IFRS 9 Expected Credit Loss for Private Credit: Beyond the Three Stages

The three-stage model, applied to something it was not designed for

IFRS 9 was drafted with bank loan books in mind. Its three-stage expected credit loss (ECL) model assumes a reasonably large population of broadly similar exposures, each with a default history that can support statistical PD estimation. Private credit — bilaterally-originated, covenant-heavy, sponsor-backed — does not fit that picture cleanly.

At initial recognition, an instrument sits in Stage 1 and attracts a 12-month ECL. It moves to Stage 2 on a significant increase in credit risk (SICR), at which point lifetime ECL applies. It moves to Stage 3 when it becomes credit-impaired. The mechanics are straightforward; the judgement is in the triggers and the inputs.

Stage 2 triggers in practice

The IASB declined to prescribe quantitative SICR thresholds. For a private credit book, the practical triggers that drive stage 2 migration are not publicly-observable events but operating indicators that the originating lender observes directly through its information rights.

• Maintenance covenant breach (leverage, fixed charge coverage, minimum EBITDA)
• Two consecutive quarters of LTM EBITDA decline greater than 15 percent
• Loss of a top-5 customer contributing more than 10 percent of revenue
• Withdrawal of sponsor support, or sponsor refusal of an equity cure where one is contractually available
• Downgrade of shadow rating by two or more notches, or crossing from B- to CCC+
• Rising PIK utilisation where a cash/PIK toggle exists

PD calibration without market proxies

The central methodological difficulty for private credit ECL is that directly-originated loans typically have no liquid market proxy from which a PD term structure can be observed. The pragmatic and now broadly-accepted approach is to anchor to a rating cohort in the public high-yield universe — commonly CCC-rated issuers for the deep middle market and B2/B3 for the upper middle market — and then apply a set of adjustments. Some practitioners and Big-firm advisory practices instead anchor to single-B equivalents, on the empirical basis that direct-lending realised default rates have tracked closer to single-B than to CCC.

These adjustments reflect the structural features of private credit that are not captured in the public proxy: stronger covenant packages, tighter information rights, smaller lender groups (so faster workout convergence), and sponsor equity support. S&P Global's and Moody's periodic private credit default studies, together with the Cliffwater Direct Lending Index, provide empirical anchors for the size of the adjustment — though both rating agencies caution that the observed default population is still small and the distribution is unlikely to be stable through cycle.

LGD across the capital structure

LGD estimation in private credit separates cleanly by seniority. Senior secured first lien positions in sponsor-backed middle market transactions have historically recovered in the 60–75 percent range, per S&P Global LossStats and PitchBook LCD recovery data, though recoveries on the 2020–2022 vintages are still maturing. Unitranche positions, which blend first and second lien economics, typically model to a 55–65 percent recovery. Mezzanine and second lien positions sit materially lower, in the 25–40 percent range, with wide dispersion driven by enterprise value outcomes.

The LGD model should be coupled to an explicit enterprise value waterfall. Applying a flat recovery rate across a portfolio — common practice as recently as 2022 — is increasingly pushed back on by auditors and by the FCA in its thematic supervisory work under the AIFMD framework.

Forward-looking macroeconomic overlays

IFRS 9 requires ECL to be probability-weighted across multiple forward-looking scenarios. The Federal Reserve's semi-annual Financial Stability Report and the FSB's Global Monitoring Report on Non-Bank Financial Intermediation are increasingly used as anchors for the macro scenario set.

A typical overlay structure uses three scenarios: a baseline (50–60 percent weight), a downside (20–30 percent weight, modelling a mild recession and 100–150bps of spread widening), and a severe downside (10–20 percent weight, modelling a sharp contraction with default rates rising to 8–10 percent in the relevant cohort). The weights are a governance decision, but the SEC has made clear in its recent examinations of private fund advisers that weights should be reviewed and documented each reporting date, not set-and-forget.

Where private credit ECL practice is heading

Three trajectories are clear. First, regulators expect PD and LGD models to be calibrated and backtested against the manager's own realised experience as the data accumulates — the CCC-anchor-plus-adjustment approach is a starting point, not an end state. Second, stage migration triggers will become more quantitative and more auditable. Third, the integration of ECL with the fair value mark will tighten: inconsistencies between the implied PD in the fair value discount rate and the PD used for ECL are increasingly untenable.