The divergence problem
In 2022, a study by MIT Sloan researchers found that the average correlation between the six major ESG rating providers was just 0.54 — compared to 0.99 for credit ratings from Moody's and S&P. For investors who grew up trusting credit ratings as near-objective facts, this gap is disorienting. A company like Tesla carries an ESG score ranging from 14 (MSCI) to 78 (FTSE Russell) depending on which provider you consult.
This is not primarily a data quality problem. It is a conceptual disagreement problem. Before any data is collected, rating providers must decide: What counts as an ESG factor? How much does each factor matter? How do you handle missing disclosures? Every one of these choices creates legitimate divergence.
The core insight
ESG scores are not measurements — they are opinions. Just as two portfolio managers can produce different fair-value estimates from identical financial statements, two ESG analysts can rationally produce very different composite scores from identical sustainability data.
Three structural causes of divergence
1. Scope divergence — what is being measured
MSCI focuses primarily on financially material risks to the company. GRI focuses on the company's impact on the world. These are philosophically opposite starting points. A company with massive environmental impact but strong risk management scores well under MSCI and poorly under GRI. Neither rating is wrong — they answer different questions.
| Provider | Primary lens | Key inclusion choices |
|---|---|---|
| MSCI | Financial materiality to investors | Industry-specific metrics, transition risk, controversy deductions |
| Sustainalytics | Unmanaged ESG risk exposure | Risk exposure × management quality scoring |
| FTSE Russell | ESG exposure + management | 200+ indicators, equal sector weighting |
| Bloomberg | Disclosure quality proxy | Scores penalise non-disclosure heavily |
| ISS ESG | Impact and best practice | Long controversy tail, supply chain scope |
| CDP | Climate/water/forest disclosure | Voluntary questionnaire — measures intent, not performance |
2. Measurement divergence — how data is collected
Even when two providers agree on what to measure, they may disagree on the metric. Carbon intensity per revenue dollar and carbon intensity per employee capture the same underlying concern but can produce opposite conclusions for a capital-light software company versus a capital-heavy manufacturer. Providers also treat self-reported data, third-party audited data, and imputed estimates very differently.
3. Aggregation divergence — how factors are combined
Once individual metrics are collected, they must be combined into a score. Do you weight E, S, and G equally? Do you apply industry-specific weights? Do you use linear or percentile normalisation? Do controversies trigger deductions or separate overlays? Each of these aggregation decisions — invisible to the end user — can swing the final score significantly.
The rater-shopping problem
Companies have learned to optimise for the methodologies of their most important ratings customers. This creates a perverse incentive: rather than improving genuine sustainability performance, some companies improve disclosure quality and stakeholder communication targeted at specific provider methodologies. This is legal and common — but it widens the gap between ratings and reality.
A practical framework for multi-provider analysis
The answer is not to pick one provider and ignore the rest. It is to triangulate. When ratings converge across providers, confidence in the underlying signal is high. When they diverge sharply, that divergence itself is informative — it tells you a genuine debate exists about what matters for this company.
- 1Start with the consensus — average or median score across providers tells you the central tendency of market opinion
- 2Identify outliers — if one provider rates a company 30 points above average, find out why. The outlier methodology often reveals something the consensus misses
- 3Map divergence to business model — a logistics company rated differently on Scope 3 methodology is not the same problem as a bank rated differently on governance controversy treatment
- 4Prefer audit-verified disclosures — companies with third-party assured reports have smaller divergence across providers because imputed estimates are replaced by verified data
- 5Track stability over time — volatile scores often signal a company that is gaming disclosure rather than improving performance
How OpenESG handles divergence
OpenESG uses a proprietary methodology that combines quantitative metrics from company filings and regulatory databases with qualitative assessment of disclosure quality and news-sourced controversy monitoring. Rather than hiding the uncertainty in a single number, we expose it — each score comes with a confidence band and a breakdown of which factors are driving the result.
Our ESG Intelligence Engine, powered by Claude Sonnet, reads the latest company disclosures, regulatory filings, and news to flag discrepancies between stated commitments and observable actions. This is particularly valuable for catching the gap between what companies claim in sustainability reports and what appears in contemporaneous news coverage.
How to use ESG scores effectively
Use them as screening tools, not verdicts. An ESG score should prompt questions, not provide answers. When a score surprises you — high or low — that is the signal to dig deeper into the underlying data, not to accept the number at face value.