IFRS 9 Expected Credit Loss: What Banks Need to Know in 2026

IFRS 9 has been mandatory for EU banks since 1 January 2018, but eight years in, supervisors are still finding significant weaknesses in how banks implement the Expected Credit Loss model. The ECB's 2025 targeted review found material deficiencies in SICR assessment at 40% of reviewed institutions, and the EBA's 2026 supervisory priorities explicitly name ECL model quality as a focus area. This isn't a new regulation — it's a maturing one, with increasingly specific expectations.

This guide covers the ECL framework as regulators expect to see it in 2026 — not the theoretical standard, but the practical implementation that will survive supervisory scrutiny.

The Three-Stage ECL Model

IFRS 9's impairment model classifies every financial asset measured at amortised cost (or FVOCI) into one of three stages, based on the change in credit quality since initial recognition:

Stage 1: Performing (12-Month ECL)

All exposures start in Stage 1 at origination. The bank recognises a provision equal to 12-month expected credit losses — the portion of lifetime expected losses arising from default events possible within 12 months, weighted by the probability of those events occurring. For a well-performing mortgage portfolio, Stage 1 ECL provisions are typically 10–30 basis points of exposure.

Key point: Stage 1 is not zero-provision. Every performing loan carries a provision from day one. This is the fundamental shift from IAS 39, where no provision was recognised until a loss event occurred.

Stage 2: Significant Increase in Credit Risk (Lifetime ECL)

When the credit risk of an exposure has increased significantly since initial recognition — but the exposure is not yet credit-impaired — it moves to Stage 2. The bank must now recognise lifetime expected credit losses, which can be substantially higher than 12-month ECL. For consumer lending, Stage 2 provisions typically run 2–8% of exposure, depending on product type, remaining maturity, and economic assumptions.

The transition from Stage 1 to Stage 2 is the most contentious area of IFRS 9 implementation. It requires banks to define what constitutes a "significant increase in credit risk" (SICR) — and this is where supervisory findings concentrate.

Stage 3: Credit-Impaired (Lifetime ECL)

Stage 3 captures exposures where a credit loss event has occurred — the borrower is in default (typically defined as 90 days past due, aligned with CRR Article 178), or other evidence of impairment exists (significant financial difficulty, breach of covenants, bankruptcy filing). Lifetime ECL is recognised, and interest revenue is calculated on the net carrying amount (gross amount minus loss allowance) rather than the gross amount.

SICR Triggers: Where Banks Get It Wrong

IFRS 9 does not define a bright-line test for SICR. The standard requires banks to consider "all reasonable and supportable information" including forward-looking data. In practice, banks use a combination of quantitative and qualitative triggers:

Quantitative SICR Triggers

• PD deterioration: The most common trigger — a relative or absolute increase in the lifetime PD since origination. For example, a relative PD increase exceeding 200% (the origination PD has tripled) or an absolute PD increase exceeding 100bps. The ECB has criticised banks that use only relative thresholds without absolute floors, as this can miss deterioration in originally high-quality portfolios.
• Days past due: IFRS 9 includes a rebuttable presumption that credit risk has increased significantly when contractual payments are more than 30 days past due. Most banks use 30 DPD as a backstop SICR trigger, though some (particularly in retail) have rebutted this presumption with evidence that 30 DPD does not always indicate SICR for certain products.
• Internal rating migration: For wholesale portfolios with internal rating systems, a downgrade of more than a specified number of notches (typically 2–3 notches on the internal scale) triggers Stage 2.
• External rating downgrade: Where external ratings are available, a downgrade below investment grade (BBB- or equivalent) from the origination rating often serves as a SICR trigger.

Qualitative SICR Triggers

• Forbearance or modification measures granted to the borrower
• Significant change in expected business or financial performance
• Watch list placement or early warning system alerts
• Adverse changes in business, financial, or economic conditions that affect the borrower
• Significant sector-specific or geographic deterioration

Common Supervisory Findings on SICR

The ECB's targeted reviews and the EBA's benchmarking exercises have consistently flagged several weaknesses:

• Over-reliance on single triggers: Banks using only 30 DPD as their SICR backstop, without forward-looking PD-based triggers, miss deterioration until the borrower is already distressed
• Stale origination PDs: If origination PDs aren't properly stored or are recalculated using current models (rather than the model in force at origination), the SICR assessment becomes unreliable
• Insufficient collective assessment: IFRS 9 requires that even when individual SICR triggers haven't fired, banks should assess whether broader portfolio or macro factors indicate SICR for segments of the portfolio. Many banks' collective assessment frameworks are underdeveloped.
• Cure periods too short: Some banks move exposures back from Stage 2 to Stage 1 too quickly — a borrower who was 35 DPD but then pays current may not have genuinely recovered in credit quality

Macro Overlays and Forward-Looking Information

The forward-looking nature of IFRS 9 requires banks to incorporate macroeconomic forecasts into their ECL calculations. In practice, this means generating multiple economic scenarios (typically three — base, upside, downside — though some banks use up to five), estimating PDs, LGDs, and EADs under each scenario, and probability-weighting the results.

Scenario Design

Each scenario should represent a plausible macroeconomic outcome with associated probability weights. A typical setup:

• Base case (50–60% weight): Central economic forecast — GDP growth, unemployment, interest rates aligned with consensus estimates
• Upside (10–20% weight): Faster-than-expected recovery, lower unemployment, stronger GDP
• Downside (20–30% weight): Recession, rising unemployment, asset price declines, credit contraction

Supervisors expect the scenarios to be internally consistent (GDP decline should correlate with rising unemployment and falling house prices), to have defined time horizons (typically 3–5 years, with reversion to long-run averages thereafter), and to be updated at least quarterly — more frequently during periods of rapid economic change.

Management Overlays

When model outputs don't fully capture current or emerging risks, banks apply management overlays — manual adjustments to ECL provisions. During COVID-19, overlays constituted a significant proportion of total ECL provisions at many banks, sometimes exceeding 30% of total allowances.

In 2026, supervisors are focused on overlay governance:

• Documentation: Every overlay must be supported by a documented rationale, quantitative methodology, and expected unwind timeline
• Approval: Overlays must be approved at an appropriate governance level — typically the bank's impairment committee or equivalent
• Review cadence: Overlays should be reviewed at least quarterly and removed when the underlying model captures the risk adequately
• Avoiding "stale" overlays: Supervisors flag overlays that persist beyond their stated rationale — a COVID overlay still in place in 2026 requires strong justification
• Two-way overlays: Banks should consider both positive and negative overlays. If a model overestimates losses in certain scenarios, a positive (reducing) overlay may be warranted

Model Validation Requirements

ECL models are subject to the same model risk management standards as regulatory capital models. The EBA's guidelines on credit risk management and the ECB's TRIM findings set clear expectations:

Backtesting

Banks must backtest ECL model outputs against realised losses. This involves comparing predicted PDs against actual default rates, predicted LGDs against realised loss severities, and predicted stage allocations against actual migrations. The ECB expects banks to perform backtesting at a granular level — by product, segment, and vintage — not just at portfolio level.

Sensitivity Analysis

Banks should test how ECL provisions change under different assumptions — alternative SICR thresholds, different macro scenarios, varying probability weights. This isn't just a model validation exercise; it's a disclosure requirement under IFRS 7.35G, which requires banks to explain how changes in economic assumptions affect ECL.

Model Inventory and Documentation

Every ECL model component — PD model, LGD model, EAD model, staging model, macro overlay methodology — should be documented in a model inventory with defined ownership, validation cycle, materiality assessment, and model risk rating. The risk management framework should explicitly cover model risk as a risk category.

2026 EBA Supervisory Priorities for IFRS 9

The EBA's 2026 supervisory priorities place particular emphasis on several IFRS 9 areas:

1. SICR calibration consistency: The EBA benchmarking exercise found significant dispersion in Stage 2 ratios across banks with similar portfolios. Banks with outlier Stage 2 ratios — either abnormally high or suspiciously low — face targeted supervisory scrutiny.
2. Forward-looking information quality: Supervisors are challenging the quality of macro scenarios, particularly whether downside scenarios are severe enough and whether probability weights reflect genuine assessment or are mechanistically set.
3. ECL interaction with capital: The interaction between IFRS 9 provisions and CRR3 capital requirements (Basel IV) remains a focus. Banks must ensure that ECL provisions and regulatory expected losses are reconciled, and that the transitional arrangements (which phased in the IFRS 9 capital impact) are properly unwound by 2028.
4. Climate risk in ECL: The ECB increasingly expects banks to incorporate climate risk into ECL assessments — either through adjusted macro scenarios that include climate transition pathways, or through sectoral overlays for carbon-intensive industries. CSRD reporting requirements amplify the need for climate-adjusted credit risk assessment.
5. Data quality: Accurate ECL depends on reliable data — origination dates, origination credit quality, historical default data, collateral valuations. Supervisors are increasingly challenging data lineage and data quality governance as part of IFRS 9 reviews.

Common Model Weaknesses Flagged by Supervisors

Based on published supervisory findings from the ECB, EBA, and Bank of England PRA, the most frequently cited model weaknesses in ECL implementations are:

• Low-default portfolios: For sovereign, bank, and large corporate portfolios where defaults are rare, PD and LGD models rely heavily on external data or through-the-cycle estimates. Supervisors flag insufficient conservatism in these models.
• Lifetime PD term structures: Converting point-in-time 12-month PDs into lifetime PD curves requires assumptions about default timing. Many banks use overly simplistic approaches (linear extrapolation) rather than survival analysis or hazard rate models.
• LGD in downturn: IFRS 9 requires LGD estimates to reflect expected (not best-case) recovery, incorporating forward-looking conditions. Supervisors frequently find that LGD models don't adequately capture downturn severity or that collateral valuations are stale.
• Prepayment and behavioural maturity: For revolving credit facilities and products with prepayment options, the effective maturity used in ECL calculations must reflect expected rather than contractual behaviour. Getting this wrong — especially for credit cards and overdrafts — can materially misstate lifetime ECL.
• Discount rate application: ECL should be discounted at the effective interest rate. Some banks use approximations (flat rate, portfolio average) that introduce systematic error, especially for long-dated exposures.

"The difference between a good IFRS 9 implementation and a poor one isn't the model sophistication — it's the governance. Models are tools. Governance determines whether those tools produce reliable, auditable, and decision-useful provisions."

Practical Implications for Compliance Teams

IFRS 9 ECL is often viewed as a finance function responsibility — and it is, primarily. But compliance teams need to understand it for several reasons:

• SREP impact: Weak ECL frameworks lead to qualitative SREP findings that affect the overall supervisory assessment, including Pillar 2 capital requirements. Compliance teams tracking regulatory relationships need visibility into IFRS 9 supervisory issues.
• Regulatory reporting: FINREP templates require detailed ECL disclosure by stage, product, and geography. Compliance oversight of regulatory reporting accuracy must cover IFRS 9 data flows.
• Conduct risk: ECL assumptions — particularly around forbearance and collections — intersect with conduct requirements. A bank that aggressively collects on distressed loans to reduce LGD assumptions may face conduct risk questions.
• Regulatory change monitoring: IASB post-implementation reviews, EBA guideline updates, and ECB supervisory letters on IFRS 9 create a continuous stream of expectations that must be tracked and implemented.

For banking compliance teams, understanding the ECL framework isn't about building models — it's about ensuring the organisation's risk governance covers provisioning adequacy, and that supervisory expectations are met proactively rather than reactively. RegPulse tracks EBA, ECB, and IASB publications related to IFRS 9 implementation, alerting compliance teams to new guidance, supervisory findings, and interpretation changes. For related regulatory frameworks affecting bank compliance, see our guides on EBA banking regulation and the operational resilience requirements that intersect with financial risk management.