Credit Risk

IFRS 9 Expected Credit Loss Governance Beyond Model Performance

By Jonas (Yonas) Mohamed Osman Abdelghafour · 8 Aug 2026

ECL is a system of judgements

IFRS 9 specifies classification, measurement and impairment requirements for financial instruments. In practice, expected credit loss combines probability of default, loss given default, exposure, staging, forward-looking scenarios, discounting and post-model adjustments. Weakness in any one component can materially distort the final estimate.

Validation should therefore assess the complete estimation process. A model can rank risk well while producing biased lifetime losses, unstable staging or implausible responses to macroeconomic scenarios.

Challenge significant increase in credit risk

The transfer from 12-month to lifetime expected loss is often the most sensitive judgement in the framework. Institutions should test the effectiveness of relative deterioration thresholds, backstops, watchlists and qualitative indicators across portfolios and economic conditions.

A sound review examines transition rates, cure behaviour, time in stage, default emergence and the interaction between automated criteria and expert interventions. Concentrating only on default-model AUC or Gini statistics leaves this central accounting judgement underexamined.

Control scenarios and overlays

Macroeconomic scenarios should be internally coherent, relevant to portfolio risk and appropriately weighted. Governance should document why variables, paths and weights are reasonable and how non-linearity enters the loss estimate. Sensitivity analysis can reveal whether small judgement changes produce disproportionate movements.

Overlays are sometimes necessary when models cannot capture a developing risk. Each overlay should have a defined rationale, transparent calculation, owner, monitoring measure and exit condition. An overlay that rolls forward indefinitely without evidence becomes an uncontrolled parallel model.

Reconcile model, ledger and disclosure

ECL control requires reconciliation between source systems, model populations, general ledger balances and financial-statement disclosures. Movements should be explained through exposure changes, stage transfers, model updates, scenario changes, write-offs and recoveries.

This decomposition turns a large accounting number into an auditable story. It also enables the audit committee and board to distinguish genuine portfolio deterioration from methodology or data effects.

Primary sources

About the author

Jonas (Yonas) Mohamed Osman Abdelghafour writes about financial risk management, quantitative modelling, actuarial science, banking risk, insurance risk, capital modelling, model validation, climate risk and geopolitical risk. His work focuses on translating complex quantitative and regulatory risk issues into practical frameworks for financial institutions. Author profile.