If you have ever looked up a company's ESG rating and found that MSCI rates it AA while Sustainalytics assigns it a high-risk score, you are not misreading the data. The disagreement is real, widespread, and structural.
Academic research by Berg, Koelbel, and Rigobon at MIT Sloan found that correlations between major ESG rating providers range from 0.38 to 0.71. For context, credit rating agencies - Moody's, S&P, Fitch - correlate at 0.99 on the same bonds. ESG ratings from different providers on the same company agree less than half the time.
This is not a bug that will be fixed with more data. It is a direct consequence of how these ratings are built.
Three Root Causes
1. There is no agreed definition of "ESG"
Environmental, Social, and Governance covers an enormous range of factors: carbon emissions, water use, biodiversity, labour practices, supply chain standards, board diversity, executive pay, lobbying disclosure, data privacy, and dozens more. Each rating provider decides which factors matter, how much each matters relative to the others, and how to measure them.
MSCI uses an industry-relative approach - it scores companies against peers in the same sector. A coal company with unusually good safety practices may score higher than a technology company with ordinary governance. Sustainalytics uses an absolute risk framework. The same company looks different through each lens because the lens itself is different.
2. Disclosure-based systems reward reporting, not performance
Most major ESG ratings rely heavily on company self-disclosure. Companies submit sustainability reports, respond to provider questionnaires, and receive scores based largely on the quality and completeness of what they report. This creates a systematic bias: companies that report more - regardless of what they actually do - tend to score higher.
Research has shown that improvements in ESG scores often follow improvements in disclosure practices rather than improvements in actual ESG outcomes. A company that hires a sustainability communications team may improve its rating without reducing a single tonne of emissions.
3. Scope and materiality differ by provider
Different providers use different industry classification systems and materiality frameworks. What counts as a financially material ESG risk for a pharmaceutical company varies significantly between MSCI's materiality map, SASB standards, and Sustainalytics' sub-industry framework. A risk that drives a score down with one provider may not even be measured by another.
What This Means in Practice
For investors, the divergence creates three practical problems:
- False precision: A single ESG score implies a level of certainty that does not exist. Two equally reputable providers may reach opposite conclusions on the same company.
- Gaming risk: Companies learn which provider's methodology they are being scored against and optimise disclosures accordingly. Better reporting replaces better performance.
- Stale data: Most ratings update annually. A company with a recent major scandal may carry a high ESG score for months before the rating reflects reality.
The Direction of Travel
The ESG assessment landscape is evolving. Regulatory pressure - from the EU's Green Claims Directive, SEC climate disclosure rules, and ISSB standards - is pushing toward standardised, audited disclosure rather than self-reported data.
But standardised disclosure still does not solve the fundamental problem. It improves data consistency; it does not verify whether companies are doing what they say.
The next step is moving from disclosure-based scoring to evidence-based verification. Instead of asking "what did the company disclose?", verification asks "what did the company actually do, and what does independent evidence say about it?"
This shift matters because it changes the incentive structure entirely. A company cannot improve its score by publishing a better sustainability report. The only path to a higher score is closing the gap between commitment and action.
That is the premise behind Walk-The-Talk Score™ - and it is why we think the ratings divergence problem, while real, is ultimately a symptom of measuring the wrong thing.