Article 5 · Commission Delegated Regulation (EU) 2023/67

Contents

Article 5Selection of a random statistical sample#

1.The audit authority shall select a random statistical sample, from the population established in accordance with Articles 3 and 4, using one of the following methods:

(a)the MUS standard approach;

(b)Simple Random Sampling (‘SRS’).

2.Where the audit authority uses the MUS standard approach, it shall select a sample using PPS.The low value units shall be selected on the basis of an interval of selection calculated using the expenditure of a low value stratum after determining an exhaustive high value stratum. All high value units above the interval of selection shall be audited, subject to exceptions provided in Article 3(6), first subparagraph, points (a) and (b), and third subparagraph.

3.Where the audit authority uses SRS, it shall select a sample using equal probability selection, with the optional use of an exhaustive stratum.

4.Annexes I and II set out sampling parameters and formulas to calculate the sample size for the methods referred to in paragraph 1, points (a) and (b), except where the capped sample size referred to in paragraph 7 is applied. Those formulas may be used with different sampling designs covering stratification or multi-period sampling or a combination of both.

5.The sample shall have at least 30 units and at least 3 units in each random stratum of a sampling period.

6.In the case of multi-period sampling, the audit authority shall apply one of the following approaches to recalculate the sample size in order to adjust to updated sampling parameters:

(a)standard recalculation of the sample size;

(b)global recalculation of the sample size.

Where the audit authority uses the approach of the standard recalculation of the sample size as referred to in point (a), the sample size of subsequent sampling period or periods shall be recalculated while maintaining the sample sizes of previous sampling periods of the accounting year.Where the audit authority uses the approach of the global recalculation of the sample size as referred to in point (b), both the total sample size and the sample size per sampling period shall be recalculated.

7.As regards programmes assessed in category 1 or 2 as set out in Table 2 of Annex XI to Regulation (EU) 2021/1060, which are not subject to enhanced proportionate arrangements pursuant to Article 83 of that Regulation, the audit authority may cap the sample size at 50 sampling units.Where the capped sample size referred to in the first subparagraph is used, it shall be applied to a sample for the whole population comprising, where applicable, more than one programme and programming periods.All units in random strata and only high value units in exhaustive strata shall count for the capped sample size.Multi-period sampling procedures may be applied with the capped sample size. In case of underestimated forecasts of population size or expenditure for the second or subsequent sampling periods, the audit authority shall take one of the following actions:

(a)increase the sample size to take into consideration the underestimated values of the forecasts;

(b)calculate the sample size in accordance with the formulas set out in Annex II.

Defined terms · 21

Show every defined term

audit population
means a sampling population composed of units with positive values where expenditure is above 0;
bilateral testing
means an approach to statistical sampling allowing the calculation of both the upper error limit and the lower error limit;
confidence interval
means the interval that contains the true value of error in the population with a certain probability known as ‘confidence level’;
confidence level
means a probability that a confidence interval includes the true value of the estimated parameter;
equal probability
means one of the random selection methods in which random numbers are used in order to select randomly the units constituting the sample with equal probabilities;
exhaustive stratum
means a part of the positive population related to the accounting year or to a sampling period for which all sampling units are audited.
lower error limit
means an error limit calculated by deduction of the ‘sampling precision’ from the ‘extrapolated error’, adjusted, where applicable, by adding delimited systemic errors and uncorrected anomalous errors;
multi-period sampling
means a sampling procedure where the audit population for an accounting year is divided into two or more sampling periods, which could be of the same or of a different length;
negative population
means a sampling population composed of units with negative values or null values where expenditure is below or equal to 0;
non-statistical sampling method
means a sampling method that does not involve the evaluation of the sampling risk and precision and is based on a random selection;
PPS
means one of the random selection methods that uses the monetary unit as an auxiliary variable for sampling where the selection of the units constituting the sample is based on a probability proportional to the monetary value of the sampling unit (higher value units have higher probability of selection) and the selection is usually based on a systematic selection with a random starting point and application of a systematic rule to select the additional units;
projected error
means the result of extrapolating the random errors found in the sample to the total population where the extrapolation/projection procedure depends on the sampling method used;
random selection
means a probabilistic selection which refers either to probability proportional to size or equal probability selection ensured by using random number generating software, specialised or not, including MS Excel;
sampling method
means a technical tool to select a sample and extrapolate results within a sampling methodology, which may be either statistical or non-statistical;
sampling methodology
means a methodology describing the main elements and steps of a sampling procedure and which encompasses a sample selection stage including subsampling and extrapolation of results;
sampling precision
means a sampling parameter measuring uncertainty in the extrapolation of sampling results to the population, and corresponding to the maximum expected deviation between the extrapolated error and the true population error which is achieved with a probability equal to the confidence level;
SRS
means a standardised statistical sampling method based on equal probability selection, compatible with different sampling designs, including stratification and ‘multi-period sampling’;
subsampling
means a two-stage or multi-stage sampling where an error for a sampling unit is established on the basis of an extrapolation from a subsample of invoices or other subsampling units;
subsampling unit
means a unit, which may be an invoice or other unit, into which a sampling unit is divided for the purpose of subsampling and which is audited exhaustively unless another level of subsampling for the subsampling unit is applied;
unilateral testing
means an approach to statistical sampling allowing the calculation of only one error limit, typically the upper error limit;
upper error limit
means the sum of the ‘sampling precision’ and the ‘extrapolated error’ and, where applicable, delimited systemic errors and uncorrected anomalous errors;