Key Points
- PD is one of three inputs to the expected credit loss calculation, alongside loss given default and exposure at default.
- Banks typically derive PD from internal rating models, while non-financial entities often rely on external credit ratings or historical loss data.
- A 12-month PD applies in Stage 1; a lifetime PD applies once the asset moves to Stage 2 after a significant increase in credit risk.
- Auditors who accept management's PD without testing the underlying assumptions face the most frequent inspection challenge on ECL files.
What is probability of default (PD)?
You open the ECL model and the first thing you see is a PD column populated with numbers that nobody on the finance team can trace to source data. In our experience, that's the single most common finding on IFRS 9 files. The number drives everything downstream, yet it's often a PIOOMA estimate dressed up with a formula reference.
IFRS 9.5 .5.9 requires an entity to measure lifetime expected credit losses (ECL) for financial instruments where credit risk has increased significantly since initial recognition, and 12-month ECL for all other instruments. PD sits at the centre of both calculations. The 12-month PD captures the probability that the borrower defaults within the next twelve months. Lifetime PD captures that probability over the remaining contractual life of the instrument, adjusted for expected prepayments.
For entities with large portfolios (trade receivables at a distributor, loan books at a bank), PD is usually estimated at the portfolio level using historical default rates grouped by shared credit characteristics. IFRS 9 .B5.5.52 permits this collective assessment. For individually significant exposures, the entity may assign a PD based on internal credit scores or external ratings mapped to default probabilities. The auditor's job under ISA 540.13 is to evaluate whether management's method for deriving PD is appropriate and whether the data inputs are reliable. That includes checking whether the forward-looking adjustments reflect current conditions rather than stale assumptions.
Worked example: Henriksen Shipping A/S
Client: Danish maritime logistics company, FY2024, revenue EUR 140M, IFRS reporter, trade receivables portfolio of EUR 23M across 312 customers.
Step 1: Segment the receivables portfolio
The finance team groups receivables into four buckets by customer credit profile: investment-grade corporates (EUR 9.1M), sub-investment-grade corporates (EUR 7.4M), SME customers with no external rating (EUR 4.8M), and government or state-owned entities (EUR 1.7M).
Step 2: Assign 12-month PD to each segment
Henriksen uses published default rate tables from Moody's for rated segments and three years of internal payment history for the unrated SME bucket. Investment-grade: 0.08%. Sub-investment-grade: 1.45%. SME (unrated): 3.2% (derived from historical write-off data). Government: 0.02%.
Step 3: Apply forward-looking adjustment
Henriksen's ECL model incorporates a macroeconomic overlay. Management applies a 15% uplift to the base PDs for the sub-investment-grade and SME segments, reflecting a projected increase in European freight insolvencies during 2025 based on industry data from the Danish Maritime Authority.
Step 4: Calculate the 12-month ECL contribution from PD
ECL for each segment equals PD x LGD x EAD. Assuming a uniform LGD of 40% across segments, the PD-driven ECL is: investment-grade EUR 2,912 (0.08% x 40% x EUR 9.1M), sub-investment-grade EUR 49,432 (1.67% x 40% x EUR 7.4M), SME EUR 70,656 (3.68% x 40% x EUR 4.8M), government EUR 136 (0.02% x 40% x EUR 1.7M). Total 12-month ECL: EUR 123,136.
The total €123K provision is small relative to the €23M portfolio, but the point is traceability. Each PD links to an external data source or internal write-off history, and the macro overlay is sourced and quantified. When a reviewer opens this file next year, they can see exactly why each number is what it is.
Why it matters in practice
Auditors frequently accept management's PD inputs without testing the underlying data. ISA 540.13 (b) requires the auditor to evaluate whether the data on which the estimate is based is relevant and reliable. For PD, that means tracing historical default rates to actual write-off records, not just reviewing the spreadsheet formula. It's tedious work, and it's tempting to treat the PD column as a given when the ECL balance looks immaterial. But that's exactly the kind of ticking and bashing that inspection teams flag.
The FRC's 2022 thematic review of ECL audits found that firms often failed to challenge the forward-looking adjustments applied to PD estimates. IFRS 9.5 .5.17(c) requires incorporation of reasonable and supportable forward-looking information, but audit teams accepted management overlays without verifying the external data sources or testing the sensitivity of ECL to alternative scenarios.
PD vs Loss Given Default (LGD)
PD and LGD answer different questions within the same ECL formula. PD measures how likely the borrower is to default. LGD measures how much the entity will lose if default occurs, expressed as a percentage of the exposure. A high PD with a low LGD (a risky borrower with strong collateral) can produce the same ECL as a low PD with a high LGD (a creditworthy borrower with no collateral).
Audit implications differ too. PD testing focuses on default rate data and rating migrations alongside forward-looking macro scenarios. LGD testing focuses on collateral valuations and historical recovery rates. Both feed into ISA 540 procedures, but the data sources and subject-matter expertise required are distinct.
Related terms
Related tools
Related reading
Frequently asked questions
Does PD apply to trade receivables, or only to banks?
PD applies to any financial asset measured at amortised cost or FVOCI, including trade receivables. IFRS 9.5.5.15 permits a simplified approach (the provision matrix) for trade receivables without a significant financing component, but the provision matrix implicitly embeds PD within the historical loss rates. Entities with significant financing components in their receivables must apply the general model with explicit PD inputs.
How do I audit the PD when the entity uses an internal model?
The auditor evaluates the model's methodology and tests the accuracy of input data, then assesses whether the model's outputs are consistent with actual default experience. ISA 540.13(a) requires the auditor to determine whether management's method is appropriate in the circumstances. For internal PD models, that includes back-testing the model's predictions against observed defaults over at least two prior periods.
When does a 12-month PD become a lifetime PD?
The shift happens when there is a significant increase in credit risk since initial recognition. IFRS 9.5.5.9 requires the entity to measure lifetime expected credit losses (using lifetime PD) once SICR is triggered. The entity must define quantitative and qualitative criteria for SICR and apply them consistently across the portfolio.