Budget execution and an error rate that crosses the threshold
You are preparing the ex-post control report for an EU agency whose 2024 budget has commitment appropriations of €84m and payment appropriations of €76m. At 31 October, commitments stand at 71% and payments at 58% of the respective appropriations. Historically, commitments made in the final two months of the year average 22% of the annual total, and the agency's internal target is 95% committed by 31 December. A €5.2m IT contract was legally committed in September, but the supplier now expects to invoice nothing before February, and the works element of the contract is 14 weeks late. Separately, your transactional sample of 60 payments, totalling €3.8m, was drawn from a population of 1,200 payments worth €76m. It disclosed five errors totalling €92,000: three ineligible staff costs, one arithmetical error and one payment to a wrong beneficiary that has not been recovered. The agency reports irregularities above a 2% materiality threshold to the authorising officer.
1. Extrapolate the error rate and conclude whether the 2% materiality threshold is exceeded. What action follows?
Read the worked answer
The sample error rate in monetary terms is €92,000 / €3,800,000 = 2.42%. Extrapolated to the €76m population, the most likely error is 2.42% × €76m ≈ €1.84m. The threshold is 2% × €76m = €1.52m. The projected error exceeds it by roughly €0.32m, and the 5-in-60 error frequency (8.3%) points the same way, so the conclusion does not depend on a single arithmetic slip.
But projection is a point estimate, not the whole picture. Add sampling risk: with only 60 items and 5 errors, the upper error limit will be well above €1.84m, so the true exposure could be materially larger. Two consequences follow. First, the control environment cannot be relied on, so the report should not say '5 errors found' and stop there - it should state the projected error and the upper limit. Second, action: recover the wrong-beneficiary payment, correct the arithmetical error, re-examine the ineligible staff costs across all staff-cost payments rather than only in the sample, and consider extending the sample to 120 payments to narrow the range before the report is signed.
If the €92,000 is not recovered, the irregular amount must be reported as such, regardless of materiality.
2. Will the agency reach the 95% commitment target, and how should the €5.2m IT contract and the late works be treated?
Read the worked answer
At 31 October, commitments are 71% × €84m = €59.64m. If the last two months follow the historic 22% pattern (€18.48m), the year-end figure is €78.12m, or 93.0% - a shortfall of €1.68m against the 95% target of €79.80m.
The €5.2m IT contract is already legally committed, so it is inside the €59.64m and must not be counted a second time. It will not generate payment this year, which matters for the payment appropriations, not for the commitment target. The real risk is the opposite: if any commitments recorded rely on legal commitments that were never signed, the 71% is overstated. The works element being 14 weeks late affects the payment forecast and may trigger a carryover request for the payment appropriations, subject to the applicable rules.
The correct response is a written forecast to the authorising officer identifying the €1.68m gap, the specific operations at risk and their status. What the agency must not do is create commitments in December to close a KPI: commitments without an underlying legal obligation or economic substance breach sound financial management and will return as decommitments or as an audit finding. Report the likely 93% outcome and its causes rather than engineering the number.
3. The director asks you to remove the extrapolation from the report because 'the errors we found are small in absolute terms'. How do you respond?
Read the worked answer
The five errors are indeed small individually - the largest component is the wrong-beneficiary payment, and no single item approaches €1.52m. That is precisely why projection is needed: sampling is designed to draw a conclusion about the population, and reporting only the absolute amount found would understate the population error by a factor of about twenty. Removing the extrapolation would misdescribe the evidence and undermine the purpose of the report, which is to inform the authorising officer, not to reassure him.
The report can and should present both perspectives fairly: the €92,000 actually identified and recoverable, and the projected €1.84m with its sampling risk. It should distinguish errors of legality and regularity (ineligible staff costs, wrong beneficiary) from a pure arithmetic error, because the former raise system questions and the latter does not. If the director disagrees with the methodology, his disagreement is recorded in the report or in the management response - the auditor does not delete the finding.
This separation matters: the auditor's conclusion, the director's comments and the corrective action plan are three distinct sections, and only the last is owned by management.