Credit risk is the risk that a borrower does not pay. At Meridian Bank it is measured, approved and provisioned by the Risk function, the second line of defence, which is independent of the business lines that originate loans. Corporate lending is still the largest consumer of Meridian’s capital: in a typical European universal bank credit risk accounts for roughly four fifths of risk-weighted assets, and the cost of risk (new provisions net of releases) is the line of the income statement that moves most in a downturn. Climate enters this function through three doors: the rating of each borrower, the expected credit loss booked in the accounts, and the limits set on sectors.
Where it sits. Under the Chief Risk Officer, the credit risk department is organised in three blocks. The counterparty credit analysts (sometimes called credit officers or “credit risk managers”) review every credit application above the business line’s delegated authority, propose or validate the internal rating, and sit in credit committees. The credit risk methodology team builds the rating models, the probability of default (PD), loss given default (LGD) and exposure at default (EAD) parameters, and the IFRS 9 expected credit loss (ECL) models; a separate model validation team, reporting independently, tests them every year. The portfolio management and reporting team aggregates exposures by sector and country, monitors limits and runs the watchlist. Next to them, the ESG risk team (also second line) owns the climate and environmental risk framework: sector heatmaps, the transition-risk questionnaire, the materiality assessment required by supervisors. In Finance, the IFRS 9 team (often called the “provisioning” or “impairment” team) books the ECL each quarter, with the ECL committee deciding on staging rules, scenario weights and overlays.
Who the clients are. The internal clients are the business lines: coverage bankers and loan structurers (see B01, B03) submit credit applications; the capital markets desks submit underwriting requests; Treasury and Finance consume the parameters for capital, pricing and accounting. External counterparts are the supervisors (the ECB’s Joint Supervisory Team for a significant institution like Meridian), the external auditors, who challenge the ECL every quarter, and the rating agencies, whose opinions on the bank’s clients are one input among others.
How it “earns” money. Risk does not earn revenue, but its outputs drive the price and the profit of every loan. The PD and LGD set the expected loss and the risk weight; the risk weight sets the capital; the capital and expected loss set the minimum margin at which a loan reaches the 12% RAROC hurdle (see E03). One notch of rating on a large corporate changes capital by 30 to 50% and the hurdle margin by tens of basis points. The ECL booked under IFRS 9 hits the income statement directly: moving a EUR 100m loan from stage 1 to stage 2 can multiply its provision by four or five in one quarter. Orders of magnitude for Meridian: a corporate book of EUR 60bn, a cost of risk of 25 to 40 bp a year in normal times, a stock of performing (stage 1 and 2) provisions of a few hundred million euros.
The team around a credit decision. For a large corporate such as the Castellan Cement case of this module, the file passes through: the coverage banker and the credit analyst in the business line (first line), who write the credit application; the counterparty credit analyst in Risk, who challenges the financial projections and the rating; the ESG risk analyst, who scores the transition and physical risk and checks the sector policy; the credit committee, chaired by a senior risk officer, which approves limits and the rating; and later the IFRS 9 team, which uses the rating and the scenarios to book provisions.
A realistic week. On Monday the counterparty analyst for building materials receives the annual review of Castellan Cement from the coverage team, with the carbon-cost projections prepared for the account plan (see B01). On Tuesday she rebuilds the projections in the bank’s spreadsheet model and asks the ESG risk team for the latest sector heatmap and the carbon-price paths approved by the climate scenario committee. On Wednesday the rating model is run three times, once per carbon-price path, and she drafts the rating rationale. On Thursday the ECL committee meets for the quarterly closing: the IFRS 9 team presents the model output and the proposed overlays, among them a climate overlay on cement and lime, and the external auditors attend as observers. On Friday the credit committee reviews Castellan: the rating is confirmed or changed, the limits renewed, and early-warning triggers set. Between these events, the analysts run the watchlist, answer requests from the supervisor’s on-site inspection on climate in credit risk, and review new applications.
IN ONE SENTENCE — An internal rating is the bank’s estimate, expressed as a grade on a master scale, of how likely a borrower is to default over the next year.
Every corporate borrower at Meridian receives a grade on the bank’s master scale. Each grade corresponds to a one-year probability of default calibrated on long-run default experience: AA 0.03%, A 0.07%, BBB 0.20%, BB 1.00%, B 4.00% (real master scales have 15 to 25 notches; this course uses five grades). The rating is produced by a rating model, reviewed by the analyst, possibly changed by an override, and approved by the credit committee. It is reviewed at least once a year and whenever a material event occurs.
A rating is used everywhere in the bank: in the IRB capital formula (see E02), in the pricing tool, in the delegation of authority (a BBB client can be approved at a lower level than a BB one), in limits (maximum exposure by grade), in the IFRS 9 staging test and in the ECL, and in the bank’s own disclosures. Because a single number feeds so many uses, its governance is heavy: the model is approved by the supervisor, validated every year, and overrides are logged and monitored.
How a corporate rating model works. Large-corporate models at European banks are usually “expert” or hybrid scorecards, because large corporates default too rarely to fit a purely statistical model. A typical structure has three modules.
Ratings are meant to be through-the-cycle: the grade should not swing with every quarter’s earnings, but should reflect the borrower’s resilience over a cycle. This matters for climate: a carbon cost that will appear in 2030 is not in today’s ratios, so a model fed only with last year’s accounts does not see it. It enters only if the analyst rates on projected financials, or through the qualitative module, or by override.
FORMULA
is the one-year probability of default given a financial score, here reduced to leverage (net debt over EBITDA); is the intercept, which sets the PD level of the portfolio; is the sensitivity, the change in the log-odds of default for one turn of leverage.
INTUITION — Think of the log-odds as a thermometer and leverage as the heating. Each additional turn of leverage raises the temperature by the same number of degrees, and the PD, which is what the thermometer reads, rises faster and faster as leverage gets high.
The three credit risk parameters combine into the expected loss, the average loss the bank expects on an exposure over one year:
FORMULA
is the expected loss in euros; the one-year probability of default; the share of the exposure lost if default occurs, after recoveries and costs; the amount owed at the moment of default, including the part of undrawn lines the borrower is expected to draw before defaulting.
PD. The probability of default comes from the rating. Two families of models produce it. Statistical scorecards fit a regression of default on ratios, as above. Structural models derive it from the market value of the firm. The best-known is Merton’s model (1974): the firm defaults when the value of its assets falls below its debt at maturity. If the asset value follows a geometric Brownian motion, the distance to default and the PD are:
FORMULA
is the market value of the firm’s assets today; the default point (debt due, in practice short-term debt plus a share of long-term liabilities); the expected return on assets; the asset volatility; the horizon in years; the standard normal distribution function.
This model is the cleanest way to see what a climate shock does to credit risk. A carbon cost that is not passed on to customers destroys part of the firm’s value. If it takes away a fraction of asset value, becomes , the distance to default shrinks and the PD rises. This is the approach of Bouchet and Le Guenedal (2020) and Reinders et al. (2023), presented in Roncalli’s notebook (section 13.1.1), where is the present value of future carbon costs over the value of the firm. Its main lesson is non-linearity: because the PD is the tail of a normal distribution, a moderate loss of value multiplies the PD.
INTERACTIVE LAB — Distance to default and a carbon shock to asset value
See how a loss of asset value caused by carbon costs moves the distance to default and multiplies the PD in Merton’s model.
| Parameter you can move | Range | Default |
|---|---|---|
| Asset value A0 | 3000 to 10000 EUR m | 6000 EUR m |
| Default point D | 1000 to 6000 EUR m | 3500 EUR m |
| Expected return on assets | 0 to 10 % | 5 % |
| Asset volatility | 10 to 40 % | 20 % |
| Horizon | 1 to 5 years | 1 years |
| Share of asset value lost to carbon costs (xi) | 0 to 30 % | 10 % |
| Result at default settings | Value |
|---|---|
| Distance to default, no shock | 2.84 |
| PD, no shock | 0.222 % |
| Distance to default after the shock | 2.32 |
| PD after the shock | 1.022 % |
| PD multiple | 4.60 x |
Try it. With Castellan-like inputs (assets EUR 6,000m, default point EUR 3,500m, volatility 20%), the PD is 0.22%, close to BBB. A 10% loss of value multiplies it by 4.6. Then raise the volatility to 25%: the same shock does less in relative terms, because the starting PD is already higher.
With an asset value of EUR 6,000m, a default point of EUR 3,500m, a 5% expected return and 20% asset volatility, the distance to default is 2.84 and the one-year PD 0.222%, in the BBB range. If carbon costs destroy 10% of the asset value, the distance to default falls to 2.32 and the PD rises to 1.022%, 4.6 times higher. Banks rarely use Merton’s model for rating decisions (it needs a listed share price and maps poorly to observed default rates without the empirical recalibration that commercial versions add), but empirical work using it is the main evidence that carbon intensity and credit risk are linked: Capasso, Gianfrate and Spinelli (2020) find, on 458 European firms over 2008 to 2017, that a higher carbon intensity is associated with a shorter distance to default, other things equal.
LGD. Loss given default depends on seniority, collateral and the value of the business at default. Under the foundation IRB approach (F-IRB) the bank does not estimate it: senior unsecured corporate exposures receive 40% under CRR3, with lower values for eligible collateral. Under the advanced approach (A-IRB) the bank estimates LGD from its own workout data, with a downturn adjustment; CRR3 has removed the advanced approach for large corporates (consolidated revenue above EUR 500m), so for a client like Castellan every bank in the EU is now on F-IRB or on the standardised approach. Climate reaches the LGD in two ways: physical damage to the assets or collateral (a flooded plant, a building in a wildfire zone), and loss of value of assets that the transition makes obsolete (a coal-fired kiln, an internal-combustion engine plant), the “stranded asset” channel.
EAD. For a drawn loan, EAD is the amount outstanding. For an undrawn commitment it is the drawn part plus a credit conversion factor (CCF) times the undrawn part: borrowers in distress draw their lines before defaulting. Under CRR3 the CCF on undrawn committed facilities is 40% for F-IRB and standardised banks, 50% on performance guarantees and similar transaction-related contingent items, and 100% on financial guarantees. A stressed borrower may draw more than usual; Roncalli’s notebook notes that climate effects on CCF are almost unexplored, and stress tests usually assume none.
The accounting standard IFRS 9, in force since 2018, requires the bank to book a loss allowance on every loan, undrawn commitment and financial guarantee measured at amortised cost, from the day it is granted. The allowance depends on the stage of the exposure.
| Stage | Situation | Allowance |
|---|---|---|
| 1 | No significant increase in credit risk since origination | 12-month ECL |
| 2 | Significant increase in credit risk (SICR) | Lifetime ECL |
| 3 | Credit-impaired (defaulted) | Lifetime ECL, interest on net amount |
FORMULA
is the probability of default in year given survival to the start of year ; the probability of surviving the first years, the product of for ; the discount factor at the effective interest rate, here taken at mid-year; the remaining contractual life; and are best estimates, not the regulatory values.
Three differences from the regulatory expected loss trip up newcomers. IFRS 9 parameters are point-in-time (they reflect current conditions and forecasts) and unbiased (no regulatory floor, no downturn LGD, no conservatism margin), whereas regulatory parameters are through-the-cycle and conservative. IFRS 9 requires a probability-weighted outcome over several forward-looking scenarios, typically a baseline, an upside and a downside with weights decided by the ECL committee. And the stage 2 cliff: lifetime ECL on a five-year loan is four to five times the 12-month ECL, so staging decisions drive the provision more than PD changes do.
PITFALL — “Stage 2 means the loan is bad.” It does not. Stage 2 means that credit risk has increased significantly relative to origination. A loan granted to a B-rated borrower that is still B is stage 1; a loan granted at A that is now BBB may be stage 2. The test is relative, which is why a slow climate drift can push large, investment-grade books into stage 2.
INTERACTIVE LAB — 12-month versus lifetime ECL on a simple loan
Compute the 12-month and lifetime ECL of a bullet loan and see why the stage 2 cliff is steep for long loans.
| Parameter you can move | Range | Default |
|---|---|---|
| Exposure at default | 10 to 500 EUR m | 100 EUR m |
| Annual PD | 0.05 to 5 % | 1 % |
| LGD (best estimate) | 10 to 70 % | 45 % |
| Remaining life | 1 to 10 years | 5 years |
| Effective interest rate | 0 to 8 % | 4 % |
| Result at default settings | Value |
|---|---|
| 12-month ECL (stage 1) | 0.441 EUR m |
| Lifetime ECL (stage 2) | 2.004 EUR m |
| Stage 2 / stage 1 | 4.54 x |
| Cumulative PD over the life | 4.90 % |
Try it. Set the life to 1 year: stage 1 and stage 2 coincide. Set it to 10 years: the cliff more than doubles again. Long transition loans are the ones most exposed to staging.
On a EUR 100m five-year loan with a 1% annual PD, 45% LGD and a 4% effective interest rate, the 12-month ECL is EUR 0.441m and the lifetime ECL EUR 2.004m, 4.54 times more. The cumulative five-year PD is 4.90%.
The credit application (or credit memo) is the central document. It contains: the request (limits by product, tenor, pricing, hold); the borrower’s business and strategy; historical financials and projections with a base case and a downside; the rating proposal and its rationale; structure and documentation (covenants, guarantees, security); the relationship economics and RAROC; the risks and mitigants; and, since 2021, a section on ESG factors (section 3). The rating sheet records the model output, the qualitative scores, any override and its justification. The credit committee minutes record the decision, its conditions (shorter tenor, lower hold, covenant, early-warning triggers) and dissent if any. For IFRS 9 the ECL memo of each quarter documents scenarios, weights, staging statistics and overlays, and is the basis of the audit file and of the IFRS 7 disclosures.
Systems. Ratings sit in a rating tool linked to the credit workflow; limits in a limit-management system; exposures, collateral and parameters in the risk data warehouse; ECL is computed by an impairment engine that runs PD term structures and macro scenarios at loan level. The BCBS 239 principles on risk-data aggregation apply: a climate field added to a client record (an emission figure, a transition score, a flood-zone flag) must have an owner, a source and a quality control, or the supervisor will not trust what is computed from it.
Climate is not a new risk category for a credit officer. It is a set of drivers that act through the existing parameters. Supervisors use this framing consistently: the Basel Committee’s principles for the management of climate-related financial risks (June 2022), the ECB Guide on climate-related and environmental risks (November 2020) and the EBA Guidelines on the management of ESG risks all ask banks to show how physical and transition drivers reach PD, LGD, EAD, correlations and collateral values, rather than to create a separate silo.
| Driver | Channel to the borrower | Credit parameter |
|---|---|---|
| Carbon price, phase-out of free allocation | Operating cost, margin | PD |
| Technology and demand shifts | Revenue, stranded assets | PD, LGD |
| Transition capex | Leverage during the build | PD |
| Acute physical events (flood, storm) | Damage, interruption | PD, LGD, EAD |
| Chronic physical change (heat, water) | Productivity, asset value | PD, LGD |
| Sector-wide shocks | Many borrowers at once | Correlation |
For a hard-to-abate industrial borrower in Europe, the dominant driver over a loan’s life is the carbon price combined with the end of free allocation under the EU Emissions Trading System. The CBAM Regulation (EU) 2023/956 and the amended ETS Directive multiply free allocation to cement, steel, aluminium, fertilisers and hydrogen by 97.5% in 2026, 95% in 2027, 90% in 2028, 77.5% in 2029 and 51.5% in 2030, down to zero in 2034. A Commission proposal of July 2026 would slow that schedule; it is under negotiation and this module uses the law in force (see C10).
EBA Guidelines on loan origination and monitoring (EBA/GL/2020/06). Applicable since mid-2021, they ask banks to include ESG factors in their credit risk policies, to assess borrowers’ exposure to climate and environmental risks at origination, especially in sectors with high exposure, and to consider the borrower’s transition plan where relevant. This is why every credit application at Meridian has an ESG section.
EBA Guidelines on the management of ESG risks (EBA/GL/2025/01). Adopted in January 2025 under Article 87a of the CRD, they apply from 11 January 2026 (from 11 January 2027 at the latest for small and non-complex institutions). For credit risk they ask banks to embed ESG factors, especially environmental ones, in internal risk classification procedures, rating and scoring models and collateral valuation; to combine exposure-based, sector-based and portfolio-alignment methods with scenario analysis; and to run the materiality assessment over short, medium and a long-term horizon of at least ten years. They also require the prudential transition plans that CRD6 introduced, consistent with the bank’s CSRD transition plan.
ECB expectations and enforcement. The ECB Guide of 2020 contains thirteen expectations; expectation 8 asks banks to consider climate risks at all stages of the credit process, including in ratings and collateral valuation. After the 2022 thematic review the ECB set deadlines ending in 2024 for full alignment and has used binding decisions with periodic penalty payments for late delivery: EUR 187,650 on ABANCA, published on 10 November 2025, for failing to complete the materiality assessment by the deadline, and periodic penalty payments of EUR 7,551,050 on Crédit Agricole, announced on 13 February 2026, for failing to complete the materiality assessment required by an ECB decision of 8 February 2024 during 75 days in 2024. The ECB’s supervisory priorities for 2026 to 2028, published in late 2025, keep climate and nature risks among the central themes and announce a review of prudential transition plans.
IRB rules. CRR3 (Regulation (EU) 2024/1623, applicable since January 2025) did not create a climate factor in the risk-weight formula. Climate enters capital only through the inputs: a lower rating raises the risk weight, a lower collateral value raises the LGD for secured exposures. The EBA’s 2023 report on the role of environmental and social risks in the prudential framework concluded that the IRB framework can capture these risks through ratings, parameters and collateral valuation, and recommended clarifications rather than new factors; the debate on supporting or penalising factors is covered in E02.
IFRS 9 and climate. IFRS 9 contains no climate rule, but its requirement to use “reasonable and supportable forward-looking information” applies. The IASB’s educational material on the effects of climate-related matters on financial statements (November 2020, updated in 2023) and the illustrative examples on disclosures about uncertainties published on 28 November 2025 make the point that material climate risks must be reflected in estimates and disclosed. The ECB’s report “IFRS 9 overlays and model improvements for novel risks” (July 2024) found that banks often handled climate, geopolitical and energy risks through top-level overlays and asked them to move towards in-model treatment where possible, with granular identification of the drivers and documented, time-limited overlays meanwhile.
Banks use three families of methods, in increasing order of granularity.
FORMULA
is the unconditional (baseline) probability of default; the realisation of the climate risk factor in the scenario, in standard deviations, with positive values meaning an adverse transition; the sensitivity (correlation) of the borrower’s asset value to that factor; and the standard normal distribution function and its inverse.
PITFALL — A conditional PD at a given is the PD in that scenario, not an expected PD. Plugging a stressed PD into an ECL as if it were the average outcome overstates the provision; IFRS 9 wants a probability-weighted mean across scenarios. The opposite error is common too: the PD at is not the baseline PD but is lower than it, because the baseline already averages over all values of .
The three global agencies say that ESG factors are part of their credit analysis when, and only when, they are material to creditworthiness, and they publish how. Fitch has published ESG Relevance Scores since 2018, on a scale of 1 to 5 that measures how relevant an E, S or G factor is to the rating, not whether the issuer is “good”. Moody’s publishes issuer profile scores for E, S and G and an overall ESG Credit Impact Score (CIS-1 to CIS-5) that states how much ESG considerations affected the rating. S&P Global Ratings published ESG credit indicators from 2021 but stopped in 2023 and now explains ESG influences in narrative paragraphs in its reports. Fitch updated its general ESG approach in January 2026.
The empirical evidence is that agencies do price transition risk, a little. Using S&P ratings of more than 2,200 firms over 2005 to 2022, Borsuk et al. (2025) find that moving from a low- to a high-emission profile is associated with about a quarter of a notch lower rating, an effect softened by strong balance sheets and credible net-zero commitments. An ECB blog of November 2025 concluded that climate considerations still explain a small share of rating changes. The EU ESG Ratings Regulation (EU) 2024/3005, which applies from 2 July 2026 and puts ESG rating providers under ESMA authorisation, does not cover credit ratings issued under the CRA Regulation, which remain under ESMA’s separate CRA supervision.
For a bank’s credit officer the practical rules are: an agency rating is an input, not a substitute for the internal rating (the EBA guidelines insist on own assessment); a vendor ESG score is not a credit score and should never feed the PD model directly (see A03); and the agencies’ horizon (two to three years of projections) is usually shorter than the life of a transition loan.
Practice among large European banks has converged on a layered approach. Almost all have sector heatmaps and a transition-risk questionnaire in the credit file. Many have added a qualitative “climate” factor or a capped override in the corporate rating model rather than recalibrating the statistical part, because default data with climate signal do not exist. A minority re-rate their largest high-emitting clients under scenarios. In IFRS 9, climate is mostly handled through post-model adjustments on exposed sectors, with a slow move towards scenario variables (carbon price, energy prices) in the forward-looking models, which the ECB pushes. The NGFS scenarios (long-term Phase V of November 2024, short-term scenarios of May 2025) are the common reference for narratives and carbon-price paths. Two open questions dominate: how to reconcile the 30-year horizon of climate scenarios with the 3 to 5 years of a corporate loan, and how much weight to give a disorderly transition when policy itself is being revised.
The “deal” of a credit risk function is the annual review of a borrower and the quarterly provisioning cycle that follows. The flow below follows Castellan Cement’s 2026 review from the request of the coverage team to the booking of the year-end ECL.
INTERACTIVE LAB — Annual review of a hard-to-abate borrower, from file to provision
Follow the documents, committees and controls that turn a carbon-price assumption into a rating, a limit and a provision.
| # | Who | Does what | Deliverable |
|---|---|---|---|
| 1 | Climate scenario committee (Risk, Economics, ESG risk) | Approves the carbon-price paths and their narratives for the year, derived from NGFS scenarios | Scenario note with three price paths and weights |
| 2 | Coverage banker and first-line credit analyst | Prepare the annual review: financials, projections, carbon-cost model, transition plan assessment | Draft credit application |
| 3 | ESG risk analyst | Updates the sector heatmap score, checks the sector policy and writes the climate and environmental opinion | ESG risk opinion, transition score |
| 4 | Counterparty credit analyst (Risk) | Rebuilds the projections under each carbon-price path and runs the rating model on each | Rating sheet with scenario ratings |
| 5 | Counterparty credit analyst (Risk) | Proposes the final rating, any override and early-warning triggers | Rating rationale and override log entry |
| 6 | Credit committee | Approves the rating, limits, conditions and the next review date | Committee minutes and limit update |
| 7 | Credit risk methodology | Translates the scenario ratings into PD term structures for the impairment engine | Scenario PD curves by counterparty |
| 8 | IFRS 9 team (Finance) | Runs the impairment engine, applies the staging rules and computes model ECL | Model ECL and staging report |
| 9 | IFRS 9 team with ESG risk | Quantifies the climate post-model adjustment for exposures the model does not capture | Overlay proposal with method and release conditions |
| 10 | Model validation | Reviews the overlay method, the scenario PDs and the sensitivity analysis | Validation note |
| 11 | ECL committee (CFO, CRO, chief accountant) | Approves scenarios, weights, staging and overlays for the closing | ECL committee minutes; booked allowance |
| 12 | External auditor and Joint Supervisory Team | Challenge the overlay and the link between climate scenarios, ratings and ECL | Audit file; supervisory findings if any |
| 13 | Finance reporting | Discloses ECL, overlays and climate sensitivities in the IFRS 7 notes and Pillar 3 | Annual report and Pillar 3 ESG tables |
Scenarios first (step 1). Ratings and provisions must use the same view of the future, or the bank says one thing to the credit committee and another to the auditors. At Meridian a climate scenario committee approves, each autumn, the carbon-price paths used across the bank, derived from the NGFS scenarios and from the forward curve. In 2026 it adopted three paths to 2030 for EU allowances: held at EUR 85 (“current policies”), rising to EUR 130 (“orderly”) and rising to EUR 180 with a demand shock (“disorderly”), with weights of 50%, 30% and 20%. The weights are a judgement documented in the note; they are the most debated number of the exercise.
The credit application (steps 2 and 3). The coverage team writes the business case; the first-line analyst writes the financial analysis. For high-emitting sectors, Meridian’s credit policy requires a carbon-cost model (emissions, free allocation, price, pass-through) in the projections and a transition plan assessment, which the coverage team already produced for the account plan (see B01). The ESG risk analyst adds an opinion: sector heatmap score, compliance with the sector policy, controversies, physical-risk screening of the main sites from their geolocation, and the transition score. The opinion is binding on policy questions (an exclusion cannot be overridden by the committee) and advisory on credit questions.
The rating (steps 4 and 5). The counterparty analyst in Risk does not accept the business line’s projections as given. She rebuilds them, runs the model on each scenario and decides which result becomes the rating. Meridian’s rating policy says the rating is assigned on the base case, that is the “current policies” path, and that the other scenarios inform two decisions: whether an override is justified, and which early-warning triggers to set. Any override is logged with its reason; validation monitors the override rate by sector, and a rate above 10% triggers a review of the model.
The committee (step 6). The credit committee approves the rating together with the limits. Its minutes record the climate discussion, which supervisors now read in on-site inspections: they look for evidence that the climate section changed a decision (a tenor, a hold, a covenant, a trigger) rather than being filled in after the decision.
From rating to provision (steps 7 to 11). Methodology turns the scenario ratings into PD term structures; the impairment engine computes model ECL each quarter. Because Meridian’s forward-looking model uses macro variables (GDP, unemployment, interest rates, house prices) that do not carry the carbon-cost cliff of a cement producer, the IFRS 9 team proposes a post-model adjustment (overlay) for exposures where the model misses the risk. The overlay needs a documented method, an owner, a validation opinion and a release condition, the four points the ECB looks for. The ECL committee approves it at the quarterly closing.
Outside eyes (steps 12 and 13). Auditors test whether the overlay is reasonable and supportable and whether it double-counts the model. The Joint Supervisory Team, in its credit risk and climate reviews, tests consistency: the same scenario in the ICAAP, the stress test, the ratings and the ECL. The bank discloses the overlay in the IFRS 7 notes, usually with a sensitivity to scenario weights.
Question. Castellan Cement’s annual review is due. Its rating, BBB, was set on 2025 accounts that contain almost no carbon cost. What happens to the rating, to the provision and to the capital if the carbon price and the free-allocation phase-out are projected over the life of Meridian’s exposures?
Data (case assumptions, as in B01, C02 and C10). Castellan is a fictional listed European cement group. All its production is in the EU.
| Item | Value |
|---|---|
| Revenue; EBITDA (2025) | EUR 4,200m; EUR 900m |
| Net debt; leverage | EUR 2,160m; 2.4x |
| Cementitious production; clinker | 20 Mt; 14.8 Mt |
| Scope 1 intensity; emissions | 590 kg/t; 11.8 Mt |
| 2030 intensity target (plan) | 480 kg/t |
| Plan capex net of grants, 2026-2030 | EUR 720m |
| Internal rating | BBB (PD 0.20%) |
Meridian’s exposures, as in the account plan, at the reporting date of 31 December 2026:
| Facility | Amount | Remaining life | EAD |
|---|---|---|---|
| RCF hold, 10% drawn | EUR 150m | 2 years | 15 + 40% x 135 = 69 |
| Performance guarantees | EUR 40m | 3 years | 50% x 40 = 20 |
| Term loan, drawn | EUR 60m | 4 years | 60 |
| Total | EUR 250m | EUR 149m |
The exposure at default is computed with the CRR3 conversion factors, which Meridian also uses as its IFRS 9 best estimate for these products (a simplification: many banks estimate behavioural drawdown rates for IFRS 9). The IFRS 9 LGD is set at 40%, equal to the F-IRB value, again a simplification stated in the ECL memo. All three facilities were granted in 2023, when Castellan was already rated BBB.
The starting point. At BBB, the regulatory expected loss is and Meridian’s IFRS 9 model, which uses macro scenarios only and keeps Castellan’s PD at 0.20% a year, books a 12-month ECL of EUR 0.117m (the difference is the discounting at the effective interest rate of 3.5%). The relationship RAROC is 9.91% (see B01 and section 6).
Reading. The rating model has seen Castellan’s 2025 accounts, in which the net carbon cost is small because free allocation still covers most emissions. The question of the review is not whether the climate matters (B01 and C10 have shown that the 2030 carbon bill is 64% of today’s EBITDA before pass-through) but whether it matters within the life of Meridian’s exposures, and through which parameter.
What the analyst asks for. Before running anything, the counterparty analyst, Hélène Marchetti, sends the coverage team a data request. She wants the verified ETS emissions of each plant for the last three years, not the group total in the sustainability report; the free allocation actually received, plant by plant, from the Union registry; the board-approved capex plan with the split between approved and unapproved items; the hedging position in allowances (C10 shows a layered forward programme with Meridian); and the clauses of the existing facilities that could be affected, such as a leverage covenant at 3.5x in the RCF. The covenant matters: it is the level at which the lenders recover control, and path 3 crosses it. The coverage team’s carbon model from the account plan is a starting point, but the analyst records every assumption she changes, because the committee will ask whose numbers they are.
Question. What are Castellan’s carbon cost, EBITDA and leverage in each year to 2030 under the three paths approved by the climate scenario committee?
Data. The carbon-cost model of B01 is reused without change: , with the 0.656 clinker benchmark applied to 14.8 Mt of clinker (free allocation of 9.709 Mt before the CBAM factor) and the CBAM factors of the law in force. Price rises linearly from EUR 85 in 2026 to the 2030 level of the path. Selling prices pass through a share of the cost increase borne by an unabated producer (one at 590 kg), because the least efficient plants set the market price and CBAM puts the same cost on imports. EBITDA is stated at the 2026 carbon cost of the unchanged plant, as in B01. Net debt grows by the plan’s net capex, EUR 144m a year, to EUR 2,880m in 2030 (a conservative simplification: operating cash flow after dividends is assumed to cover maintenance capex and interest only).
| Path | 2030 price | Pass-through | 2030 intensity | Capture opex |
|---|---|---|---|---|
| 1 Current policies | EUR 85 | 75% | 480 kg (plan delivered) | EUR 45m |
| 2 Orderly | EUR 130 | 75% | 480 kg (plan delivered) | EUR 45m |
| 3 Disorderly | EUR 180 | 50% | 505 kg (capture delayed) | none |
The disorderly path combines a sharp late tightening of carbon policy with a construction downturn that weakens pricing power, and a delay of the capture unit, whose storage chain Castellan does not control. The plan’s other levers (clinker substitution, fuels) are delivered in all three paths, which is why the intensity reaches 505 kg rather than 480 kg.
FORMULA
is EUR 900m; the pass-through rate; the carbon cost of an unabated producer at 590 kg in year ; Castellan’s own carbon cost; the operating cost of transporting and storing captured CO2 (EUR 45m a year from 2030 if the unit runs); the net debt and the leverage.
Calculation (2030). Castellan’s net short is Mt in paths 1 and 2 and Mt in path 3. The unabated producer’s short is 6.800 Mt in all paths. In path 2, at EUR 130: Castellan’s carbon cost is EUR 598.0m; the unabated cost is EUR 884.0m, so the price increase passed through is ; EBITDA is . In path 3, at EUR 180: Castellan pays EUR 918.0m; pass-through is ; EBITDA is .
INTERACTIVE LAB — Castellan’s leverage under three carbon-price paths
Project carbon cost, EBITDA and leverage to 2030 for each path and read the scenario rating from the peak leverage.
| Parameter you can move | Range | Default |
|---|---|---|
| Path 1: 2030 allowance price | 40 to 250 EUR/t | 85 EUR/t |
| Path 1: pass-through | 0 to 100 % | 75 % |
| Path 1: 2030 intensity | 480 to 590 kg/t | 480 kg/t |
| Path 2: 2030 allowance price | 40 to 250 EUR/t | 130 EUR/t |
| Path 2: pass-through | 0 to 100 % | 75 % |
| Path 2: 2030 intensity | 480 to 590 kg/t | 480 kg/t |
| Path 3: 2030 allowance price | 40 to 250 EUR/t | 180 EUR/t |
| Path 3: pass-through | 0 to 100 % | 50 % |
| Path 3: 2030 intensity | 480 to 590 kg/t | 505 kg/t |
| Plan capex net of grants, 2026-2030 | 0 to 1500 EUR m | 720 EUR m |
| Result at default settings | Value |
|---|---|
| 2030 carbon cost, path 1 | 391.0 EUR m |
| 2030 carbon cost, path 2 | 598.0 EUR m |
| 2030 carbon cost, path 3 | 918.0 EUR m |
| 2030 EBITDA, path 1 | 947.1 EUR m |
| 2030 EBITDA, path 2 | 969.6 EUR m |
| 2030 EBITDA, path 3 | 693.2 EUR m |
| 2030 leverage, path 1 | 3.04 x |
| 2030 leverage, path 2 | 2.97 x |
| 2030 leverage, path 3 | 4.15 x |
| Model PD at peak leverage, path 1 | 0.379 % |
| Model PD at peak leverage, path 2 | 0.353 % |
| Model PD at peak leverage, path 3 | 1.145 % |
| Scenario grade, path 1 (1 AA, 2 A, 3 BBB, 4 BB, 5 B) | 3 |
| Scenario grade, path 2 | 3 |
| Scenario grade, path 3 | 4 |
| Leverage at the BBB/BB boundary | 3.21 x |
| Headroom to the boundary, path 1 | 0.17 x |
| EBITDA headroom to the boundary in 2030, path 1 | 49.1 EUR m |
| 2030 net short, path 1 | 4.600 Mt |
| 2030 net short, path 3 | 5.100 Mt |
| 2030 price increase passed through, path 2 | 514.2 EUR m |
| 2030 price increase passed through, path 3 | 512.8 EUR m |
| 2030 carbon cost of an unabated producer, path 2 | 884.0 EUR m |
| 2030 carbon cost of an unabated producer, path 3 | 1,224.0 EUR m |
| 2027 leverage, path 1 | 2.52 x |
Try it. Put path 3 back on plan (480 kg, 75% pass-through): even at EUR 180 Castellan stays BBB. Then set path 1 to 590 kg with zero capex, the ‘do nothing’ company of B01: at EUR 85 it looks better in 2030 (2.68x) than the company that invests.
Result. In 2030 the carbon cost is EUR 391.0m on path 1, EUR 598.0m on path 2 and EUR 918.0m on path 3. EBITDA is EUR 947.1m, EUR 969.6m and EUR 693.2m; leverage 3.04x, 2.97x and 4.15x. In all three paths 2030 is the peak year: the CBAM factor falls from 77.5% to 51.5% that year, while the plan’s capex has been spent.
Reading. Three findings, two of them counter-intuitive. First, a higher carbon price is not bad for Castellan as long as the plan is delivered and prices pass through: path 2 leaves Castellan stronger than path 1, because the price of cement is set by the less efficient producers and Castellan’s cost rises less than its price. Second, the damage in path 3 comes from the combination of three adverse events (price, weaker pass-through, delayed capture), not from the price alone. Third, as B01 already showed, the company that invests is more leveraged in 2030 than the one that does nothing (3.04x against 2.68x at EUR 85): the credit benefit of a transition plan comes after the loan life of most of Meridian’s exposures.
PITFALL — A climate stress that only shocks the carbon price, holding everything else constant, gets the sign wrong for efficient producers in a sector with pass-through. The variables that drive the result are the client’s intensity relative to the marginal producer and the pass-through rate. A credit analyst should ask for both before reading any “carbon-price sensitivity”.
Question. Which grade does Meridian’s rating model give Castellan on each path, and which one becomes the rating?
Data. Meridian’s large-corporate model for building materials (fictional, reduced here to its dominant driver) maps leverage to a PD through the logistic function of section 2.1, calibrated so that 2.4x gives 0.20% (BBB), with a sensitivity of 1.0 per turn of leverage. The model PD is mapped to a grade at the geometric mid-points of the master scale: BBB up to 0.447%, BB up to 2.0%. The rating policy rates on the peak leverage of the projection over the life of the exposures, and assigns the rating on the base case (path 1). The analyst may override by one grade with a written justification.
INTERACTIVE LAB — Meridian’s leverage-to-PD model and the master scale
Map a leverage ratio to a model PD, a grade on the master scale and an IRB risk weight, with or without an analyst override.
| Parameter you can move | Range | Default |
|---|---|---|
| Net debt / EBITDA | 0.5 to 6 x | 3.04 x |
| Sensitivity of log-odds to one turn of leverage | 0.5 to 1.5 | 1 |
| Analyst override (grades, + is worse) | -1 / 0 / 1 | 0 |
| Result at default settings | Value |
|---|---|
| Model PD | 0.379 % |
| Model grade (1 AA, 2 A, 3 BBB, 4 BB, 5 B) | 3 |
| Final grade after override | 3 |
| Master-scale PD of the final grade | 0.20 % |
| F-IRB risk weight (LGD 40%, M 2.5) | 39.02 % |
Try it. Raise the sensitivity to 1.5: the same 3.04x now gives 0.52% and the grade falls to BB. The calibration of beta, which nobody can estimate from climate defaults, decides the rating.
Calculation. Path 1: peak leverage 3.04x, model PD 0.379%, grade BBB. Path 2: 2.97x, 0.353%, BBB. Path 3: 4.15x, 1.145%, BB. The leverage at which Castellan would cross into BB is 3.21x.
Result. The base-case rating stays BBB, with a PD of 0.20% on the master scale and an F-IRB risk weight of 39.02%. On path 3 the model gives BB, whose master-scale PD is 1.00% and risk weight 82.06%.
The rating decision. The analyst considers an override to BB and rejects it, for reasons she writes in the rationale. The base case already includes the plan’s capex and the phase-out schedule of the law in force; path 3 requires three adverse events at once; and the headroom to the boundary on the base case is 0.17 turn of leverage, or EUR 49.1m of 2030 EBITDA. She instead proposes early-warning triggers, which the committee will adopt: a rating review if the final investment decision on the capture unit is not taken by 31 December 2027, if projected 2030 leverage on the base case exceeds 3.2x at any quarterly update, or if the 2028 to 2030 CBAM schedule is changed. She also marks the outlook as negative in the rating tool, a field that Meridian uses for watchlist reporting but not for capital.
The rationale as written. The rating sheet’s free-text box, which validation reads when it samples overrides, says in substance: “Model grade BBB on the base case (peak leverage 3.04x in 2030, after EUR 720m of net transition capex). Orderly path BBB (2.97x): the borrower’s intensity is below the marginal producer’s and benefits from pass-through. Disorderly path BB (4.15x): requires a carbon price of EUR 180, pass-through falling to 50% and a delay of the capture unit. No override: the downside needs three concurrent adverse events, two of which are observable in advance. Negative outlook; triggers set. Next review brought forward.” Four sentences that a reader in two years can check against what happened.
Reading. The 2030 peak leverage on the base case sits at 0.17x from the boundary. A small change in the pass-through assumption or in the sensitivity , both judgemental, would flip the grade. This is normal for a borrower at the edge of a grade, but it shows where the model risk lies: not in the carbon arithmetic, which is mechanical, but in the two parameters that link it to default.
Question. Is the bottom-up result plausible against the top-down method that Meridian applies to the rest of its cement book?
Data. Meridian’s sector heatmap scores cement “very high” on transition sensitivity. For portfolio stress testing and ECL overlays the methodology team translates this score into a climate factor realisation of in the disorderly scenario, with a sensitivity of asset values to that factor (case assumptions; Roncalli’s notebook shows tables with from 1 to 3 and from 20% to 50%).
INTERACTIVE LAB — Scenario-conditional PD: shifting the climate factor
Compute the PD conditional on a climate factor realisation and compare it with the bottom-up re-rating of Castellan.
| Parameter you can move | Range | Default |
|---|---|---|
| Baseline PD | 0.03 to 5 % | 0.2 % |
| Sensitivity to the climate factor (rho) | 5 to 50 % | 20 % |
| Climate factor realisation X | -1 to 3 sd | 2 sd |
| Bottom-up stressed PD to match | 0.05 to 10 % | 1.145 % |
| Result at default settings | Value |
|---|---|
| Conditional PD | 1.328 % |
| Multiple of baseline PD | 6.64 x |
| X implied by the bottom-up PD | 1.89 sd |
Try it. Set X to 0: the conditional PD (0.065%) is below the baseline 0.20%. The baseline is the average over all values of X, not the PD in a ‘neutral’ year.
Calculation. For a BBB borrower at 0.20%, the conditional PD at is , 6.64 times the baseline. Castellan’s bottom-up path-3 PD of 1.145% corresponds to . As a check of the implementation, a PD of 0.22% at gives 0.364%, the value shown in Roncalli’s table 13.7 for BBB with the same parameters.
Reading. The two methods agree on the order of magnitude: the sector shift gives 1.33%, the counterparty re-rating 1.15%. The top-down method is slightly harsher because it treats Castellan as an average cement producer, while its intensity is below the sector’s (590 kg against 630 kg for the rest of Meridian’s book) and its plan cuts it further. That is the argument for bottom-up re-rating of the largest names: it is the only way a credible plan is rewarded. The weakness of the top-down method is also its strength: it needs no projections, so it can cover forty names in the time one is re-rated.
Question. Does the climate reach Meridian only through the PD, or also through exposure and loss severity?
EAD. In path 3 Castellan’s EBITDA falls by a quarter in a year. A borrower in that position draws on its revolving facility to fund working capital and the carbon bill, which is paid in allowances surrendered by 30 September each year. Meridian’s methodology team applies, in path 3 only, an add-on of 20 percentage points to the 40% conversion factor of the undrawn RCF, the form proposed in Roncalli’s notebook (). The RCF’s EAD rises from EUR 69m to .
LGD. Castellan’s loans are senior unsecured, so the LGD depends on the value of the group at default. In a disorderly transition the buyers of a defaulted cement group would value its kilns at their carbon-adjusted cash flows: the plants without capture are worth less. Meridian sets the IFRS 9 LGD at 45% in path 3, five points above the 40% used elsewhere (case assumption). The regulatory F-IRB LGD stays at 40% whatever the scenario: it is fixed by CRR3 for senior unsecured claims.
Physical risk. The ESG risk analyst screens the geolocation of Castellan’s eleven plants and four main quarries against flood, heat-stress and water-stress maps. One plant, on a river port, lies in a zone with a 1-in-100-year flood risk; two quarries are in areas of medium water stress, which matters for dust suppression and for the permit disputes already noted in B01. The analyst’s conclusion, written in the ESG section, is that physical risk is not material to the rating over the life of the exposures: no plant represents more than 12% of capacity, the river plant is insured, and cement demand rises after floods (reconstruction). This is a typical result for diversified industrial groups; physical risk is usually material for real estate, agriculture, utilities’ networks and single-asset project finance (see B06, B08).
NOTE — Physical and transition risks are modelled separately at most banks, which misses their interaction. A construction downturn caused by a disorderly transition can coincide with flood damage; the supervisory stress tests are only starting to combine them (see E08).
Question. Should Castellan move to stage 2, and how much ECL should Meridian book?
Data. The methodology team turns each path into a PD term structure: the PD of each year from 2027 to 2030 is the model PD at that year’s projected leverage. Scenario weights are 50%, 30% and 20%. Meridian’s SICR rule: an exposure moves to stage 2 if its lifetime PD over the remaining life, as estimated today, is at least twice the lifetime PD that was expected for the same period at origination, and at least 0.50 percentage point higher. At origination in 2023 the expected PD was 0.20% a year, so the expected four-year cumulative PD was 0.798%. Meridian’s policy applies the test to the probability-weighted PD curve. Model ECL keeps the PD at 0.20% a year; EIR 3.5%.
FORMULA
is the PD of year in scenario ; the scenario weight; the weighted PD curve used for the staging test; the ECL computed with scenario ’s PD curve, EAD and LGD, over 12 months in stage 1 or over the life in stage 2.
INTERACTIVE LAB — Castellan: staging test and probability-weighted ECL
Build PD curves from the three carbon-price paths, run the SICR test, and compare the climate-weighted ECL with the model ECL to size the overlay.
| Parameter you can move | Range | Default |
|---|---|---|
| Weight of path 1 (current policies) | 0 to 100 % | 50 % |
| Weight of path 2 (orderly) | 0 to 100 % | 30 % |
| Model sensitivity beta | 0.5 to 1.5 | 1 |
| IFRS 9 LGD, paths 1 and 2 | 20 to 60 % | 40 % |
| IFRS 9 LGD, path 3 | 20 to 70 % | 45 % |
| CCF add-on on the undrawn RCF, path 3 | 0 to 60 pp | 20 pp |
| SICR threshold (lifetime PD multiple) | 1.5 to 4 x | 2 x |
| Effective interest rate | 1 to 6 % | 3.5 % |
| Staging method (0 weighted PD, 1 scenario by scenario) | 0 / 1 | 0 |
| Result at default settings | Value |
|---|---|
| Year-1 PD, path 1 | 0.226 % |
| 2030 PD, path 3 | 1.145 % |
| Term loan: 4-year PD expected at origination | 0.798 % |
| Term loan: weighted 4-year PD | 1.272 % |
| Term loan: ratio to origination (weighted) | 1.60 x |
| Term loan: ratio to origination (path 3 alone) | 2.46 x |
| Model ECL (macro only) | 0.117 EUR m |
| 12-month ECL, path 1 | 0.133 EUR m |
| 12-month ECL, path 2 | 0.132 EUR m |
| 12-month ECL, path 3 | 0.183 EUR m |
| Lifetime ECL, path 3 | 0.761 EUR m |
| Lifetime ECL, weighted | 0.492 EUR m |
| ECL booked under the staging method | 0.143 EUR m |
| Overlay on Castellan | 0.025 EUR m |
| ECL if downgraded to BB and in stage 2 | 1.640 EUR m |
| Additional ECL on a downgrade to BB | 1.523 EUR m |
| ECL on a downgrade as a multiple of model ECL | 14.0 x |
| RCF EAD in path 3 | 96.0 EUR m |
Try it. Switch the staging method to scenario by scenario: the term loan moves to stage 2 in path 3 only, and the overlay is multiplied by more than four. Then set the weight of path 3 to 100%: only the term loan crosses the threshold, because the two shorter facilities mature before the 2030 cliff.
Calculation, PD curves. On path 1, the PD for 2027 is 0.226% (leverage 2.52x), rising to 0.379% in 2030. On path 3, the 2030 PD is 1.145%. The weighted four-year PD of the term loan is 1.272%.
Calculation, staging. For the term loan the weighted lifetime PD is times the PD expected at origination: below the multiple of 2, so the loan stays in stage 1. The RCF and the guarantees, with two and three years left, mature before the 2030 cliff and are even further from the threshold. On path 3 alone the term loan’s ratio would be 2.46, above the threshold.
Calculation, ECL. In stage 1, the 12-month ECL is EUR 0.133m on path 1, EUR 0.132m on path 2 and EUR 0.183m on path 3 (higher EAD and LGD). The weighted ECL is EUR 0.143m against EUR 0.117m from the model: the overlay on Castellan is EUR 0.025m.
Result. Castellan stays in stage 1 and the climate overlay on the name is EUR 0.025m. If Meridian staged scenario by scenario (a method used by some banks), the term loan would be in stage 2 in path 3, the ECL would be EUR 0.227m and the overlay EUR 0.110m. If Castellan were actually downgraded to BB, all three facilities would move to stage 2 (a fall from BBB to BB multiplies the PD by five) and the ECL would be EUR 1.640m, fourteen times the current model figure.
The ECL memo. The IFRS 9 team records Castellan’s overlay in the year-end memo with four elements: the driver (transition risk from the free-allocation phase-out, not captured by the macro model), the method (scenario PD curves from the counterparty re-rating, weighted 50/30/20), the measurement (weighted ECL minus model ECL, by facility), and the sensitivity (with a 100% weight on path 3, the term loan alone crosses the threshold and the ECL is EUR 0.604m). The memo also states what would release the overlay: the commissioning of the capture unit or a re-rating that incorporates the carbon cost in the model’s own inputs. The external auditor’s question at the closing meeting is predictable: why 20% for the disorderly path? The answer, documented in the scenario note, is that the committee’s weights are reviewed annually, are consistent with those used in the ICAAP, and that the disclosure shows the result under alternative weights so that readers can form their own view.
Reading. The climate overlay on one BBB name is small; the stage 2 cliff is not. What drives the provision is not the carbon price but the timing of its effect relative to the remaining life of the exposures: the cliff of 2030 falls within the life of the term loan only. A seven-year transition loan granted in 2027 (B01’s EUR 60m) would carry the whole cliff, and its staging would be decided by the same test. This is the argument for setting early-warning triggers now (stage 5.3): a downgrade to BB would add EUR 1.523m of provisions in one quarter.
PITFALL — Mixing climate scenarios and macro scenarios in the same weights double-counts. Meridian’s macro downside scenario already includes a recession; path 3 includes a construction downturn. The ECL memo must explain how the overlay is measured on top of the model (here: against the model’s own Castellan ECL, which uses a flat 0.20%), or the auditor will challenge it, rightly.
Question. What climate overlay should Meridian book on the rest of its cement book, which nobody has re-rated name by name?
Data. Outside Castellan, Meridian’s cement and lime book has an EAD of EUR 825m across about forty borrowers, with an average intensity of 630 kg per tonne (see B01). The ESG risk team classifies their transition plans with the scorecard of B01: EUR 525m to borrowers whose plans are “Credible” or “Developing”, average model PD 0.40%; EUR 300m to borrowers whose plans are “Insufficient” or absent, average PD 1.00%. The overlay uses the sector shift of stage 5.4 (, ) with the same 20% weight as path 3, and 80% on the unchanged model PD. Borrowers with insufficient plans are moved collectively to stage 2, as IFRS 9 permits when the information needed to assess SICR is available only at portfolio level; their average remaining life is three years. LGD 40%, EIR 3.5%.
INTERACTIVE LAB — Climate overlay on the rest of the cement book
Size a top-down post-model adjustment by sub-book: weighted conditional PD, collective staging, and overlay over the model ECL.
| Parameter you can move | Range | Default |
|---|---|---|
| EAD, plans credible or developing | 0 to 1500 EUR m | 525 EUR m |
| EAD, plans insufficient | 0 to 1500 EUR m | 300 EUR m |
| Model PD, first sub-book | 0.05 to 3 % | 0.4 % |
| Model PD, second sub-book | 0.05 to 5 % | 1 % |
| Climate factor X in the stress scenario | 0 to 3 sd | 2 sd |
| Sensitivity rho | 5 to 50 % | 20 % |
| Weight of the stress scenario | 0 to 100 % | 20 % |
| LGD | 20 to 60 % | 40 % |
| Remaining life, second sub-book | 1 to 7 years | 3 years |
| Effective interest rate | 1 to 6 % | 3.5 % |
| Collective stage 2 for insufficient plans (1 yes, 0 no) | 0 / 1 | 1 |
| Result at default settings | Value |
|---|---|
| Weighted PD, first sub-book | 0.814 % |
| Weighted PD, second sub-book | 1.894 % |
| Model ECL, first sub-book | 0.826 EUR m |
| Model ECL, second sub-book | 1.180 EUR m |
| Climate ECL, first sub-book | 1.680 EUR m |
| Climate ECL, second sub-book | 6.359 EUR m |
| Sector overlay | 6.034 EUR m |
| Coverage before overlay | 24.3 bp |
| Coverage after overlay | 97.4 bp |
Try it. Switch off the collective stage 2: the overlay falls from EUR 6.0m to EUR 1.9m. Most of a sector overlay is a staging decision, which is why auditors ask for its criteria first.
Calculation. Weighted PDs: for the first sub-book ; for the second, 1.894%. Model ECL: EUR 0.826m and EUR 1.180m. Climate ECL: EUR 1.680m in stage 1 for the first sub-book; EUR 6.359m in stage 2 over three years for the second.
Result. The sector overlay is EUR 6.034m. Coverage of the cement book outside Castellan rises from 24.3 bp to 97.4 bp. Without the collective move to stage 2 the overlay would be EUR 1.909m.
Reading. The overlay is dominated by one decision, collective staging of borrowers without a credible plan, and by one parameter, the weight of the stress scenario. Both are judgements, so the ECL committee asks for a sensitivity table and a release condition: the overlay on a borrower is released when it is re-rated bottom-up with scenario projections, as Castellan was. This creates the right incentive inside the bank: the coverage team that wants a lower provision on its client must bring the data for a bottom-up review.
The validation note. Model validation’s review of the sector overlay raises three points, typical of the genre. The PDs of the two sub-books are averages, so the conditional PD of the average understates the average of the conditional PDs, because the Vasicek function is convex in the PD: a loan-by-loan calculation would give a somewhat larger overlay, which the team must compute or bound. The 20% weight is borrowed from the counterparty exercise, but the forty smaller names have less hedging, less diversification and weaker access to markets than Castellan, which argues for a higher sensitivity rather than the same one. And the collective stage 2 transfer relies on the transition-plan classification, a tool built for steering (B01), not for accounting: validation asks that the criteria be written into the IFRS 9 policy so that a client cannot leave stage 2 simply because a banker re-scores its plan. The ECL committee accepts the first and third points and keeps the weight, recording the disagreement.
Question. What must the ESG section of Castellan’s credit application contain for the committee, the auditors and the supervisor?
The committee of 4 December 2026 receives a file of twenty pages; the ESG section takes two. Its content follows Meridian’s template, built on the EBA loan-origination and ESG-risk guidelines.
| Item | Castellan content |
|---|---|
| Sector and policy | Cement, heatmap “very high”; sector policy met, one controversy escalated (B01) |
| Emissions and data quality | Scope 1 11.8 Mt, scope 2 0.9 Mt; verified ETS data, limited assurance |
| Transition plan | Target 480 kg by 2030; class “Developing”, 65% (B01) |
| Carbon-cost exposure | Net short 4.600 Mt in 2030 on plan; 6.800 Mt without |
| Scenario projections | Three paths; peak leverage 3.04x, 2.97x, 4.15x |
| Scenario ratings | BBB, BBB, BB; base rating BBB, no override |
| Physical risk | One plant in flood zone; not material |
| IFRS 9 | Stage 1; overlay EUR 0.025m |
| Mitigants and conditions | Triggers; ring-fencing of new money (B01) |
The discussion. The chair, Meridian’s deputy chief risk officer, asks three questions that show what a committee does with a climate section. First: what would make path 3 the base case? The analyst names the two observable events, the capture decision and a change in the CBAM schedule, which are the triggers. Second: is the bank’s exposure growing into the cliff? Yes, if the EUR 60m transition loan proposed in the account plan is approved: its seven-year life covers 2030 and the entire free-allocation phase-out, so its own application must carry the scenario analysis, and the committee asks that its margin step-up and drawdown conditions (B01) be linked to the same triggers. Third: are the projections the client’s or the bank’s? The bank’s: the analyst used Castellan’s capex plan but her own price, pass-through and leverage assumptions, and says so.
The business line’s view. The head of industrials coverage, who attends without a vote, argues that the negative outlook will be read by Castellan’s treasurer as a signal if it ever leaks into pricing discussions, and that the bank’s own account plan supports the client’s transition. The chair’s answer is that the outlook is an internal monitoring field with no effect on price or capital, and that supporting the transition is exactly why the triggers are tied to the capture decision rather than to the carbon price: the bank is telling the client, through the conditions on new money, which event would improve its view.
The decision. The committee confirms BBB with a negative outlook, renews the limits for one year, adopts the three early-warning triggers, and requires the next review before 30 June 2028 at the latest, whatever happens. The minutes record that climate was discussed and what changed as a result: the review date, the outlook, the triggers and the conditions on new money.
WHO DOES WHAT — A weak ESG section has three signs: it describes the client’s sustainability report instead of the client’s credit risk; it has no numbers linking climate to cash flow; and nothing in the committee’s decision depends on it. A supervisor reading the file asks one question: “If this section had been blank, would the decision have been different?”
| Stage | Key figure | Value |
|---|---|---|
| 5.1 | EAD; regulatory EL; model ECL | EUR 149m; EUR 0.119m; EUR 0.117m |
| 5.2 | 2030 carbon cost, paths 1/2/3 | EUR 391.0m; 598.0m; 918.0m |
| 5.2 | 2030 EBITDA, paths 1/2/3 | EUR 947.1m; 969.6m; 693.2m |
| 5.2 | 2030 leverage, paths 1/2/3 | 3.04x; 2.97x; 4.15x |
| 5.3 | Model PD at peak, paths 1/2/3 | 0.379%; 0.353%; 1.145% |
| 5.3 | Scenario grades; rating retained | BBB, BBB, BB; BBB |
| 5.3 | BBB/BB boundary; risk weight BBB/BB | 3.21x; 39.02%/82.06% |
| 5.4 | Sector-shift PD; X implied by path 3 | 1.328%; 1.89 |
| 5.5 | Path 3 RCF EAD; LGD | EUR 96m; 45% |
| 5.6 | Term loan PD ratio, weighted; path 3 | 1.60x; 2.46x |
| 5.6 | ECL booked; overlay on Castellan | EUR 0.143m; EUR 0.025m |
| 5.6 | ECL if downgraded to BB | EUR 1.640m |
| 5.7 | Sector overlay; coverage before/after | EUR 6.034m; 24.3/97.4 bp |
| 5.8 | Committee decision | BBB, negative outlook, three triggers |
Castellan is a large corporate under CRR3: Meridian uses the foundation IRB approach, with its own PD, the regulatory LGD of 40% for senior unsecured claims, the regulatory conversion factors and an effective maturity of 2.5 years. The risk weight depends on the grade only. At BBB (PD 0.20%) it is 39.02% and Castellan’s EUR 149m of EAD generates EUR 58.14m of credit RWA. At BB (PD 1.00%) the risk weight is 82.06% and the RWA EUR 122.27m: one grade more than doubles the capital. Under the standardised approach, if Castellan carries an agency rating in the BBB range the risk weight is 75%, and 100% in the BB range; the output floor, at its final calibration of 72.5% of standardised RWA from 2030, makes the standardised figure a lower bound on what an IRB bank holds for the portfolio as a whole. The floor dampens the capital effect of a climate downgrade for floor-bound banks: their marginal capital comes from the standardised weight, which moves from 75% to 100%, not from 39% to 82% (see E02).
Climate does not change the regulatory LGD or the conversion factors under F-IRB: the path-3 increases of section 5.5 are IFRS 9 best estimates, not capital parameters. It enters capital through the PD, that is through the rating decision of section 5.3. This is why supervisors look at rating overrides and at the way climate enters the qualitative module: it is the only lever by which an IRB bank’s capital reflects transition risk for unsecured corporate lending. In Pillar 2, the ECB can and does add capital for climate risk governance shortcomings, but not for individual exposures.
The relationship economics are those of B01. Lending revenue net of funds transfer pricing is EUR 1.179m (drawn RCF margin, commitment fee net of the liquidity charge, upfront fee, term loan margin, guarantee fees), ancillary revenue EUR 1.15m, operating costs at 40% and 60% of the two, expected loss on EAD of EUR 149m, ancillary revenue carrying EUR 2.5 of RWA per euro.
| Line | BBB | BB |
|---|---|---|
| Revenues (lending + ancillary) | 2.329 | 2.329 |
| Operating costs | -1.162 | -1.162 |
| Expected loss (PD x 40% x 149) | -0.119 | -0.596 |
| Pre-tax result | 1.048 | 0.571 |
| Credit RWA + ancillary RWA | 58.14 + 2.88 | 122.27 + 2.88 |
| Allocated capital (13%) | 7.93 | 16.27 |
| RAROC (after 25% tax) | 9.91% | 2.63% |
INTERACTIVE LAB — Castellan relationship RAROC by grade
Show how the rating of the borrower drives risk weight, capital, expected loss and RAROC for an unchanged relationship, and compare provisions with the regulatory expected loss.
| Parameter you can move | Range | Default |
|---|---|---|
| Grade | A / BBB / BB / B | BBB |
| Ancillary revenue | 0 to 4 EUR m | 1.15 EUR m |
| IFRS 9 allowance on Castellan | 0 to 3 EUR m | 0.143 EUR m |
| Result at default settings | Value |
|---|---|
| F-IRB risk weight | 39.02 % |
| Credit RWA | 58.14 EUR m |
| Allocated capital (credit and ancillary) | 7.93 EUR m |
| Regulatory expected loss | 0.119 EUR m |
| Relationship RAROC | 9.91 % |
| Provisions minus expected loss (negative = shortfall) | 0.024 EUR m |
Try it. Select BB and raise ancillary revenue: even at EUR 4m a year the relationship does not reach the hurdle. A climate downgrade cannot be compensated by cross-selling.
Reading. The climate re-rating is the largest single variable in the relationship’s economics, larger than the whole pipeline of B01. At BBB the relationship returns 9.91%; at BB 2.63%; at A, the grade Castellan might reach after a successful transition, 18.71%. The re-rating also changes the incentives of the business line. A coverage banker whose client falls to BB sees the RAROC of the relationship collapse even though nothing has defaulted, and the reflex is to reduce exposure, which is the “credit supply” channel documented in the literature reviewed by Roncalli (section 13.1.2): banks that price transition risk lend less to exposed borrowers. Whether that helps the climate depends on who replaces the lender. Meridian’s answer, in B01, is to condition new money on the transition rather than to exit.
The sustainable feature of the relationship, the margin step-up on the planned transition loan and the sustainability-linked RCF, moves revenue by a few tens of thousands of euros (B01, section 6.2). The rating moves the capital by EUR 8.34m. A sustainability-linked ratchet does not change the rating; a credible plan, by keeping the base-case leverage below the boundary, does.
The funds transfer price of a loan does not depend on the borrower’s rating at Meridian: Treasury charges the liquidity premium by tenor and the contingent liquidity cost of undrawn lines (see E04). Climate reaches liquidity through the undrawn commitments. In the liquidity coverage ratio, committed credit facilities to non-financial corporates carry a 10% outflow rate, and Meridian holds high-quality liquid assets against them at a carry cost of 40 bp. The path-3 conversion-factor add-on of section 5.5 is the credit-risk version of the same phenomenon: a stressed sector draws its lines at the same time. If the cement book’s undrawn commitments are a material share of a sector concentration, the internal liquidity stress test (ILAAP) should include a sector drawdown scenario; Meridian’s does for real estate and energy, and adds cement and steel in 2026. Long transition loans pay more FTP (6 bp per year of tenor), which pushes business lines towards shorter tenors exactly when the borrower needs long money.
Booking. The ECL of Castellan’s drawn loans reduces their carrying amount; the ECL of the undrawn RCF and of the guarantees is a provision on the liability side. The changes go through the income statement as cost of risk. The overlays of sections 5.6 and 5.7 (EUR 0.025m and EUR 6.034m) are booked at the year-end closing as post-model adjustments, allocated to the exposures for the purposes of stage reporting, with the collective stage 2 move shown in the stage-transfer tables.
Disclosure. IFRS 7 requires the bank to explain the inputs, assumptions and techniques of ECL, including forward-looking information, and the reasons for significant changes in the allowance. Meridian discloses the climate overlay separately, with the scenario weights and a sensitivity (the ECL if path 3 had a weight of 100%). The IASB’s illustrative examples of November 2025 encourage exactly this kind of disclosure when climate-related uncertainty affects significant estimates.
The IRB comparison. Under CRR Article 159, an IRB bank compares its total regulatory expected loss with its credit risk adjustments (provisions) on the IRB portfolio. A shortfall is deducted from CET1; an excess can be added to Tier 2 capital up to 0.6% of credit RWA. On Castellan at BBB, provisions of EUR 0.143m exceed the expected loss of EUR 0.119m by EUR 0.024m. If Castellan were downgraded to BB and moved to stage 2, provisions of EUR 1.640m would exceed the expected loss of EUR 0.596m by EUR 1.044m. Lifetime provisions on long exposures are therefore partly “capital in the accounts”: a climate overlay increases the excess and, within the cap, Tier 2.
Classification. Castellan’s loans pass the SPPI test and are at amortised cost. A margin step-up linked to the capture unit’s commissioning, or a sustainability-linked ratchet, is a contingent feature; the IFRS 9 amendments effective from 1 January 2026 confirm that such features can be consistent with basic lending when their effect is not significant (see E05).
Pillar 3 ESG. Castellan appears in template 1 of the EBA’s Pillar 3 ESG disclosures (banking book, climate transition risk), under the cement NACE code, with its gross carrying amount, the share in stage 2, the accumulated impairment, the financed emissions and the maturity bucket. This template makes visible what the credit function does: a collective stage 2 move on the cement book appears in the bank’s public disclosure as a jump of stage 2 exposures in a carbon-intensive sector, which analysts and supervisors read as a signal of transition risk being recognised, or of its previous under-recognition.
Financed emissions. The PCAF attribution of Castellan’s emissions to Meridian depends on the drawn amount and the client’s enterprise value (see B01, E07), not on the rating. A downgrade changes nothing in financed emissions; a fall in Castellan’s share price raises them, because enterprise value falls.
Green Asset Ratio. The ratio is unaffected by ratings and provisions, except that the GAR is computed on gross carrying amounts, before ECL. The two systems, risk and taxonomy, measure different things: Castellan’s term loan may be partly taxonomy-aligned through its capex and still carry a climate overlay.
The ICAAP and the materiality assessment. The cement book’s re-rating and overlay feed the bank’s climate materiality assessment, which the EBA guidelines require over at least ten years, and the transition-risk chapter of the ICAAP. The prudential transition plan required by CRD6 uses them as evidence that the bank identifies the borrowers whose plans are insufficient, and states what it does about them: collective stage 2, a growth cap, conditions on new money.
Climate in credit risk raises a particular kind of greenwashing: not a false green label, but a risk function that appears to integrate climate (a questionnaire, a heatmap, a paragraph in every file) while nothing in ratings, limits or provisions depends on it. Supervisors call this a “tick-box” approach. The opposite failure exists too: a climate overlay so large and so loosely documented that it serves as a general reserve.
| Risk | Example | Control | Owner |
|---|---|---|---|
| Climate absent from rating | Rating on historical accounts only | Projections with carbon cost for sensitive sectors | Credit policy, Risk |
| Unjustified override | Downgrade without numbers, or none despite them | Override log, rate monitoring | Model validation |
| Model risk in the link | Beta, pass-through set to fit a result | Sensitivity analysis, benchmarking | Methodology, validation |
| Double counting | Climate and macro downturn both in ECL | Overlay measured against model output | IFRS 9 team |
| Perpetual overlay | Overlay never released or reviewed | Release conditions, quarterly review | ECL committee |
| Inconsistent scenarios | ICAAP, stress test and ECL use different paths | Single scenario committee | CRO office |
| Client data accepted as given | Client’s carbon price and pass-through | Bank’s own assumptions, stated | Counterparty analyst |
| Horizon mismatch | Cliff after the rating horizon ignored | Rate on peak over exposure life | Credit policy |
| Data quality | Emissions estimated, not reported | Data-quality score, PCAF levels | ESG risk, data office |
| Stage 2 cliff surprise | Sudden collective transfer | Early-warning triggers, watchlist | Portfolio management |
Red-flag checklist for a credit file on a carbon-intensive borrower.
Regulation and supervision (dated).
Scenarios. NGFS long-term scenarios Phase V (November 2024) with a new damage function that raises physical-risk estimates; NGFS short-term scenarios (May 2025), the first designed for a 5-year horizon closer to credit risk needs; updated NGFS guide to climate scenario analysis for supervisors (November 2025).
Evidence on pricing and ratings. Capasso et al. (2020): higher carbon intensity, shorter distance to default for European firms. Borsuk et al. (2025): high-emission firms rated about a quarter-notch lower by S&P. Studies of loan pricing reviewed by Roncalli (section 13.1.1) generally find a carbon premium on loan spreads that is real but small, a few basis points to a few tens of basis points depending on sample and period, larger after the Paris Agreement and for firms without disclosure. The order of magnitude is consistent with the Castellan case: for a borrower that delivers its plan, the climate effect within the loan life is a fraction of a grade; for one that does not, it is a grade or more.
Practice and open debates.
INTERACTIVE LAB — Credit risk and climate: check your understanding
Test the mechanics of ratings, PD, ECL, staging and the climate channels covered in the module.
Q1. Through which parameter does climate mainly enter the regulatory capital of an unsecured large-corporate loan under F-IRB?
Answer: B. Under F-IRB the LGD and conversion factors are regulatory values and CRR3 has no climate factor; climate enters through the PD, that is the rating.
Q2. In Merton’s model, a carbon cost that destroys 10% of a BBB firm’s asset value typically…
Answer: C. The PD is the tail of a normal distribution: a moderate fall in the distance to default multiplies it, 4.6 times in the module’s example.
Q3. Which statement on IFRS 9 staging is correct?
Answer: B. The test is relative to origination; 30 days past due is only a backstop, and stage 2 uses lifetime ECL.
Q4. Why is Castellan’s term loan, not the RCF, the facility closest to the stage 2 threshold?
Answer: B. The SICR test compares lifetime PDs over the remaining life; a two-year facility does not see the 2030 jump in PD.
Q5. On path 2 (EUR 130 in 2030), Castellan’s 2030 leverage is lower than on path 1 (EUR 85). Why?
Answer: B. With 75% pass-through of an unabated producer’s cost increase, an efficient producer gains margin as the price rises.
Q6. A conditional PD computed with the Vasicek formula at X = 2 should be used in the ECL as…
Answer: B. It is the PD in a given state of the climate factor; IFRS 9 wants a probability-weighted average across scenarios.
Q7. Which agency stopped publishing numerical ESG credit indicators in 2023?
Answer: C. S&P replaced its ESG credit indicators with narrative paragraphs; Fitch keeps ESG Relevance Scores and Moody’s its CIS scores.
Q8. What does the EU ESG Ratings Regulation (EU) 2024/3005 do to credit ratings?
Answer: B. The regulation authorises and supervises ESG rating providers from 2 July 2026; credit ratings are outside its scope.
Q9. Which horizon do the EBA Guidelines on ESG risk management require for the materiality assessment?
Answer: C. The guidelines ask for short, medium and a long-term horizon of at least ten years.
Q10. The sector overlay of the case falls from EUR 6.0m to EUR 1.9m when one decision is reversed. Which one?
Answer: B. Lifetime ECL on the ‘insufficient’ sub-book is most of the overlay; staging decisions dominate provisions.
Q11. Under CRR Article 159, if an IRB bank’s provisions exceed its regulatory expected loss, the excess…
Answer: B. A shortfall is deducted from CET1; an excess is eligible for Tier 2 within the 0.6% cap.
Q12. A credit file’s ESG section is weakest when…
Answer: C. Supervisors test whether climate analysis changed a decision; a section that could have been blank is the tick-box approach.
Q13. Why does the transition plan raise Castellan’s 2030 leverage compared with doing nothing at EUR 85?
Answer: B. Debt rises with capex during the build; the benefits accrue mostly after 2030.
Q14. What is the ECB’s main criticism of climate overlays in IFRS 9, in its 2024 report on novel risks?
Answer: B. The ECB asks banks to identify drivers and transmission channels and to calibrate in-model where possible, with governed overlays meanwhile.
A steel producer has net debt of EUR 1,500m and EBITDA of EUR 600m. Meridian’s model is the one of the case: with ; grade boundaries at 0.447% (BBB/BB). (a) Compute leverage and PD today. (b) A carbon-price path reduces EBITDA by EUR 120m in the peak year: re-rate. (c) Same with a reduction of EUR 150m.
Solution. (a) Leverage is x. The odds are , so the PD is 0.221%: BBB. (b) EBITDA falls to EUR 480m and leverage rises to 3.125x. The odds become and the PD is 0.412%: still BBB, but close to the boundary. (c) EBITDA of EUR 450m gives 3.333x, odds , PD 0.507%: BB. A EUR 30m difference in the projected carbon cost, 5% of EBITDA, changes the grade, the risk weight from 39.02% to 82.06%, and the capital on the exposure. The exercise shows why the analyst must document the pass-through and intensity assumptions behind the EBITDA reduction.
A EUR 50m bullet loan with five years left was granted at A (PD 0.07% a year). After a sector review its PD is now 0.25% a year. LGD 40%. Ignore discounting. Meridian’s SICR rule: lifetime PD at least twice the PD expected at origination for the same period and at least 0.50 percentage point higher. (a) Stage? (b) ECL before and after the transfer.
Solution. (a) Expected at origination over five years: . Today: . The ratio is 3.56 and the increase 0.89 percentage point: both criteria are met and the loan moves to stage 2, although 0.25% is still an investment-grade PD. (b) In stage 1 the ECL would be . In stage 2 it is , five times more (without discounting, the lifetime ECL is EAD times LGD times the cumulative PD). The relative nature of the test is what makes investment-grade books sensitive to a gradual climate re-rating.
Meridian’s steel book has EUR 200m of EAD at an average model PD of 1.00%, LGD 45%, all in stage 1. The heatmap gives steel in the disorderly scenario with , and the ECL committee gives that scenario a weight of 25%. Ignore discounting. Compute the conditional PD, the weighted PD and the overlay.
Solution. . The conditional PD is . The weighted PD is . Model ECL: . Climate ECL: . The overlay is EUR 0.548m, 27 bp of the book. If the committee also moved a part of the book to stage 2, the overlay would be several times larger, as section 5.7 shows: the method’s real lever is staging, not the PD shift.
1. How does climate risk enter a corporate PD? Through the borrower’s cash flows and balance sheet, then through the rating model. A carbon cost, a loss of demand or a transition capex programme changes projected EBITDA and leverage; the rating model, if it rates on projections, translates that into a PD. Where the model uses historical accounts, climate can only enter through the qualitative module or an override. Banks also apply sector-level shifts in stress tests and ECL overlays. The key is to state the transmission channel in numbers, not to add a generic “ESG score” to the PD.
2. Is a high carbon price bad for the credit quality of a cement producer? Not necessarily. It depends on the producer’s intensity relative to the marginal producer that sets the price, and on pass-through, which CBAM is designed to preserve. In the case, an efficient producer that delivers its plan is stronger at EUR 130 than at EUR 85. A producer above the sector average, or one in a market with imports outside CBAM, suffers. The answer is always relative.
3. What is the difference between the regulatory expected loss and IFRS 9 ECL? The regulatory EL uses through-the-cycle PD, downturn or regulatory LGD and a one-year horizon; it serves capital. IFRS 9 ECL uses point-in-time, unbiased parameters, forward-looking scenarios with probability weights, and a lifetime horizon for stage 2. The two are compared under Article 159 of the CRR: a shortfall of provisions is deducted from CET1. The same borrower can therefore have a small EL and a large ECL when its exposure is long and in stage 2.
4. What is a SICR and how would you design the test? A significant increase in credit risk since initial recognition, which moves an exposure to lifetime ECL. A good test compares lifetime PD over the remaining life with the PD expected at origination for the same period, with a relative threshold and an absolute floor to avoid noise at very low PDs. It is complemented by qualitative triggers (watchlist, forbearance) and the 30-days-past-due backstop. For climate, the question is whether scenario-weighted PDs or scenario-by-scenario staging is used, which changes the result materially.
5. Why do banks use overlays for climate, and what makes a good one? Because their ECL models are driven by macro variables that do not capture sector-specific transition shocks, and data to recalibrate the models do not exist. A good overlay has a documented method linked to identified drivers, is measured on top of the model without double counting, has an owner, a validation review and a release condition, and is disclosed with a sensitivity. A bad overlay is a round number decided at the closing.
6. How do rating agencies treat ESG? They include ESG factors when material to credit quality and explain their influence. Fitch publishes ESG Relevance Scores, Moody’s issuer profile and credit impact scores, S&P narrative paragraphs since dropping its indicators in 2023. Studies find a small effect of emissions on ratings, around a quarter-notch for high emitters. A bank should use agency views as an input and run its own assessment, as the EBA guidelines require.
7. You are the credit analyst on a cement company. What do you want to see in the ESG section of the file? The carbon-cost projection with the bank’s own price paths, free allocation schedule and pass-through; the client’s intensity against the sector and the Taxonomy threshold; the transition plan with approved and unapproved capex; the effect of the plan on leverage during the build; physical-risk screening of the main plants; controversies; and the result: scenario ratings, triggers and any overlay. I want each element to lead to a number or a decision.
8. Does a climate downgrade help the climate? Not by itself. It raises the cost of credit and reduces lending to exposed borrowers, which protects the bank and may push capital away from high emitters. But if it also starves credible transition projects of finance, real emissions may not fall, and the borrower may refinance elsewhere. The constructive answer is to differentiate within the sector: reward credible plans through bottom-up re-rating and conditions on new money, as in the Castellan case, rather than downgrading the whole sector.
Regulation and supervision
Secondary sources used for dates and practice
Academic and textbook references