An ESG rating tells you how well a company manages its environmental, social and governance risks — the three areas ESG stands for. A high rating means the ESG rating provider finds the company well prepared for the ESG risks it faces. A low rating means it finds risks the company has left unaddressed. It is a risk assessment, not a measure of the good or harm a company does in the world.
Reading a rating starts with knowing whose rating it is. MSCI scores a company against others in its industry, on a letter scale from AAA down to CCC, where AAA is best.
Sustainalytics ignores the industry comparison and counts the risk a company has left unmanaged, on a number scale where lower is better. A company can look strong on both — as an MSCI AA and a Sustainalytics 12 — but the two ratings are working within different frameworks.
ESG rating providers do not even agree on what an ESG rating is for. MSCI and Sustainalytics both measure risk to the company. S&P Global states that its ESG Score applies double materiality, treating an issue as material when it affects society or the environment as well as shareholder value. LSEG, which replaced its Refinitiv methodology outright in 2026, says its scores are best read within a double materiality framework. Each rating makes sense on its own terms. What you cannot do is compare one provider's rating to another's, because they are not measuring the same thing on the same scale.
Four ESG rating providers account for most of the scores in use today: MSCI, Morningstar Sustainalytics, S&P Global and LSEG. Each is owned by a large financial data business, and each runs a scale that behaves differently from the others.
Two ESG rating providers cited throughout the academic literature have since stopped rating companies: KLD, discontinued in 2017 after passing to MSCI, and Moody's ESG, which closed its ratings operation. Moody's ESG was built on Vigeo Eiris, the provider most explicitly oriented toward a company's impact on the world rather than risk to the company.
ESG ratings from different providers agree far less than most investors assume. The landmark study on this is Aggregate Confusion by Florian Berg, Julian Kölbel and Roberto Rigobon, published in the Review of Finance. Across six ESG rating providers, they found an average pairwise correlation of 0.54, with individual pairs ranging from 0.38 to 0.71. Credit ratings, by comparison, correlate at 99%.
The pillar-level results are more surprising than the headline. Environmental ratings agree most closely, at an average correlation of 0.53. Social ratings average 0.42. Governance — the pillar built on the most standardized, most publicly filed data — agrees least, at 0.30.
These figures come with a caveat. The study's baseline year is 2014, and two of its six raters no longer operate in that form. The direction of the finding has held up across replications, but the specific correlations describe a provider landscape that has since consolidated.
The disagreement is not theoretical. On 20 August 2026, three constituents of the Dow Jones Industrial Average carried ratings pointing in opposite directions depending on who published them.
Boeing shows the widest spread. LSEG ranks it third of 124 peers, near the top of its group. Sustainalytics places it in the High risk band. S&P scores it 34 and marks the score under review.
Chevron carries the only Severe rating in the index, at 42.39. LSEG ranks it 45th of 257 peers, roughly the top fifth of its group.
Amazon holds the lowest S&P score in the Dow at 20 out of 100. Sustainalytics rates its unmanaged risk Low, and LSEG ranks it sixth of 95.
None of these is an error. LSEG scores publicly disclosed policies and management practices, so a company that documents its processes thoroughly ranks well. Sustainalytics counts the risk left unmanaged once those practices are accounted for. S&P scores relative to industry peers and cautions against reading its number across industries. Three questions, three answers, one company.
For what this divergence does to studies of ESG fund performance, see do ESG funds outperform.
Divergence comes from three sources, and the largest is not the one most people expect. Berg, Kölbel and Rigobon decompose it: measurement contributes 56%, scope 38%, and weight just 6%.
Scope is which issues get counted at all. Measurement is how a given issue is turned into a number — two providers can both assess labor practices and score the same company differently because they use different proxies for it. Weight is how much each issue contributes to the total. Weighting, the thing ESG methodology debates usually focus on, turns out to matter least.
The authors also identify a rater effect: a provider's overall view of a company influences how it scores that company on individual categories. Ratings are not fully independent measurements bolted together.
MSCI scores each company on a small subset of issues chosen for its industry. From 33 Key Issues across three pillars, a company is evaluated on two to seven Environmental and Social Key Issues, selected by the extent to which its industry generates large environmental or social externalities. Key Issues and weights are set across 163 GICS sub-industries and reviewed annually.
Every company is additionally assessed on the Governance Pillar, which carries a weight floored at 33% and runs on a deduction model — each company starts at 10 and loses points against best practice.
Each Key Issue combines an Exposure Score and a Management Score, so higher exposure demands stronger management to reach the same result. Controversies deduct from the Management Score, up to five points for a very severe case judged to reflect structural problems.
The individual scores roll into a Weighted Average Key Issue Score, which is then normalized against the company's industry peer set to produce the Industry-Adjusted Score on a 0–10 scale. That score maps to the seven-band letter rating from AAA down to CCC.
MSCI ratings are explicitly not absolute. MSCI states that its assessments are intended to be interpreted relative to a company's industry peers. A fossil fuel producer is rated as a fossil fuel producer. It can hold a top rating by managing industry risks better than other fossil fuel producers, and the rating is doing exactly what it was built to do when that happens.
Sustainalytics inverts the scale most people expect. Its ESG Risk Rating runs from 0 upward, and lower is better, because the number measures how much risk is left unmanaged. This is the most common misreading of any ESG rating in circulation.
The construction is subtraction. Sustainalytics estimates a company's exposure to each material ESG issue at the sub-industry level, assesses how much of that exposure management has addressed, and treats the remainder as unmanaged risk. A separate category, unmanageable risk, captures exposure inherent to the industry that no company action can remove — the emissions of an oil producer being the standard case. Summing unmanaged risk across every material issue produces the score.
Scores fall into five absolute bands: negligible below 10, low from 10 to 20, medium from 20 to 30, high from 30 to 40, and severe at 40 and above. Because the bands are absolute, they are comparable across sectors in a way MSCI's letter ratings are not.
The exclusions are specific to each ESG rating provider, which is why treating "the ratings" as one object with one blind spot repeats the error this page is arguing against.
MSCI selects Key Issues by industry, so an issue material to mining is simply not scored for a software company. Where a company does not disclose on a performance indicator, MSCI does not assume worst-case; it assigns a below-average score in the industry context.
Sustainalytics separates exposure from management, so a company with high exposure can still score well by managing it, and inherent industry exposure is carved out as unmanageable rather than counted against the company.
S&P Global runs off the Corporate Sustainability Assessment, a set of 62 industry-specific questionnaires. Roughly 70% of CSA questions require publicly available data to score any points, which rewards disclosure quality directly. For companies that do not participate, S&P applies statistical imputation to estimate performance rather than assigning zero — the S&P Global ESG Score includes that modeling, while the CSA Score is the same assessment without it.
LSEG takes the opposite approach to gaps. It states that estimated data, non-public information and AI are not used, and that a company failing to publish against a metric scores zero for it. This structurally advantages companies that report more. LSEG also applies capping metrics: a company cannot exceed 3 out of 5 on a theme without meeting a designated gateway indicator, which is intended to prevent strong disclosure from masking a missing fundamental.
The upshot is that a company can fall outside one provider's concern set entirely while being penalized by another for the same activity. Product harm is the clearest case — depending on the provider and the industry, it may be scored as a Key Issue, absorbed into a controversy deduction, or not assessed at all.
An ESG rating can be high and accurate while the company behind it does things the investor holding it would rather not fund. This is not a failure of the rating. It is the rating answering its own question correctly.
The mechanism is straightforward. Two companies in the same industry face the same industry-selected issue set. Both disclose thoroughly, both hold recognized certifications, both maintain board oversight and published targets, and neither carries a severe controversy. Both rate well. Nothing in that assessment asks what share of revenue either derives from a harm category, how much either emits per dollar of revenue, or what temperature pathway either is aligned to. Those figures exist. They are simply not what the rating aggregates.
Most investors searching for real-world impact end up reading an ESG rating instead — not because they chose it, but because it is what the market makes available. Holding-level impact data changes what is answerable. Ziggma applies it to whole portfolios.
Four checks catch most errors, and they take about a minute.
Check the scale direction. MSCI runs AAA to CCC with AAA best. Sustainalytics runs 0 upward with lower better. Getting these backwards inverts the conclusion entirely.
Check what the company was scored against. MSCI is peer-relative, so the rating describes standing within an industry. Sustainalytics bands are absolute and cross-sector comparable.
Check what is material in that industry. A company is scored on the two to seven issues its industry generates, not on everything a reasonable person might care about.
Check that you are not reading it as an impact score. If the question is what the company does to the world, the rating is not the instrument, however well it answers its own question.
Impact data answers the outward-looking question an ESG rating leaves open. Ziggma's Impact Score, built on ACA Ethos data, assesses each holding across Climate Action, Resource Use, Fair Labor and Accountability. A portfolio's Global Warming Potential reports the temperature pathway its holdings are aligned to, revenue exposure to harm categories is assessed per company, and a Controversy Score tracks incidents separately.
This is impact materiality, not double materiality — what a company does to the world, not both directions at once. It complements an ESG rating rather than replacing it. You can apply either view to holdings you already own with a free Portfolio Checkup, or screen on impact metrics in the Ziggma screener.
