Climate Stress Testing for Loan Portfolios
In our previous article, Managing Geopolitical Sanctions Risk in Trade Finance, we argued that a control is only as strong as the evidence a bank can show behind it. Climate stress testing puts the same question to a loan book. Can the lender show which borrowers would struggle after a flood or a sharp rise in carbon costs, and what that would do to its capital? For an SME, the answer arrives as new questions from its bank: where are your sites, how much energy do you use, and what is your plan?
Climate risk can reach a loan portfolio through two types of risk driver. Physical risk arises from acute events such as floods, or from chronic change such as rising temperatures. These can disrupt a borrower’s business, cause gradual deterioration, or affect the value of its collateral. Transition risk arises from the shift to a low-carbon economy and may include policy, technology and market change, such as carbon pricing. This can raise a borrower’s costs or change demand for its products (EBA, 2025b).
For bank Chief Risk Officers (CROs), the question is which method to use and how far its results can be trusted. For SME Chief Financial Officers (CFOs), it is what data their bank will ask for and how a stress test may change the credit on offer. We separate regulatory requirements, research findings and our own advisory judgment, which we mark with phrases such as “in our view” and “we recommend”.
In our view, the starting point was weak. In its 2022 supervisory climate stress test, the European Central Bank (ECB) found that around 60% of banks did not yet have a climate risk stress-testing framework, and that just 20% considered climate risk as a variable when granting loans. It reported that banks’ projected credit and market losses came to around EUR 70 billion in aggregate for the 41 banks in the bottom-up test. This covered the short-term disorderly transition and two physical risk scenarios. The ECB stated that this figure significantly understates the actual climate-related risk (ECB, 2022).
Meanwhile, borrowers already feel the physical side. In the EIB Investment Survey 2025, 68% of EU firms reported costs due to extreme events such as flooding, wildfires or storms (EIB, 2025).

What EU Rules Now Require, and When
Regulatory and supervisory baseline (EU)
The European Banking Authority (EBA) Guidelines on the management of ESG risks (EBA/GL/2025/01) apply from 11 January 2026, and at the latest from 11 January 2027 for small and non-complex institutions. They expect institutions to embed ESG risks in their risk appetite, internal controls and capital planning (ICAAP), and to set early warning indicators and thresholds with procedures to escalate breaches (EBA, 2025a).
The Guidelines on environmental scenario analysis (EBA/GL/2025/04) apply from 1 January 2027. For institutions using the Internal Ratings-Based (IRB) Approach, they fulfil the mandate in Article 177(2a) of the Capital Requirements Regulation. They specify how physical and transition risks in particular should be reflected in the scenarios used for credit risk stress testing (EBA, 2025b).
CRD VI requires the EBA to issue guidelines on criteria for setting ESG scenarios (European Parliament & Council of the European Union, 2024a).
Proportionality
In our view, two provisions shape the method. Institutions should run a thorough gap analysis of their stress testing models. Where possible, and taking into account their materiality assessment, they should apply environmental shocks directly at exposure level (EBA, 2025b).
Proportionality is explicit. Where justified by the materiality of the risks, small and non-complex institutions may rely on a predominantly qualitative approach. On the same condition, other institutions that are not large may use sensitivity analysis to test their short-term financial resilience (EBA, 2025b). Size alone is no exemption: smaller institutions are not immune, for example where exposures concentrate in ESG-sensitive sectors or in areas prone to physical risks (EBA, 2025a).
Supervisory direction (euro area)
In January 2026 the ECB said it will assess banks’ prudential transition plans and further analyse their capabilities to tackle physical risk. Its supervisory follow-up includes binding decisions where necessary (ECB, 2026).
What this means for SMEs: the EBA states that its guidelines do not penalise SME financing, that data collection efforts are targeted towards large corporates, and it added a reference to a voluntary reporting standard for SMEs (EBA, 2025a).
What the Evidence Says About Credit Losses
ECB analysis built on the 2025 EU-wide stress test finds that green investment adds to leverage and default probabilities in high energy-intensive firms. Under the stress scenario, over 2024–27, their leverage rises by an average of 4.33 percentage points and their default probabilities by a median of 91%. At system level, the extra credit losses lower banks’ CET1 ratio by 74 basis points over 2025–27 under the adverse scenario. In a separate sensitivity analysis, extreme flood events deplete CET1 by 77 basis points more than the EBA’s adverse scenario alone (Abbondanza et al., 2025).
The “Fit-for-55” exercise of the European Supervisory Authorities and the ECB concluded that, under its scenarios, transition risks alone are unlikely to threaten financial stability. For banks, first-round losses were 6.7% of starting-point exposures in the first adverse scenario, rising to 10.9% when combined with adverse macroeconomic conditions, with a large margin of uncertainty (ESAs & ECB, 2024).
At the other end, an International Monetary Fund (IMF) working paper subjects more than 2.5 million borrowers of the euro area’s largest banking groups to two carbon pricing scenarios. It finds notable variation within and between sectors, but in the more realistic scenario the aggregate impact on firms’ creditworthiness is not material. The study does not consider macroeconomic transmission channels (Löschenbrand et al., 2025).
In our view, these studies do not conflict, because they test different things. A carbon price on its own does little harm to the average borrower under realistic assumptions. The damage grows under harsher assumptions, or when other shocks are added, such as green investment that raises firms’ debt, floods or a weak economy. So always say how a figure was produced, and remember that an average can hide the borrowers most at risk.
Floods, SME Defaults and the Collateral Question
In our view, physical risk is local, which makes it an SME question. A study of securitised SME loans in six European countries uses a discrete-time survival model. It finds that coastal, flash and river floods, taken together, significantly increase SME default probabilities in Spain and France. In the Netherlands the effect is not significant, and in Italy, Belgium and Portugal default becomes less likely after a flood. The authors suggest this reflects flood infrastructure or financial support (Chen et al., 2026).
ECB analysis of six major euro area flood events since 2018 finds that loans to flooded firms are 0.7 percentage points more likely to become non-performing in the next quarter than loans to unaffected neighbouring firms. After the 2024 Valencia flood the effect reached up to 2.0 percentage points in the next two quarters (Abbondanza et al., 2025).
Collateral is the second channel. The Financial Stability Board (FSB) sets out how a physical shock to real estate could spread if insurance becomes less available, shifting risk to households, businesses or governments (FSB, 2025). For the Netherlands, IMF researchers find a non-linear relationship between flood damage and bank capital depletion, and that stronger flood defences can mitigate some impacts (Lepore & Mok, 2024).
What this means for SMEs: in our view, lenders are likely to ask more often for the address of each site and key asset, its flood exposure and the insurance in place. We also believe that an SME that can show those three items, and what it has done to adapt, gives its bank better evidence of its own flood risk than location alone.
Stress Tests Change Lending, Including to Smaller Borrowers
Climate stress tests can also change behaviour at some banks. An ECB working paper uses loan-level euro area data on the ECB’s 2022 climate stress test. Participating banks significantly reduced credit to greenhouse gas-intensive industries relative to non-participants, although the test carried no capital implications or publication of individual results. The authors say this mainly reflects differences between participating and other banks, not a clear causal effect. Among affected firms, smaller borrowers were more negatively affected in the main sample, but not in a more comparable bank sample. Only the best-performing banks in the test significantly reduced brown credit (De Cicco et al., 2025).
A study of 230 large listed European banks from 2017 to 2022 finds no average effect on borrowers’ loan financing or investment. Supervisory climate stress tests raised transparency mainly at banks already committed to climate issues, and the authors point to unintended substitution towards less committed and less tightly regulated banks (Bischof et al., 2025).
EBA guidelines list client dialogue on resilience, including any transition plan, among banks’ risk tools, with ending a relationship as a last resort (EBA, 2025a). In our view, a sector exit that removes risk from the book without reducing it in the economy should be a deliberate, documented decision, not a side effect. SME CFOs in emissions-intensive sectors should expect closer questions first, and prepare a credible transition plan before the bank asks.
Know the Limits of the Models
Three limits belong in every board paper. First, data: in 2022 the ECB found that banks often rely on proxies for emission-intensive exposures and need to step up customer engagement to obtain better data (ECB, 2022). The FSB still records gaps in data coverage, granularity, consistency, comparability and quality (FSB, 2025).
Second, models: a Bank for International Settlements (BIS) working paper identifies a major obstacle: no generally accepted industry credit risk models adjust for physical risk. It proposes extending the one-factor Vasicek model (Pozdyshev et al., 2025).
Third, scenarios: the Network for Greening the Financial System (NGFS) warns that the physical risk estimates in its Phase V long-term scenarios rest on a paper that has received academic critiques (Kotz et al., 2024, as cited in NGFS, 2025b). It also notes that the economic effects of climate change might turn out more severe than its scenarios anticipate NGFS, 2025b). Its first short-term scenarios run to 2030 and, the NGFS notes, do not alleviate banks’ responsibility to design and implement their own risk management frameworks (NGFS, 2025a).
In our view, these limits are a reason to record assumptions and model risk explicitly, not a reason to wait.
A Proportionate Path for Lenders and SMEs
For lenders, we suggest this sequence; the steps draw on the sources cited, the order is ours.
Start with materiality: identify the segments materially exposed to physical and transition risk, and record the evidence for that conclusion (EBA, 2025a).
Match the method to materiality first, then to your size: a predominantly qualitative approach, sensitivity analysis or full scenario analysis, as the proportionality provisions allow (EBA, 2025b).
Geocode borrower sites and collateral and link them to hazard data. The guidelines ask for geographical location in stress test models and, where possible, shocks at exposure level (EBA, 2025b). In our view, physical risk shocks at exposure level need this data.
Run the gap analysis of your stress testing models, and prioritise the credit risk models (EBA, 2025b).
Connect the results to the ICAAP, credit policy and early warning indicators with thresholds and escalation (EBA, 2025a).
Record model risk: proxies used (EBA, 2025b), the NGFS caveat on Phase V physical risk estimates (NGFS, 2025b).and, in our view, how each result is used in decisions.
Illustrative key risk indicators, with thresholds set by each board: the share of exposure geocoded to a hazard score; the share in high flood-hazard areas, split by insured and uninsured collateral; and the stressed default probability uplift by segment. The sources give no numeric thresholds.
For SMEs, we propose three horizons, sized to the firm:
Next 30 days: list your sites and key assets with addresses; check their flood exposure and insurance cover; estimate your energy use; and ask your bank what climate data it will need.
1–6 months: write a one-page adaptation and transition note covering what you have done, what is planned and at what cost; and run a simple cash scenario, such as a two-week site closure or a sharp rise in energy costs.
6+ months: update the note after any event or major investment, and review it with your bank at each credit review.
Conclusion
Climate stress testing is no longer an exercise for the largest banks alone. The EBA guidelines set dates, methods and proportionality. The evidence on losses is mixed for good reason: carbon pricing alone has little aggregate effect, while energy intensity, floods and macroeconomic stress together move capital and default rates materially. The models, data and scenarios all carry acknowledged limits.
In our view, this argues for starting small and documenting everything. A lender should know which segments matter, where its borrowers and collateral are, and which method it may use, and should watch how results feed into credit policy. An SME should prepare the few facts its bank will ask for before the question arrives.
Key Takeaways
EBA/GL/2025/01 applies from 11 January 2026 (small and non-complex institutions by 11 January 2027); EBA/GL/2025/04 applies from 1 January 2027,
Proportionality is explicit: qualitative approaches and sensitivity analysis are allowed, but size alone is no exemption,
Quote loss estimates with their method: 74 plus 77 basis points of CET1 (Abbondanza et al., 2025) against no material aggregate effect from carbon pricing alone (Löschenbrand et al., 2025),
Floods can push borrowers into default: map where your most concentrated exposures and their collateral sit,
Stress tests change lending: participating banks cut credit to greenhouse gas-intensive industries, and smaller borrowers were hit harder,
SMEs: keep site addresses, flood and insurance details, energy use and a one-page transition note ready
Strategic Implication
For boards and credit committees, the decision is not whether to run a climate stress test but whether the institution can explain its result. A practical board question is: which segments of our loan book are materially exposed, on what evidence, and what would we change in origination, pricing or capital if the stressed default probability for one of them breached its tolerance?
The answer should name the method used, its limits and the person who owns the scenario assumptions. In our view, a bank that can show this is better placed with its supervisor and with the SMEs that most need finance to adapt.
Ready to pressure-test your risk framework?
Amaranth Brose helps lenders and SMEs build proportionate climate scenario analysis, key risk indicators and lender-ready borrower files.
What’s Next
On 24 November 2026 we continue the Lending edition with “Circular Economy Financing - Assessing Creditworthiness of Recycling Ventures” which evidence lets a lender assess a business model that conventional credit data describes poorly?


Comments