Over the past few weeks, I compiled ESG rating data, financed emissions disclosures, and verified Sustainalytics pillar-level data for Indonesia's four largest banks. All four are constituents of the IDX ESG Leaders index and use PCAF methodology, and all have now disclosed their financed emissions data. A consistent pattern emerges, one that deserves closer attention than it has so far received.
The pillar data
Below are the environmental pillar scores and financed emissions data:
| E Score | S Score | G Score | Financed Emissions | |
|---|---|---|---|---|
| Bank A | 0.79 | 3.54 | 5.14 | ~18.1M tCO₂e (FY2023, 44% portfolio) |
| Bank B | 2.63 | 8.21 | 4.68 | ~37.4M tCO₂e (FY2024, 100% productive) |
| Bank C | 3.07 | 5.54 | 4.67 | 12.4M tCO₂e (FY2025, SBTi scope) |
| Bank D | 3.51 | 5.94 | 4.81 | 44.6M tCO₂e (FY2025, productive credit) |
Pillar scores from rating summaries and bank disclosures, latest as of mid-2026 (banks anonymized). Emissions data from published sustainability reports.
On Sustainalytics' scale, lower scores represent lower risk. All four banks carry Environmental scores ranging from 0.79 to 3.51, on a scale from negligible to very low risk. Environmental is the strongest pillar across the board. In contrast, Social and Governance consistently score higher, indicating higher risk.
| Category | Score Range |
|---|---|
| Negligible | 0 to 9.99 |
| Low | 10 to 19.99 |
| Medium | 20 to 29.99 |
| High | 30 to 39.99 |
| Severe | 40+ |
Source: Morningstar Sustainalytics, "Overview of Sustainalytics' ESG Risk Ratings."
Meanwhile, financed emissions across all four banks range from 12 to 44 million tonnes of CO₂ equivalent, a spread of nearly four times between the smallest and largest figure.
Across these four banks, the Environmental score bears no visible relationship to the scale of financed emissions. Bank D carries both the highest Environmental score (3.51) and the highest reported emissions (44.6 million tonnes). To be precise about what this shows: with four institutions and disclosure boundaries that differ across them (a point I return to below), this is an illustration, not a statistical test. But the illustration does a specific job. It shows that a bank can sit anywhere in this emissions range and still receive an Environmental score between negligible and very low. The pillar and the carbon flowing through the lending portfolio appear to measure two different things.
The clearest case
Bank A offers the most detailed trajectory. Its ESG Risk Rating improved from 28.45 in 2024 to 17.5 in January 2025, then to 9.8 by August 2025: Negligible Risk, ranked 31st out of 989 banks globally, best in ASEAN. The Environmental pillar drove much of this improvement: 0.79, effectively negligible.
This is not a hollow result. The bank implemented a wide-ranging sustainability strategy, grew its sustainable financing portfolio, reduced financed emissions intensity by roughly 20%, strengthened climate risk management, and committed to Net Zero by 2060. These actions are real, and the methodology appropriately recognized them.
But approximately 18 million tonnes of CO₂ in financed emissions did not disappear because they are well-managed. Under the risk-rating architecture, a pillar score is unmanaged risk: exposure minus managed risk. For banks, exposure is set largely at the subindustry level, so the absolute volume of carbon in the lending book does little to move it. What a bank can move is the managed-risk side (policies, frameworks, targets, disclosures), and that is exactly what these four banks have moved. The methodology does not directly weight financed carbon the way it would weight, say, a coal plant's stack emissions for an energy company. Across all four banks, management is winning. The exposure remains.
The intensity distinction
Much of the improvement banks report is framed in intensity terms: emissions per unit of lending. Bank A specifically cites intensity reduction as a rating driver. Intensity reductions constitute real progress. But a bank can reduce intensity while absolute emissions stay flat, or even rise, if the lending portfolio grows. Both statements can be true at once. The scoring methodology captures the first. Whether it adequately reflects the second is where the methodology reaches its boundary, and that boundary holds across all four banks in the dataset.
The risk-rating defense
There is a defensible answer to all of this, and it deserves to be stated in its strongest form. Sustainalytics' product is a risk rating, not an impact rating. It measures financially material ESG risk to the enterprise (the risk that environmental issues damage the bank's own economic value), not the bank's footprint on the climate. On that definition, financed emissions are an impact metric. The methodology is not failing to capture them; it was never designed to.
That defense holds only if financed emissions are not themselves a channel of financial risk. I do not think that position survives contact with the transition scenarios regulators now ask banks to model. Carbon embedded in a lending portfolio is exposure to carbon pricing, to taxonomy-driven repricing, and to the credit deterioration of high-emission borrowers. That is transition risk, and transition risk is squarely financial. If tens of millions of tonnes of financed carbon can coexist with a negligible Environmental risk score, the question is not whether the rating measures impact. It is whether it measures the risk it claims to measure.
The comparability problem
Attempting to compare financed emissions across the four banks surfaced a second problem, one that is itself revealing.
Each bank calculates financed emissions using different boundaries. One covers business loans and project finance. Another covers only productive lending. A third spans five asset classes: listed equity, bonds, project finance, commercial real estate, and corporate lending. The fourth reports on roughly 44% of its total loan portfolio. Data vintages span FY2023 to FY2025. One bank labels its figure "Scope 3 emissions" rather than "financed emissions."
A bank reporting 44 million tonnes with full coverage of productive lending is not necessarily financing more carbon than a bank reporting 18 million on 44% of its portfolio. It may simply be disclosing more broadly.
This cuts both ways. It constrains the scoring system, and it constrains my own comparison above, which is why I treat the four-bank pattern as an illustration of a level mismatch, not as a cross-sectional finding. I examine how Indonesia's disclosure regime produced this state of affairs in a companion note.
This matters because it exposes a double gap in the measurement architecture. The first gap sits between Environmental pillar scores and the financed emissions they do not directly weight. The second sits between the financed emissions data that now exists and the ability to use it comparably. Even if a rating provider wanted to incorporate financed emissions into the Environmental pillar, the current state of disclosure (different scopes, coverages, vintages, and labels) would make consistent calibration across banks difficult.
The first gap is a question of scoring methodology. The second is a question of disclosure standardization. Both sit in the pipeline between what banks now report and what investors actually see.
The structural concern
There is a growing concern, one I share from practitioner experience, that sustainable finance has built a considerable structure on foundations that have not been examined proportionately. The analogy to pre-crisis credit ratings has been drawn by others, and it is imperfect, but the structural observation holds: a measurement system can be internally consistent, widely adopted, and technically defensible at every step, while still producing outcomes that do not reflect the underlying risk it was designed to capture.
Every component of the Indonesian banking ESG system functions correctly. Banks disclose. PCAF is adopted. The rating methodology is applied consistently. Fund managers follow their mandates. The IDX ESG Leaders index selects constituents according to its rules. No one is making false claims.
And yet: all four banks carry Environmental scores below 3.6 while collectively financing tens of millions of tonnes of carbon. The environmental risk signal reaching investors, the signal that drives index inclusion and capital allocation, does not reflect the climate exposure embedded in these lending portfolios.
Alignment without engagement
Much of the ESG investment infrastructure in Indonesia operates on an alignment model: it channels capital toward banks that already score well. The IDX ESG Leaders index selects constituents based on composite ESG scores. Banks that meet the criteria earn inclusion and attract the mandate-constrained capital that follows.
But if sustainable investing is fundamentally about managing financial risk, and I believe it should be, then the relevant question is whether the criteria being satisfied correspond to actual risk reduction. Environmental scores between 0.79 and 3.51 tell investors that environmental risk ranges from negligible to very low. The financed emissions data tells a different story about climate transition exposure.
The risk does not disappear because the score is low. It becomes invisible to the metrics investors rely on. That is a different problem from greenwashing, and in some ways a more structural one, because it requires no misleading claim from anyone.
The design questions
Should the Environmental pillar for banks incorporate financed emissions volume as a direct input, not only as a measure of how well that exposure is managed? Should absolute emissions be weighted alongside intensity metrics? Should pillar-level scores be disclosed publicly as standard practice, rather than locked behind expensive terminals, so investors can see that the Environmental signal is consistently the lowest-risk pillar across all four banks? And should financed emissions methodology be standardized enough to support meaningful comparison?
These are design questions: testable, answerable, and increasingly urgent as the capital flowing through ESG infrastructure continues to scale.
The Indonesian banking sector (where disclosure is advancing, methodology is not yet standardized, scoring has not kept pace, and the gap between ESG labels and climate exposure is directly observable) is a setting where deeper, data-driven research can make a real difference.
The banks are disclosing. The pillar data is there. Whether the scoring system is built to see what the banks are now showing is the question I keep returning to.
Related Reading
Sautner, Z., van Lent, L., Vilkov, G., and Zhang, R. (2023). Firm-Level Climate Change Exposure. Journal of Finance, 78(3), 1449–1498.
Berg, F., Kölbel, J. F., and Rigobon, R. (2022). Aggregate Confusion: The Divergence of ESG Ratings. Review of Finance, 26(6), 1315–1344.
Giglio, S., Kelly, B., and Stroebel, J. (2021). Climate Finance. Annual Review of Financial Economics, 13, 15–36.