Defined Benefit Obligation

Where DBO audits go wrong

On most pension engagements, the DBO section of the file reads like this: actuary's report attached, discount rate noted, mortality table referenced, conclusion that the obligation is reasonable. No sensitivity analysis. No record of what the team actually tested. It's a tick box exercise dressed up as audit evidence, and inspectors have been flagging it for years.

The FRC's 2023/24 inspection cycle found the same pattern it found in 2022 and 2021: engagement teams accepted actuarial assumptions without testing whether alternative assumptions within a reasonable range would produce a materially different number. At firms like ours, the honest reason is that most audit seniors do not have the actuarial training to challenge a discount rate derivation on its merits, so the file defaults to confirming the actuary's credentials rather than evaluating the actuary's work. ISA 500.8 explicitly requires the latter. The gap between what the standard demands and what teams actually do on pension files is one of the widest in practice.

A defined benefit obligation is the present value (PV) of future pension payments an employer owes its employees for service already rendered, measured using actuarial assumptions that project cash outflows over decades. IAS 19.67 requires entities to use the projected unit credit method, which assigns a slice of the total expected benefit to each year of employee service. The calculation rests on demographic assumptions (mortality rates, expected turnover) and financial assumptions (the discount rate, projected salary increases).

The discount rate is the single most sensitive input. IAS 19.83 requires it to reflect high-quality corporate bond yields matching the currency and estimated duration of the obligation. But two actuaries working with the same workforce and the same duration profile can produce discount rates 30 basis points apart depending on which bond universe they use and what extrapolation method they apply to maturities beyond the observable curve. That 30-basis-point gap can shift a EUR 80M DBO by EUR 3M or more.

What actually happens on most engagements is that the team receives the actuary's report two weeks before sign-off, runs a brief reasonableness check against a published yield curve and moves on. Nobody rebuilds the calculation. Nobody asks why the actuary chose one bond index over another. The file should tell a story of independent evaluation. Usually it tells a story of reliance.

What the standard requires

Under ISA 540 , the DBO is an accounting estimate with high estimation uncertainty. The auditor must obtain an understanding of the entity's process for making the estimate (including the role of the actuary), identify the significant assumptions and evaluate whether those assumptions are reasonable given the entity's specific circumstances. Separately, the auditor must test the data inputs that feed the actuarial model.

ISA 500.8 adds a layer specific to management's experts. The auditor evaluates whether the expert has the competence and objectivity to produce reliable work. Then the auditor evaluates the relevance and reasonableness of the expert's findings. These are two separate steps. In practice, teams collapse them into one ("the actuary is qualified, so the report is sufficient"), which is exactly the shortcut inspectors keep citing.

What actually happens is that the standard's requirements are technically clear but practically difficult because the auditor is asked to exercise independent judgment over inputs that require specialist knowledge they typically don't have, and the standard resolves this tension by saying the auditor may use their own expert (a reviewing actuary) but does not require it.

The discount rate problem

IAS 19.83 says "high-quality corporate bonds." It does not say which index or which extrapolation method. In markets with deep corporate bond data (the UK, the eurozone), the choice of reference portfolio alone can produce discount rates that differ by 20 to 40 basis points for the same duration. In markets without a deep corporate bond market, IAS 19.84 requires the use of government bond yields instead, which introduces a different set of judgment calls about credit spreads.

I believe the discount rate is the area where audit firms most consistently under-invest in challenge, because testing it properly requires either a reviewing actuary (expensive) or a team member who understands yield curve construction (rare). A senior who can reconcile a bank confirmation to a sub-ledger cannot necessarily evaluate whether the iBoxx EUR Corporates AA 10+ index is a better reference than the Mercer yield curve for an obligation with a 14-year duration. So the file documents what is easy to document (the rate used, the index referenced) and skips what is hard (why this index, why this extrapolation, what the DBO would be under a different defensible choice).

DBO vs plan assets

The DBO measures what the employer owes. Plan assets measure what has been set aside to pay it. The net defined benefit liability (or asset) is the difference, and that net number is what hits the balance sheet under IAS 19.63 .

There is a legitimate disagreement among practitioners about where audit effort should concentrate. One view is that the DBO deserves the majority of testing time because it carries more estimation uncertainty (assumptions about mortality, salary growth and discount rates over decades). The other view is that plan asset valuation is where misstatements actually hide, because unquoted investments and property held in the fund are often valued using models that receive less scrutiny than the actuarial report itself. I lean toward the second view, because we've seen engagements where the team spent 80% of pension hours on DBO assumptions and missed a EUR 2M write-down on an illiquid property fund sitting in plan assets.

Worked example: Gruber Maschinenbau GmbH

Client: German industrial manufacturer, FY2024, revenue EUR 185M, IFRS reporter, 1,200 employees.

Gruber operates a final-salary pension plan closed to new entrants since 2015. It covers 480 active members and 320 retirees currently drawing benefits (total promised annual pensions for retirees: EUR 4.8M).

The actuary applied a discount rate of 3.45%, derived from the iBoxx EUR Corporates AA 10+ index. Weighted average duration of the obligation: 14.2 years. The audit team confirmed the rate against the Mercer yield curve for the same duration. Salary growth was set at 2.5% (consistent with the collective bargaining agreement through 2027) and mortality was based on the Heubeck 2018G tables (standard for German plans). So far, this is SALY work. Every assumption matches prior year methodology, every source is standard. The team verified salary growth against the CBA document and confirmed no updated mortality tables had been published.

Result: DBO of EUR 78.3M (prior year: EUR 81.1M). The EUR 2.8M decrease came from the 35-basis-point discount rate increase during FY2024. Sensitivity analysis showed a 50-basis-point decrease in discount rate would push the DBO up by EUR 5.9M (7.5%), confirming the materiality of this single input.

Where it gets complicated

Two weeks before sign-off, the team discovers that Gruber's HR department agreed to a supplementary pension top-up for 35 long-service employees as part of a retention package negotiated in Q4. The top-up adds approximately EUR 1.2M to the DBO, but it was agreed after the actuary's valuation date and is not reflected in the actuarial report.

The engagement partner now faces a choice. Ask the actuary to rerun the valuation (which will delay the audit by at least a week and cost the client an additional actuary fee). Or argue that the top-up relates to future service and doesn't affect the DBO at year-end. The second option is wrong (the top-up was agreed in the reporting period as compensation for past service), but it is the option that requires the least effort and creates the least friction with management.

We asked the actuary to rerun. The revised DBO came in at EUR 79.4M. Without the rerun, the file would have contained a EUR 1.1M misstatement on a line item where performance materiality was EUR 1.5M. That is uncomfortably close. And the point isn't that we caught it. The point is that we nearly didn't, because the standard workflow treats the actuary's report as a finished product rather than a starting point, and adjusting events that fall between the valuation date and the audit report date get lost in the gap.

The independence illusion

There is a structural contradiction sitting underneath every DBO audit. The standard says the auditor must exercise independent judgment over the actuarial assumptions. But the actuary is a specialist and the auditor is not. The auditor is asked to challenge work they could not reproduce, on assumptions whose sensitivity they can measure but whose derivation they often cannot evaluate at a technical level.

This creates a dependency the standard doesn't acknowledge. The auditor "evaluates" the discount rate by comparing it to a published yield curve. But the yield curve itself is a product of choices (which bonds to include and how to extrapolate beyond the longest observable maturity) that the auditor accepts without examination. The evaluation is really a comparison of one expert's output against another expert's output, with the auditor serving as referee in a match whose rules they learned from a training slide rather than from practice.

I think firms should be honest about this. If the DBO is material and the assumptions are sensitive, the engagement needs a reviewing actuary. Not as a luxury but as a prerequisite for the file to contain genuine evidence of challenge rather than a record of the team confirming that the number looked about right. The cost of a reviewing actuary on a material pension engagement is trivial relative to the inspection risk and the potential misstatement, yet we've seen it treated as an optional extra on engagements where the pension liability exceeds EUR 50M.

One consequence that rarely gets discussed: when audit firms systematically under-challenge actuarial assumptions, entities learn that the actuarial report will pass review without friction, and the incentive to appoint a conservative actuary (one who pressure-tests assumptions rather than confirming management's preferences) weakens. Over time, the quality of the inputs degrades because nobody in the chain has a reason to push back.

Key standard references

• IAS 19.67 -69: Measurement of the DBO using the projected unit credit method
• IAS 19.83 : Discount rate requirements (high-quality corporate bond yields matching the currency and duration of the obligation)
• IAS 19.84 : Government bond yield fallback where a deep corporate bond market does not exist
• IAS 19.63 : Net defined benefit liability or asset recognised on the balance sheet (DBO minus plan assets)
• ISA 540 : Auditing accounting estimates, including the framework for evaluating the DBO as a high-estimation-uncertainty item
• ISA 500.8 : Evaluating the work of management's expert (the actuary), covering both competence and reasonableness of findings

Related terms

Related reading

Related tools

Frequently asked questions