Understanding ECL during crises: A Financial Safety Net

Why ECL during crises matters for your institution

Crises—whether banking panics, sovereign stress, or sharp commodity shocks—can change expected credit losses rapidly. For institutions applying IFRS 9, failure to reflect these changes promptly and transparently can cause misstated provisions, regulatory pushback, erosion of capital buffers, and loss of stakeholder confidence. You must answer three key questions quickly: how much additional lifetime loss is now expected, which exposures move from Stage 1 to Stage 2 or 3 under Three‑Stage Classification, and whether your data and governance support a robust, testable change in assumptions.

Regulators and auditors will scrutinize both the numbers and the process; see guidance on regulatory challenges for ECL that commonly arise in crisis periods. This matters not only for compliance: accurate, forward‑looking ECL estimates inform lending decisions, pricing, capital planning and strategic response.

Core concepts: definitions, components, and examples

What “ECL during crises” means

ECL during crises is the forward‑looking estimate of credit losses that explicitly incorporates adverse macroeconomic scenarios, heightened PDs, increased LGDs from fire‑sale collateral, and longer cure periods. Practically, it means: updating scenario probabilities, recalibrating PD curves, applying overlays where models underperform, and re‑running staging assessments across the book.

Three‑Stage Classification: practical refresher

Under IFRS 9’s Three‑Stage Classification, exposures are classified by credit deterioration since initial recognition. In a crisis: many previously Stage 1 accounts may meet the “significant increase in credit risk” threshold and move to Stage 2, increasing 12‑month to lifetime ECL recognition. Provide a clear, quantitative staging policy (e.g., 30% increase in 12‑month PD or specific covenant breach triggers) and document exceptions.

Sensitivity Testing and scenario design

Sensitivity Testing is essential to show how ECL responds to plausible shocks. Typical approach: baseline + two stress scenarios (adverse and severe adverse). Example: for a retail mortgage portfolio with average PD 0.5%, an adverse scenario could double PDs to 1.0% and increase LGD by 10 percentage points; compute resulting ECL uplift and provide a sensitivity table in Risk Committee Reports.

Risk Model Governance and model adjustments

Risk Model Governance must accelerate during crises. Actions include expedited model validation cycles, documented overlays where models lag (e.g., sudden unemployment spike), and clear sign‑off paths. Create a short governance playbook: change request, validation checklist, executive summary, and final approval. This will be referenced in Risk Committee Reports and audit trails.

IFRS 7 Disclosures and presentation

Disclosure expectations increase in a crisis. You must expand narrative explanations, sensitivity tables, and reconciliation of changes in ECL balances. For actionable guidance on disclosure format and examples, review best practices for presenting ECL in statements to ensure clarity and consistency across financial reports.

Historical Data and Calibration

Historical Data and Calibration are the backbone for stress response. If historical crises are limited, you must supplement with proxy events or international benchmarks and use conservative calibration techniques. Use segmented vintage analysis where possible (e.g., cohorts by origination year, LTV buckets, or industry) to estimate how PD and LGD reacted historically and calibrate your current model accordingly.

Practical use cases and scenarios

Below are recurring situations and how to approach them.

1. Sudden macro shock (e.g., commodity price collapse)

1. Immediate triage: identify top 20% of exposures by exposure‑at‑default (EAD) and industry concentration.
2. Run quick scenario re‑scoring: increase PDs by calibrated factors (e.g., commodity‑exposed corporate PD × 2.5), compute lifetime ECL deltas.
3. Document provisional overlays for sectors with no direct data and plan for calibration update within 30–60 days.

2. Sovereign or systemic banking stress

Systemic events require broader macro overlays and coordination with capital management. Consider wider macro indicators and the institution’s liquidity profile. Use your stress framework to derive scenario weights and monitor market indicators for prompt recalibration.

3. Localized defaults (e.g., industry collapse)

For concentrated exposures, deploy granular vintage and sector models, apply targeted restructuring PD assumptions, and prepare focused Risk Committee Reports to request provisioning authority or capital relief actions.

Data gaps and proxies

When internal loss history is sparse, lean on external data. Our recommendations for assembling credible proxies and constructing conservative overlays are informed by principles in key ECL data sources and the methods summarized later in this article.

Impact on decisions, performance and reporting

Accurate crisis ECL affects:

• Profitability: higher ECL reduces retained earnings; quantify expected OCI vs P&L impacts.
• Capital planning: provisioning consumes CET1; stress scenarios inform management buffer decisions.
• Risk appetite: changing ECL alters lending capacity and pricing strategies across segments.
• Stakeholder confidence: transparent disclosures and clear Risk Committee Reports reduce market uncertainty.

Use ECL as a strategic tool — for example, by integrating ECL scenario outputs into pricing models so higher expected losses are compensated by margin or covenants. For a broader view on how ECL interacts with macro policy and stability, see our piece on ECL and financial stability and how ECL can be used as part of systemic risk monitoring in severe downturns. Likewise, when you produce regular management packs, tie ECL changes to forward‑looking capital and liquidity plans so that boards can act decisively.

From a reporting perspective, ensure consistency between how you derive numbers and how they are communicated — read about practical considerations for presenting ECL in statements if you need formatting examples and disclosure language.

Common mistakes and how to avoid them

• Waiting for “perfect” data: Crises require timely decisions. Use best‑available proxies and document assumptions. Avoid paralysis; apply conservative overlays and plan for iterative refinement.
• Ignoring staging dynamics: Misapplication of the Three‑Stage Classification can understate lifetime ECL. Define quantitative staging triggers and stress‑test them.
• Poor governance during speed changes: Rapid model changes without validation invite regulatory scrutiny. Keep short, auditable decision records and involve independent validation functions as soon as possible.
• Insufficient sensitivity testing: Present discrete scenario outputs and sensitivity tables. Don’t rely on a single “most likely” scenario.
• Weak disclosure linkage: Disclosures should explain the modelling and economic rationale — link assumptions to observable indicators to avoid being challenged in audit or supervisory reviews. Guidance on presenting ECL in statements is useful here.
• Overfitting historical calibrations: When calibrating with limited crisis data, emphasize reasonableness and apply conservative margins rather than mechanical extrapolation. For further techniques on sourcing and validating data, consult our discussion of data for ECL estimation.

Practical, actionable tips and checklists

Use this rapid response checklist in crisis mode.

1. Activate ECL crisis protocol: timeline, owners, data sources, and decision gates (Day 0 = triage, Day 7 = provisional estimates, Day 30 = calibrated update).
2. Run a minimum of three scenarios: baseline, adverse, severe adverse. Apply Sensitivity Testing around key drivers (PD, LGD, cure rates).
3. Quantify staging movements by segment and document threshold logic for Three‑Stage Classification changes.
4. Apply temporary overlays where models are known to lag; clearly document triggers to remove overlays.
5. Prepare Risk Committee Reports with five‑page executive summary, top 10 exposures, scenario tables, and recommended provisioning actions.
6. Ensure Risk Model Governance: expedited validations, sign‑off trail, independent review, and auditor engagement plan.
7. Expand IFRS 7 Disclosures to describe scenario design, sensitivity ranges and management judgement — consult internal reporting templates and regulators’ expectations.
8. Plan iterative calibration using Historical Data and Calibration approaches: vintage analysis, stress multipliers, and proxy benchmarking.

Operational tip: create templated report packs (Excel + narrative Word) so you can populate scenarios rapidly and ensure consistent presentation in board materials and external filings.

KPIs / success metrics

• Timeliness: provisional ECL scenarios published within 7 business days of trigger event.
• Coverage: 100% of top 50 exposures by EAD assessed for staging movement within Day 3.
• Model responsiveness: percentage change in model PD explained by scenario variables ≥ 80% (target metric for calibration effectiveness).
• Governance: 0 unresolved model change exceptions at next Risk Committee meeting.
• Disclosure completeness: all required IFRS 7 Disclosures updated and reconciled to the financial statements (audit sign‑off).
• Backtesting: variance between realized losses and forecasted ECL within targeted bands over successive rolling 12‑month windows after recalibration.

FAQ

Reference pillar article

This piece is part of a wider content cluster. For foundational concepts and broader policy context, see the pillar article: The Ultimate Guide: Introduction to Expected Credit Losses (ECL).

Additional reading in this cluster includes practical discussions on macro relationships such as ECL as macro risk tool, the role of ECL in systemic oversight described under ECL and financial stability, and the analyses of economic challenges in ECL for modelers and finance teams. For disclosure‑focused issues see presenting ECL in statements and technical advice on data for ECL estimation.