Most audit tests look at less than the whole population. When the auditor selects fewer than 100% of the items and uses the results to conclude on the entire population, that is audit sampling, and AU-C 530, Audit Sampling applies. The standard does not require statistical methods. It does require the auditor to design the sample so it can provide a reasonable basis for conclusions, to select items so that each sampling unit has a chance of selection, and to project and evaluate the results.
This guide covers when sampling applies, the main methods (attribute sampling, monetary unit sampling and nonstatistical sampling), the drivers of sample size, selection, and how to project and evaluate misstatements. Table values shown are those published in the AICPA audit sampling guide; use the tables in your firm's methodology for actual engagements. The guide paraphrases AU-C 530.
When is a test audit sampling?
A test is audit sampling when the auditor applies a procedure to less than 100% of a population and evaluates the results to draw a conclusion about the whole population.
These are not audit sampling:
- Testing all items in a population, or all items in a defined subset, such as every invoice over $50,000. If the remaining items are not tested or are tested separately, the conclusion covers only what was tested.
- Selecting key items because they are large, unusual or high risk, without projecting to the rest.
- Walkthroughs to understand a process.
- Analytical procedures, which are not sampling.
A common design is a combination: test all items above a threshold 100%, then sample the remainder. Only the remainder is subject to AU-C 530. Summary: AU-C 530 explained.
Statistical vs nonstatistical sampling
Under AU-C 530, either approach is acceptable when properly applied.
| Statistical sampling | Nonstatistical sampling | |
|---|---|---|
| Selection | Random selection of sample items | Haphazard, systematic or random |
| Evaluation | Probability theory to measure sampling risk | Auditor judgment to evaluate sampling risk |
| Sample size | From formulas or tables | Judgment, often guided by firm tables built on statistical models |
| Main benefit | Sampling risk is quantified and defensible | Simpler; flexible for small populations |
AICPA audit sampling guidance expects a nonstatistical sample to be of a size comparable to a well-designed statistical sample for the same objective. A sample of 10 items chosen only because "10 seems enough" has no documented basis.
Sampling risk and nonsampling risk
Sampling risk is the risk that the auditor's conclusion based on a sample differs from the conclusion that would be reached if the whole population were tested. It has two forms:
- In tests of controls, concluding controls are more effective than they are (risk of overreliance), or less effective than they are.
- In tests of details, concluding a material misstatement does not exist when it does (risk of incorrect acceptance), or that one exists when it does not.
The first form in each pair affects audit effectiveness and is the one sample size is designed to control. The second affects efficiency.
Nonsampling risk is everything else: using the wrong procedure, misreading evidence, failing to recognize an error. Larger samples do not fix it. Supervision and review do.
Attribute sampling for tests of controls
Attribute sampling estimates the rate of deviation from a control in a population. It answers the question: is the deviation rate low enough to rely on the control?
Inputs
| Input | Effect on sample size |
|---|---|
| Confidence level (the complement of the acceptable risk of overreliance) | Higher confidence, larger sample |
| Tolerable deviation rate | Lower tolerable rate, larger sample |
| Expected deviation rate | Higher expected rate, larger sample |
| Population size | Little effect for populations over a few thousand items |
Commonly published sample sizes
| Confidence | Tolerable rate | Expected rate | Sample size |
|---|---|---|---|
| 95% | 5% | 0% | 59 |
| 95% | 5% | 1% | 93 |
| 90% | 5% | 0% | 45 |
| 90% | 10% | 0% | 22 |
The audit sample size calculator computes the sample size and the achieved upper deviation limit from your inputs.
Evaluating the result
After testing, compute the sample deviation rate and the upper deviation limit at the chosen confidence level. If the upper limit is below the tolerable rate, the sample supports reliance. If it exceeds the tolerable rate, the auditor reduces reliance on the control and increases substantive testing.
Each deviation also needs qualitative evaluation: was it an error or intentional, and does it indicate a broader problem? A single deviation that looks deliberate can matter more than a rate.
For small, infrequent controls, such as a quarterly or monthly review, firms usually use fixed sample sizes from their methodology (for example, a small number of months for a monthly control) rather than attribute tables, because the population is too small for sampling formulas to apply sensibly.
Monetary unit sampling for tests of details
Monetary unit sampling (MUS), also called probability-proportional-to-size (PPS) sampling, treats each dollar in the population as a sampling unit. A dollar is selected and the item containing it, such as an invoice or a customer balance, is tested. Larger items are more likely to be selected, and every item at or above the sampling interval is selected with certainty.
MUS fits tests for overstatement, such as existence of receivables or validity of recorded expenses. It does not test well for understatement or for zero and negative balances, which have no chance of selection. Use a different design for completeness.
Sample size
A common formula:
Sample size = (book value × reliability factor) ÷ (tolerable misstatement − expected misstatement × expansion factor)
The sampling interval is book value divided by sample size.
Published values at 95% confidence: a reliability factor of 3.0 for zero misstatements and an expansion factor of 1.6. At 90% confidence, the zero-error reliability factor is 2.31 and the expansion factor 1.5.
Worked example (illustrative)
Accounts receivable of $5,000,000. Tolerable misstatement $150,000. Expected misstatement $0. Confidence 95%.
- Sample size: (5,000,000 × 3.0) ÷ 150,000 = 100
- Sampling interval: 5,000,000 ÷ 100 = $50,000
Using systematic selection with a random start, the auditor selects every 50,000th dollar. Every balance of $50,000 or more is selected.
Testing finds one misstatement: a balance recorded at $10,000 should be $8,000. The balance is below the interval, so the auditor calculates a tainting percentage and projects it.
| Step | Calculation | Amount |
|---|---|---|
| Tainting | (10,000 − 8,000) ÷ 10,000 | 20% |
| Projected misstatement | 20% × 50,000 | $10,000 |
| Basic precision | 3.0 × 50,000 | $150,000 |
| Incremental allowance | (4.75 − 3.0 − 1.0) × 10,000 | $7,500 |
| Upper misstatement limit | 150,000 + 10,000 + 7,500 | $167,500 |
The 4.75 is the commonly published reliability factor for one misstatement at 95% confidence. The upper misstatement limit of $167,500 exceeds tolerable misstatement of $150,000, so the sample does not support the conclusion that the balance is fairly stated at the desired confidence. Options include asking management to investigate and correct the population, extending testing, or performing other procedures. The known $2,000 misstatement goes to the summary of misstatements regardless.
If a misstatement is found in an item at or above the sampling interval, the actual misstatement is used, not a projection, because every such item was tested.
Run the arithmetic with the monetary unit sampling calculator.
Nonstatistical sampling for tests of details
Many small and mid-size firms use nonstatistical samples for substantive tests, often sized from a firm table driven by:
- the amount of the population not covered by key items
- tolerable misstatement for the account
- the assessed risk of material misstatement
- the assurance obtained from other procedures, such as substantive analytics or tests of controls
A common firm approach expresses sample size as (population subject to sampling ÷ tolerable misstatement) × an assurance factor that rises with risk. Whatever the method, document how the size was determined.
Projecting misstatements in a nonstatistical sample
AU-C 530 requires the auditor to project misstatements found in the sample to the population. Two common methods:
- Ratio method: misstatement in the sample ÷ book value of the sample × book value of the population sampled.
- Difference method: average misstatement per item × number of items in the population.
Illustrative example: a nonstatistical sample of 40 expense items with a book value of $400,000 is drawn from a population of $3,200,000 (excluding items tested 100%). Misstatements found total $6,000. Ratio projection: 6,000 ÷ 400,000 × 3,200,000 = $48,000. Add misstatements found in the 100% tested items to get the total for the account.
Compare the projected misstatement with tolerable misstatement. If it is close, the risk that actual misstatement exceeds tolerable is high, and the auditor considers more work. The projected misstatement is recorded on the summary of misstatements under AU-C 450 as a projected (not factual) misstatement.
Selection methods
AU-C 530 requires items to be selected so that each sampling unit in the population has a chance of selection.
| Method | How it works | Notes |
|---|---|---|
| Random | Random number generator mapped to the population | Common for statistical sampling; systematic selection with a random start is also used |
| Systematic | Every nth item or nth dollar from a random start | Used for MUS; watch for patterns in the population |
| Haphazard | Auditor selects without conscious bias | Nonstatistical only; avoid unconsciously skipping hard-to-find or small items |
| Block | A contiguous group of items | Rarely appropriate as a sampling method on its own |
Before selecting, confirm the population is complete: tie the listing to the general ledger or trial balance, and check that it includes all items in the period. A perfect sample from an incomplete population proves nothing about the missing items.
What to document
To meet AU-C 230 and AU-C 530, firms typically document:
- the objective of the test and the assertion addressed
- the population, how its completeness was tested, and the sampling unit
- the definition of a deviation or misstatement
- the confidence level or risk, tolerable rate or tolerable misstatement, and expected rate or misstatement
- the sample size and how it was determined
- the selection method and the identifying characteristics of each item tested
- the results, including each deviation or misstatement and its investigation
- the projection and evaluation, and the conclusion
Common problems
- Sample sizes with no documented basis.
- Projected misstatements omitted from the summary of misstatements.
- Populations not reconciled to the trial balance before selection.
- Deviations explained away without a documented, convincing basis.
- MUS used to test completeness.
- Key items and the sampled remainder mixed, with the conclusion stated for the whole account.
Frequently asked questions
What is the difference between attribute sampling and monetary unit sampling?
Attribute sampling estimates the rate of deviations in a population and is used for tests of controls. Monetary unit sampling estimates the dollar amount of misstatement and is used for substantive tests of details, especially for overstatement.
Does AU-C 530 require statistical sampling?
No. AU-C 530 allows statistical or nonstatistical sampling. Both require a sample that provides a reasonable basis for conclusions, selection that gives every item a chance, and projection of misstatements.
What drives the sample size in an audit test?
The main drivers are the acceptable sampling risk or confidence level, tolerable misstatement or tolerable deviation rate, and expected misstatement or deviation rate. Population size has little effect once the population is large.
Do I have to project a misstatement found in a sample?
Yes. AU-C 530 requires projection of misstatements found in a sample to the population, and the projected amount is evaluated against tolerable misstatement and accumulated under AU-C 450.
Is testing every item over a threshold considered sampling?
No. Testing all items above a threshold is not sampling. If the remaining items are sampled and the results projected, only that remainder is subject to AU-C 530.
Sources
- AICPA, AU-C Section 530, Audit Sampling
- AICPA, Audit Guide: Audit Sampling (attribute sampling sample-size tables, monetary unit sampling reliability and expansion factors, nonstatistical sample-size approach)
- AICPA, AU-C Section 450, Evaluation of Misstatements Identified During the Audit
- AICPA, AU-C Section 230, Audit Documentation
- AICPA, AU-C Section 330, Performing Audit Procedures in Response to Assessed Risks and Evaluating the Audit Evidence Obtained
CORAA prepares ready-to-review audit files for US CPA firms from QuickBooks books; you review and sign. See /us.
Part of CORAA University · Evidence and audit procedures: related guides, calculators and templates.
AU-C references are to AICPA Professional Standards, paraphrased here; read the standard itself before relying on a requirement. This page is guidance, not professional advice. Report an error