## 1. Key Considerations in Designing a Bank Diagnostic

## Source details

**Canonical URL:** [1. Key Considerations in Designing a Bank Diagnostic](https://www.imf.org/-/media/files/publications/tnm/2019/tnmea2019004.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tnm/2019/tnmea2019004.pdf.md)
- [Structured JSON version](/-/media/files/publications/tnm/2019/tnmea2019004.pdf.json)

---

### I. Rationale and Scope
- Purpose:
  - In times of macrofinancial stress, diagnostics of the condition of banks provide critical input to resolving problems, providing direction, and informing decisions.
  - Diagnostics centered around an AQR by independent experts should inform:
    - (i) how much bank capital is needed and how those needs will be met;
    - (ii) which banks receive public support, which can survive, and which should be closed; and
    - (iii) a communication strategy.
- Core focus:
  - Quantify size and distribution of problem assets and estimate current and future losses, including additional losses from:
    - (i) the cost of financing nonperforming assets;
    - (ii) further NPL formation and deterioration in loan recovery values;
    - (iii) excessive operational costs;
    - (iv) losses from extending fresh credit to keep problem borrowers from bankruptcy; and
    - (v) fraud and asset stripping.
- Common shortcomings in available information:
  - loan misclassifications;
  - overvalued loan collateral;
  - underestimated recovery costs;
  - overvalued securities;
  - uncertainty from deteriorating economic conditions.
- Diagnostics are ongoing processes; identifying capital shortfalls is necessary but often not sufficient—restructuring is required for viability.
- Local stakeholders often have incentives to underplay problems (bank managers/shareholders, supervisors, political leaders).
- In Fund-supported programs, diagnostics are commissioned by local authorities and must provide information beyond capital shortfalls for viability assessments.

### II. Design Considerations

- Building Block Approach (prioritization of activities):
  - Asset quality (good, bad, problematic; losses and provisions) and adequacy of capital;
  - Liquidity, including expected flows and management;
  - Sufficiency and quality of earnings (pre-provision profits net of accrued and capitalized interest on NPLs);
  - Capacity for loan workout and recovery;
  - Business base (clients, products, people, network, costing, pricing);
  - Financial group structure and ultimate ownership;
  - Scope and effects of intragroup and related-party transactions;
  - Adequacy of internal governance and Board processes;
  - Management quality and integrity of controls and systems.
- Risk Assessment:
  - Design diagnostics based on risk assessment covering related-party exposures, NPL recovery capacity, business-line profitability, intra-group relationships, sovereign exposure, and hedging/FX risk.
  - Initial ‘stocktaking’ should cover accounting practices, supervision reliability, judicial effectiveness, diagnostic components, sequencing, obstacles to resolution, information gaps, back-of-the-envelope financial needs, and technical assistance needs.
- Timing and relation to program financing envelope:
  - Financing envelope can be estimated ahead of a comprehensive diagnostic.
  - Risk-targeted AQR or desk-based stress test/sensitivity analysis can generate usable capital-need estimates while detailed reviews proceed.
  - Risk-targeted triage into:
    - (i) banks that can overcome stress with own resources;
    - (ii) banks requiring restructuring with public resources;
    - (iii) banks that should be closed.
  - Diagnostics can be phased; enhanced assisted inspections follow for restructuring plan validation.

- Ensuring Credibility:
  - Comprehensive TORs with detailed scope and methodologies.
  - Identify accounting and regulatory standards to be applied; ensure legal basis for enforcement.
  - Realistic macroeconomic forecasts for affordability and collateral values.
  - Disciplined project management with explicit governance.
  - Use external experts where local credibility is weak; local participation to enhance ownership.
  - Ex ante identification of remediation strategy and financing sources.

- Terms of Reference (TORs) key elements:
  - Scope (e.g., BSA and off-balance sheet plus profit and loss versus limited AQR).
  - Information to be provided (records, data tapes, macro scenarios).
  - Reporting standards (e.g., International Standards of Auditing).
  - Referent accounting standards (local GAAP or IFRS such as IFRS 9) and any required modifications.
  - Sampling and examination methodologies.
  - TORs must address reliability of loan classification and valuation, internal credit rating systems, provisioning, collateral valuation (independent valuations using recognized approaches), and creditor-rights effectiveness.

### BOX 2 — Work Streams in Full-Scope Bank Diagnostics (overview and sequencing)
- Typical work streams (examples):
  - (1) full-scope onsite inspections; (2) specialized audit reviews.
  - Modular European crisis approach: (1) balance sheet assessments (BSA); (2) asset quality reviews; (3) data integrity verifications; (4) loan-loss forecasts; (5) risk-weighted asset reviews; (6) loan-loss provision reviews; (7) forward-looking regulatory solvency stress tests.
  - Other work streams: distressed credit operations reviews; operating profit analysis.

### III. Key Technical Features

- Sampling:
  - Loan-file review and independent collateral valuation are most time-consuming.
  - Practical approach: individually review largest and less common loan files; segment remaining portfolios into homogeneous pools and sample.
  - Sampling considerations:
    - Weaker control environments require larger samples.
    - Loan reviews must trace loan proceeds and servicing to detect related-party issues and evergreening.
    - Sampled loans are re-underwritten individually (not via models).
  - TORs should be explicit on sampling methodology; extrapolation requires stratification.
  - Bank exam policies favor large coverage (above 40 percent of total exposure) for problem banks and material portfolios.

- Loss Dynamics and Loan-Loss Forecasts:
  - Point-in-time diagnostics give a static picture; forward-looking loss flows are critical.
  - Sources of losses beyond NPLs: large deposit withdrawals; continued lending to bad borrowers; forced deleveraging; operational cash-flow losses; asset stripping.
  - Loan-loss estimate perspectives:
    - Point-in-time to identify ‘clean balance sheet’.
    - One-year forward-looking estimates for regulatory capital.
    - Lifetime loss estimates for asset transfers to AMCs and under IFRS 9 expected credit-loss approach.
  - AQRs typically cause a step-increase in PDs and LGDs; forecasting approaches must address this.

- Accounting Considerations:
  - Accounting standards can be backward-looking or principle-based, allowing discretion and optimistic loss recognition.
  - Issues:
    - Accrual of unearned interest on bad loans distorts income and capital metrics.
    - Securities classification arbitrage (e.g., ‘held to maturity’) can hide mark-to-market losses.
    - Government bonds held at face value can defer loss recognition but may mask insolvency if market values reflect fiscal deterioration.
  - TORs should set conservative, rule-based criteria where local GAAP is insufficient.
  - Calibration trade-offs:
    - Excessively optimistic assessments → successive losses.
    - Excessively pessimistic assessments → undue fiscal burden.
  - Loan-loss provisions:
    - Standards vary; collective provisions often optimistic.
    - TORs should ensure representative loss-forecasting assumptions and sufficient collective provisioning, including IFRS 9 expected credit loss models.
    - A ‘central model’ can benchmark general provisions.

- Assessing Bank Viability:
  - Viability is assessed via a dual market and financial test:
    - Market test: tangible business base able to recover full cost allocation in normal conditions.
    - Financial test: reasonable certainty on asset quality and adequate provisioning; sustainable profitability; positive pre-provision profits to absorb normal credit costs net of accrued income on NPLs.
  - Viability drivers: management renewal; commercial franchise; balance sheet restructuring scope; ability to fund operations and recover assets; cost rationalization; appropriate recapitalization level.

### BOX 3 — Bank Viability Assessments (highlights)
- Business plans should be quantified, time-bound, and include:
  - moderate Texas ratio (comfortably below 100 percent);
  - capacity to underwrite/disburse new loans and recover/reduce NPLs;
  - no large mismatches or excessive concentrations;
  - ability to return to wholesale funding markets;
  - adequate capitalization and ability to attract private investors.
- Benchmarks:
  - Market average ROE of circa 8 percent (European crisis longer-term benchmark).
  - ECB target range of 6% to 10% ROE to assess viability.
- Stress testing:
  - Complements but does not substitute for AQR or business-model assessment.
  - Bottom-up preferred when banks cooperate; top-down applies uniform assumptions.
  - Regulatory capital horizon usually one year; supervisors often apply two- or three-year horizons.
- Liquidity:
  - Diagnostics focus on capital but must consider liquidity; IMF programs address liquidity via monitoring, lender-of-last-resort facilities, and bank-level funding plans.
- Communications:
  - Form intra-agency coordination; authorities should “speak with one voice”.
  - Disclosure strategy: maximize disclosure consistent with financial stability; detailed disclosure when credible backstops exist.
  - Internal/external channels: daily technical discussions, weekly management updates, shared Q&A, publish methodology and assumptions at the start.
- Post-diagnostic actions:
  - Supervisors enforce findings via regulatory solvency adjustments; banks submit capital action plans (convert subordinated debt, issue capital, or deleverage).
  - Public solvency support only under strict conditions; ideally limited to systemically important viable banks after upfront loss recognition and shareholder losses.

### BOX 4 — Funding the Irish Bank Diagnostic Exercise and Enhanced Capacity Development (selected figures and lessons)
- Central Bank of Ireland (CBI) spent about €31 million (0.7 bps of bank assets) over three and a half years on the Irish Financial Measures Program.
  - About one half spent in first year on AQR and loan-loss forecasts.
  - Other half spent over 2012–13 on follow-up activities (distressed credit operations review; Mortgage Arrears Review and Solution; development of loan-loss forecast models for 12 asset classes; PRISM implementation).
- ECB comprehensive assessment (2014) cost almost €23 million for Irish banks.
- 2015 report: total of €152 million (3.5 bps of bank assets) paid in consultancy fees related to clean-up by:
  - CBI: €85 million
  - National Treasury Management Agency: €41 million
  - Department of Finance: €22 million
  - National Pension Reserve Fund: amount not separately listed.
- Cost breakdown by phase (figures in €m):
  - Restructuring, recapitalization, and guarantees — Total: 59.0 (2008: 6.5; 2009: 14.2; 2010: 12.0; 2011: 18.2; 2012: 3.4; 2013: 3.5; 2014: 1.2)
  - Financial measures program — (2009: –28.1; 2010: 1.6; 2011: 1.1; 2012: 0.3; 2013: 31.1)
  - ECB comprehensive assessment — 2014: 22.6 22.6
  - Residential mortgage arrears — Total: 17.1 (2009: –0.0; 2010: 7.0; 2011: 8.7; 2012: 1.4)
  - Other — Total: 22.1 (2008: 0.1; 2009: 6.5; 2010: 2.7; 2011: 8.4; 2012: 1.6; 2013: 1.8; 2014: 1.1)
  - Total costs by year:
    - 2008: 6.5
    - 2009: 20.7
    - 2010: 14.8
    - 2011: 54.8
    - 2012: 13.5
    - 2013: 15.1
    - 2014: 26.6
    - Grand total: 151.9
- Outcomes: Irish banks restructured, regained market access, reduced NPLs, and returned to profitability; state retains ownership in some banks.
- Financing options for diagnostics:
  - Banks pay for exercise (e.g., special audit or fee in connection with business plans).
  - Supervisor pays and applies one-off levy on banks.
  - Donor funding in low-income countries.
  - Matching portion of a program loan with an IFI and local government funds.
- Governance and project management:
  - Independent third parties should play an active role; declare conflicts of interest.
  - Validation by independent quality assurance advisers is required.
  - PMO framework needed; PMO staffed with professional project managers; resident advisor beneficial.
- Lessons:
  - Depth of review and staff determine cost and credibility.
  - Avoid cost-minimization that undermines governance and quality.

### BOX 7 — The ECB/SSM Approach (selected features and parameters)
- CA phases: Phase 1 risk assessment; Phase 2 BSA (AQR); Phase 3 stress test; exercise took 12 months starting November 2013.
- Coverage:
  - Credit-file reviews covered 58 percent of credit risk-weighted assets for 128 banks.
  - Reviews covered loans, advances, financial leases, off-balance sheet items, letters of credit, shipping and project finance.
- Sampling and extrapolation:
  - Loan tapes used for stratification; findings extrapolated to homogeneous exposures.
  - Collateral revaluation for sampled debtors where recent third-party valuation absent.
- Challenger model:
  - ECB used a central challenger model to test collective provisions.
- Loan-loss forecasts and interest treatment:
  - Suspend interest service on impaired loans for forecasts; interest on forborne loans and NPLs not classified as impaired included in full.
- Appendix II — Sampling practice:
  - Coverage often targets 20 percent to 40 percent of total loan portfolio exposure.
  - Precision levels for proportional sampling: 5 percent to 20 percent; reliability levels: 80 percent to 95 percent (both in 5 percent intervals).
  - Numerical sampling precision: common levels 5 percent or 10 percent; reliability levels 90 percent or 95 percent.
  - ECB 2014 specifics: target sampling error < 5 percent of total post-adjusted provisions at 90 percent confidence, assuming adjustment ~25 percent; typical files sampled per portfolio: 250 to 450 files.

### APPENDIX V — Challenging Collective Loan Provisioning Inputs under IFRS (IAS 39) (key formulas and parameter guidance)
- Collective provisioning algorithm:
  - Provision = Exposure * [1 - (1-PD)
EP
* (1-CR) * LGL.
- Parameter definitions:
  - PD = probability of default (applies to performing and nonperforming loans; PD=1 for NPLs).
  - EP = emergence period for incurred but not reported provisions.
  - CR = cure rate (proportion of loans that default then return to performing).
- LGL for real-estate-backed loans:
  - LGL = C * (1-PTT + Current_REPI) * (1 - FSD) * DF * (1 - WOC).
  - Components:
    - C = collateral value at origination.
    - PTT = forecast peak-to-trough variation since origination.
    - REPI = current Real Estate Price Index.
    - FSD = fire sale discount.
    - DF = time discount value.
    - WOC = work-out costs to cash recovery.
- Diagnostic guidance and conservative parameter-setting:
  - Use conservative, justifiable, macro-consistent inputs; when empirical evidence lacking, use benchmark (e.g., fourth quartile) or rule-based parameters.
  - Tenets for TORs:
    - Segregate forborne/refinanced/restructured loans for PD and CR calculation.
    - Key stratification attribute: ability to service fully from own resources.
    - EP guidance:
      - EP of 12 months may be conservative and IFRS compatible.
      - EP may exceed 12 months where internal loan review cycles are longer or ineffective.
    - CR guidance:
      - Filter false cures from re-aging/refinancing.
      - Condition cure on sustained performance (example: full service of interest for 24 months with 10 percent to 20 percent principal reduction).
    - FSD and WOC guidance:
      - If doubtful, set conservative rule-based haircuts; absent justifiable estimates, set at the 50 percent to 30 percent level.
    - DF and TTL:
      - TTL may exceed six years in countries with protracted judicial procedures; set TTL to phase out collateral about five years from first 90 days past due.
    - Discount rate:
      - Ideally the effective loan interest rate; at minimum, not lower than average cost of marginal funding source.

*Source: tnmea2019004 — Technical Notes and Manuals 19/04 | 2019 (International Monetary Fund).*

### 1. Key Considerations in Designing a Bank Diagnostic ...................................................................

### 1. Key Considerations in Designing a Bank Diagnostic

### I. Rationale and Scope
- In times of macrofinancial stress, diagnostics of the condition of banks provide critical input to resolving problems, providing direction, and informing decisions.
- In countries that seek IMF support to correct macro-financial imbalances, meeting the ultimate program goals of restoring depositor and market confidence, financial stability, and conditions for growth, often requires restructuring and recapitalization of the banking system.
- The reliability of the information available on the condition of banks and their borrowers is usually insufficient to lay out a roadmap at the start of the IMF program; common shortcomings include:
  - loan misclassifications;
  - overvalued loan collateral;
  - underestimated recovery costs;
  - overvalued securities;
  - uncertainty from deteriorating economic conditions.
- Standard accounting disclosures are usually inadequate for forming a view on the true economic condition and prospects of institutions in situations of economic stress.
- Bank diagnostics, centered around an AQR process and undertaken by independent experts, provide input for a comprehensive strategy that should address:
  - (i) how much bank capital is needed and how those needs will be met;
  - (ii) a method for deciding which, if any, banks will receive public support, which could survive on their own, and which are not viable and should be closed; and
  - (iii) a communication strategy.
- The core focus of a diagnostic is to quantify the size and distribution of problem assets in banks and to estimate their current and future losses. Sources of potential additional losses include:
  - (i) the cost of financing nonperforming assets;
  - (ii) further nonperforming loan (NPL) formation and deterioration in loan recovery values;
  - (iii) excessive operational costs;
  - (iv) losses from extending fresh credit to keep problem borrowers from bankruptcy; and
  - (v) fraud and asset stripping.
- Diagnostics form part of an ongoing process rather than a point-in-time event. Identifying and remedying capital shortfalls is necessary but typically not sufficient; restructuring of fundamental bank problems is required for viable banks to support recovery.
- Local stakeholders are often incentivized to underplay problems: bank managers and shareholders, supervisors, and political leaders may all have motives to delay recognition.
- Diagnostics involve considerable expert judgement; valuation is a moving target, performed under pressure with limited data and resources. Assigning diagnostics to independent experts and carefully considering accounting conventions, valuation practices and assumptions, and local features is advisable.
- In the context of a Fund-supported program, diagnostics are commissioned by local authorities, must provide information beyond a capital shortfall, and provide input for assessments of each bank’s viability.

### II. Design Considerations

#### A. Building Block Approach
- Diagnostics should aim to identify underlying causes of banks’ problems and can be arranged as building blocks to provide needed information at each stage.
- A robust diagnostic should consider and undertake discrete activities prioritized by levels of identified risk in these areas:
  - Asset quality (good, bad, and problematic, including losses and provisions) and adequacy of capital;
  - Liquidity, including expected flows of funds, and how these are managed;
  - The sufficiency and quality of earnings (pre-provision profits net of accrued and capitalized interest on NPLs);
  - Capacity to work out and to recover problem loans (and good new loans);
  - The business base (clients, products, people, network, costing, and pricing);
  - The financial group structure (parent and holding companies above and subsidiaries below the bank) and its (ultimate) ownership structure;
  - Scope and (potential) effects of intragroup and related-party transactions;
  - The adequacy of internal governance and Board processes; and
  - Management quality and the integrity of controls and systems.
- In Fund programs where diagnostics are critical, they have generally prioritized asset valuation and bank solvency to gauge total financing needs of the program, assess fiscal options, and assemble a consistent financial envelope and overall program.
- Onsite reviews typically focus on valuing illiquid assets and addressing solvency-related issues to ensure banks can meet long-term financial commitments.

#### B. Risk Assessment
- The design of bank diagnostics should be based on an assessment of risks. Critical risk factors vary across countries and may include related-party exposures, banks’ capacity to recover value from NPLs, business line profitability, intra-group relationships, exposure to a weak sovereign, and poorly hedged security or foreign risk.
- The assessment should:
  - Identify early areas where problem assets may accumulate to calibrate scope and intensity of diagnostic reviews (Appendix I);
  - Evaluate the perimeter of entities under review to ensure intragroup relationships, related parties, and insider activities are covered;
  - Consider off-balance sheet contingencies (see Appendix I).
- In Fund programs, staff should undertake initial ‘stocktaking’ based on existing information, known gaps, and local issues to inform diagnostic design. The stocktaking should cover:
  - The role and quality of local factors, e.g., accounting practices, the reliability of supervision, and the effectiveness of the judicial system in executing creditors’ rights;
  - Identification of essential components of the diagnostic to calibrate scope and extension, and cover all key relevant risk areas (Appendix I);
  - Sequencing and early planning on additional steps following the diagnostic, e.g., validation of business plans and assessment of viability;
  - Consideration of potential obstacles to resolution, and any information gaps;
  - An early analysis and estimation (a back-of-the envelope calculation) of financial needs and possible capital requirements, and technical assistance needs.

BOX 1. Key Considerations in Designing a Bank Diagnostic Exercise
- Incentives: Key stakeholders are usually not interested in revealing losses. Bankers try to preserve shareholder value, maintain market liquidity, avoid revealing weak or corrupt lending practices, and keep their jobs through a “hide, hold, and hope for recovery” strategy; auditors may focus on retaining their client; supervisors on not losing face, while senior officials and policymakers may want to deny or minimize potential fiscal costs, playing for time and hoping to maintain relationships with powerful bank owners and managers.
- Standing and probity of ultimate owners: Shareholders that wish to retain control must show capacity and willingness to support the bank and act as a source of strength. Diagnostics must take account of efforts to obscure ultimate ownership, dubious financial standing or self-financed capital, and the possibility of underlying fraud. Bankers may also try to use political connections to weaken the scrutiny.
- Integrity and professionalism of Boards of Directors and senior management: Boards and managers may lack experience, display recalcitrant attitudes, and fail to respond diligently to the requests of supervisors, or benefit from inside lending, abuse, and fraud. Their record in placing the interests of depositors and creditors ahead of shareholders is critical to earning continued trust.
- Financial standing of the sovereign: Large exposures to a weak sovereign can represent a substantial risk to banks, while sovereign debt restructuring can potentially render whole banking systems insolvent. Debt sustainability concerns can limit the scope for using public resources for recapitalization.
- Resolvability: Impediments to resolution can include complex corporate structures (including cross-border), opaque intergroup relationships, undeclared related-party exposures, operational interdependencies, reputational risks, and hidden intra-group guarantees and shared fund flows. Legal or institutional impediments to foreclosure may erode the loss-absorbing capacity of collateral.
- Bank key control functions: Asset quality and the validity and reliability of data and information depend on control functions. Risks include mis-recorded or unrecorded assets and liabilities, structural gaps in credit underwriting, lack of reliability of loan review functions, and stratagems to hide losses or engineer false gains.
- Ineffective audit and onsite inspection: Weak external oversight through insufficient onsite testing of information and systems, lack of independence, stature, resources, and incentives of supervisors and auditors, and political capture or corruption, resulting in poor or unreliable information.
- Loan classification and provisioning rules: Evergreening (re-aging and refinancing); lender forbearance; inconsistent application (e.g., loose overdraft rules); arbitrary and optimistic general provisioning model inputs (including under International Financial Reporting Standards such as IFRS 9); lack of empirical loss evidence; and optimistic collateral valuations are all potential problem areas.
- Infrastructure (completeness and efficiency of systems): Litigation risks; inefficient collateral enforcement; unreliable judicial systems and property registration; lack of ancillary professions, such as valuation experts or bailiffs; or politically protected or loss-making public companies warrant consideration.
- Ineffective regulation: Inappropriate risk weights from internal-ratings-based models can underreport regulatory risk; nonconsolidated regulatory reporting can allow delinquent assets to be hidden from view in nonbank subsidiaries; and weak regulatory capital definitions can result in inability to absorb loss (e.g., excessive reliance on hybrid instruments or deferred tax assets).
- Macroeconomic expectations: Material expected changes in future exchange rates, interest rates and collateral value need to be factored in, as well as any expectation for continued economic deterioration that will impact on debt affordability.

#### C. Timing of Bank Diagnostics and Relation with Program Financing Envelope
- The financing envelope of an IMF program can be estimated ahead of completion of a comprehensive bank diagnostic.
- Bank diagnostics take time; supervisory and financial resources limitations, time constraints, and the urgency of finalizing program design sometimes justify settling on a risk-targeted AQR to produce an estimate of capital needs for program financing.
- A well-done expert estimation, possibly including a desk-based stress test or sensitivity analysis, can generate a usable estimate of capital needs and financial implications for the program while a detailed review is being completed.
- A risk-targeted approach can be used to map or triage banks into three groups:
  - (i) those that can overcome the stress with their own resources;
  - (ii) those that will require restructuring with public resources; and
  - (iii) those that should be closed.
- Enhanced assisted bank inspections can follow to provide efficient follow-up on restructuring plans, ensure greater ownership and political acceptance, and help build capacity.
- A more comprehensive suite of diagnostics may be necessary over the course of the program if local supervisors and auditors lack credibility, prudential standards are weak, tighter norms are used to estimate true bank conditions, or the banks have suffered a large shock affecting many business lines. Diagnostics can be phased as several basic building blocks through the program timeline (Box 2).

*Technical Notes and Manuals 19/04 | 2019 — IMF*

### BOX 2. Work Streams in Full-Scope Bank Diagnostics

### BOX 2. Work Streams in Full-Scope Bank Diagnostics

### Overview
- Diagnostics consist of several work streams sequenced over time.
- Comprehensive approaches followed in the past include:
  - (1) full-scope onsite inspections; and
  - (2) specialized audit reviews.
- More recently, during the European crisis, approaches have been divided into smaller blocks:
  - (1) balance sheet assessments (BSA);
  - (2) asset quality reviews;
  - (3) data integrity verifications;
  - (4) loan-loss forecasts;
  - (5) risk-weighted asset reviews;
  - (6) loan-loss provision reviews; and
  - (7) forward-looking regulatory solvency stress tests.
- Other relevant work streams include:
  - (1) distressed credit operations reviews (to assess the capacity of loan work-out and recovery activities); and
  - (2) operating profit analysis (to evaluate the profitability of the business model, and the capacity of pre-provision profits to absorb and normalize credit costs).

### D. Ensuring Credibility
- The bank diagnostic approach must be credible. Practices that help ensure credibility include:
  - Comprehensive terms of reference (TOR) addressing the scope and assessment methodologies, with sufficient detail on all essential elements, such as loan classification and provisioning to ensure objectivity, transparency, and rigor.
  - Identifying the accounting and regulatory standards that will be applied, which may involve prior passage of regulatory reforms, ensuring that there is an adequate legal basis for assessment and, critically, enforcement of capital requirements or other actions to remedy the shortcomings found by the review.
  - Realistic macroeconomic forecasts on variables related to affordability (e.g., GDP growth, unemployment, exchange rates, interest rates) and collateral values (residential real estate index, commercial real estate index, exchange and interest rates) to ensure consistent collateral values are applied and future environmental risks are fully captured.
  - Disciplined project management framework with an explicit governance structure (credible oversight).
  - Use of external experts if the evaluation, including the possibility of international firms or experts if the credibility of local audit and supervisory standards is in question, with local participation to bring in-depth knowledge and to enhance ownership.
  - Allocate sufficient resources (and political will) to review and redo work if gaps are identified.
  - Ex ante identification of the strategy and financing sources to remedy the problems found in the exercise; lack of clarity regarding the possible sources of funding for recapitalization can undermine the credibility of the exercise and lead to accusations that results have been “tailored” to fit within the resource envelope available.

### E. Terms of Reference
- TORs for the bank diagnostic set out the scope, depth, timing, and sequencing of the exercise. They are technical Memoranda of Understanding among stakeholders on how the diagnostic will be conducted.
- In a program context, this document is drafted by the authorities and their advisors in consultation with IMF staff. It is sometimes produced with the assistance of third-party consultants.
- The TOR should be designed to enable a rigorous assessment of key variables similar to supervisory CAMEL ratings:
  - (i) net Capital;
  - (ii) Asset quality;
  - (iii) Management;
  - (iv) Earnings; and
  - (v) Liquidity flows.
- The typical structure of the TORs will include details on:
  - The scope of the diagnostic (e.g., BSA and off-balance sheet plus profit and loss versus limited AQR);
  - The information to be provided to the reviewers (e.g., records, data tapes, macroeconomic scenarios, etc.);
  - The reports, opinions, and information to be delivered by the reviewers;
  - The standards of field work and reporting (say, International Standards of Auditing);
  - The referent standards of accounting (say, local Generally Accepted Accounting practices (GAAP), or International Financial Reporting Standards (IFRS), such as IFRS 9);
  - Any modification required to the above as per detailed agreed-upon procedures; and
  - The specific methodologies and procedures requested for sampling and examination.
- TORs must define clearly the scope of the diagnostic. For a BSA, the scope should include assurances that all liabilities and contingencies are accounted for.
- Key issues and prescriptions in the TOR:
  - Reliability of loan classification and the valuation of loans is almost always the key issue.
  - Exclusions should be justified based on the initial risk assessment (e.g., cash balances due from banks or in correspondent accounts may need direct verification).
  - Performance of internal credit rating systems and the adequacy of loan-loss buffers should always be a key focus.
  - TORs should include assessment of credit risk infrastructures, objective criteria and risk drivers for grading loans and provisioning.
  - Loss estimates should be based on conservative assumptions, with recovery values based on the expected value at the time of liquidation, taking account of forecast changes in collateral market prices and currency fluctuations, as well as conservative discounts for forced-sale and time-value of money effects.
  - Independent valuation of collateral is usually necessary and the TOR should require independent valuation by accredited agents, using an internationally recognized valuation approach (such as International Valuation Standards). Valuations should be limited to the income (for assets with cashflows) or market approach in distressed environments.
  - The diagnostic needs to consider the effectiveness of creditor rights and institutional frameworks and their impact on collateral value through the ease and speed of foreclosure, seizure, and sale.
  - The assessment must test the reliability and credibility of methodologies used to estimate loss provisions (individual and general, i.e., collective, “incurred but not reported,” and lifetime expected credit loss). Assessment of credit information, loan underwriting, internal loan review and risk controls, and policies and capacity for loan recovery and debt restructuring should ideally be covered.

### III. KEY TECHNICAL FEATURES

#### A. Sampling
- Loan file review and independent valuation of collateral are the most time-consuming and resource-intensive tasks of a bank diagnostic.
- Practical approaches:
  - Largest and less common loan files are typically reviewed individually.
  - Other loan portfolios are often segmented into pools of homogeneous risk for sampling; similar approach applied to collateral valuations.
  - Segmentation groups loans with similar risk characteristics—adjustments to the sample set are extrapolated to the rest of the segment.
- Sampling considerations:
  - Appropriate sampling depends on effectiveness of banks’ internal controls and reliability of data. The weaker the control environment, the larger the sample size should be, especially if forensic reviews are performed.
  - Reading a loan file is usually insufficient; tracking loan proceeds and sources of loan servicing is essential to uncover problem debtors or related-party issues.
  - Extra time must be factored to trace transactions and discover insider transactions.
  - Sampled loans are re-underwritten one by one, not with models. Loan reviews are the only way to identify forbearance and evergreening practices and to identify NPLs whose interest revenue is accounted for but never paid in cash.
  - TORs must decide whether to extrapolate from samples or to achieve high loan coverage with a bias toward higher direct loan coverage.
- Sampling methodology:
  - TORs should be explicit regarding sampling methodology and selection technique. There are three basic approaches to sampling (Appendix II).
  - Extrapolation requires stratification into homogeneous pools.
  - Bank examination policies favor large coverage (above 40 percent of total exposure) for problem banks and across all credit portfolios, including retail portfolios identified as material during the initial risk assessment.

#### B. Loss Dynamics and Loan-Loss Forecasts
- Point-in-time diagnostics provide a static “picture” but understanding future flows (sources of losses) is critical.
- Risks if underlying sources of loss are not addressed:
  - Heavy losses on existing defaulted loans may be concealed if future declines in collateral values and adverse changes in foreign exchange and interest rates are not fully considered.
  - Further deterioration in economic conditions could lead to significant new NPL formation.
  - An AQR plus a loan-loss forecast will not be enough if sources of loss are not addressed through restructuring of bank operations to ensure long-term viability.
- Sources of bank losses beyond NPLs:
  - Large deposit withdrawals;
  - Continued lending to bad borrowers;
  - Forced deleveraging;
  - Operational cash-flow losses; and
  - Other forms of asset stripping.
- Viability notes:
  - A bank may be insolvent but still have access to funding and thus continue to operate and generate new losses.
  - A recapitalized bank with stable liquidity may still be making losses and may not be viable long term if operating costs exceed revenues.
- Loan-loss estimates perspectives:
  - Point-in-time estimate often used to identify the ‘clean balance sheet’ before a solvency stress test and to clarify if past financial reports need restatement.
  - Estimates of future losses are needed for regulatory capital, which is based on a one-year forward-looking estimate.
  - Lifetime loss estimates used as input to identify transfer values when moving assets to an asset management company (AMC) and are becoming more common with IFRS 9 expected credit-loss approach.
- AQR design implications:
  - AQRs typically lead to adjustment of loan classifications and losses at the reference time point, resulting in a step-increase in probabilities of default (PDs) and losses given default (LGDs) that presents technical challenges for loss forecasting.
  - The step-increase can be overcome by adjusting historically reported NPL levels or careful selection of the NPL forecasting approach.
  - Selection of approach depends on information availability; different forecasting models are often applied to different asset classes.

#### C. Accounting Considerations
- Accounting standards provide scope for optimistic views on losses; A key objective of AQRs is to obtain a reliable estimate of problem assets while undertaking a prudent estimation of loss.
- Accounting standards characteristics and implications:
  - Are often backward-looking (for example, International Accounting Standard (IAS) 39 allows provisioning only for incurred losses). This can result in concealment of losses in weak macroeconomic environments as collateral values lag. This feature led the European Union to go beyond accounting standards and consider expected losses in the Comprehensive Assessment. Even if forward looking (e.g., IFRS 9), loss forecasts may not be based on consistent and conservative macroeconomic forecasts.
  - Are principle-based, providing discretion in loss recognition, particularly in the absence of supervisory guidance.
  - Favor income accrual over cash revenue recognition for problem assets. Accrual of unearned interest income on bad loans can significantly distort interest income, loan-loss provisioning coverage ratios, and overall capital levels.
  - Authorities normally set out little guidance in terms of implementation (see Appendix V on collective provisions under IFRS).
- Valuation of securities and trading assets depends on portfolio classification. Potential accounting arbitrages include:
  - Recording assets as ‘held to maturity’ to avoid mark-to-market losses.
  - Recycling portfolios across portfolios and entities to place latent losses outside the review scope.
- Government bonds:
  - Government bonds held as bank assets are a special case; market value drops during fiscal deterioration can make the banking system insolvent if market values are used.
  - Allocating bonds to the investment account at face value could defer future loss recognition but is sustainable only if financing cost is affordable.
  - One approach is segregating existing holdings (marked to market) and new recapitalization bonds (held at face value), acknowledging liquidity disadvantages.
  - Requirement to mark-to-market should be seen in context of overall strategy for the banking system.
- Local accounting norms should be interpreted conservatively. If local GAAP does not resolve valuation issues, TORs should provide precise ad hoc rule-based criteria to appraise bank portfolios (example: Spanish AQR in 2012 using stricter criteria for refinancing and forbearance, including rule-based haircuts).
- Calibration trade-offs:
  - Excessively optimistic assessments can lead to successive losses and insufficient remedial measures.
  - Excessively pessimistic assessments can lead to unnecessarily high burdens on banks and unwarranted public recapitalizations with negative fiscal, credit, and economic effects.
  - Appendix V shows an example using eight different parameters, each of which can be “optimized” to bias the result.
- Loan-loss provisions:
  - Accounting provisions are critical to loss recognition, but standards and oversight vary considerably across national borders.
  - Provision creation typically leads to a loss in banks’ net asset value through reduction in measured loan values.
  - Given lack of international agreement on provisioning rules, there is substantial scope for variation and optimistic assumptions.
  - General provisions are usually based on limited and optimistic data.
  - TORs should specify an approach to ensure loan classifications and loss-forecasting assumptions are representative, and that collective provisioning is sufficient (including expected credit loss models under IFRS 9).
  - In retail segments, provisioning is based largely on modeling algorithms using roll rates, migration analysis, aging of arrears, loan-to-value ratios (LTVs), and other techniques for pool provisioning; these need careful assessment.
  - Comparing deviations across banks and products helps visualize conservatism in model inputs and management estimates. This requires investigating re-aging and refinancing practices.
  - Use of a ‘central model’ can provide a benchmark to test sufficiency of general provisions (example: Central Bank of Ireland in 2013; ECB generalized this approach in 2014 comprehensive assessment and developed a challenger model approach).

#### D. Assessing Bank Viability
- Restructuring problem banks to restore viability is a key challenge; achieving viability is critical to restoring financial stability.
- Viability is the central consideration in deciding the future of a bank (whether it is closed or restructured).
- Key factors in restoring viability include:
  - Renewal of the management team;
  - The bank’s underlying commercial franchise and service levels;
  - Scope for balance sheet restructuring;
  - Ability to fund operations and to recover problem assets;
  - Scope for cost rationalization; and
  - An appropriate recapitalization level (Box 3).

*Source: tnmea2019004 - BOX 2. Work Streams in Full-Scope Bank Diagnostics*

### BOX 3. Bank Viability Assessments

### BOX 3. Bank Viability Assessments

### Definition of viability
- A bank is considered to be viable if it passes a dual test of market and financial viability.
- Market test: demonstrate a tangible business base of clients and products sufficient—in normal conditions—to recover full cost allocation (including expected credit costs and the cost of capital).
- Financial test: reasonable certainty regarding asset quality and adequate provisioning, with actual performance (or within a reasonable timeframe) demonstrated by:
  - Sustainable profitability, with growing net interest margin and return on equity normalizing to market rates.
  - Positive pre-provision profits to absorb normal credit costs net of accrued income on NPLs.

### Business plans and diagnostics
- Common practice: combine bank diagnostics with bank-provided business plans, recognizing plans are generally optimistic.
- Business plans should include the following metrics:
  - A moderate Texas ratio (i.e., comfortably below 100 percent).
  - Capacity to underwrite and disburse new loans, to recover problem loans and to reduce NPLs.
  - No large balance sheet mismatches or excessive concentrations.
  - Ability to return to wholesale funding markets.
  - Adequate capitalization and ability to attract private investors.
- Business plan content requirements:
  - Quantified, time-bound, with performance indicators.
  - (1) Line activity: market segment shares, turnover, number of clients and accounts served, and expected volumes.
  - (2) Financial, capital and liquidity plans.
  - (3) Forward-looking projections under alternative scenarios (base and adverse).
- Benchmarks and examples:
  - Market average ROE of circa 8 percent used in the European crisis as a longer-term benchmark.
  - ECB target range of 6% to 10% ROE to assess viability.
- Business-plan assessment should:
  - Quantify impact of recapitalization instruments (e.g., cash, bonds, hybrid instruments, deferred tax assets) on profitability, cash flows, and liquidity.
  - Track evolving balance sheet structure (divestitures, deleveraging) and impact on earnings capacity.
  - Monitor pace and effectiveness of problem loan workouts.
  - Reduce cost base by aligning staff and branch network to realized business volumes.
- Aggregate coherence: benchmark assumptions across banks to ensure deposit and lending growth rates are consistent with macroeconomic outlook.

### Stress testing and regulatory solvency
- Role and limits:
  - Stress testing models future loss potential and assesses resilience; complements AQRs but cannot substitute for an AQR or business-model assessment.
  - Stress tests can assist triage, threshold-setting for public funds, initial bank classification, and crisis-management structure.
- Approaches:
  - Bottom-up: exercises conducted by individual banks under supervisor oversight; when rigorous, provide more granular and accurate results and are preferable when banks cooperate fully.
  - Top-down: supervisor-designed high-level exercises applying same assumptions and models to all banks.
  - Dynamic balance sheet approach: accounts for restructuring-driven changes in profit and loss, revenue, costs, and regulatory risk weights.
  - Static balance sheet approach: assumes maturing assets/liabilities are replaced with similar instruments and business mix/model is maintained during the horizon.
- Scenarios, horizons, and thresholds:
  - Tests often apply a baseline and an adverse scenario (adverse typically applies a shock such as two-standard deviations or ninety-fifth percentile shock to GDP growth).
  - Regulatory capital horizon usually one year (as per Basel), but supervisors typically apply a two- or three-year horizon.
  - Pass/fail thresholds applied to regulatory capital ratios in both baseline and adverse scenarios; severity defines conservatism.
- Inputs and governance:
  - Loan-loss forecast inputs should be produced by independent experts (typically those engaged to undertake the AQR) to ensure prudent and consistent treatment and provide legal protection for supervisors.
  - Stress-test results should be considered alongside other diagnostics; stress tests alone may create wrong incentives or fail to challenge input data or capture nonlinear shock effects.

### Liquidity
- Bank liquidity is a critical assessment factor; while diagnostics focus on capital, liquidity failures often trigger bank failures.
- In IMF programs, liquidity issues are typically addressed with:
  - Comprehensive monitoring frameworks.
  - Lender-of-last-resort facilities.
  - Bank-level funding plans.

### Communications
- Coordination and governance:
  - Effective coordination among national authorities is required before diagnostics begin; form an intra-agency group of senior central bank and government representatives to develop the financial sector strategy and identify creditor loss-bearing and support modalities.
- Messaging and disclosure:
  - Articulate a clear narrative on why the exercise was needed, approach taken, findings, and remediation.
  - Authorities should “speak with one voice” to avoid conflicting messages.
  - Aim for maximum disclosure consistent with financial stability; limited disclosure may be necessary where credible capital and liquidity backstops do not exist.
  - When credible backstops are in place and understood, detailed disclosure of methodology and bank-specific results can aid investors and counterparties.
  - Recent European practice: ECB and EBA publication of full AQR and stress-testing methodologies and granular bank-condition information.
- Internal and external channels:
  - Formalize supervisor–bank communications during the exercise via daily technical discussions, weekly management updates, and a shared Q&A document.
  - Supervisors typically publish methodology and assumptions at the start and expect banks to publish disclosures when supervisor publicly communicates results.

### Post-diagnostic actions
- When diagnostics identify regulatory capital shortfalls:
  - Supervisors use regulatory solvency to enforce diagnostic findings; adjustments to asset values lead to expected upward loan-loss provisions and downward regulatory capital.
  - Regulatory adjustments can be made in advance through Pillar II add-ons.
  - Banks must submit capital action plans to address shortfalls via converting subordinated debt to equity, issuing fresh capital, or deleveraging.
  - Non-compliant banks face resolution or, in exceptional circumstances (e.g., high spillovers), restructuring with state support.
- Public solvency support conditions:
  - Provide only under strict conditions that maximize burden sharing, minimize moral hazard, and protect taxpayers.
  - Ideally limited to systemically important banks that are deemed viable, after upfront loss recognition and losses borne by existing shareholders.
  - Systemic-importance assessments based on size, interconnectedness, substitutability, and complexity.
  - Viability assessments require judgment on bank business plans including earnings, liquidity, and capital projections.

### Implementation issues: timing and costs
- Timing:
  - Bank diagnostics during a Fund-supported program should start as early as possible.
  - Staff should aim for diagnostics finalized before the half-way point of the program and agree scope, timeline, and TORs ex ante.
  - Agree in advance how diagnostics will be paid for and establish robust governance and oversight.
- Cost estimates:
  - No precise cost estimates; depending on circumstances, total costs may be in the range of 2 bps to 7   bps of the total nominal value of the balance sheets under review.
  - Cost drivers include nature of investigations (forensic vs going-concern), asset-portfolio size, sampling requirements, head-count, and staff specialization.

*Source: tnmea2019004 - BOX 3. Bank Viability Assessments*

### BOX 4. Funding the Irish Bank Diagnostic Exercise and Enhanced Capacity

### BOX 4. Funding the Irish Bank Diagnostic Exercise and Enhanced Capacity Development

### Funding and costs: Ireland case
- The Central Bank of Ireland (CBI) spent about €31 million (0.7 bps of bank assets) over three and a half years to design and implement the Irish Financial Measures Program.
- About one half of the total was spent in the first year, performing the AQR and loan-loss forecasts, including associated data integrity verifications.
- The other half was spent over 2012–13 on follow-up activities, including:
  - (1) a distressed credit operations review;
  - (2) a Mortgage Arrears Review and Solution program;
  - (3) development of loan-loss forecast models at the CBI for 12 asset classes and implementing a new risk-based supervisory framework (PRISM).
- The investment enhanced the overall quality of program implementation and its credibility, including reinforcing the CBI’s overall capacity.
- The subsequent ‘comprehensive assessment’ diagnostic of the ECB, completed in 2014, cost almost €23 million for the Irish banks.
- A 2015 report from the Irish Comptroller and Auditor General states that a total of €152 million (3.5 bps of bank assets) was paid out in consultancy fees related to clean-up of the Irish banking system, to various law firms, accountants, and banking experts by:
  - the CBI: €85 million
  - the National Treasury Management Agency: €41 million
  - the Department of Finance: €22 million
  - and the National Pension Reserve Fund (amount not separately listed in the table).

### Cost breakdown by phase (figures in €m)
- Restructuring, recapitalization, and guarantees:
  - 2008: 6.5
  - 2009: 14.2
  - 2010: 12.0
  - 2011: 18.2
  - 2012: 3.4
  - 2013: 3.5
  - 2014: 1.2
  - Total: 59.0
- Financial measures program:
  - 2009: –28.1
  - 2010: 1.6
  - 2011: 1.1
  - 2012: 0.3
  - 2013: 31.1
- ECB comprehensive assessment:
  - 2014: 22.6 22.6
- Residential mortgage arrears:
  - 2009: –0.0
  - 2010: 7.0
  - 2011: 8.7
  - 2012: 1.4
  - Total: 17.1
- Other:
  - 2008: 0.1
  - 2009: 6.5
  - 2010: 2.7
  - 2011: 8.4
  - 2012: 1.6
  - 2013: 1.8
  - 2014: 1.1
  - Total: 22.1
- Total costs by year:
  - 2008: 6.5
  - 2009: 20.7
  - 2010: 14.8
  - 2011: 54.8
  - 2012: 13.5
  - 2013: 15.1
  - 2014: 26.6
  - Grand total: 151.9

### Outcomes and lessons from the Irish experience
- The Irish banks successfully restructured, regained market access, reduced NPLs to manageable levels and have returned to profitability.
- Although the state retains ownership in the banks, the experience demonstrates that targeted measures combined with effective implementation justify the upfront costs.

### Ways to finance a bank diagnostic (options listed)
- Requiring the banks to pay for the exercise, for example in connection with presentation of business plans or through an expanded (special) audit based on agreed-upon procedures and standards.
- Supervisor pays for the reviews and applies a special one-off levy on the banks concerned to recover costs.
- For low-income countries, requesting donors to fund the exercise.
- Matching a portion of a program loan with an IFI with local government funds for implementation tasks.

### Cost and credibility considerations
- Depth of review and the staff that perform it are key drivers of the overall cost envelope.
- The budget dedicated to paying for bank diagnostics determines timeliness, quality, and reliability of the exercise.
- The standing of firms responsible for the review influences credibility of results.
- In systemic crises, explicit PMO and quality assurance budget allocations are needed to enhance credibility and oversight.
- High fees charged by specialist firms or international audit companies complicate convincing authorities to undertake diagnostics.
- It is advisable not to let cost minimization drive decisions; reducing costs or shortcuts on governance have often led to bad outcomes.

### Governance of bank diagnostics
- Independent third parties (independent of banks and supervisors) should have an active role throughout the diagnostic to provide impartial technical expertise, operational capacity, and credibility.
- Local third-party expertise often needs supplementation by external experts when local expertise is not considered independent or credible.
- Third-party vendors should declare direct and indirect conflicts of interest at the outset.
- Validation should be undertaken by an independent party in a quality assurance advisory role, providing guidance on:
  - (i) methodologies;
  - (ii) sampling techniques;
  - (iii) oversight of diagnostic reviews.
- The validator needs close engagement with work and access to portfolio categories, credit files, and bank data tapes; material issues identified must be addressed before finalizing results.
- External parties should be part of a broader governance structure to promote transparency and rigorous implementation; they should opine on compliance with best practice, TORs, and agreed guidance and methodologies.

### Role of banking supervision
- Local authorities/supervisors play an important role; participation expands coverage and increases credibility.
- Supervisory resources can be augmented by private contractors or the exercise can be fully outsourced, but assisted onsite inspections often balance local ownership and efficiency.
- TORs for execution, including a specific roadmap, should be agreed under strong coordination and rigorous oversight.
- Inclusion of local supervisors enhances ownership and brings local knowledge; excluding them risks undermining ownership of restructuring programs except in cases of egregious integrity failures.
- Supervisors typically hold the best available financial information and market intelligence and understand local regulatory and market nuances.
- Assisted inspection programs provide opportunities to transfer know-how and build local capacity and to advocate diagnostic-based reform measures with government stakeholders.

### Project management
- Intensive diagnostics of multiple banks require a disciplined PMO framework.
- PMO should be attributed to a competent official agency unit and staffed with professional or certified project managers.
- Technical assistance or direct funding from IFIs or other sources may be required to support an effective PMO.
- Placing a resident advisor in the PMO unit is beneficial if resources permit.

### International experience overview (high-level points)
- Bank diagnostics have been recommended by the Fund and World Bank since the 1980s and were essential in official sector support programs.
- Early diagnostics faced challenges due to lack of explicit PMO and dedicated quality assurance, complicating oversight.
- Diagnostics evolved to include “institutional review” and management contracts in transition economies, and the “business plan” diagnostic model in later crises.
- Common complications historically:
  - (i) public scrutiny and high expectations;
  - (ii) high costs and difficulties in monitoring and quality control;
  - (iii) large inaccuracies in estimates of capital needs;
  - (iv) dependence on accounting criteria;
  - (v) difference between recapitalization needs and overall financing envelope;
  - (vi) difficulties valuing assets in unstable conditions;
  - (vii) static, point-in-time estimates rather than forward-looking viability assessments.
- Euro area crisis diagnostics combined AQRs with loan-loss forecasts, stress testing, wider public exposure, intensive use of private contractors, reinforced PMO, and stronger governance focused on quality assurance.
- The ECB’s 2014 Comprehensive Assessment (CA) comprised:
  - (i) a risk assessment to decide which portfolios to scrutinize;
  - (ii) an AQR applying a standard manual;
  - (iii) stress tests undertaken with the EBA.
- The CA publicly disclosed process, timeline, manuals, and methodologies and published comprehensive results in advance.

*Source: BOX 4. Funding the Irish Bank Diagnostic Exercise and Enhanced Capacity Development (Technical Notes and Manuals 19/04).*

### BOX 7. The ECB/SSM Approach

### BOX 7. The ECB/SSM Approach

### Overview
- The comprehensive assessment (CA) began in November 2013 and covered all large European Union banks that would be supervised by the SSM. The CA comprised three phases: Phase 1 — risk assessment; Phase 2 — an AQR referred to as a balance sheet assessment or BSA; and Phase 3 — a stress test. The exercise took 12 months to complete.
- Objectives: provide the SSM with assurances that supervised banks were adequately capitalized, viable, and reporting accurate information; foster transparency; and build market confidence.
- The BSA manual provided a consistent basis for national supervisors and advisory firms to conduct the AQR in a consistent manner.
- The BSA manual was developed with the assistance of independent experts and was heavily influenced by prior AQRs in Ireland and Spain.

### Credit-file review and coverage
- Credit files were sampled from a “loan tape” provided by each bank containing loan information such as segment, classification, status, and credit performance data; data reliability was assessed.
- Field attributes in the loan tapes enabled stratification of portfolios into pools of assets with similar characteristics to build loan samples.
- The reviews covered 58 percent of credit risk-weighted assets for the 128 banks subject to the comprehensive assessment.
- The credit file reviews covered all loans, advances, financial leases, and other off-balance sheet items, including letters of credit and specialized asset finance such as shipping and project finance.
- Loan sampling was employed because of large data volumes; sample size depended on portfolio and individual loan characteristics.
- National competent authority bank teams verified correct classification of credit exposures.
- Expected future losses were estimated for incorporation into the stress test.

### Collateral, valuation, and extrapolation
- Collateral valuation was a key factor. Generally, collateral was revalued for all debtors selected in the sample where a third-party valuation less than a year old was not available; national competent authority bank teams carried out revaluations and fed results into other exercise components.
- Findings of the credit file review were extrapolated to the wider portfolio to assess adequacy of specific provisions. Projections from samples were applied to homogeneous exposures (in line with standard audit guidelines).

### Challenger model and provisioning
- The ECB used a central “challenger model” to challenge sufficiency and sensitivity of collective provisions.
- The approach reviewed each bank’s collective provision model for adherence to accounting principles; differences between a bank’s estimation and the challenger model provision were analyzed and reconciled.

### Loan-loss forecasts and interest treatment
- Loan-loss forecasts and stress tests performed with the BSA provided a forward-looking view of earnings.
- Forecast instructions: suspend interest service on impaired loans (contrary to IAS 39 which forces accrued interest on impaired loans to be unwound ex post); interest on forborne loans and NPLs not classified as impaired was included in full.

### Appendix I — Balance sheet assessment: issues of concern (selected)
- Group issues / Out of scope activities
  - Screen consolidation perimeter for groups and type of group under review (all key units included).
  - Possible impacts: losses from assets and liabilities left out of review scope (insurance, AMCs); unrecorded intra-group guarantees; contingent liabilities erode bank capital.
- Securities / Investment
  - Valuation: valuation basis not shown on a mark-to-market basis (e.g., government securities); banks transferring securities from trading to held-for-investment or off-balance-sheet vehicles to avoid marking-to-market.
  - Possible impacts: additional provisions needed or hidden reserves; additional loss to provision; effect of sovereign downgrades.
- Placements with banks
  - Liquidity and valuation concerns: need to screen placements for lending by local parents to stressed subsidiaries; extended maturity terms concealing losses.
  - Possible impacts: liquidity pressure if subsidiaries cannot repay parent loans; additional loss to provision; hidden encumbrance of correspondent and due-from bank accounts.
- Loans
  - Valuation concerns: amounts of classified loans transferred to unconsolidated subsidiaries; rollovers used to conceal impairment; overdrafts used to rollover unpaid balances.
  - Possible impacts: additional NPLs and losses to provisions on consolidation; overdrafts keeping NPLs current; distortions in provisioning models (forbearance; cure rates; emerging period adjustments).
  - Collateral issues: ascertain existence and legal perfection of collateral liens; high potential for additional losses once LTV goes beyond 100%.
- Liquidity issues
  - Ascertain share of residential mortgages pledged to covered bonds and default triggers; possible forced transfer of good residential mortgages to secure collateral shortfalls.
- Consolidation and collateral
  - Unrecognized hidden losses in subsidiaries reduce parent equity; inflated collateral values without PV discount and related costs; losses at foreclosure unaccounted for.
- Macro and SME/retail exposure
  - Impact of slowdown on SME and retail portfolios and possible sharp increase in NPLs and additional provisions.
- Equity investments, accruals, and other assets
  - Valuation of investments in subsidiaries; write down of goodwill and investments reduce capital.
  - Suspension of interest accruals on NPLs: accounting location and revenue impact could result in write-offs, reduction in income, and lower pre-provision profits.
  - Valuation and location of other real estate owned, deferred tax assets, goodwill, bad/aged inter-branch items, letters of credit, unfunded loan commitments, and derivative exposures could require additional provisions or charge-offs.

### Appendix II — Sampling techniques and practice (key points and parameters)
- Asset-quality reviews typically apply three sampling methods: (i) judgmental (non-statistical); (ii) statistical proportional; and (iii) statistical numerical.
- Overall coverage often targets from 20 percent to 40 percent of the total loan portfolio exposure; considerations in determining coverage include quantity/quality/nature of population, the bank’s risk management systems, risk appetite, net benefits of methods, purpose/objective, and resource constraints.
- Precision and reliability levels affect sample size:
  - For proportional sampling, reasonable precision levels vary from 5 percent to 20 percent, and reliability levels from 80 percent to 95 percent; both in 5 percent intervals.
  - For numerical sampling, common precision levels are 5 percent or 10 percent, and reliability levels of 90 percent or 95 percent.
- Judgmental sampling has no statistical measurement and cannot statistically relate sample results to the entire population.
- Use cases:
  - Proportional sampling: appropriate when total amount of items is relevant (discover additional classified loans in commercial portfolios; test population items for monetary or compliance errors).
  - Numerical sampling: defines sample by number of items; used to reveal presence of a defined characteristic and to determine frequency of errors when dollar amount is not relevant.
  - Proportional or numerical sampling adequate to validate internal loan reviews, evaluate credit quality compliance with underwriting standards, and accuracy of internal risk rating.
  - Numerical sampling adequate to test accuracy of management information systems, renewals/deferrals/extensions compliance, score overrides compliance, and recently extended loans compliance.
- ECB 2014 AQR sampling specifics:
  - Sampling approach based on statistical techniques minimizing sample size through stratification subject to a low sampling error expected to be less than 5 percent of total post-adjusted provisions, with a level of assurance of at least 90 percent.
  - After deducting high-quality exposures, the ECB mandated a sample for each portfolio divided into 49 strata differentiated by size of debtor-level exposure and by risk.
  - A sample is selected from each stratum by (1) exposure buckets set based on the bank’s portfolio composition; and (2) risk buckets set using basic risk indicators (e.g., past due).
  - The number of files selected from each stratum was typically set to target a maximum error of 5 percent in post-adjustment provisions at a 90 percent confidence level, assuming an adjustment to provisions of around 25 percent.
  - The number of files sampled from each stratum varies depending on concentration of portfolio, number of observations in the strata, and riskiness of the strata.
  - In general, the number of files expected to be sampled for each portfolio typically ranges between 250 and 450 files.

### Spain AQR (2012) — sampling and scope highlights
- Originally planned as a stand-alone AQR, Spain’s 2012 bank diagnostic included a solvency stress test, which forced increased sampling including random sampling in portfolios.
- Not a full balance sheet review: deferred tax assets, security investments, goodwill and exposures to the public sector and other financial institutions were excluded.
- Key features:
  - Limited to an assessment of credit risk only.
  - Included the large exposures in all portfolios.
  - Included the 50 largest NPLs (doubtful exposures), the 50 largest new entrants, and the 50 largest existing (reclassified or repaid) exposures in that category.
  - Instructions to review at least 500 loan files per bank, with the “group” or “unit of risk” concept resulting in review of many more individual loan files corresponding to obligors linked in groups.
  - A total of 117,000 loan files were reviewed for the 17 target banking groups (6,888 files on average).
  - Appraisals of collateral for real estate were also sampled, and 1.7 million items were re-based.
  - Random sampling expanded to facilitate stress tests undertaken by Oliver Wyman to enable extrapolation of reclassifications and refinanced loans (to recalibrate PDs and LGD).
  - The Bank of Spain coordinated the AQR and resolved auditor queries.
  - The AQR applied loan classification criteria in Bank of Spain Circular CBE 4/2004.
  - Random sampling in residential mortgage and SME portfolios used “monetary unit sampling”; accounting reconciliation and loan-by-loan data integrity validation of sampled files were undertaken.

*Source: BOX 7. The ECB/SSM Approach (tnmea2019004).*

### APPENDIX V. CHALLENGING COLLECTIVE LOAN PROVISIONING INPUTS

### APPENDIX V. CHALLENGING COLLECTIVE LOAN PROVISIONING INPUTS UNDER IFRS (IAS 39)

### Purpose and context
- IFRS 9 is a principle-based standard that provides bankers with disproportionate discretion in selecting model inputs for provisioning methodologies.
- For diagnostic exercises, it is necessary to understand how local banks implement IFRS 9 to:
  - ascertain the implications of local practice;
  - determine any needs to set criteria for the purposes of the diagnostic in the TORs;
  - foresee the need for standardizing a local rules-based standard benchmark.
- Empirical loss evidence must be available to justify the selection of input values, including for “modeled” parameters (no actual experience) or parameters “imported” from another jurisdiction.
- If modeled or imported, consider using a benchmark combination that fulfills three conditions (e.g., fourth quartile in the system) or set rule-based parameters for banks lacking sufficient empirical loss evidence, consistent with the ECB “challenger” approach described in its AQR Manual 2014.

### Collective provisioning algorithm (structure and key parameters)
- Provision = Exposure * [1 - (1-PD)
EP
* (1-CR) * LGL.
- Key parameters:
  - PD is the probability of default. This algorithm applies to both performing (PD<1), and nonperforming loans (PD=1).
  - EP is the emergence period used to estimate an Incurred but Not Reported provision for loans that defaulted but have not still been identified as such.
  - CR is the cure rate, the proportion of loans that default but then return to performing status, as per observed number of cases.

### Loss-given-loss (LGL) for loans backed by real estate collateral
- LGL = C * (1-PTT + Current_REPI) * (1 - FSD) * DF * (1 - WOC).
- Component definitions:
  - C is the collateral value when the loan was originated.
  - PTT is the forecast peak-to-through (negative) variation in the collateral price since loan origination.
  - REPI is the current relevant Real Estate Price Index.
  - FSD is the fire sale discount.
  - DF is the time discount value.
  - WOC are the work-out-costs to cash recovery.

### Diagnostic guidance, conservative parameter setting, and sensitivity considerations
- The sensitivity of provisioning estimates to simultaneous changes in the eight parameters can shift loss estimates significantly under IFRS 9 Expected Credit Loss or an Incurred but Not Reported standard.
- For the diagnostic, inputs must be conservative, justifiable and consistent with macroeconomic forecasts.
- Recommended approaches when banks lack sufficient empirical evidence:
  - use a benchmark combination that fulfills the three conservatism/justifiability/consistency conditions (e.g., fourth quartile in the system); or
  - set rule-based parameters for those banks that do not fulfill the three conditions.
- Tenets to clarify in TORs (to be applied unless justifiably deviated):
  - Segregate forborne, refinanced, and restructured loans from relevant loan pools to calculate PD and CR separately from fully performing loans.
  - Among stratification variables (LTV, days in arrears, internal rating, among others), a key attribute should be whether the client is able to serve a fully interest income from its own resources.
  - Emergence period (EP) guidance:
    - EP should be based on an assessment of the effectiveness and reliability of internal credit controls (especially the internal loan review function).
    - An EP of 12 months may be seen as conservative, IFRS compatible, as it approximates a one-year expected loss for non-defaulted loans.
    - If the cycle of internal loan review is higher than a year or seen as not effective and/or non-reliable, the EP could increase above 12 months to account for hidden potential losses.
  - Cure rate (CR) guidance:
    - Filter false cures from re-aging and refinancing ex ante.
    - If modeled, there must be sufficient observations of economically reasonable explanatory variables.
    - Condition the declaration of cure to a period of sustained performance, for example, full service of interest due for 24 months with a 10 percent to 20 percent reduction in principal through continued debt service.
  - Fire sale discount (FSD) and work-out costs (WOC):
    - If there are doubts about FSD and WOC, set conservative rule-based haircut discounts.
    - Short of a justifiable estimation, these parameters should be set at the 50 percent to 30 percent level.
  - Discount factor (DF) and time-to-liquidation (TTL):
    - DF is dependent on TTL; in countries with protracted judicial procedures, TTL may exceed six years on average.
    - TTL should be set to fully phase out collateral in about five years from the date of the first 90 days past due to account for reduced effectiveness of eligible collaterals.
  - Discount rate:
    - The discount rate should ideally be the effective loan interest rate as required under accounting rules.
    - At a minimum, the discount rate should not be lower than the average cost of funds of the marginal source of funding.

*International Monetary Fund — Technical Notes and Manuals 19/04 | 2019 (APPENDIX V).*

---


_Source: https://www.imf.org/-/media/files/publications/tnm/2019/tnmea2019004.pdf_