How to build an IFRS 9 expected credit loss model for an SME lender
An IFRS 9 expected credit loss (ECL) model for an SME lender places each loan in Stage 1, 2 or 3 according to how much its credit risk has increased, estimates probability of default, loss given default and exposure at default, and weights the result across forward-looking economic scenarios. Stage 1 loans carry 12-month expected losses; Stages 2 and 3 carry lifetime expected losses.
SME lenders and financing funds face a particular IFRS 9 problem: the standard expects forward-looking, probability-weighted estimates, but a young portfolio rarely has years of default history. A well-built model handles that openly rather than hiding it.
What does IFRS 9 require?
Impairment is based on expected losses, not incurred losses. For every financial asset measured at amortised cost, the lender recognises a loss allowance equal to either 12-month expected credit losses or lifetime expected credit losses, depending on its stage.
How should loans be staged?
- Stage 1: performing loans with no significant increase in credit risk since origination. Allowance: 12-month ECL.
- Stage 2: loans with a significant increase in credit risk. IFRS 9 presumes this once payments are more than 30 days past due, but earlier indicators such as restructuring or watch-list status should also move loans here. Allowance: lifetime ECL.
- Stage 3: credit-impaired loans, presumed at more than 90 days past due. Allowance: lifetime ECL, with interest recognised on the net carrying amount.
How do you estimate probability of default with limited data?
Common approaches, often combined:
- roll-rate or transition matrices built from monthly delinquency buckets
- vintage analysis of loans grouped by origination period
- external benchmarks such as rating agency default studies, adjusted for the SME segment
- a documented management overlay where internal data is too thin
How are LGD and EAD estimated?
Loss given default reflects what is recovered after default: collateral, guarantees (for example, portfolio guarantee schemes for SMEs), and the time and cost of recovery, discounted at the effective interest rate. Exposure at default for term loans follows the amortisation schedule; for facilities with undrawn limits it includes an expected drawdown.
How are forward-looking scenarios applied?
Build at least three scenarios, such as base, upside and downside, each with projected macroeconomic variables that correlate with your defaults. In Gulf economies these often include non-oil GDP growth, oil prices and interest rates. Adjust PDs under each scenario and weight the resulting ECL by scenario probability.
What do auditors look for?
A written methodology, reconciliation of model inputs to the loan ledger, evidence for staging thresholds, back-testing of past estimates against actual defaults, and governance showing who approved the assumptions and overlays.
How much loan history is needed for an IFRS 9 model?
Three to five years of monthly loan-level data is ideal. With less, combine internal roll rates with external benchmarks and document a management overlay, then refine the model as your own history builds.
How often should the ECL model be updated?
Staging and balances are refreshed at every reporting date. Macroeconomic scenarios and weights should be reviewed at least annually and whenever the economic outlook changes materially.