Expected credit loss (ECL) / IFRS 9

Expected credit loss (ECL) is the impairment measure required by IFRS 9, representing the probability-weighted estimate of credit losses a lender expects to incur over the life of a financial asset, discounted to present value. It is a forward-looking measure: a provision is recognised from the moment a loan is originated, before any evidence of borrower distress exists.

IFRS 9 Financial Instruments replaced IAS 39 for annual periods beginning on or after 1 January 2018. It covers classification and measurement, impairment and hedge accounting; ECL is the impairment component, and the part with the largest practical effect on lenders.

Why the Model Changed: Incurred Loss vs Expected Loss

Under the previous IAS 39 incurred loss model, a lender could only recognise impairment once a loss event had already occurred — typically missed payments. The consequence was that provisions lagged the credit cycle, arriving after deterioration was already visible. This "too little, too late" recognition was a widely criticised feature of accounting through the 2008 financial crisis.

IFRS 9 reverses the timing. A provision is recognised on day one for every loan, and the amount increases sharply if credit risk deteriorates — rather than waiting for default to become probable.

Comparison: IAS 39 vs. IFRS 9

• Trigger for provision: Observed loss event (IAS 39) vs. Recognised at origination (IFRS 9)
• Outlook: Historical and current (IAS 39) vs. Historical, current, and forecast (IFRS 9)
• Timing: Lagging (IAS 39) vs. Forward-looking (IFRS 9)
• Effect on provisions: Lower, back-loaded (IAS 39) vs. Higher, front-loaded (IFRS 9)
• Volatility: Lower (IAS 39) vs. Higher, sensitive to macroeconomic forecasts (IFRS 9)

The IFRS 9 Three-Stage Model

Every in-scope financial asset sits in one of three stages, determined by how much its credit risk has changed since origination.

• Stage 1 (Performing): - Condition: No significant increase in credit risk since initial recognition.

  • Loss allowance: 12-month ECL.
  • Interest revenue basis: Gross carrying amount.

• Stage 2 (Underperforming / Watchlist): - Condition: Significant increase in credit risk (SICR), but not credit-impaired.

  • Loss allowance: Lifetime ECL.
  • Interest revenue basis: Gross carrying amount.

• Stage 3 (Credit-Impaired / Non-Performing): - Condition: Objective evidence of default.

  • Loss allowance: Lifetime ECL.
  • Interest revenue basis: Net carrying amount (gross carrying amount less allowance).

12-month ECL is not the loss expected over the next twelve months. It is the lifetime loss on the portion of exposure where a default event is possible within the next twelve months — a distinction frequently misstated.

Lifetime ECL covers losses from all possible default events over the remaining life of the instrument.

The practical consequence is the stage 1 to stage 2 cliff: moving a loan into stage 2 replaces a 12-month allowance with a lifetime allowance, which can multiply the provision several times over on a single loan. This makes SICR assessment the single most consequential judgement in the model.

Significant Increase in Credit Risk (SICR)

IFRS 9 does not prescribe a definition of SICR. It requires an assessment of the change in the risk of default over the remaining life relative to what was expected at initial recognition — a relative measure, not an absolute one. A loan originated as high risk has not experienced a SICR simply by remaining high risk.

Indicators commonly used include:

• Days past due — IFRS 9 contains a rebuttable presumption that a SICR has occurred once payments are more than 30 days past due. This is a backstop, not the primary test; relying on it alone means the model is not genuinely forward-looking.
• Internal risk grade migration beyond a defined threshold.
• Forbearance or restructuring granted for credit reasons.
• Watchlist placement or covenant breach.
• Behavioural signals — declining transaction volumes, repeated late payments within tolerance, deteriorating collections performance.
• Borrower-specific events — loss of a major customer, business closure, illness, or loss of employment.
• Sector or geographic deterioration affecting a defined borrower segment.

Curing. A loan moves back to stage 1 when the SICR conditions no longer apply. Institutions typically apply a probation or cure period requiring a minimum number of consecutive on-time payments before reclassification, to avoid oscillation.

Low credit risk exemption. An entity may assume no SICR has occurred where the instrument has low credit risk at the reporting date. This is designed for investment-grade instruments and is generally not available for microfinance or SACCO loan portfolios.

Defining Default

IFRS 9 does not define default either, but includes a rebuttable presumption that default occurs no later than 90 days past due. The definition applied must be consistent with that used for internal credit risk management purposes.

Other default indicators typically include: the borrower being unlikely to pay in full without recourse to collateral realisation, bankruptcy or insolvency, distressed restructuring, and write-off.

The definition must be applied consistently across staging, PD estimation and disclosure.

How ECL Is Calculated

The standard calculation decomposes ECL into three parameters:

ECL = PD × LGD × EAD, discounted at the original effective interest rate

PD — Probability of Default. The likelihood the borrower defaults within a given horizon (12 months for stage 1, remaining lifetime for stages 2 and 3). Estimated from historical default rates by segment, adjusted for forward-looking information.

LGD — Loss Given Default. The proportion of exposure expected to be lost after default, net of expected recoveries from collateral, guarantees and collections effort, and net of the costs of obtaining them. Expressed as a percentage: LGD = 1 − recovery rate.

EAD — Exposure at Default. The expected outstanding balance at the point of default, including accrued interest and expected drawdowns on undrawn commitments.

EIR — Effective Interest Rate. The discount rate. ECL must be measured at present value, so expected cash shortfalls are discounted from the date they are expected to occur back to the reporting date at the loan's original EIR.

Worked example

A loan with an outstanding balance of 10,000, a 12-month PD of 8%, and an LGD of 45%:

ECL = 0.08 × 0.45 × 10,000 = 360

If the loan subsequently experiences a SICR and moves to stage 2, the lifetime PD — say 22% over the remaining term — replaces the 12-month PD:

ECL = 0.22 × 0.45 × 10,000 = 990

The exposure has not changed and no payment has yet been missed. The allowance has nearly tripled on a change of stage alone.

The Three Measurement Requirements

IFRS 9 requires that every ECL estimate be:

1. Unbiased and probability-weighted — not a single best-estimate outcome. The measurement must reflect a range of possible outcomes, typically implemented as weighted base, upside and downside macroeconomic scenarios.
2. Discounted for the time value of money at the original effective interest rate.
3. Based on reasonable and supportable information available without undue cost or effort — including historical experience, current conditions and forecasts of future economic conditions.

The forward-looking requirement is what distinguishes ECL from historical provisioning. Institutions must establish a defensible link between macroeconomic variables — GDP growth, inflation, exchange rate, commodity prices, unemployment, rainfall in agricultural portfolios — and observed default behaviour, then apply forecasts of those variables to the PD estimate.

The Simplified Approach and the Provision Matrix

For trade receivables, contract assets and lease receivables, IFRS 9 permits (and in some cases requires) a simplified approach: lifetime ECL is recognised at all times, with no staging assessment.

The common implementation is a provision matrix — historical loss rates applied to an aging schedule, adjusted for forward-looking information.

• Current: 1.0% historical loss rate + 0.2% forward-looking adjustment = 1.2% applied ECL rate
• 1–30 days: 3.0% historical loss rate + 0.5% forward-looking adjustment = 3.5% applied ECL rate
• 31–60 days: 12.0% historical loss rate + 1.5% forward-looking adjustment = 13.5% applied ECL rate
• 61–90 days: 30.0% historical loss rate + 3.0% forward-looking adjustment = 33.0% applied ECL rate
• 91–180 days: 60.0% historical loss rate + 5.0% forward-looking adjustment = 65.0% applied ECL rate
• 180+ days: 100.0% historical loss rate + 0.0% forward-looking adjustment = 100.0% applied ECL rate

Illustrative rates only. A provision matrix must be built from the institution's own loss history, segmented by portfolio, and supported by documented forward-looking adjustments.

Many lenders adapt this logic for loan portfolios via migration or roll-rate analysis: measuring the historical rate at which balances move from each aging bucket to the next, and compounding those rates through to write-off to derive a lifetime loss estimate per bucket.

Collective vs Individual Assessment

Individual assessment is applied to exposures that are large, unique, or where borrower-specific information is available — typically stage 3 corporate or SME loans, where ECL is calculated from expected recovery cash flows on that specific borrower.

Collective assessment groups exposures with shared credit risk characteristics and applies segment-level parameters. It is the default for high-volume, homogeneous portfolios such as microloans, group loans and consumer credit.

Segmentation drives model quality. Common dimensions include product type, loan cycle number, group versus individual methodology, tenor, sector, geography, and collateralisation. Segments must be granular enough that loss behaviour within each is genuinely homogeneous — but not so granular that individual segments lack sufficient loss history to estimate parameters reliably.

IFRS 9 also requires collective assessment where a SICR has occurred at a segment level but is not yet observable in individual loans — for example, a shock affecting one crop or one trading corridor.

Write-Off and Modification

Write-off. The gross carrying amount is written off when there is no reasonable expectation of recovery — in whole or in part. Write-off is a derecognition event, not a provisioning judgement, and does not extinguish the legal right to pursue recovery. Subsequent recoveries are recognised in profit or loss.

Modification. Where contractual cash flows are renegotiated, the entity must determine whether the modification is substantial. A substantial modification results in derecognition of the original asset and recognition of a new one; a non-substantial modification requires a recalculation of the gross carrying amount using the original EIR, with a modification gain or loss recognised immediately.

Restructuring for credit reasons is a strong SICR indicator, and a distressed restructuring is generally a default indicator. Rescheduling a struggling loan does not reset its stage.

POCI assets. Financial assets that are purchased or originated credit-impaired are measured differently: no loss allowance is recognised at initial recognition, a credit-adjusted EIR is used, and only cumulative changes in lifetime ECL since initial recognition are recognised thereafter.

ECL in Microfinance and SACCO Portfolios

The model was designed with large banks and long-dated instruments in mind, and several practical tensions arise in small-loan portfolios.

• Short tenors compress the staging distinction. On a four-month loan, 12-month ECL and lifetime ECL converge, because the remaining life is shorter than the twelve-month horizon. The stage 1 to stage 2 gap narrows considerably, though staging still affects interest recognition and disclosure.
• Days-past-due backstops dominate in practice. Where behavioural data and risk grading are limited, the 30- and 90-day presumptions become the de facto staging model. This is permissible but weakens the forward-looking intent, and auditors increasingly challenge it.
• Loss history may be thin. Institutions with short operating histories or recently launched products lack the default observations needed for credible PD estimation, and must rely on proxy portfolios with documented justification.
• Macro linkage is hard to evidence. Establishing a statistically supportable relationship between national macro variables and default in a small, geographically concentrated portfolio is often not achievable. Qualitative, documented overlays are the standard fallback.
• Group lending complicates EAD and LGD. Where a guarantee, group fund or joint liability exists, expected recoveries must reflect what is realistically collectible from those sources, not their nominal value.
• Cycle number is a strong risk driver. First-cycle borrowers default at materially higher rates than repeat borrowers, which makes cycle number one of the more useful segmentation variables available in a microfinance portfolio.

ECL vs Regulatory Provisioning

IFRS 9 ECL is an accounting measure. Many central banks and regulators separately prescribe minimum provisioning rules based on fixed percentages applied to classification bands — for example, prescribed rates for substandard, doubtful and loss categories defined by days past due.

These two figures rarely agree. Where the regulatory minimum exceeds the ECL allowance, the common supervisory treatment is to require the shortfall to be appropriated from retained earnings to a regulatory reserve within equity, rather than adjusting the income statement. Institutions in this position must maintain both calculations and reconcile between them.

Requirements differ by jurisdiction, and institutions should confirm the applicable treatment with their own regulator.

Key Disclosure Requirements

IFRS 7 requires extensive disclosure supporting the ECL figures, including:

• A reconciliation of the loss allowance from opening to closing balance, showing transfers between stages, new originations, derecognitions and write-offs.
• Gross carrying amounts by credit risk grade and by stage.
• The inputs, assumptions and estimation techniques used, including macroeconomic scenarios and their weightings.
• How SICR and default are defined, and the basis for any rebuttal of the 30-day or 90-day presumptions.
• Write-off policy and amounts still subject to enforcement activity.
• Sensitivity of the allowance to changes in key assumptions.

Frequently Asked Questions

Does IFRS 9 apply to microfinance institutions and SACCOs? Yes, wherever IFRS is the applicable reporting framework. Smaller entities reporting under IFRS for SMEs follow different requirements, so the applicable framework should be confirmed.

How does IFRS 9 differ from CECL? CECL is the US GAAP equivalent under ASC 326. The principal difference is that CECL requires lifetime expected losses on all assets from origination, with no staging or 12-month category.

This page is a general explanation of accounting concepts and is not accounting, audit or regulatory advice. Application of IFRS 9 depends on entity-specific facts and the requirements of your reporting framework and regulator.