Monetary unit sampling evaluation: tainting, projected misstatement and the upper misstatement limit
To evaluate a monetary unit sample you work out the tainting (the percentage misstatement) of each item with a difference, project it over the sampling interval, and add basic precision and an incremental allowance. The total is the upper misstatement limit: if it does not exceed tolerable misstatement, the population is accepted. In the example below, four overstatements and one understatement give an upper limit of 117,812 against a tolerable misstatement of 120,000.
Updated 10 November 2026
What you need to evaluate the sample
The evaluation uses the planning inputs plus the result for every item tested:
- The sampling interval you selected with. Every projection is built on it. If you selected every 32,000, you evaluate with 32,000, even if the theoretical sample size had decimals.
- The confidence level and tolerable misstatement set at planning.
- The book value (BV) and audited value (AV) of every sample item, including those without differences.
The example continues the one in the monetary unit sampling calculator guide: sales of 4,812,350, tolerable misstatement 120,000, expected misstatement 15,000 and 95% confidence. Items were selected with a 32,000 interval. Testing turns up five differences.
Tainting: the misstatement percentage of each item
Tainting = (book value − audited value) ÷ book value
MUS samples monetary units, not invoices. Each selected unit stands for its whole interval, so what you project is not the difference in currency but the share of the item that is wrong. An invoice booked at 8,000 that should have been 6,000 has a difference of 2,000 and a tainting of 25%. That 25% applies to the 32,000 interval it represents.
Projected misstatement and the most likely misstatement
For each item smaller than the interval, projected misstatement is tainting × interval. Add up the projections and the actual differences in the top stratum and you have the most likely misstatement in the population, your best point estimate of the error it contains. A 25% tainting with a 32,000 interval projects 8,000.
Top-stratum items
An item equal to or larger than the interval is always selected, so that part of the population has been tested in full. Its difference counts at its actual amount, with no projection and no incremental allowance, because there is no sampling risk on it. In the example, a 58,400 invoice with a 2,500 difference contributes exactly 2,500.
Basic precision
Basic precision = CF(0 misstatements) × interval = 3.00 × 32,000 = 96,000
This is the upper limit you would have if the sample had no misstatements at all. It covers the risk that the population holds errors the sample did not hit. With zero errors the upper limit is therefore 96,000, not zero.
Incremental allowance: ranking the taintings
Each additional misstatement widens the allowance for sampling risk by more than its own projection. Overstatements in sampled items (not top-stratum items) are ranked from the largest tainting to the smallest, and each gets the increase in the confidence factor minus one:
Incremental allowance = (CF(i) − CF(i−1) − 1) × projected misstatement
The order matters: the largest increments go to the largest taintings, which gives the most prudent result. These are the AICPA confidence factors for 0 to 4 misstatements and their increments:
| Misstatements | CF at 90% | Increment at 90% | CF at 95% | Increment at 95% |
|---|---|---|---|---|
| 0 | 2.31 | – | 3.00 | – |
| 1 | 3.89 | 1.58 | 4.75 | 1.75 |
| 2 | 5.33 | 1.44 | 6.30 | 1.55 |
| 3 | 6.69 | 1.36 | 7.76 | 1.46 |
| 4 | 8.00 | 1.31 | 9.16 | 1.40 |
Source: AICPA, Audit Guide: Audit Sampling. Poisson factors rounded up to two decimals, as the guide prints them.
Worked example: evaluating a monetary unit sample
Interval 32,000, 95% confidence. The five differences, in the order they were found:
| Item | Book value | Audited value | Difference | Tainting | Projected | CF increment | Allowance |
|---|---|---|---|---|---|---|---|
| INV-1043 | 10,000 | 9,400 | 600 | 6% | 1,920 | 1.55 | 1,056 |
| INV-2217 | 8,000 | 6,000 | 2,000 | 25% | 8,000 | 1.75 | 6,000 |
| INV-3380 | 2,400 | 2,280 | 120 | 5% | 1,600 | 1.46 | 736 |
| INV-0561 (top stratum) | 58,400 | 55,900 | 2,500 | – | 2,500 | – | – |
| INV-4102 (understatement) | 1,500 | 1,650 | −150 | −10% | −3,200 | – | – |
Ranked, the overstatement taintings are 25%, 6% and 5%, and they take the increments 1.75, 1.55 and 1.46. INV-2217 adds (1.75 − 1) × 8,000 = 6,000; INV-1043 adds (1.55 − 1) × 1,920 = 1,056; INV-3380 adds (1.46 − 1) × 1,600 = 736.
| Component | Amount |
|---|---|
| Basic precision (3.00 × 32,000) | 96,000 |
| Projected misstatement (1,920 + 8,000 + 1,600 + 2,500) | 14,020 |
| Incremental allowance (6,000 + 1,056 + 736) | 7,792 |
| Upper misstatement limit | 117,812 |
| Tolerable misstatement | 120,000 |
The upper limit (117,812) does not exceed tolerable misstatement (120,000): at 95% confidence, sales are not overstated by more than 120,000. The population is accepted, with little headroom, which deserves a line in the working paper. The differences found (5,070 net) go to the summary of uncorrected misstatements at their projected amount, not only the actual one.
Understatements
INV-4102 is understated: it projects 3,200 in the other direction. In this method it is not deducted from the overstatement limit; it is reported separately. Net projected misstatement would be 14,020 − 3,200 = 10,820, but offsetting understatements against overstatements would shrink the limit without enough evidence behind it. Some firm methodologies allow controlled netting or compute a separate understatement limit; follow yours and document it.
MUS is also weak at finding understatements: an omitted item, or one booked at zero, cannot be selected. If completeness is the main risk, test it from a different population.
What to do when the upper limit exceeds tolerable misstatement
Suppose INV-2217 had been audited at 4,800 instead of 6,000. Its tainting becomes 40% and it projects 12,800; projected misstatement rises to 18,820, the incremental allowance to 11,392 and the upper limit to 126,212, which is 6,212 over tolerable misstatement. The sample does not support a conclusion that the population is free of material misstatement. Sensible responses:
- Look at the nature and cause of each misstatement. If they share a cause (one customer, one period, one process), testing that subset in full may work better.
- Ask management to investigate and correct the population, then test again. Correcting only the sample items does not remove the projection.
- Extend the sample or perform other substantive procedures. Extending only helps if the extra items come back clean; with a genuinely high error rate the limit will not come down.
- Revisit the risk assessment and whether related controls can be relied on.
- If the misstatement is not corrected, consider the effect on the auditor's report.
Under ISA 530, treating a misstatement as an anomaly (not representative of the population) is only possible in extremely rare cases and with a high degree of certainty. It is not a way to make the result fit.
Documenting the conclusion
A reviewer should be able to re-perform the evaluation from the working paper: interval and random start used, list of items tested, each difference with its tainting, the ranking, the factors applied, the upper limit and the conclusion against tolerable misstatement. If you work in a spreadsheet, the guide to audit sampling in Excel shows how to set up those columns so they hold together.
Frequently asked questions
What is tainting in monetary unit sampling?
It is the misstatement percentage of an item: its difference divided by its book value. An invoice booked at 8,000 and audited at 6,000 has a 25% tainting. That percentage is projected over the sampling interval.
How do you calculate the upper misstatement limit?
Add basic precision (the confidence factor for zero misstatements times the interval), the projected misstatement, and the incremental allowance for the overstatements ranked by tainting. If the total does not exceed tolerable misstatement, accept the population.
In a monetary unit sample, what is the most likely misstatement in the population equal to?
The sum of the projected misstatements: tainting times the sampling interval for every sampled item with an error, plus the actual misstatement in top-stratum items. It excludes basic precision and the incremental allowance.
For monetary unit sampling, what does a sampling interval of 400 mean?
One monetary unit is selected in every 400 of the cumulative listing. Every item of 400 or more is certain to be selected, and an item of 100 has a 1 in 4 chance. If that item of 100 has a 25% tainting, it projects 0.25 × 400 = 100.
What if the upper misstatement limit exceeds tolerable misstatement?
Do not accept the population. Analyse the cause of the misstatements, ask management to investigate and correct the population, extend testing or perform other procedures, and if it is not corrected consider the effect on your report.
Are understatements deducted from the upper limit?
Not in the prudent AICPA approach: they are projected and reported separately. Deducting them would lower the overstatement limit without enough support. Some firms allow controlled netting; follow your methodology.
- AICPA, Audit Guide: Audit Sampling
- IAASB, ISA 530 Audit Sampling
- AICPA, AU-C 530 Audit Sampling
- ICAC, NIA-ES 530 Muestreo de auditoría
The evaluation without hand formulas
AuditSampler Pro records the audited value of each item, works out tainting, incremental allowance and the upper limit with the AICPA factors, and produces the working paper as a PDF.