and signals exclusively on the actual ESG behaviour of a company as reported by more than 100,000 public sources.
Dr Heiko Bailer (pictured), Head of Quantitative Investments at RepRisk, a leading ESG data science firm, has been driving the systematic integration of actionable ESG signals across the investment process. Most importantly, he comments: “for successful integration into a wide variety of investment processes, ESG signals must be based on a consistent and relevant data-capturing process, be frequently updated, have sufficient history, and be transparent and customisable.”
RepRisk, in contrast to many ESG rating providers, does not consider company filings (self-reports), their product lines, or how they are internally structured (e.g. how many women are on the board of directors). The purpose of RepRisk’s dataset is to systematically identify and assess material ESG risks related to any public and private company worldwide. Since 2006, RepRisk has built its research and metrics exclusively on more than 100,000 media and stakeholder sources in 20 languages, such as NGO and regional government reports, regulators, and news websites. AI and machine learning algorithms filter, classify, and auto-link more than 500,000 documents daily, after which a global team of 85+ analysts curate and analyse the data according to a rules-based methodology. The output of the research is available via an online due diligence platform, as well as data feeds for seamless integration into internal systems.
This leads to daily-updated and highly relevant data points specifically designed to measure the ESG behaviour of companies and their reputational risk that can impact future financial performance.
Consistent and relevant data capturing and classification processes
Since 2006, RepRisk has used a consistent data-capturing process using the unique combination of AI and machine learning (ML) with a large team of analysts who ensure the depth and integrity of the data and feed their mappings back into the ML processes. As new and relevant themes such as epidemics/pandemics, lobbying, and racism/racial inequality arise, RepRisk expands its research scope. It also adds new languages on a regular basis.
Graph 1 shows the topic tag of Racism / Racial Inequality, which was recently added to RepRisk’s research scope, and the tag’s number of violations as monitored in the top 5 sectors in North America.
Traders and data scientists who want to successfully test trading strategies need datasets that are updated daily to match the pace of the markets. Asset managers also expect the signal frequency for portfolio construction, monitoring, and labelling to match their rebalancing horizon. Most investors need more than annual datapoints. And even if rebalancing is scheduled quarterly or monthly, investors can implement a special threshold-based rebalance that is triggered when a company approaches high-risk exposure.
For ongoing risk monitoring, daily updates can be crucial. Graph 2 shows how RepRisk’s G-score for Wirecard AG quickly captured its worsening reputational risk exposure and could have prevented investors from the ensuing fall-out.
Another important point is that the data must be based on a methodology that is consistent over time – and have a long enough data history to capture various market conditions. Since 2006, RepRisk has created its datasets point-in-time using the same rules-based methodology to identify and assess ESG risks, while extending its research scope and language coverage.
Transparency and customisation
There are a variety of strategies and signals which determine how trading decisions are made – strategies can differ by style, e.g. qualitative or quantitative, by universe, e.g. equities, bonds, private equity, or by trading frequency. However, the transparency of the data source and the ability for the strategist to customise the signals to the investment case are the two most important factors.
Graph 3 shows the rising ESG scores for Volkswagen AG during the “Dieselgate” scandal in the fall of 2015. Clearly, the social S-score was not driving the scandal, but rather the environmental E-score and the governmental G-score. To integrate actionable ESG signals into various use cases, they must be transparent and customisable – be it a G-score for the banking sector, a S-score for the retail sector, or with respect to increasing standards and regulations, such as a S-UNGC score (based on the 10 Principles of the UN Global Compact) or an E-SASB score (based on the Sustainability Accounting Standards Board materiality map).
As Dr. Bailer phrases it: “Doing good while doing well requires investors to reward companies that respect ESG standards, while punishing violators – all without sacrificing performance. Shareholder activism can be done at shareholder meetings or simply through investing and divesting, thereby stirring capital flows. Either way, the underlying data must be reliable and based on a systematic, outside-in view of a company’s business conduct – which is the approach RepRisk takes.”
Heiko Bailer, PhD, Head ESG Quantitative Investments, RepRisk ESG Data Science
Heiko Bailer Head ESG Quantitative Investments, RepRisk, has many years of experience as a quantitative investment manager working for hedge funds, family offices, and firms such as Credit Suisse, Deutsche Bank, and ABN AMRO, in London, New York, Tokyo, and Zurich. He has a PhD in Statistics as well as a Computational Finance degree from the University of Washington.