ESG Consensus Ratings – The Key to Asset Owner Oversight


Utilizing Artificial Intelligence and Machine Learning to Overcome the Subjectivity of ESG Rating Vendors


Steve Glass


Abel Noser Holdings / Zeno AN Solutions




Download Whitepaper PDF >

View Related Tabb Forum Post >


September, 2022




The centrality of ESG-related issues to investors is borne out by the rapid integration and growth of ESG-oriented retail and institutional investing. In this regard, ESG-related initiatives are ultimately about managing risk. For example, as noted by the Organization for Economic Co-operation and Development (OECD), a poor environmental record may make a firm vulnerable to legal or regulatory fines/sanctions; socially, mistreatment of workers and dissatisfied employees may lead to higher absenteeism, lower productivity and weaker client servicing/relationships; and weak corporate governance may incentivize and/or enable unethical behaviors related to pay, accounting irregularities and even fraud.  For all these reasons, identifying and addressing the various ESG issues germane to a corporation (i.e., “material”) is a quintessential exercise in risk management – for company management, investment managers thinking about holding that security in their investment portfolio, and asset owners concerned whether the manager is acting in accordance with fund policies.


However, even for practitioners who care deeply about ESG, a key obstacle they face is that almost all ESG-related data consists of raw metrics. Consequently, the user must have the internal resources to both collect and aggregate the data, as well as sufficient internal subject-matter expertise to consume the raw data and assess whether follow up is warranted. The one exception is ESG ratings.


ESG ratings purport to assess and rank the degree to which each company manages its respective ESG risks.  Today, several vendors offer services that assess company compatibility with ESG-related ideals and provide associated ESG ratings/scores. In theory, practitioners thereby need not consume the raw data themselves. Rather, they can utilize these ratings/scores to determine which companies are best managing their respective material ESG risks; and focus their analysis on monitoring company ESG ratings/scores.


In this regard, it is easy to see how ESG ratings/scores can have applicability at every stage of the investment cycle: asset allocation, investment universes, portfolio construction, investment selection, risk management, and regulatory/client reporting. And ideally, a better ESG rating will correlate over time to higher returns.


Unfortunately, as discussed below, the firms that provide ESG ratings/scores often yield very different rating assessments. Indeed, it’s not uncommon for a company to be rated very highly by one rating provider while simultaneously being rated very poorly by another. This lack of consensus has historically represented a significant challenge for practitioners hoping to develop policies and investment decisions based on those ratings/scores. Happily, recent developments in the ESG ecosystem offer a solution to this problem.


The whitepaper is divided into two sections. The first section explains in greater detail why individual providers of ESG ratings/scores, even with the best of intentions, often assess the same company very differently. The second section describes how artificial intelligence and machine learning have helped provide an elegant quantitative solution to this problem.







A crowd’s collective intelligence [when independent and diverse]

will produce better outcomes than a small group of experts…

James Surowiechi, The Wisdom of the Crowd, 2004