IFRS quantitative finance series
IFRS 9 Expected Credit Loss Modelling for Banks: A Quantitative Guide
IFRS 9 converts credit-risk expectations into an accounting estimate that must be probability-weighted, forward-looking and discounted. For a bank, the difficult part is not writing ECL as PD × LGD × EAD. It is building a controlled measurement chain that remains faithful to contractual cash flows, staging rules and reporting evidence.
The measurement architecture
A robust engine separates exposure data, staging, scenario generation, parameter estimation, cash-flow projection, discounting and accounting output. Twelve-month ECL is not the cash loss expected over the next twelve months; it is the lifetime cash shortfall arising from defaults that are possible during that period. Lifetime ECL expands the default window, not the definition of loss.
Probability weighting and time value
The estimate should reflect a range of possible outcomes rather than a single central forecast. Scenario weights, conditional PD paths, LGD and exposure profiles must be internally coherent. Expected cash shortfalls are discounted to the reporting date using the relevant effective interest rate or its permitted approximation.
From model result to financial statement
The final allowance must reconcile from instrument-level calculations to the general ledger and IFRS 7 tables. Stage transfers, originations, derecognitions, repayments, write-offs, recoveries and model changes need separate movement attribution so users can understand why the allowance changed.
Quantitative expression
Implementation controls
- Reconcile source balances to the ledger before modelling.
- Approve scenarios and weights independently of model production.
- Trace every disclosed allowance movement to controlled data.
- Backtest components and total ECL at portfolio level.