## tnmea2026004

## Source details

**Canonical URL:** [tnmea2026004](https://www.imf.org/-/media/files/publications/tnm/2026/english/tnmea2026004.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tnm/2026/english/tnmea2026004.pdf.md)
- [Structured JSON version](/-/media/files/publications/tnm/2026/english/tnmea2026004.pdf.json)

---

### Executive summary — overview and key benefits
- IFRS 9 Financial Instruments became effective in January 2018.
- IFRS 9 components:
  - classification and measurement aligned with the entity’s business model;
  - forward-looking expected credit loss (ECL) model with three stages replacing the incurred-loss approach under IAS 39;
  - hedge accounting model.
- Benefits for banking soundness:
  - timelier loan-loss provisions from forward-looking ECLs;
  - greater transparency via extended disclosures;
  - strengthened governance and credit-risk-management practices (data infrastructure, ECL estimates, staging breakdowns);
  - enhanced resilience to economic shocks through robust provisioning and scenario analysis;
  - support for forward-looking, risk-based banking supervision with richer supervisory data.
- Implementation remains challenging in several jurisdictions, particularly low-income countries.

### Key features of IFRS 9 classification, measurement, and ECL staging
- Single classification approach based on:
  - the entity’s business model for managing financial assets; and
  - contractual cash flow characteristics.
- Primary categories:
  - Amortized cost: “hold to collect” business model and solely payments of principal and interest.
  - FVOCI: “hold to collect and sell”.
  - FVTPL: all other assets, trading assets, and derivatives not designated as hedges.
- Fair value option: designation to FVTPL to address accounting mismatches.
- ECL impairment model:
  - Three stages: Stage 1 (initial recognition), Stage 2 (significant increase in credit risk, SICR), Stage 3 (credit impaired).
  - Stage 1: Twelve-month ECL recognized in profit or loss; interest revenue on gross carrying amount.
  - Stage 2: Lifetime ECL recognized; interest revenue same as Stage 1.
  - Stage 3: Lifetime ECL recognized; interest revenue based on amortized cost (gross carrying amount less loss allowance).

### Key implementation challenges
- New concepts and models: fair value, expected losses, forward-looking information, valuation and ECL models.
- Data challenges:
  - limited availability and reliability of historical and forward-looking data, including macroeconomic variables;
  - need for multiple macroeconomic scenarios with assigned probabilities and numerous variables that may be unavailable.
- Capacity constraints in EMDEs:
  - limited technical expertise in credit risk modeling;
  - legacy IT systems incompatible with IFRS 9 requirements.
- Risks from managerial discretion: potential understatement of ECLs, delayed loss recognition, overstated regulatory capital, reduced comparability.
- Complexity in developing, validating, monitoring, and recalibrating lifetime ECL models.

### Supervisory role, tools, and transition considerations
- Supervisory priorities:
  - introduce requirements and expectations to ensure consistent, high-quality implementation;
  - verify bank compliance and adequacy of asset valuations and ECL provisions from a prudential perspective;
  - take corrective measures when outcomes fall short.
- Tools and measures:
  - broad supervisory and corrective powers for loan classification and provisioning concerns;
  - maintain regulatory provisioning systems (prudential backstops) at least temporarily;
  - consider improving consistency between prudential and accounting frameworks while recognizing distinct objectives.
- Transition and sequencing:
  - consider a multiyear transition period with appropriate sequencing of tasks;
  - use parallel runs and potential transitional arrangements to smooth day-one CET1 effects and limit “cliff effects”;
  - encourage banks to develop capabilities to assess climate-related financial risks (CFRs) in balance-sheet valuations.

### Proportionality and procyclicality — high-level points
- Proportionality:
  - implementation should be proportionate to bank size and complexity; simplified approaches may be acceptable under safeguards and documentation without undermining prudence.
- Procyclicality:
  - ECL is in principle less procyclical than the incurred-loss model but does not eliminate procyclicality;
  - risks include spikes in provisioning at downturn onset and persistence of procyclicality if SICR is not identified timely;
  - supervisors should ensure timely loss recognition and capital build-up during upswings.

### Implementation guidance and safeguards (concise)
- Supervisory actions:
  - set clear requirements and expectations; verify compliance; deploy corrective measures; consider temporary regulatory backstops; sequence multiyear transitions; promote inter-authority and auditor cooperation; ensure proportionality without sacrificing prudence; support capacity building in EMDEs (data, IT, modeling).

---

### BOX 2 — Approaches to Measure Expected Credit Losses (ECL): definition, measurement, and operational implications
- Definition and core approach:
  - ECLs are probability-weighted estimates of credit losses (difference between contractual cash flows and expected receipts, discounted at original effective interest rate).
  - Common four-component approach: Probability of default (PD), Loss given default (LGD), Exposure at default (EAD), Discounting.
  - Measurement horizon depends on whether credit risk has significantly increased since origination; ECLs measured individually or collectively when borrower-specific information is limited.
- IFRS 9 forward-looking requirements:
  - probability-weighted estimates across scenarios with reasonable and supportable information available without undue cost or effort;
  - calculate ECL under each scenario and weight by likelihood;
  - Twelve-month ECLs: portion of lifetime ECLs from defaults within the 12 months after reporting date.
  - Lifetime ECLs: present value of losses if default occurs at any time during remaining life.
- Operational, IT, and human-resource challenges:
  - need for upgraded core banking and IT systems to track, report, and recalibrate ECLs in frequent intervals;
  - increased scope for judgment (no prescriptive definition of “default” or SICR indicators);
  - implementation costs: initial (IT, data, model development, governance, staff) and ongoing (data collection, recalibration, reporting, monitoring).
  - supportive public infrastructure (valuation professionals, public statistics, robust external audit) can alleviate costs.
- Effects on profitability, capital, and procyclicality:
  - potentially significantly higher allowances where many exposures are Stage 2;
  - short-term effects: reduced profitability and potential adverse effects on CET1 capital ratios;
  - IFRS 9 reduces procyclicality versus IAS 39 in principle, but point-in-time estimates and staging shifts can increase provisioning in downturns.
- Supervisory directions and preparatory measures:
  - introduce requirements (default definition, SICR criteria, macroeconomic scenarios, collateral valuation);
  - verify compliance and consider transition/parallel runs and potential progressive recognition of day-one CET1 effects.
- Governance, methodologies, IT, and data expectations for banks:
  - documented methodologies, senior responsibilities, robust IT and controls, independent validation, and strict conditions for third-party models (ownership, documentation, supervisory access).
- Engagement with auditors and post-implementation review:
  - close engagement on SICR, forward-looking elements, data, and IT; regular post-implementation reviews; ensure ECL metrics align with risk management and stress tests.
- Supervisory challenger models:
  - can identify outliers using loan-by-loan data but mainly trigger further investigation; require technical capacity and IT infrastructure.
- Coverage beyond amortized cost:
  - supervise FVTPL and FVOCI classification and valuation (Level 3 inputs per IFRS 13); review reclassifications and hedge accounting documentation.
- Climate-related financial risks:
  - CFRs affect SICR, impairment, collateral values, and ECLs; phased capability development encouraged, especially in EMDEs.
- Supervisory corrective measures and prudential backstops:
  - range of powers: require policy revisions, increase provisioning, capital add-ons; maintain prudential backstops during transition; benchmark internal ECL to regulatory parameters.
- Disclosure expectations:
  - banks should disclose implementation intent, key concepts, modeling techniques, macro assumptions, judgmental adjustments, and risk-management organization.
- Prudential vs accounting alignment opportunities:
  - accounting ECLs cover expected losses (past, current, forward-looking); regulatory capital covers unexpected losses via stress scenarios; aligning default definitions would be beneficial.

*Source: BOX 2, IFRS 9 Implementation from the Perspective of Banking Supervisors (Technical Notes and Manuals).*

---

### BOX 3 — Interaction with the Basel Framework: capital, provisioning, and policy implications
- Interaction overview:
  - Accounting provisions reduce retained earnings and CET1.
  - Under the Basel standardized approach, exposures are risk-weighted net of specific provisions and gross of general provisions.
  - Regulatory concepts of “specific provisions” and “general provisions” do not map to accounting standards; some banks treat Stage 1/2 ECLs as general provisions and Stage 3 as specific provisions.
- Findings on standardized approach and proposed adjustment:
  - Under standardized approach, no recognition that general provisions reduce exposure values; suggested prudential adjustment: deduct all accounting provisions from exposure value for risk-weighting.
- IRB approach: shortfall and excess treatment:
  - Compare total eligible accounting provisions with regulatory expected loss (EL = PD × LGD × EAD).
  - Shortfall between accounting provisions and regulatory EL is fully deducted from CET1 (no tax consideration).
  - Excess is added to Tier 2 up to a limit of 0.6 percent of credit risk-weighted assets under IRB.
  - Shortfall treatment prevents double counting by ensuring regulatory EL acts as a floor.
- High-level differences between EL and ECL:
  - Default definition: Accounting — none provided; Prudential — default when past due more than 90 days or obligor unlikely to pay.
  - PD measurement period: Accounting — 12 months/lifetime; Prudential IRB — 12 months.
  - PD position in cycle: Accounting — point-in-time; Prudential IRB — long-run average (PIT, through-the-cycle, or hybrid).
  - ECL measurement: Accounting — asset-specific, neutral estimate with forward-looking factors; Prudential IRB — probability-weighted scenarios, economic loss including collection costs, downturn LGD, UL for workout.
  - Discount rate: Accounting — effective interest rate; Prudential IRB — no mandatory rate.
  - Data observation periods: Prudential IRB expects PD five years minimum; LGD and EAD five years for retail and seven years for corporate exposures.
  - Internal validation/governance: Prudential IRB has minimum requirements; accounting has no specific guidance.
- Double counting and supervisory implications:
  - Double counting can occur if accounting ECL reduces CET1 and supervisors also raise capital via Pillar 2/stress tests; supervisors should identify and address double counting (example: adjustment to hurdle rates by some supervisors).
- Policy recommendations and safeguards:
  - Proportionate IFRS 9 implementation reflecting local circumstances;
  - define conditions for simplified approaches with documentation and supervisory oversight;
  - strengthen supervisory monitoring and reassess simplified approaches periodically.

*Source: Basel Framework.*

---

### BOX 4 — Implementation of Simplified Approaches to Measure ECLs
- Simplified approaches banks may use:
  - Historical loss rate approach with adjustments for current and forecast conditions; still assess SICR;
  - Internal benchmarking to similar portfolios when data lacking;
  - Simple extrapolation for lifetime PDs rather than transition matrices;
  - Portfolio-average LGD components instead of transaction-level drivers;
  - Collective grouping for exposures with similar risk characteristics;
  - Simplified staging criteria (fewer macro variables, greater qualitative reliance);
  - Basic historical-cycle analysis to estimate ECLs under different forecast points using multiple scenarios/sources.
- Governance and controls:
  - Simplified approaches require sound governance and effective controls; enhancements must not be compromised.
- Jurisdictional tailoring examples:
  - Canada: tailoring of supervisory expectations to size, nature, complexity.
  - Brazil: standardized rules for smaller institutions; simple extrapolation and benchmarking where data are lacking; larger institutions use internal models with prudential backstops.
- Procyclicality: nature and persistence
  - Accounting rules reflect real-economy cyclicality; ECL may be less procyclical than incurred loss in principle, but empirical evidence (COVID-19) is inconclusive.
  - Procyclical aspects remain due to point-in-time PD parameters and potential failure to anticipate downturns or identify SICRs timely.
- Factors delaying ECL recognition:
  - excessive reliance on practical expedients (past-due triggers, “low credit risk” simplification, 12-month PD in place of lifetime PD);
  - limited forecasting capacity for abrupt shocks;
  - management incentives to defer SICR recognition.
- Tools to reduce procyclicality:
  - robust ECL implementation and governance;
  - ensure consistency between ECL provisioning and risk management;
  - use prudential tools (releasable buffers, positive neutral capital countercyclical buffer) to smooth behavior.

*Source: IFRS 9 Financial Instruments and GPPC “The Implementation of IFRS 9 Impairment Requirements by Banks.”*

---

### Country experiences and supervisory recommendations (selected EMDE examples)
- Nigeria:
  - required forward-looking credit-risk assessment, independent external auditor verification, implementation plans, impact assessments, and a parallel run ahead of IFRS 9;
  - allowed a four-year transitional arrangement for day-one CET1 effects;
  - challenges: large increase in impairment allowances, implementation costs, insufficient historical default and granular borrower/sector data; largest banks invested heavily in data systems and external consultants.
- Ghana:
  - 2018 adoption increased loss allowances and compressed profits; Domestic Debt Exchange Programme 2022–23 caused substantial impairment on government securities;
  - Bank of Ghana actions: reporting templates, reconciliation requirements between IFRS impairments and prudential provisioning, targeted reviews, independent validation, plans for supervisory challenger model, and detailed impairment data submissions.
- Philippines:
  - BSP required adoption of PFRS 9 in 2018 with proportionality for simpler banks; actions included impact assessments, model documentation, enhanced supervisory returns, on-site examinations, independent model validation, top-down benchmarking, and model-risk-management guidance.
- Chile:
  - transitioned in 2021; CMF developed prudential criteria superseding IFRS 9 for loans at amortized cost and standardized methodologies for minimum provisions; noted minimal aggregate effect on capital adequacy in early reports; actions included more detailed reporting, independent validation, and targeted model reviews.
- IMF-recommended supervisory measures:
  - issue detailed guidelines for IFRS 9 transition; require recognition of excess regulatory provisioning needs as nondistributable “credit risk reserve”; introduce model-risk-management guidelines; amend prudential returns to capture ECL details; require model documentation and validation reports.
  - supervisory activities: build baseline ECL challenger models; proactive engagement with banks and auditors; build supervisory expertise and a core IFRS 9 team; develop toolkits for ECL assessment; require banks to maintain internal capacity and audit of outsourced processes.

### Supervisory review tasks and validation expectations
- Data collection and benchmarking: collect economic scenarios, ECL estimates, backtesting results; require notifications for material methodology changes.
- Key review tasks:
  - assess classification and measurement policies, SICR criteria, data quality, ECL computation engines, coherence of assumptions and forward-looking scenarios, backtesting and model recalibration, PMAs governance, sample credit-file reviews, independence and adequacy of validation units.
- Validation and independence:
  - ensure periodic independent assessment by model validation units and internal audit; supervisors should assess internal control frameworks and validation independence.

*Source: IMF technical note on IFRS 9 implementation and supervisory measures in emerging market and developing economies (excerpt).*

---

### Annex 3 — Lessons from the COVID-19 pandemic
- Exceptional regulatory measures were widely used to preserve safety and soundness, maintain lending, and mitigate procyclicality; supervisors used flexibility within existing frameworks rather than relaxing standards.
- Guidance during the pandemic clarified:
  - payment deferrals do not automatically trigger SICR or migration to Stage 2;
  - banks should not apply standards mechanistically and must weigh long-term trends and public support measures in forward-looking inputs;
  - holistic assessments are required when borrower-specific information is limited.
- ECL model lessons:
  - models could not fully capture unprecedented uncertainty due to nonrepresentative historical data;
  - postmodel overlays were widely used to capture public support effects and other gaps, but diverse practices and limited transparency raised concerns;
  - limited sector-specific segmentation and inadequate model validation and controls were common issues.
- Supervisory recommendations going forward:
  - provide more detailed guidance on overlays and scenarios, with strict governance, controls, documentation, and disclosure requirements;
  - ensure downside scenarios are sufficiently severe to capture tail risks;
  - investigate diverse provisioning practices and differences in modeled provisions.

*Source: tnmea2026004 - ANNEX 3. Lessons from the COVID-19 Pandemic.*

*tnmea2026004*

### Executive Summary 2

### Executive Summary 2

### Overview
- International Financial Reporting Standards 9 (IFRS 9) Financial Instruments became effective in January 2018.
- IFRS 9 includes:
  - a classification and measurement approach for financial assets that reflects the entity’s business model;
  - a forward-looking expected credit loss (ECL) model with three stages, which replaced the previous incurred-loss approach under IAS 39 Financial Instruments: Recognition and Measurement;
  - a hedge accounting model.
- Most advanced economies and many emerging market and developing economies (EMDEs) have implemented or plan to implement IFRS 9, but implementation remains challenging in several jurisdictions, particularly low-income countries.

### Key benefits from a banking soundness perspective
- More accurate classification of credit quality and timelier recognition of loan-loss provisions due to the forward-looking ECL framework.
- Greater transparency and reliability in banks’ financial statements, especially regarding credit risk, supported by extended disclosure requirements.
- Improved governance and credit-risk-management practices, including enhanced data infrastructure, ECL estimates, and staging breakdowns.
- Enhanced resilience against economic shocks because robust provisioning and forward-looking scenarios better shield capital during rapid asset-quality deterioration.
- Support for more effective, forward-looking, risk-based banking supervision by providing supervisors with richer data to assess risk profiles and capital sufficiency.

### Key components of IFRS 9 (summary)
- Single classification approach based on:
  - the entity’s business model for managing financial assets; and
  - the contractual cash flow characteristics of those assets.
- Three primary classification categories:
  - Amortized cost: “hold to collect” business model and cash flows that are payments of principal and interest.
  - Fair value through other comprehensive income (FVOCI): “hold to collect and sell” business model.
  - Fair value through profit or loss (FVTPL): assets not meeting amortized cost or FVOCI criteria, held for trading, or derivatives not designated as hedging instruments.
- Fair value option: a financial asset can be designated as FVTPL to eliminate or significantly reduce an accounting mismatch.
- ECL impairment model:
  - Three-stage model reflecting credit quality deterioration: Stage 1 (initial recognition), Stage 2 (significant increase in credit risk), Stage 3 (credit impaired).
  - Stage 1: Twelve-month ECL recognized in profit or loss; interest revenue on gross carrying amount.
  - Stage 2: Full lifetime ECL recognized; interest revenue calculation same as Stage 1.
  - Stage 3: Full lifetime ECL recognized; interest revenue calculated based on amortized cost (gross carrying amount adjusted for loss allowance).

### Key challenges to implementing IFRS 9
- New concepts and models: fair value, expected losses, forward-looking information, valuation and ECL models, and new data sources.
- Data challenges:
  - Availability and reliability constraints for historical data and forward-looking elements, including macroeconomic variables.
  - Need for multiple alternative macroeconomic scenarios with assigned probabilities, requiring numerous variables and data points that may be unavailable.
- Resource and capacity constraints in EMDEs:
  - Lack of technical expertise in credit risk modeling.
  - Legacy information technology systems incompatible with IFRS 9 classification and measurement requirements.
- Risk of opportunistic use of judgment and managerial discretion, potentially resulting in underestimated ECLs, delayed loss recognition, overstated regulatory capital, and reduced reliability and comparability across banks.
- Complexity of developing and validating lifetime ECL models, including continuous monitoring and recalibration.

### Role of banking supervisors (summary of responsibilities and tools)
- Supervisors should focus efforts in three directions:
  - Introduce requirements and expectations to ensure consistent and high-quality implementation.
  - Verify bank compliance to ensure asset valuations and ECL provisions are adequate from a prudential perspective.
  - Take corrective measures when banks’ approaches do not deliver expected supervisory outcomes.
- Supervisory actions and considerations:
  - Use a broad range of supervisory and corrective powers when loan classification and provisioning raise prudential concerns.
  - Maintain regulatory provisioning systems where provisioning parameters or minimum provisioning percentages are imposed by regulation and serve as a backstop, at least temporarily.
  - Consider improving consistency between prudential and accounting frameworks while recognizing their distinct objectives; full convergence is not necessarily required.
  - Promote cooperation among supervisors, other domestic authorities, and external auditors to ensure consistent implementation and transparency.
- Transitional and preparatory measures:
  - Authorities should consider a transition period with appropriate sequencing of key tasks to provide sufficient preparation time for banks and supervisors.
  - A well-sequenced multiyear transition period can help supervisors complete preparatory steps and encourage banks to progressively build ECL framework components.
  - After IFRS 9 becomes binding, transitional arrangements could be introduced to smooth effects on banks’ regulatory capital.
  - Supervisors should raise awareness and encourage banks to develop capabilities to assess climate-related financial risks’ effects on balance sheet valuations.

### Proportionality and procyclicality (high-level points)
- Proportionality:
  - IFRS 9 should be implemented in a proportionate and sound manner; scope, implementation length, acceptable methodologies, and prudential requirements may be adjusted to reflect specific circumstances.
  - Required sophistication is not the same for all banks and portfolios; simplified approaches and proxies can be accepted under certain conditions but must not undermine prudence.
  - Supervisors should ensure simple approaches are well justified, documented, and subjected to adequate safeguards.
- Procyclicality:
  - The ECL approach is, in principle, less procyclical than the incurred-loss approach, but procyclicality does not disappear.
  - Forward-looking provisioning should avoid a spike in provisioning at the onset of a downturn and enable accumulation of provisions to absorb losses in bad times and support continued lending.
  - Procyclicality can persist if banks fail to anticipate downturns or cannot/will not identify significant increases in credit risk timely; it can be reduced using a combination of tools.
  - Supervisors should ensure banks do not delay loss recognition and build enough capital during upswing phases.

### Implementation guidance and safeguards (concise)
- Realizing IFRS 9 benefits requires careful preparation and consistent implementation across banks to avoid increased managerial discretion, high costs, unreliable classification and ECL estimates, and lower comparability.
- Supervisors should:
  - set clear requirements and expectations;
  - verify compliance and adequacy of provisioning from a prudential perspective;
  - deploy corrective measures and supervisory tools where necessary;
  - consider temporary regulatory backstops for provisioning;
  - sequence a multiyear transition and consider post-adoption transitional arrangements for regulatory capital smoothing;
  - promote inter-authority and external auditor cooperation;
  - ensure proportionality without sacrificing prudence;
  - support capacity building in EMDEs, including data, IT, and modeling expertise.

*tnmea2026004 - Executive Summary 2.*

### BOX 2. Approaches to Measure Expected Credit Losses

### BOX 2. Approaches to Measure Expected Credit Losses

### Definition and core measurement approach
- Expected credit losses (ECLs) are probability-weighted estimates of credit losses.
- A credit loss is the difference between the cash flows due under the contract and the cash flows the entity expects to receive, discounted at the original effective interest rate.
- Time value of money must be taken into account; banks should discount expected cash flows at the effective interest rate determined at initial recognition.
- Common four-component approach to estimating ECLs:
  - Probability of default (PD): how likely a default event is to happen.
  - Loss given default (LGD): how much loss may result in the event of default.
  - Exposure at default (EAD): outstanding balance when default occurs.
  - Discounting: the time value of money.
- Components are measured based on risk of default over one of two time horizons depending on whether credit risk has significantly increased since origination; assessment relies on quantitative and qualitative elements and backstop indicators.
- ECLs are generally measured individually; can be measured on a collective basis when borrower-specific information is limited.

### IFRS 9 requirements and forward-looking information
- IFRS 9 requires probability-weighted estimates using a range of possible outcomes and reasonable and supportable information available without undue cost or effort.
- ECL determination requires calculating ECL under each economic scenario and weighting by the likelihood of occurrence.
- Information to consider includes past events, current conditions, and forecasts of future economic conditions.
- Twelve-month ECLs:
  - Portion of lifetime ECLs from possible defaults within the 12 months after the reporting date.
  - Effect of entire credit loss weighted by probability that loss will occur in the next 12 months.
- Lifetime ECLs:
  - Present value of losses if borrower defaults at any time during remaining life of the financial instrument.
- Incorporating forward-looking information is key; estimating lifetime ECLs (Stage 2) and recoverable amounts for credit-impaired transactions (Stage 3) requires assumptions and judgment on short- and long-term outlooks.
- Banks typically develop statistical models based on historical relationships between key variables (for example, GDP growth, unemployment rate, and PD parameters). Historical relationships may not hold under extreme conditions (for example, COVID-19).

### Operational, IT, and human-resource challenges
- Reliable ECL models usually require significant technical and operational changes, including upgrades to core banking systems and IT infrastructure.
- Legacy systems may not be compatible with IFRS 9 classification and measurement requirements; banks may need to invest in new systems or significant upgrades.
- Systems should be capable of tracking and reporting ECLs in real time and supporting frequent model recalibration.
- IFRS 9’s principles-based nature increases the scope for judgment and managerial discretion relative to IAS 39:
  - IFRS 9 does not define "default" or prescribe SICR indicators; does not prescribe a specific approach for quantifying ECL.
  - Incorporating forward-looking information and assumptions requires expert judgment.
- Implementation costs:
  - Initial costs: IT systems, data collection and analysis, ECL model development, upgrades in risk-management and governance frameworks, staff capacity building.
  - Ongoing costs: regular data collection and review, model recalibration, reporting and disclosures, monitoring credit risk on a forward-looking basis.
- Supportive public infrastructure (competent valuation professionals, public statistics, robust external audit) would alleviate implementation costs and challenges.

### Effects on profitability, capital, and procyclicality
- Implementation may require significantly higher allowances for ECLs, especially if a high proportion of exposures are classified in Stage 2.
- Short-term effect: reduced profitability and potential adverse effect on CET1 capital ratios if profitability is already shallow.
- Banks should analyze quantitative effects and engage with supervisors on transitional arrangements to limit cliff effects.
- IFRS 9 reduces procyclicality relative to IAS 39 but procyclical effects remain:
  - Moving loans from Stage 1 to Stage 2 and using point-in-time estimates can increase ECLs in downturns, affecting profitability and capital and potentially tightening credit supply.
  - Supervisory monitoring and macroprudential tools may be needed to address procyclical behavior.

### Supervisory challenges and recommended supervisory actions
- Supervisory challenges: ensuring consistency of IFRS 9 implementation across banks; assessing first-time adoption effects; accompanying banks during transition; mitigating procyclical effects.
- Key supervisory directions:
  - Introduce requirements and expectations (definition of default, SICR criteria, macroeconomic scenarios, collateral valuation).
  - Verify bank compliance to ensure asset valuations and ECL provisions are adequate from a prudential perspective.
  - Take corrective measures when banks’ approaches do not deliver expected supervisory outcomes.
- Preparatory and transition measures:
  - Consider a transition period and parallel run (reports under both pre-IFRS 9 and IFRS 9) to prepare banks and supervisors and smooth CET1 effects.
  - Potential transitional arrangements to phase in day-one CET1 effects (for example, progressive recognition to limit cliff effects).
- Road map elements (generic):
  - Preparatory phase: introduce requirements and expectations; develop supervisory capabilities; engage with banks and auditors; improve credit risk examination procedures.
  - Implementation: supervise internal lines of defense (validation, internal controls); continue developing in-house expertise to review ECL models.
  - Post-implementation: conduct post-implementation reviews; collect information regularly; on-site inspections or deep dives.
- Supervisory specification and guidance to ensure comparability:
  - Align definition of default for accounting with regulatory default definitions and supplement with objective criteria (for example, credit-impairment triggers).
  - Introduce minimum SICR assessment criteria (credit risk ratings, past due status, PD, watch lists, forbearance indicators); specify exit criteria and observation periods for restructurings.
  - Provide guidelines on macroeconomic scenarios: design, range, severity, horizon, number of scenarios to capture nonlinearities, and weights for adverse scenarios.
  - Provide supervisory guidance on collateral valuation (frequency, independent experts, haircuts).
  - Specify conditions for classifying exposures as "low credit risk" at reporting date.
  - Define acceptable methodologies for hedge effectiveness assessment.

### Governance, methodologies, IT, and data expectations for banks
- Requirements and expectations for banks:
  - Sound credit methodologies and documented instances of expert judgment and postmodel adjustments.
  - Assign senior responsibilities for IFRS 9 implementation; document accounting policies and credit risk methodologies.
  - Maintain IT systems to handle large volumes of staging and quantification data; introduce controls to verify data completeness and accuracy.
  - Ensure sound governance across departments (risk, finance) for consistent application.
  - Internal audit and external auditors should independently verify IFRS 9 implementation and ECL model reliability.
  - Conditions for using consultants or third-party models: banks must retain ownership, ensure independence and competence of consultants, maintain documentation, ensure supervisory access, validate model performance, and secure knowledge transfer.
- Data quality and availability:
  - Supervisors should ensure banks understand data needs and have plans to handle data gaps.
- ECL methodologies and models:
  - Segment credit portfolios; select modeling approaches; develop and validate ECL models; document methodologies and performance indicators.
- IT systems, processes, and controls:
  - Systems must be largely automated, process vast information, and deliver reliable results at least semiannually or quarterly.
  - Key judgments (economic scenarios, SICR definition, postmodel adjustments) should be governed and overseen.

### Engagement with external auditors and post-implementation review
- Supervisors should engage closely with external auditors on high-risk areas: SICR criteria, forward-looking elements, data availability, IT adequacy.
- Conduct regular post-implementation reviews to assess progress, lessons learned, and introduce adjustments; communicate thematic review results and best practices.
- Ensure ECL metrics are used in risk management and aligned with other risk methodologies (for example, stress-test scenarios).

### Supervisory tools: challenger models, scope, and limitations
- Supervisors may consider developing a supervisory "challenger" model to identify potential material misstatements using bank-provided data (loan-by-loan).
- A single challenger model mainly identifies outliers and triggers further investigation rather than determining optimal provisioning levels.
- Implementing challenger models requires strong technical capabilities, IT infrastructure, and expertise; supervisors should weigh pros and cons and introduce tools after gaining IFRS 9 experience.

### Coverage beyond amortized cost: fair value categories and hedging
- Supervisors must oversee classification and measurement for FVTPL and FVOCI categories; misclassification or overvaluation can artificially inflate CET1 capital.
- Verify compliance with business model and solely-payments-of-principal-and-interest tests; upgrade regulatory reports to collect data on classifications; investigate material FVTPL/FVOCI instances and ensure independent validation of valuation models.
- Ensure approaches to estimate interest rates and fair value at initial recognition are reasonable, especially when Level 3 inputs are used (unobservable inputs per IFRS 13).
- Review reclassification between IFRS 9 categories; reclassifications should be rare and due to business model changes and must be prospective.
- Hedge accounting:
  - Review banks’ hedge accounting documentation, methodologies for hedge effectiveness, and list of acceptable hedging instruments.
  - Ensure cash flow hedge reserve treatment for prudential CET1 calculation is applied (derecognition for hedging of items not fair valued on the balance sheet).
  - Supervisors should verify disclosure adequacy regarding hedge accounting and valuation of derivatives in FVTPL.

### Climate-related financial risks (CFRs)
- CFRs can affect classification and provisioning, including SICR assessment, impairment determinations, collateral values, and ECL measurements.
- Use of forward-looking data for CFRs is nascent even in large banks; improvements in data and modeling may take years.
- Supervisors should encourage phased action plans:
  - Assess whether internal corporate credit ratings should reflect transition or physical risks.
  - Identify collateral types at risk of value loss or inaccessibility.
  - Identify loan portfolios most sensitive to CFRs and adapt scenarios and weightings for ECL calculations accordingly.
- For EMDEs, it may be premature to require valuation and provisioning adjustments for CFRs but supervisors should encourage capability development.

### Supervisory corrective measures and prudential backstops
- Supervisors have a range of powers when classification or provisioning is inadequate:
  - Require banks to revise policies, increase provisioning, impose capital add-ons, strengthen internal validation, or impose other remedial measures.
- The supervisory response depends on available technical capabilities and whether regulatory provisioning frameworks exist.
- Maintain a prudential backstop (regulatory provisioning) at least temporarily during transition:
  - Backstop helps ensure adequate provisioning while banks and supervisors gain IFRS 9 experience.
  - Supervisors may require periodic benchmarking between internal ECL estimates and provisions calculated using regulatory parameters.
- Regulatory provisioning frameworks can be backward-looking and are not a panacea; authorities should review and improve prudential benchmarks to include forward-looking indicators.
- Cooperate with accounting enforcers, market regulators, external auditors, and other supervisors to ensure high-quality, consistent IFRS 9 implementation and transparency.

### Disclosure expectations
- Supervisors should build on IFRS disclosure requirements and clarify expectations to improve comparability:
  - Banks should disclose how they intend to implement IFRS 9; how they apply key concepts; credit loss modeling techniques; forward-looking macroeconomic assumptions; material judgmental adjustments; and organization of risk-management functions.

### Prudential vs accounting frameworks and alignment opportunities
- Accounting ECLs estimate probability-weighted expected losses using past, current, and forward-looking information; regulatory capital covers unexpected losses via stress scenarios.
- Opportunities to improve consistency:
  - The distinction between specific and general provisions under the Basel standardized approach is less relevant under IFRS 9 and could be simplified.
  - Using the same definition of default for accounting and prudential purposes would be beneficial.

*Source: BOX 2, IFRS 9 Implementation from the Perspective of Banking Supervisors (Technical Notes and Manuals).*

### BOX 3. Interaction with the Basel Framework

### BOX 3. Interaction with the Basel Framework

### Interaction overview
- When a bank recognizes accounting provisions, retained earnings and, therefore, Common Equity Tier 1 (CET1) capitals are reduced accordingly.
- Under the standardized approach for credit risk under the Basel framework, exposures are risk-weighted net of specific provisions and gross of general provisions.
  - Specific provisions include “provisions ascribed to identified deterioration of particular assets or known liabilities, whether individual or grouped” that are excluded from general provisions.
  - General provisions are “provisions or loan-loss reserves held against future, presently unidentified losses and are freely available to meet losses which subsequently materialize [...].”
- The capital requirement is based on the exposure value gross of general provisions, even though CET1 capital has already been reduced by these general provisions; there is no recognition in the Basel framework that exposures may already be partially covered by general provisions.
- The regulatory concepts of specific provisions and general provisions do not exist in accounting standards. Anecdotal evidence suggests some banks may consider ECL allowances for assets in Stages 1 and 2 as general provisions, and Stage 3 ECL allowances as specific provisions, implying a likely higher amount of general provisions when banks use ECL approaches.

### Findings on standardized approach and proposed adjustment
- Under the standardized approach:
  - Exposures are risk-weighted net of specific provisions and gross of general provisions.
  - There is no recognition that general provisions reduce exposure values for risk-weighting.
- Suggested prudential adjustment:
  - Remove the distinction between specific provisions and general provisions in the standardized approach and apply the same prudential treatment by giving consideration to adjust the exposure value to be risk weighted by deducting all accounting provisions.

### IRB approach: shortfall and excess treatment
- Under the internal ratings based (IRB) approaches:
  - Banks compare total eligible accounting provisions with the regulatory measure of expected loss (EL) calculated as PD × LGD × EAD.
  - Any shortfall between total eligible accounting provisions and regulatory EL is fully deducted, without considering tax effects, from CET1 capital.
  - Any excess is added to Tier 2 capital, up to a limit of 0.6 percent of credit risk-weighted assets calculated under the IRB approaches.
- Rationale and effects:
  - The shortfall treatment mitigates double counting by only deducting the amount by which the regulatory EL exceeds accounting allowances, rather than deducting the full regulatory EL in addition to the accounting-based CET1 capital reduction.
  - The shortfall treatment ensures regulatory EL acts as a floor, so a minimum amount of losses is reflected in regulatory capital even if accounting-expected losses are lower.

### Differences between EL and ECL (high-level comparison points from Table 1)
- Default definition:
  - Accounting Framework: No definition is provided.
  - Prudential Framework: Default is considered to have occurred when either or both events have taken place: (1) The obligor is past due more than 90 days on any material credit obligation to the bank; (2) The bank considers that the obligor is unlikely to pay its credit obligations.
- PD estimates — Measurement period:
  - Accounting Framework: 12 months (Stage 1) / lifetime (Stages 2 and 3).
  - Prudential Framework (Standardized Approach): Not applicable.
  - Prudential Framework (IRB Approach): 12 months.
- PD estimates — Position in the cycle:
  - Accounting Framework: PIT, considering forward-looking information and macroeconomic factors.
  - Prudential Framework (IRB Approach): Long-run average of one-year default rates (can be PIT, through the cycle, or hybrid).
- ECL measurement:
  - Accounting Framework: Only asset-specific loss; neutral estimate, considering forward-looking information and macroeconomic factors.
  - Prudential Framework (IRB Approach): Probability-weighted scenarios; Economic loss, including material direct and indirect costs associated with collecting on the exposure; Downturn estimate for LGD; For defaulted exposures sum of the best estimate of EL under current economic conditions and UL estimate for workout period.
- Discount rate:
  - Accounting Framework: Effective interest rate, where applicable.
  - Prudential Framework (IRB Approach): No mandatory discount rate, flexibility but must include material discount effects.
- Exposure amount:
  - Accounting Framework: Both on- and off-balance sheet positions are measured net of specific provisions or partial write-offs.
  - Prudential Framework (IRB Approach): EAD: expected exposure upon default gross of provisions, partial write-offs, and discounts, floor at current drawn amount.
- Undrawn commitments:
  - Accounting Framework: Reflect future expected drawdowns and early repayments over the life of the financial instruments.
  - Prudential Framework (IRB Approach): Reflect the current drawn amount and likely future drawdowns up to and after default.
- Data observation period:
  - Accounting Framework: No specific guidance.
  - Prudential Framework (IRB Approach): PD: five years minimum, the data should include a representative mix of good and bad years. LGD and EAD: five years minimum for retail exposures and seven years for corporate exposures.
- Internal validation and governance:
  - Accounting Framework: No specific guidance.
  - Prudential Framework (IRB Approach): Specific minimum requirements for review and validation of internal estimates; independent credit risk control unit responsible; internal audit or an equally independent function must review at least annually the estimation of PD, LGD, and EAD parameters.

### Double counting, supervisory implications, and examples
- Issue:
  - Double counting of credit losses may arise where the same ECLs based on forward-looking information are captured in accounting provisions (reducing regulatory CET1 capital) and additional capital is required under Pillar 2 (for example, through the supervisory review process or stress testing that also considers expected losses).
  - Worsening macroeconomic expectations could raise both Pillar 2 requirements and accounting ECL provisions.
- Supervisory actions:
  - Banking supervisors should identify potential double-counting issues, assess implications, and address them.
  - Example: Some supervisors (for example, the Prudential Regulation Authority in the United Kingdom) have adjusted the hurdle rates used for supervisory stress tests to reflect the effect of earlier loss recognition under IFRS 9.

### Policy recommendations and supervisory safeguards
- Consider proportionate implementation of IFRS 9 that reflects local circumstances without diluting prudence.
- Use simplified approaches where justified, provided they are no less prudent and subject to safeguards, documentation, and supervisory oversight.
- Supervisory safeguards for simplified approaches:
  - Define conditions under which simple approaches can be used, based on firm-level and portfolio-level factors (size, complexity, economic significance, risk profile, materiality of credit portfolios, historical loss volatility).
  - Require strict justification and documentation for simplified approaches.
  - Regularly reassess appropriateness of simplified approaches and upgrade methodologies as data and expertise improve.
  - Strengthen supervisory monitoring, including use of qualitative assessments, loss rate approaches based on past losses, and other simple approaches subject to supervisory review.

*Source: Basel Framework.*

### BOX 4. Implementation of Simplified Approaches to Measure ECLs

### BOX 4. Implementation of Simplified Approaches to Measure ECLs

### Simplified approaches banks may consider
- Using a historical loss rate approach rather than measuring the PD and LGD separately, with loss rates adjusted for current conditions and forecast for future economic conditions; banks should still assess changes in credit risk for staging assessment.
- Using internal benchmarking to a similar risk portfolio to quantify PD parameters when historical data are not available.
- Using a simple extrapolation technique to determine lifetime PD parameters rather than a more sophisticated approach with transition matrices.
- Using portfolio averages for some components of the LGD rather than considering the full range of LGD drivers at a transaction level.
- Grouping exposures for collective assessment, limited to exposures that share similar risk characteristics.
- Using simplified criteria for the staging assessment, for example:
  - Fewer macroeconomic variables or simplified risk metrics for triggering transfers between stages.
  - Greater reliance on qualitative elements (for example, exposures on watch list, forbearance measures) when increases in credit risk are not well reflected in PD models.
- Performing a basic analysis of historical relationships between credit losses and the position within the economic cycle to estimate ECLs at different future points depending on economic forecasts, while using different scenarios and several sources of information rather than relying only on one forecast.

### Governance and controls
- Sound governance and effective controls are critical; implementation of IFRS 9 requires substantial changes to credit-risk-management processes and procedures, financial reporting systems, and internal controls.
- Enhancements to governance and controls should not be compromised when implementing simplified approaches.
- An effective and strong governance and controls framework is required before, during, and after the transition to IFRS 9; supervisors should ensure banks’ internal controls are effective.

### Jurisdictional tailoring examples (as discussed)
- Requirements and supervisory expectations may be tailored to the size, nature of operations, and complexity of banks.
- Example jurisdictional practices cited:
  - In Canada, although IFRS applies to all banks, requirements and supervisory expectations are tailored to the size, nature, and complexity of banks; guidance applicable to standardized deposit-taking institutions is scaled accordingly relative to IRB deposit-taking institutions.
  - In Brazil, smaller institutions are subject to standardized rules to calculate provisions established and the PD and loss given default parameters; using simple extrapolation techniques to derive lifetime PD parameters from 12-month PD parameters; using internal benchmarking to a similar risk portfolio when data are lacking; or analyzing the past relationship between historical credit losses and the position in the credit cycle to estimate ECLs at different estimated credit points in the future.
  - In Brazil, larger and more complex institutions use internal models under IFRS 9 with prudential backstops.

### Procyclicality: nature, risks, and drivers
- Accounting rules are cyclical by design; they reflect the cyclicality of the real economy.
- Using different approaches to account for credit losses does not change the total amount of credit losses to be recognized, but affects timing of recognition.
- ECL is conceptually no more nor less cyclical than the incurred-loss model; the key question is whether ECL provisioning may exacerbate procyclicality by affecting banks’ behavior (including lending decisions), thereby amplifying business and financial cycle swings.
- In principle, the ECL approach is less procyclical than the incurred-loss approach because forward-looking provisions are recorded before actual default occurs, allowing accumulation of provisions over time to absorb losses in bad times and help sustain lending.
- Empirical evidence from the COVID-19 period did not conclusively corroborate lower procyclicality under ECL; the COVID-19 crisis was an exceptional event accompanied by various support measures, so lessons from that single episode should be considered carefully.

### Persistence of procyclical effects under ECL
- Some procyclical aspects remain because:
  - ECLs are quantified using point-in-time PD parameters (the PD under current economic conditions), so ECLs fluctuate over the economic cycle as projected economic conditions change PD parameters for exposures in Stage 1 and Stage 2.
  - If banks do not anticipate a downturn, credit losses will not be effectively distributed over time before the downturn materializes.
  - Failure or unwillingness to identify SICRs on a timely basis can produce significant “cliff effects” when stage allocations shift from Stage 1 to Stage 2 or Stage 3 (from 12-month PD to lifetime PD parameters), potentially causing a sharp increase in provisioning and leading banks to reduce lending.
  - Banks close to regulatory minima may have stronger incentives to deleverage, including by reducing lending.

### Factors that can delay recognition of ECL provisions
- Excessive reliance on practical expedients, for example:
  - Using past due information, the “low credit risk simplification,” or the use of a 12-month PD instead of a lifetime PD to determine what is an SICR.
- Limited forecasting capacity or inability to anticipate sudden, abrupt shocks (for example, COVID-19), which prevent effective anticipation of changes in credit risk.
- Management incentives and governance dynamics, for example:
  - Recognizing an SICR may affect results, management bonuses, capital position, and share prices, creating temptation to delay loss recognition and reclassify exposures to Stage 2 or Stage 3 only when the crisis hits and payments are missed.

### Tools and supervisory measures to reduce procyclicality
- Supervisory focus should be on ensuring timely loss recognition and sufficient capital build-up in the upswing to absorb losses and support lending in downturns; accounting rules aim to preserve market trust and transparency, not to reduce procyclicality per se.
- Measures to minimize cliff effects and ensure adequate loss absorbency include:
  - Robust implementation of the ECL approach, building on sound credit risk practices and methodologies, to help attenuate cliff effects.
  - Supervisors should ensure banks consider all relevant information, including information processed by risk-management units, and maintain consistency between ECL provisioning and risk-management decisions.
  - Use of prudential tools to smooth procyclical behavior, such as releasable buffers in the downturn phase of the cycle.
  - Introducing a positive neutral capital countercyclical buffer to allow banks to hold capital buffers that can be released in the event of sudden shocks, including those unrelated to the credit cycle (for example, the COVID-19 pandemic).

*Source: IFRS 9 Financial Instruments and GPPC “The Implementation of IFRS 9 Impairment Requirements by Banks.”*

### 9. It emphasized the need for a clear governance and project management structure to oversee the transi-

### IFRS 9 Implementation from the Perspective of Banking Supervisors

### Implementation challenges and country experiences
- Nigeria
  - Emphasized need for a clear governance and project management structure to oversee the transition.
  - Required forward-looking credit risk assessment integrating multiple data sources and economic scenarios, rigorous validation and documentation of models and assumptions (including through independent external auditor verification), and comprehensive reporting and transparency.
  - Required institutions to submit implementation plans and conduct impact assessments (with transition reports showing ECL effects on capital and earnings).
  - Central Bank of Nigeria introduced a parallel run system ahead of IFRS 9 implementation and allowed a four-year transitional arrangement with respect to the day-one effect of IFRS 9 adoption on banks’ regulatory capital.
  - Reports indicated Nigerian banks have seen a large increase in impairment allowances in recent years, putting pressure on some banks’ regulatory capital despite transitional arrangements.
  - Key challenges: implementation costs; data and modeling constraints due to insufficient historical default data and lack of granular borrower- and sector-level data.
  - Responses: banks (especially the largest) invested heavily in data systems to build or purchase credit risk models and partnered with global consultants for IFRS 9 model calibration and validation.
  - Difficulty estimating macroeconomic risks and projecting potential scenarios because of fluctuations in exchange rates, inflation, and oil prices.

- Ghana
  - Implementation of IFRS 9 in 2018 led to materially increasing loss allowances and compressed reported profit margins.
  - Domestic Debt Exchange Programme in 2022–23 led to recognition of substantial impairment losses on banks’ large portfolios of government securities.
  - Key challenges mirrored Kenya and Nigeria: insufficient historical default data to estimate PD and LGD parameters, leading to reliance on proxies, external vendor data, or management overlays.
  - Fragmented IT and risk systems required heavy investment in IT upgrades and hiring experienced staff and external consultants, increasing implementation costs.
  - Bank of Ghana (BOG) actions:
    - Published a guide prescribing reporting templates and disclosure requirements.
    - Required reconciliation between IFRS impairment losses and prudential provisioning amounts and justification of differences.
    - Required submission of documentation of model methodologies and assumptions; conducted targeted reviews; engaged independent consultants to validate or audit banks’ ECL models.
    - In its 2023 financial stability review, announced plans to develop an independent challenger model to benchmark and test banks’ ECL outcomes and staging sensitivities.
    - Required regular, detailed impairment data and conducted reviews to examine staging migration patterns, sensitivity of ECLs to macroeconomic assumptions, and governance oversight for management overlays.

- Philippines
  - Bangko Sentral ng Pilipinas (BSP) required adoption of the Philippines Financial Reporting Standard 9 in 2018 (equivalent of IFRS 9).
  - BSP issued guidelines on classification and measurement of financial instruments and recognizing ECL impairments, adopting proportionality for banks with simpler credit operations.
  - Challenges: need for further historical loss data; variations in definitions of SICR, default, staging, and weighting of macro scenarios, causing differing provisioning effects across banks; investments needed in risk and credit resources and IT systems.
  - BSP supervisory actions:
    - Asked banks to submit impact assessment reports, full model documentation, and sensitivity analyses and scenario results.
    - Enhanced supervisory returns to capture PFRS 9 metrics to identify unusual patterns and outliers.
    - Conducted on-site examinations focusing on governance and model development and validation.
    - Published a draft model-risk-management guideline and requires independent model validation.
    - Used top-down analyses and benchmarking to review provisioning levels and detect outliers; reviewed reasonableness of assumptions in banks’ ECL models during examinations.

- Chile
  - Chilean banks transitioned to IFRS 9 in 2021 under oversight of Comisión para el Mercado Financiero (CMF).
  - CMF introduced prudential criteria superseding IFRS 9 for regulatory provisioning regarding loans at amortized cost; fully applies IFRS 9 for other categories including investments in bonds where ECL models are used.
  - CMF developed standardized methodologies for calculating provisions for loans at amortized cost including PD and LGD parameters for minimum required provisions when internal methodologies not approved.
  - CMF noted in its 2019–20 reports that aggregate effect on capital adequacy was minimal; most banks comfortably maintained regulatory capital ratios under limited adoption.
  - Challenges: lack of long default and recovery histories (especially retail and SME portfolios); need to upgrade and integrate IT and risk-management systems.
  - CMF actions: revising information systems manual for more detailed reporting; monitoring application of IFRS 9 to investments in bonds to detect unusual trends; requiring submission of provisions under standard and internal methodologies to detect outliers and validate local models; requiring independent model validation and conducting targeted model reviews in large banks.

### Recommended supervisory measures (based on IMF technical assistance in EMDEs)
- Supervisory expectations, requirements, and guidelines
  - Supervisor should issue detailed supervisory guidelines outlining expectations for IFRS 9 transition and implementation.
  - Where prudential and accounting rules differ on provisioning, supervisor should require banks to recognize any excess in regulatory provisioning needs (compared with IFRS 9 ECL provisions) as an extra credit loss deducted from retained earnings and classified as a “credit risk reserve,” which is a nondistributable reserve and not taken into account for purposes of calculating regulatory capital and other capital adequacy ratios.
  - Supervisor should consider introducing guidelines on model risk management covering design of ECL models; factors for PD, LGD, and exposure at default; minimum standards for forward-looking information; frequency and breadth of model performance tests; and submission of independent model validation reports.

- Supervisory reporting
  - Supervisor should amend or develop prudential returns to capture data on ECL estimations and details on main ECL parameters and staging of exposures.
  - Require banks to submit thorough documentation of ECL models, including inputs, assumptions, estimates, judgment, and modeling techniques.
  - Require submission of internal validation reports of the ECL model, its governance framework, and related internal audit reports.

- Supervisory activities
  - Develop a baseline ECL challenger model to review banks’ ECL models and adequacy of provisions to address potential ECL underestimation.
  - Proactively engage with banks on IFRS 9 implementation plans, progress, challenges, expectations, and remediation paths.
  - Share findings of ECL model reviews with banks’ external auditors to ensure proper adjustments to ECL models and provisions where needed.
  - Build supervisory expertise in IFRS 9 through additional and targeted training; consider establishing a core IFRS 9 team to lead reviews and transfer knowledge to other supervisors.
  - Develop a toolkit guiding staff on evaluating banks’ ECL models and provisions, including detailed procedures for assessment of inputs, assumptions, estimates, modeling techniques, and a baseline ECL model to challenge banks’ models and provisions.
  - Consider leveraging work of banks’ internal and external auditors in supervisory assessments.
  - Require banks to have measures to improve technical capacity, enhance accountability and understanding of model frameworks, and maintain internal knowledge of outsourced processes; technical capacity should extend to internal audit responsible for internal assessments of risk models.

### Supervisory approaches and tools relative to staging and ECL measurement
- Supervisory objectives and scope
  - Supervisors should adjust methodologies, procedures, and processes because IFRS 9 introduces new rules for classification (staging) and ECL measurement with increased reliance on models and quantification.
  - Although supervisors do not set accounting standards, they should review and scrutinize banks’ ECL models to ensure sound, prudent approaches with adequate governance and risk-management processes.

- Data collection and benchmarking
  - Collect information to ensure consistency of banks’ approaches and identify outliers through desk-based analysis and peer benchmarking; likely requires enhanced supervisory reporting.
  - Typical data collection: banks’ economic scenarios, ECL estimates, and comparison of estimated ECL with actual outcomes (“backtesting” results).
  - Consider requiring banks to notify supervisors of any material changes to ECL methodology or levels of provisioning.
  - Regular discussions with banks and their external auditors to understand material changes in forward-looking scenarios, modeled outputs, and key judgments, including postmodel adjustments (PMAs).

- Key supervisory review tasks
  - Collect information on policies for classifying and measuring financial assets and review banks’ booking policies.
  - Evaluate adequacy of the bank’s credit-risk-management policies, practices, and processes for early identification of problem exposures.
  - Review compliance with credit risk classification staging, including SICR criteria and curing requirements:
    - Ensure SICR criteria are documented, adequate to business model, and effectively implemented, with comprehensive quantitative and qualitative indicators (avoid excessive reliance on a single criterion).25
    - Be alerted to recurrent exceptions or indicators signaling inadequate classification (for example, a large proportion of Stage 1 exposures moving directly to Stage 3).
    - Verify that SICR triggers defined in absolute terms do not ignore credit quality at origination.
  - Identify and verify data quality and model inputs: specify data sources; introduce cartography of data sources and information systems; implement controls (completeness checks, tolerance thresholds for missing values); maintain an audit trail; introduce performance indicators; interface risk and accounting systems.
  - Review ECL computation methodology and integrity of calculation engine; ensure simplifications do not introduce bias.
  - Assess coherence of model assumptions and incorporation of forward-looking information; understand approaches used for forward-looking scenarios and challenge severity of downward scenarios; ensure banks use several forward-looking scenarios.
  - Verify model output coherence and ensure key risk parameters and modeled outputs or overlays are backtested.
  - Ensure backtesting results lead to concrete follow-up actions and model improvements.
  - Review PMAs: provide best-practice examples; understand reasons behind PMAs; ensure documentation, strict governance oversight; use PMAs as temporary solutions when models have limitations and recalibration cannot be completed quickly; pay particular attention to permanent overlays as they may indicate failure to improve models.
  - Review representative sample of credit files to reassess classification and provisioning.
  - Evaluate adequacy and independence of model validation process and challenge model-risk-management framework and its interaction with risk appetite and capital adequacy assessment processes.
  - Ensure internal controls mitigate misstatement risk and exceptions to internal policies are documented and reviewed.
  - Ensure risk-management function provides the board with critical information on impairments and model performance, including changes since last reporting period, performance of ECL models, key trends of PD and LGD, and remaining issues to be fixed.

- Validation, internal audit, and independence
  - Ensure independent, periodic assessment of ECL models by banks’ model validation unit as second line of defense.
  - Internal audit should ensure validation policies, processes, and internal controls are effective.
  - Supervisors should assess adequacy of internal control frameworks for ongoing validation and testing of models, including independence of validation unit and internal audit, sufficient resources, and adequate expertise.

*Source: IMF technical note on IFRS 9 implementation and supervisory measures in emerging market and developing economies (excerpt).*

### ANNEX 3. Lessons from the

### ANNEX 3. Lessons from the COVID-19 Pandemic

### Exceptional regulatory measures and overall approach
- Exceptional regulatory measures were adopted in many countries to:
  - maintain banks’ safety and soundness,
  - ensure that they can continue lending to households and corporations, and
  - mitigate excessive procyclicality.
- Those exceptional regulatory measures were generally consistent with the core standards implemented after the global financial crisis and the interpretations offered by the standard-setting bodies.
- Rather than relaxing accounting and prudential standards, banking supervisors have largely used the flexibility embedded in the regulatory and accounting frameworks.

### Guidance on treatment of extraordinary support, payment moratoria, and ECL measurement
- Regulators provided guidance on:
  1. how extraordinary support measures should be treated,
  2. how payment moratoria need to be taken into account for asset classification and provisioning, and
  3. how expected credit losses (ECLs) should be measured in accordance with existing standards.
- Clarifications issued by several supervisors included:
  - The use of payment deferral programs should not result in an automatic trigger for a significant increase in credit risk and, thus, would not necessarily require a migration of loans from Stage 1 to Stage 2.
  - Banks should not apply the standard mechanistically. They should instead use the flexibility inherent in International Financial Reporting Standard (IFRS) 9, for example, to give due weight to long-term economic trends when estimating long-term expected losses.
  - Forward-looking information used to incorporate the effect of COVID-19 into the ECL estimate should be reasonable and supportable, taking into account the expected nature of the shock as well as the effect of the economic support measures, and the scarcity of available and reliable information.
  - Banks should make holistic assessments where they have limited borrower-specific information to assess whether exposures subject to payment deferrals have suffered a significant increase in credit risk, are credit impaired, or have defaulted.

### Lessons on ECL model limitations and practices observed during the pandemic
- The pandemic did not highlight systematic ECL model limitations, but several lessons emerged:
  - Changes in economic conditions should be reflected in macroeconomic scenarios applied by banks and in their weightings.
  - The unprecedented level of uncertainty during the COVID-19 pandemic could only be partially captured in ECL estimates because historical information was not representative of the future economic outlook (that is, there were no historical data for similar events that could have been used to calibrate the probability of default and loss given default parameters).
  - As the effect of COVID-19 was not reflected in models, postmodel overlays or adjustments were largely used to compensate for the lack of historical data, with multiple approaches being used by banks. For example, overlays have been used to reflect the effect of public support measures.
  - The IFRS paper on application of IFRS 9 in light of the COVID-19 uncertainty notes that IFRS 9 does not prohibit the use of postmodel adjustments, if doing so is consistent with the impairment requirements of the accounting standard. However, the diversity in practice within banks and the lack of transparency surrounding postmodel adjustments were identified as primary concerns. The different approaches adopted in various banks may have resulted in unjustified differences in terms of provisioning levels.
  - Although banks’ forward-looking scenarios were able to capture shocks at the macroeconomic level, they were not intended to reflect vulnerabilities in specific sectors or segments.
  - The lack of segmentation in models and data to reflect sector-specific risks was a common limitation in firms’ ECL methodologies.
  - Independent validation, model monitoring, and internal controls were not always performed during the pandemic.

### Recommendations and supervisory focus going forward
- The experience with COVID-19 underscores the importance of more detailed guidance on overlays and scenarios, as well as a review of provisioning practices:
  - Overlays are inevitable in some instances, but strict governance processes, internal controls, documentation, and disclosures are needed. Guidance on whether overlays should be applied at the risk parameter level or when the final output is calculated would also be useful.
  - The downside scenarios used to calculate ECLs should be sufficiently severe to capture tail risks.
  - The diversity in provisioning practices and differences in modeled provisions warrant further investigation.

*Source: tnmea2026004 - ANNEX 3. Lessons from the COVID-19 Pandemic.*

---


_Source: https://www.imf.org/-/media/files/publications/tnm/2026/english/tnmea2026004.pdf_