A comprehensive explainer of the IFRS 9 staging framework, ECL methodology, SICR triggers, macroeconomic overlays, and forbearance treatment — with worked examples in the Irish bank context.
IFRS 9 replaced IAS 39 for accounting periods beginning on or after 1 January 2018. Its most significant change was replacing the incurred loss model — where provisions were only raised once a loss event had occurred — with a forward-looking expected credit loss model that requires banks to provision for losses before they materialise.
Under IAS 39, banks only recognised credit losses when there was objective evidence that a loss had already been incurred. The 2008 financial crisis demonstrated this produced provisions that were too small and arrived too late — banks were still reporting thin provisions while their loan books were deteriorating rapidly.
IFRS 9 requires banks to estimate and provision for credit losses that are expected to occur — even before any payment is missed. For loans that have deteriorated significantly, the full lifetime expected loss must be provisioned from the moment of deterioration, not just a 12-month view.
IFRS 9 classifies every financial asset into one of three stages based on its credit quality relative to origination. The stage determines both the horizon of the ECL calculation (12-month or lifetime) and how interest income is recognised on the P&L.
| Concept | Definition | Key Point |
|---|---|---|
| Expected Credit Loss (ECL) | Probability-weighted estimate of credit losses, discounted at the effective interest rate | Reflects multiple scenarios — not just the most likely outcome |
| 12-Month ECL | ECL from default events expected within 12 months of the reporting date | Applied to Stage 1 — performing assets with no significant deterioration |
| Lifetime ECL | ECL from all possible default events over the remaining life of the instrument | Applied to Stage 2 and Stage 3 — significantly deteriorated or credit-impaired |
| SICR | Significant Increase in Credit Risk since origination — the Stage 1 to Stage 2 trigger | One of the most judgemental areas in IFRS 9 application |
| Forward-Looking Information (FLI) | Macroeconomic forecasts incorporated into ECL estimates | Must be reasonable and supportable; cannot rely solely on historical data |
| Effective Interest Rate (EIR) | Rate used to discount ECL and recognise interest income over the life of the asset | Stage 3: interest applied to net carrying amount (gross less ECL) |
Every financial asset measured at amortised cost or fair value through other comprehensive income (FVOCI) must be assigned to one of three stages at each reporting date. The stage determines the ECL horizon, the interest income recognition basis, and the provisioning impact on the P&L.
ECL Horizon: 12-month ECL — expected losses from default events within the next 12 months only.
Interest income: Recognised on gross carrying amount at the effective interest rate.
Entry condition: All loans start here at origination. Asset returns here from Stage 2 when SICR conditions reverse and cure period is met.
Typical coverage: 0.1–0.5% of exposure for a well-performing loan book.
ECL Horizon: Lifetime ECL — expected losses from all possible default events over the remaining contractual life.
Interest income: Still recognised on gross carrying amount — the same as Stage 1. This is an important distinction from Stage 3.
Entry condition: SICR since origination — deterioration in credit quality relative to the position at origination, even if the asset is still performing.
Typical coverage: 1–5% of exposure; highly portfolio and macro dependent.
ECL Horizon: Lifetime ECL — same horizon as Stage 2 but typically higher severity estimates reflecting confirmed impairment.
Interest income: Recognised on the net carrying amount (gross carrying amount less ECL provision) — a significant P&L reduction vs. Stages 1 and 2.
Entry condition: Objective evidence of credit impairment — broadly aligned to regulatory default under CRR Art. 178. Default Explainer
Typical coverage: 25–60%+ of exposure depending on collateral and portfolio type.
Loans can move in both directions through the stages. Transfers are assessed at each reporting date (typically quarterly for Irish banks). The asymmetry in transfer rules is deliberate — entry to a higher stage is faster than exit.
| Transfer | Trigger | ECL Change | Exit Condition |
|---|---|---|---|
| Stage 1 → Stage 2 | SICR since origination — significant deterioration in PD or qualitative indicators | 12-month ECL → Lifetime ECL. P&L charge increases significantly | SICR conditions reverse; typically requires a period of sustained improvement |
| Stage 2 → Stage 3 | Objective evidence of credit impairment — broadly aligned to regulatory default | Lifetime ECL increases as severity assumptions move to impaired levels; interest on net carrying amount | Impairment evidence resolves; probation period met (aligned to regulatory cure — see Default Explainer) |
| Stage 3 → Stage 2 | Impairment evidence resolves; cure conditions met | ECL reduces; interest returns to gross carrying amount basis | Return to Stage 1 requires further sustained improvement and no residual SICR |
| Stage 2 → Stage 1 | SICR fully reversed; credit quality returned to origination level | Lifetime ECL → 12-month ECL. Significant provision release | No further conditions — asset remains in Stage 1 unless SICR re-emerges |
ECL is not a single number — it is a probability-weighted average of credit losses across multiple economic scenarios, discounted to present value. The three core inputs are PD, LGD, and EAD, each of which must incorporate forward-looking information.
For IFRS 9 purposes, PD must be point-in-time (PiT) — reflecting current and expected future conditions — rather than the through-the-cycle (TTC) PD used in IRB regulatory capital models. This is a critical difference.
Averages default rates across an entire economic cycle — deliberately smoothed to avoid procyclical capital swings. Stable over time; does not respond to short-term economic conditions. Used for Pillar 1 capital calculation.
Reflects current economic conditions and forward-looking forecasts. Rises in economic downturns, falls in upturns. More volatile than TTC PD — produces the procyclicality that IFRS 9 deliberately introduces to accounting (but regulators cap in capital). Banks typically derive PiT PD from TTC PD with a macro adjustment overlay.
IFRS 9 LGD must reflect expected future conditions at the time of default — not just historical averages. For collateralised exposures, this means using forward-looking collateral valuations rather than current market prices.
| Exposure Type | Key LGD Drivers | Irish Bank Consideration |
|---|---|---|
| Residential mortgage | House price forecasts at expected time of default; cost of repossession; time to resolution (18–36 months in Ireland) | Slow Irish repossession process increases LGD materially vs. other EU jurisdictions; courts-based system adds uncertainty |
| Buy-to-let mortgage | Rental yield; vacancy assumptions; house price forecasts; landlord-specific legal environment | Higher LGD than primary residence; tenant-landlord law reform adds uncertainty to resolution timelines |
| SME / Corporate | Business asset realisations; personal guarantee value; sector-specific distress discount; insolvency costs | Irish examinership process can preserve value; recovery rates vary significantly by sector |
| Unsecured retail | Borrower income and asset recovery; debt sale market pricing | Irish unsecured debt sale market has developed significantly since 2015; influences LGD calibration |
EAD is the expected balance outstanding at the time of default. For drawn term loans, this is relatively straightforward. For revolving facilities, the key challenge is estimating the credit conversion factor (CCF) — how much of an undrawn commitment will be drawn before default occurs.
SICR is the gateway from Stage 1 to Stage 2. It is assessed by comparing the credit risk of the instrument at the reporting date to its credit risk at origination. IFRS 9 provides no fixed definition — it is one of the most judgemental areas of the standard, and a primary focus of ECB and auditor scrutiny.
Most banks implement a primary quantitative test based on PD movement since origination. The most common approaches are:
SICR is triggered when the lifetime PD at the reporting date exceeds the origination PD by a fixed number of basis points — e.g. SICR if current lifetime PD > origination lifetime PD + 0.50%. Simple and auditable, but may be too blunt for high-grade portfolios where small absolute PD moves are meaningful.
SICR is triggered when lifetime PD has increased by more than a defined multiple of the origination PD — e.g. SICR if current PD > 2× origination PD. More sensitive for low-PD assets (e.g. prime mortgages) where even a small absolute increase represents a significant relative deterioration.
IFRS 9 creates a rebuttable presumption that SICR has occurred when a financial asset is more than 30 days past due. Banks can rebut this presumption only if they have reasonable and supportable information demonstrating that the 30+ DPD position does not represent a significant increase in credit risk — for example, a confirmed administrative error or a very short-term technical delay.
IFRS 9 explicitly requires that ECL estimates incorporate forward-looking information — including macroeconomic forecasts — that is reasonable and supportable without undue cost or effort. This is one of the most complex and judgemental aspects of IFRS 9 implementation, and a key area of focus for auditors and the ECB.
Because ECL is a probability-weighted estimate, banks must consider multiple economic scenarios and weight them by their probability of occurrence. A single base case is not sufficient. The typical approach used by Irish banks:
Favourable macroeconomic conditions — lower unemployment, strong GDP growth, rising property prices. Produces lower PDs and LGDs. Typically weighted 20–30%.
Central forecast from the bank's economics team or an external provider (e.g. Consensus Economics, central bank projections). Typically weighted 40–60% — the highest single weight.
Adverse macroeconomic conditions — higher unemployment, GDP contraction, property price decline. Produces materially higher PDs and LGDs. Typically weighted 20–30%.
| Variable | Link to ECL | Irish Bank Sensitivity |
|---|---|---|
| Residential house prices | Directly drives LGD on mortgage portfolios via expected collateral value at default | Very high — Irish banks have large residential mortgage books; a 10% house price fall can increase mortgage ECL significantly |
| Unemployment rate | Key driver of PD for retail/mortgage portfolios; job loss is the primary cause of consumer default | High — particularly for tracker mortgage borrowers on fixed income-sensitive repayments |
| GDP growth | Broad driver of SME and corporate default rates; revenue and cashflow sensitivity | Ireland's GDP is heavily distorted by MNC activity — modified GNI (GNI*) is a more representative indicator for domestic portfolio sensitivity |
| ECB / market interest rates | Affects debt serviceability for variable-rate borrowers; rising rates increase PD particularly for leveraged SMEs | Significant — Irish tracker mortgage books directly pass through ECB rate changes to borrowers; rapid rate rises increase default risk |
| Commercial property prices | Drives LGD on CRE-secured loans; also affects SME collateral values | Moderate — Irish CRE market has recovered strongly but remains vulnerable to office vacancy trends (remote working) and rate sensitivity |
When quantitative ECL models do not capture identifiable risks — because the risk is too recent to be reflected in historical data, or because model limitations are known — banks apply management overlays (also called post-model adjustments or PMAs). These are judgemental top-up provisions applied to model output.
Where a specific sector faces elevated risk not yet captured by borrower-level data — e.g. hospitality during COVID, office CRE facing vacancy risk — management applies a sector-wide stage migration or ECL uplift to all exposures in that sector.
Where models are known to underestimate loss — e.g. because they were calibrated in a low-rate environment and have not been updated to reflect a high-rate regime — management adds an overlay to compensate until the model is recalibrated.
For risks that are foreseeable but not yet observable in borrower behaviour — such as the lagged impact of interest rate rises on variable-rate mortgage borrowers — management may apply forward-looking overlays based on affordability analysis rather than waiting for arrears to emerge in the data.
Forbearance occurs when the bank grants a concession to a borrower experiencing or about to experience financial difficulty. Under both IFRS 9 and EBA guidelines, forborne exposures receive special treatment — they remain flagged for an extended period even after the immediate difficulty resolves, and their staging and provisioning reflects the elevated risk of re-default.
| Forbearance Type | Stage on Granting | Exit Conditions | Minimum Watch Period |
|---|---|---|---|
| Non-distressed (performing borrower) | Stage 2 — SICR triggered by the forbearance measure itself, even if borrower was Stage 1 | Demonstrated repayment ability on modified terms; no evidence of further deterioration | Minimum 12 months observation |
| Distressed (default-triggering concession) | Stage 3 — objective evidence of impairment. Regulatory default also triggered simultaneously Default Explainer | Must meet regulatory cure conditions; full probation period completed; ECL reassessed | Minimum 12 months (regulatory) + IFRS 9 judgement |
| Post-cure forborne — performing again | Stage 2 — reclassified from Stage 3 but remains flagged as forborne; not yet back to Stage 1 | Sustained repayment on modified terms; further 12-month probation to exit forborne status; then SICR reassessment for Stage 1 | Further 12 months minimum in Stage 2 before Stage 1 possible |
When the bank modifies a financial asset — whether distressed or not — it must assess whether the modification results in derecognition of the original asset (and recognition of a new one) or continuation of the original asset with an adjusted carrying amount. For most forbearance measures on performing books, derecognition does not occur. Instead, a modification gain or loss is recognised immediately in P&L:
Two illustrative Irish exposures — a residential mortgage and an SME term loan — traced through IFRS 9 staging, ECL calculation, and the impact of a macroeconomic deterioration. All figures are illustrative.
Lifetime PD = 0.4%. ECL = 12-month PD × LGD × EAD. Assume 12-month PD = 0.08%, LGD = 12% (low LTV, base case house prices), EAD = €280,000. 12-month ECL = 0.08% × 12% × €280,000 = €269. Provision: €269.
Lifetime PD has risen to 0.9% (rate rises, reduced disposable income). Relative PD increase: 0.9% / 0.4% = 2.25× origination PD. Bank's SICR threshold: 2× origination PD. SICR triggered → Stage 2. No payment missed; not 30 DPD.
Three scenarios applied across 24 remaining years. See ECL table below.
| Scenario | Weight | Lifetime PD | LGD (house price adj.) | EAD (avg.) | Scenario ECL | Weighted ECL |
|---|---|---|---|---|---|---|
| Upside | 25% | 0.6% | 8% (prices +5%) | €195,000 | €936 | €234 |
| Base | 50% | 0.9% | 12% (prices flat) | €195,000 | €2,106 | €1,053 |
| Downside | 25% | 2.1% | 22% (prices −15%) | €195,000 | €8,967 | €2,242 |
| Total Lifetime ECL (probability-weighted, discounted) | €3,529 | |||||
12-month ECL: 12m PD 0.5% × LGD 35% × EAD €750,000 = €1,313. Provision: €1,313.
Current lifetime PD = 3.6% (2× origination PD). Bank's SICR threshold: 2× origination PD. SICR triggered → Stage 2. Lifetime ECL raised: 3.6% × 35% × €750,000 × discount ≈ €8,100 (base case, 3-year remaining life).
Credit assessment confirms distressed restructuring — concession would not be offered to a performing borrower; NPV of modified cashflows below gross carrying amount. Regulatory default triggered (UTP). IFRS 9 Stage 3 simultaneously. Default Explainer — Tab 4
See calculation below. Interest income now on net carrying amount only.
| Scenario | Weight | Lifetime PD | LGD | EAD | Scenario ECL | Weighted ECL |
|---|---|---|---|---|---|---|
| Upside | 25% | 28% | 22% | €750,000 | €46,200 | €11,550 |
| Base | 50% | 42% | 35% | €750,000 | €110,250 | €55,125 |
| Downside | 25% | 65% | 52% | €750,000 | €253,500 | €63,375 |
| Total Stage 3 Lifetime ECL | €130,050 | |||||
IFRS 9 implementation had a significant impact on Irish banks given their large legacy mortgage books, residual forbearance portfolios, and the ECB's focus on provisioning adequacy as part of supervisory reviews.
AIB's large residential mortgage book — including a substantial tracker book — creates significant sensitivity to house price and unemployment scenarios in ECL models. The bank uses a three-scenario framework with explicit macro variable paths. Post-2018, AIB has focused on SICR threshold calibration to avoid excessive cliff-effect volatility.
BOI's dual portfolio (Irish and UK) requires separate macro scenario frameworks for each jurisdiction, with different house price indices, unemployment drivers, and base rate paths. Consolidation of two separate ECL models into group-level disclosures adds complexity to investor communications around provision drivers.
PTSB's balance sheet carried a high proportion of restructured and forborne mortgages post-crisis. Under IFRS 9, many of these sit in Stage 2 (performing forborne) requiring lifetime ECL — driving provision coverage ratios higher than peers despite improving payment performance. Progress in resolving the forborne book has been a key investor focus.
The COVID-19 pandemic was the first major test of IFRS 9 in practice, and it exposed several tensions in the standard's application:
EBA guidance confirmed that government-backed blanket payment moratoriums did not automatically constitute forbearance under IFRS 9 — individual borrower assessment was still required, but the systemic nature of the shock allowed some relief from automatic SICR triggers. Irish banks had to document individual borrower assessments at scale.
GDP and unemployment forecasts changed dramatically and rapidly in 2020. Banks had to update scenario weights and severities each quarter — in some cases monthly. The volatility of ECL charges in H1 2020 (large provisions) followed by significant releases in H2 2020 (as recoveries exceeded expectations) drew significant scrutiny from investors and the ECB.
The scale of uncertainty meant model outputs were considered unreliable for entire sectors (hospitality, retail, aviation). Irish banks applied large management overlays across these sectors with explicit documentation of methodology and exit criteria. Auditors required banks to demonstrate the overlays were additive to, not substitutes for, model-based ECL.
| Disclosure | What it shows | Where to find it |
|---|---|---|
| Stage migration table | Movements of gross carrying amounts and provisions between Stages 1, 2, 3 during the period. Reveals whether deterioration is accelerating. | Credit risk note in Annual Report; also in Pillar 3 report |
| ECL sensitivity analysis | How much ECL would change if macro scenarios shift — e.g. ECL if unemployment +2pp, house prices −10%. Quantifies model sensitivity. | IFRS 9 methodology note; risk section of Annual Report |
| Scenario weights and paths | The GDP, unemployment, and house price paths used in each scenario, with their weights. Allows comparison across banks. | IFRS 9 note — usually disclosed in summary form; full paths sometimes in investor presentations |
| Management overlay disclosure | Quantum and rationale of overlays applied. Increasingly required by auditors to be disclosed explicitly. | Risk section; sometimes in Half-Year Report commentary |
| Forborne exposure tables | Volume of forborne exposures by stage; performing vs. non-performing forborne split. Key indicator of legacy credit quality. | NPE and forbearance note in Annual Report; EBA templates in Pillar 3 |
| Coverage ratio by stage | ECL provisions as % of gross carrying amount by stage. Comparability across banks is limited by different portfolio mixes and macro scenario choices. | Credit risk summary tables; investor presentations |
A Probability of Default (PD) model estimates the likelihood that a borrower will default within a defined time horizon. For IFRS 9, we need both a 12-month PD and a full lifetime PD curve — reflecting the changing probability of default at each point over the remaining life of the loan.
Most retail and SME PD models are built using logistic regression — a statistical technique that combines borrower characteristics to produce a probability score. The model is trained on historical data linking borrower attributes to observed default outcomes.
A simplified Irish mortgage PD model might combine the following factors. Each factor has a coefficient estimated from historical defaults:
| Risk Factor (Xᵢ) | Description | Sign of Effect | Example Value |
|---|---|---|---|
| LTV ratio | Loan-to-value at origination or current | + (higher LTV → higher PD) | 0.75 |
| Debt-to-income ratio (DTI) | Total debt obligations / gross income | + (higher DTI → higher PD) | 3.2 |
| Employment status | Binary: 1 = employed, 0 = self-employed / variable | − (employed → lower PD) | 1 |
| Bureau score (normalised) | Credit bureau score standardised 0–1 | − (higher score → lower PD) | 0.72 |
| Loan seasoning (years) | Years since origination | − (older loan → generally lower PD) | 4 |
| Macro index (PiT adjustment) | Composite of unemployment + house prices (positive = stress) | + (macro stress → higher PD) | 0.15 |
Averages default rates across a full economic cycle. Deliberately smoothed to prevent capital requirements swinging with the economy. Stable year-on-year for a given rating grade.
A BB-rated SME might have a TTC PD of 2.5% regardless of whether we are in a boom or recession — the long-run average is what matters for capital.
Reflects current and expected future economic conditions. Rises in recessions, falls in upturns. The same BB-rated SME might have a PiT PD of 1.2% in a boom and 5.8% in a severe recession.
Banks typically derive PiT PD by applying a macro adjustment to the TTC PD: PiT PD ≈ TTC PD × Macro Scalar. The macro scalar is driven by the scenario variables (unemployment, GDP, house prices).
For Stage 2 and Stage 3 assets, IFRS 9 requires lifetime PD — the probability of default at any point over the remaining life of the loan. This is expressed as a PD curve: a series of conditional (marginal) PDs for each future period, given the borrower has survived to that point.
PD curves typically follow a characteristic shape for mortgages and term loans: marginal PD is higher in the early years (seasoning effect — new borrowers face higher default risk), reduces as the loan seasons, then rises again in later years as the borrower ages and macro uncertainty compounds. For revolving facilities the shape is flatter.
Adjust the borrower characteristics below to see how the model score and 12-month PiT PD change. Based on a simplified logistic regression model for an Irish mortgage borrower.
Loss Given Default (LGD) is the proportion of the outstanding exposure that the bank expects to lose if default occurs, after accounting for recoveries from collateral realisation, borrower assets, and any other sources. For IFRS 9 it must be forward-looking — reflecting expected future conditions at the time of default, not just historical recovery rates.
LGD is modelled as the complement of the Recovery Rate (RR): LGD = 1 − RR. Recovery is estimated by working through the sources of cash the bank expects to receive after default, netted against costs and discounted for the time it takes to receive them.
| Driver | Effect on LGD | Irish-Specific Consideration |
|---|---|---|
| Current LTV | Dominant driver — higher LTV means less collateral coverage after haircut and costs | Irish property prices highly cyclical; current LTV can move significantly between origination and default |
| House price forecast at default | Forward-looking path from now to expected resolution date, not current price | Scenario-dependent; downside scenario applies stressed HPI path (e.g. −15% to −30%) |
| Time to recovery | Longer resolution → higher discounting of recovery cashflows → higher LGD | Irish repossession process is among the slowest in Europe — courts-based, PDH protections mean 3–7 years is common for primary dwellings |
| Forced sale discount (haircut) | Discount applied to market value to reflect urgency of sale and limited buyer pool | Typically 10–15% for Irish residential; higher for less liquid areas (rural, specific property types) |
| Legal and selling costs | Reduce net recovery; higher as % for lower-value properties | Typically 4–8% of property value in Ireland; solicitor fees, receiver costs, estate agent fees |
| PDH vs. BTL / investment | PDH has additional regulatory protections; BTL can be repossessed faster | Irish PDH protections (Dunne judgment, CCMA) extend timelines materially vs. BTL; separate LGD models typically maintained |
Example: An Irish primary dwelling mortgage defaults. Outstanding balance €280,000. Current property value €350,000 (LTV 80%).
| Step | Base Case | Downside |
|---|---|---|
| Current property value | €350,000 | €350,000 |
| HPI adjustment to expected default date (+18m) | ×1.02 (+2%) | ×0.88 (−12%) |
| Expected value at default | €357,000 | €308,000 |
| Forced sale haircut (12%) | −€42,840 | −€36,960 |
| Expected gross proceeds | €314,160 | €271,040 |
| Legal / selling costs (5%) | −€14,000 | −€14,000 |
| Net recovery (before discounting) | €300,160 | €257,040 |
| Discount factor (EIR 3.5% × 3yr to recovery) | ×0.903 | ×0.903 |
| Present value of recovery | €271,044 | €232,107 |
| EAD | €280,000 | €280,000 |
| Recovery Rate (PV recovery / EAD) | 96.8% | 82.9% |
| LGD = 1 − Recovery Rate | 3.2% | 17.1% |
Adjust the inputs below to see how LGD changes across different property value scenarios and recovery assumptions for an Irish residential mortgage.
EAD is the expected outstanding balance the bank is exposed to at the point a borrower defaults. For term loans it is largely mechanical, but for revolving and committed facilities it requires modelling how much an obligor will draw down before default occurs — making it one of the more nuanced ECL inputs.
The amount already drawn and outstanding at the reporting date. For a term loan with no revolving feature this is straightforward — the amortising balance. For an overdraft or revolving credit facility it changes daily.
The amount available to draw but not yet drawn — e.g. an undrawn revolving credit facility, an overdraft limit not fully utilised, or a mortgage offer not yet drawn. This requires a Credit Conversion Factor (CCF) to estimate how much will be drawn before default.
Guarantees, letters of credit, and other contingent liabilities are not yet on the balance sheet but represent a real credit exposure. EAD includes these weighted by their probability of being called, which is itself modelled separately from PD.
| Facility Type | EAD Approach | Typical CCF | Irish Bank Consideration |
|---|---|---|---|
| Residential mortgage (term) | Amortising balance per schedule, adjusted for expected prepayments. No undrawn element once fully drawn. | N/A | EAD decreases over time as principal repays; prepayment modelling important for tracker books where overpayments are limited |
| Offset / flexible mortgage | Balance net of offset savings; revolving re-draw feature means EAD can increase. CCF applied to available re-draw headroom. | 40–70% | Re-draw features mean EAD is less predictable than standard amortising mortgages |
| Revolving credit facility (corporate) | Drawn balance + CCF × undrawn. Most significant EAD modelling challenge. Obligors in distress draw heavily on committed revolvers before default. | 60–90% | Irish SME corporates commonly have revolving working capital facilities alongside term loans. CCF materially increases EAD vs. current drawn balance |
| Overdraft | Current balance + CCF × headroom to limit. Highly variable intra-month usage makes modelling complex. | 50–75% | Overdraft usage patterns vary significantly by sector; retail overdrafts tend to be near-limit when obligor in financial difficulty |
| Credit cards | Current balance + CCF × available credit. Significant drawdown behaviour observed pre-default. | 75–95% | High CCFs observed — cardholders exhaust available credit before default in the majority of cases |
| Loan commitments (undrawn) | CCF applied to the full committed but undrawn amount. If the loan has not yet been drawn at all, EAD = CCF × commitment. | 30–70% | Important for construction and development loans where draw-down is phased; CCF reflects expected phasing to completion at time of default |
| Financial guarantees | Full amount of the guarantee × probability of being called. Separate assessment from PD of the guaranteed party. | Bespoke | Intragroup guarantees are common in Irish SME structures; default of one entity may call guarantees across the group |
For Stage 2 and Stage 3 assets requiring lifetime ECL, EAD must be projected across each future period — not just taken as today's balance. This means modelling the expected balance at each point in the remaining life of the instrument.
EAD reduces over time following the amortisation schedule, adjusted for expected prepayments (which reduce EAD faster) and any expected drawdowns on revolving elements. For a standard Irish mortgage, EAD in year 10 is simply the expected outstanding balance at that point given the repayment schedule.
For revolving facilities with no fixed maturity, the expected lifetime is modelled behaviourally — how long is this type of facility typically maintained? EAD in each future period reflects the expected average utilisation of the facility at that time, which may differ from today's utilisation.
An Irish SME has two facilities. The bank is calculating EAD for ECL purposes at the current reporting date.
| Facility | Limit | Drawn Balance | Undrawn | CCF | EAD |
|---|---|---|---|---|---|
| 5-year term loan | €800,000 | €620,000 | €0 | N/A | €620,000 |
| Revolving credit facility | €250,000 | €110,000 | €140,000 | 75% | €215,000 |
| Total EAD | €730,000 | €835,000 |
Calendar provisioning is an ECB supervisory expectation that requires banks to provision for non-performing exposures (NPEs) according to a time-based schedule — regardless of the individual loan's expected recovery. It acts as a backstop against banks holding under-provisioned NPEs indefinitely.
The ECB's calendar provisioning schedule defines minimum provision coverage levels as a function of how long an exposure has been classified as non-performing. The schedule differs by whether the exposure is secured or unsecured — reflecting the additional time needed to realise collateral.
| Years in NPE Status | Unsecured — Min. Coverage | Secured (other) | Secured (immovable property) |
|---|---|---|---|
| Year 1 | 0% | 0% | 0% |
| Year 2 | 35% | 25% | 25% |
| Year 3 | 100% | 35% | 25% |
| Year 4 | 100% | 55% | 35% |
| Year 5 | 100% | 70% | 55% |
| Year 6 | 100% | 80% | 70% |
| Year 7 | 100% | 100% | 80% |
| Year 9 | 100% | 100% | 100% |
At each reporting date, the bank compares its IFRS 9 ECL provision for each NPE against the calendar provisioning minimum for that exposure's age in NPE status. An NPE that has been non-performing for 4 years secured by property requires 35% coverage under the calendar schedule.
If the IFRS 9 ECL provision (say 18% of EAD) is below the calendar minimum (35%), there is a shortfall of 17% of EAD. The bank is under-provisioned relative to the supervisory expectation — not against IFRS 9, which might support the 18% figure based on expected recovery.
The ECB includes the shortfall as a Pillar 2 requirement (P2R) in the bank's SREP decision. The bank must hold additional CET1 capital equal to the provision shortfall — it does not affect the income statement directly but does reduce distributable capital and CET1 ratios.
As the NPE ages, the calendar minimum ratchets up — increasing the Pillar 2 capital charge year by year. This creates a direct financial incentive to resolve NPEs through sale, write-off, or restructuring, rather than holding them indefinitely. This was the ECB's explicit policy intent.
| Item | Amount | Notes |
|---|---|---|
| EAD | €320,000 | Outstanding balance at reporting date |
| IFRS 9 ECL provision (12%) | €38,400 | Model-based expected loss — well-secured, LTV 65%, base case house prices |
| Calendar provisioning minimum (35%) | €112,000 | Year 4 minimum for immovable property-secured NPE |
| Provision shortfall | €73,600 | = €112,000 − €38,400 |
| Pillar 2 capital charge (CET1) | €73,600 | Bank must hold €73,600 additional CET1 against this shortfall — equivalent to the shortfall amount, not a % of it |
| Effective total capital consumption | €73,600 + IRB capital on EAD | The P2R capital charge is on top of, not instead of, the existing Pillar 1 defaulted exposure capital requirement |
Calendar provisioning made holding aged NPEs increasingly expensive from a capital perspective. For AIB and BOI, this accelerated the decision to sell NPL portfolios to distressed debt funds from 2020 onward — even at prices that crystallised accounting losses — because the ongoing capital drag of the calendar schedule exceeded the cost of an immediate loss.
The calendar schedule's 9-year timeline for immovable property acknowledged Irish legal realities — but still imposed capital costs on banks that couldn't repossess PDH properties quickly. This created political tension between the ECB's supervisory objectives and domestic Irish housing policy, particularly around the pace of repossession enforcement.
For NPEs originated after 26 April 2019, the CRR2 Article 47c backstop applies — a binding regulatory requirement (not just Pillar 2 expectation) that deducts the shortfall directly from CET1. The timelines are similar but the mechanism is harder, applying as a Pillar 1 deduction rather than a Pillar 2 add-on. This makes resolution of post-2019 NPEs even more urgent.