The signal-to-noise problem in ESG
Rees Calder · 19 April 2026 · 6 min read
If you've ever bought an "ethical" index fund, you've trusted a small number of rating agencies to separate the good companies from the bad. That trust is doing more work than the evidence supports.
In 2022 Berg, Kölbel and Rigobon, three researchers at MIT and the University of Zurich, ran the definitive analysis of ESG ratings. They compared ratings from six major agencies (KLD, Sustainalytics, Moody's ESG (formerly Vigeo-Eiris), S&P Global (RobecoSAM), Refinitiv (Asset4) and MSCI) on a common set of companies.
The pairwise correlations ranged from 0.38 to 0.71, averaging roughly 0.54.
For context, credit ratings from Moody's and S&P correlate at roughly 0.99. ESG ratings correlate at a level closer to "mild agreement" than "measuring the same thing." If the agencies were measuring fundamentally the same underlying reality, this couldn't happen.
What that range means
A correlation around 0.5 doesn't mean "mostly agree with some noise." It means that for a meaningful fraction of companies, one rating agency will label them as ESG leaders while another will label them as laggards.
The Berg et al. paper tracks specific examples. ExxonMobil, for instance, received widely divergent ratings depending on which agency you asked. Tesla ranked very differently on environmental metrics versus social metrics, and different agencies weighted those categories differently. The same underlying company, rated by methodologies that all claim to measure "sustainability," got scores that would put it in different quartiles.
This isn't a story about which agency is right. It's a story about the field not having converged on what it's measuring.
Where the disagreement comes from
Berg et al. decomposed the disagreement into three sources.
Scope divergence (38% of disagreement): different agencies use different attribute sets. One might weight carbon emissions heavily and ignore water use. Another might weight supply-chain labour conditions but skip governance. The rating is less a measurement of sustainability than a measurement of the attributes that agency chose.
Measurement divergence (56%): even within the same attribute, agencies use different indicators. Two agencies both measuring "water stewardship" might use completely different proxy variables (total withdrawals vs. withdrawals in water-stressed regions, for example) and produce very different scores.
Weights divergence (6%): how much weight each attribute gets in the final score. Surprisingly, this is the smallest source of disagreement. The bigger issue is that agencies are measuring different things at the foundational level.
Why this matters for anyone allocating money
Three material consequences.
One: "ESG investing" as a category is noisier than its marketing. If you buy an ESG-branded fund, you're exposed to one agency's view of which companies are good, which may be radically different from another agency's view. You might be holding shares in companies that a competing methodology would classify as laggards.
Two: companies can game it. Once you know the methodologies disagree, you can design a corporate sustainability strategy that optimises for whichever rater you care about. A 2023 paper by Gibson Brandon, Krueger and Schmidt in Financial Analysts Journal found evidence that firms do strategically target specific raters to boost their ESG scores, without underlying behavioural change.
Three: flows get misdirected. Trillions of dollars now allocate to ESG. If the signals are unreliable, those flows aren't necessarily supporting the companies doing the most to change the underlying system. They're supporting the companies best at getting rated well by the specific agency whose fund you bought.
The deeper problem
ESG ratings were supposed to be a substitute for the hard work of knowing what a company actually does. They are, as currently constructed, a substitute that's only slightly better than not having one.
The honest frame for a donor or investor is that the category label "ESG" does not carry enough information to act on. You have to go down a level. "Does this fund screen out fossil fuels" is a knowable question. "Does this fund invest in the clean-energy transition" is a knowable question. "Is this fund a good ESG fund" is currently an unanswerable question, because the term doesn't carry a stable meaning.
What actually works
Three moves that get you better signal than a ticker-level ESG score.
Look at what the fund actually holds. If the top 10 holdings contain fossil fuel majors (and many ESG funds do), you're not in the fund you thought you were. Fund fact sheets disclose this. Ten minutes of reading beats a thousand-word marketing page.
Pick funds screened for specific policies, not broad ESG scores. A fossil-fuel-free fund is a concrete screen. An "ESG leaders" fund is a composite score from a single agency that disagrees meaningfully with every other agency.
Shift some of the capital to direct action. If the thing you actually want is "my money supports the clean-energy transition," a donation to a climate-policy organisation or an investment in a specific clean-energy fund does that more reliably than hoping an ESG index represents your values.
The ESG disclosure regime will probably improve. EU SFDR and ISSB standards are converging methodologies, and the 0.61 figure may be higher in five years. But right now, an ESG rating is a signal in the same way that weather reports from six different agencies, none of whom have agreed on what weather is, would be a signal. Useful in aggregate, terrible to act on specifically.
Sources used: Berg, Koelbel & Rigobon, Aggregate Confusion: The Divergence of ESG Ratings, Review of Finance (2022), Gibson Brandon, Krueger & Schmidt, ESG Rating Disagreement and Stock Returns, Financial Analysts Journal (2023), MSCI/Sustainalytics/Refinitiv methodology documents (2023-2024), EU SFDR disclosure framework, ISSB Standards documentation (2024). Full links in the planning doc.