Variables Sampling Explained: A Practical Guide for Internal and External Auditors
How Auditors Use Statistical Sampling to Estimate Monetary Misstatements and Evaluate Financial Statement Balances
Auditors rarely have the resources to examine every transaction in a large financial population.
A company may have 100,000 inventory items, 50,000 customer invoices, or millions of dollars in accounts receivable.
The auditor must determine how to obtain sufficient appropriate evidence without examining every item.
Variables sampling provides a statistical method for estimating monetary values and misstatements based on a representative sample.
It is especially useful when an auditor needs to determine whether a recorded financial statement balance is reasonably stated.
However, variables sampling is not simply a matter of selecting 30 transactions, calculating the average error, and multiplying the result by the population.
An effective sampling application requires careful planning, appropriate selection, reliable measurement, statistical evaluation, and professional judgment.
This article explains the major variables sampling techniques, their advantages and limitations, and how auditors can use them effectively.
1. What Is Variables Sampling?
Variables sampling is an audit sampling approach used to estimate a numerical characteristic of a population.
In financial auditing, that characteristic is commonly a monetary amount or misstatement.
For example, an auditor may use variables sampling to estimate:
The audited value of inventory.
The total amount of accounts receivable misstatement.
The value of inventory pricing errors.
The monetary amount of payroll overpayments.
The amount of incorrect depreciation expense.
The estimated misstatement in a population of invoices.
Variables sampling differs from attribute sampling because it measures amounts rather than simply whether a characteristic is present.
An auditor using attribute sampling might determine the percentage of purchase orders lacking approval.
An auditor using variables sampling might estimate the dollar amount of overpayments in a population of purchases.
Both methods can be valuable, but they answer different questions.
2. Variables Sampling Versus Attribute Sampling
Characteristic | Attribute sampling | Variables sampling |
Primary objective | Estimate frequency of deviations | Estimate monetary values or misstatements |
Measurement | Yes/no or pass/fail | Numerical amount |
Common application | Tests of controls | Substantive tests of details |
Example | Percentage of invoices lacking approval | Estimated dollar amount of invoice errors |
Typical conclusion | Estimated deviation rate and sampling allowance | Estimated amount and precision or misstatement limit |
Consider a population of 10,000 vendor payments.
Attribute sampling may estimate how frequently payments lacked the required authorization.
Variables sampling may estimate the total amount by which payments were incorrect.
An unauthorized payment is not necessarily an overpayment, and a correctly authorized payment is not necessarily accurately recorded.
The audit objective determines which method is appropriate.
3. The Two Major Statistical Approaches to Monetary Sampling
Auditors commonly use two broad approaches to statistical sampling for substantive tests of details.
Classical variables sampling
Classical variables sampling uses statistical estimation techniques based on the numerical characteristics of sampled items.
The principal methods are:
Mean-per-unit estimation.
Difference estimation.
Ratio estimation.
Monetary unit sampling
Monetary unit sampling (MUS) uses probability-proportional-to-size selection, generally giving larger recorded monetary amounts a greater probability of selection.
MUS is frequently useful for testing overstatements in recorded asset balances.
Both approaches can provide appropriate audit evidence when properly designed and evaluated.
The choice depends on the audit objective, population characteristics, expected misstatements, and sampling risks.
4. Mean-Per-Unit Estimation
Mean-per-unit estimation uses the average audited value of sampled items to estimate the total audited value of a population.
Suppose an auditor is examining inventory.
The population contains 10,000 inventory items.
The auditor selects a statistically appropriate random sample of 100 items and independently determines their audited values.
The total audited value of the sampled items is $48,000.
The average audited value is: $48,000 ÷ 100 = $480 per item
The estimated total population value is: 10,000 × $480 = $4,800,000
If the recorded inventory balance is $5,000,000, the point estimate suggests a $200,000 overstatement.
However, this is only a point estimate.
The auditor must also calculate the appropriate sampling precision or confidence bound before determining whether the recorded balance is acceptable.
When is mean-per-unit estimation useful?
Mean-per-unit estimation may be appropriate when the auditor wants to estimate the value of a population independently of recorded amounts.
For example, it may be useful when independently estimating inventory values or another numerical population characteristic.
ts efficiency depends heavily on the variability of the audited values.
A population containing both very small and very large items may require a larger sample unless the auditor uses an appropriate stratification strategy.
5. Difference Estimation
Difference estimation focuses on the difference between recorded and audited amounts.
The auditor determines the misstatement for each sampled item and uses those differences to estimate the total population misstatement.
Suppose an auditor examines a population of 5,000 invoices.
A random sample of 100 invoices is selected.
The auditor identifies the following results:
Sample result | Amount |
Total recorded value of sampled invoices | $250,000 |
Total audited value of sampled invoices | $247,000 |
Net sample overstatement | $3,000 |
Average overstatement per sampled invoice | $30 |
The estimated population overstatement is: 5,000 × $30 = $150,000
This provides a point estimate of the population misstatement.
The auditor must then calculate an appropriate allowance for sampling risk.
When is difference estimation useful?
Difference estimation may be effective when the auditor expects measurable differences between recorded and audited values.
Its efficiency depends on the variability of those differences.
If most differences are relatively consistent, the method may provide a useful estimate.
If differences are highly variable or occur very infrequently, the statistical precision may be poor, and the auditor may need another method.
6. Ratio Estimation
Ratio estimation uses the relationship between audited and recorded values in the sample to estimate the audited value of the population.
Suppose a company reports inventory of $8,000,000.
The auditor selects a representative sample with a recorded value of $400,000.
After performing audit procedures, the audited value of the sampled items is $392,000.
The audited-to-recorded ratio is: $392,000 ÷ $400,000 = 0.98
Applying the ratio to the population gives: $8,000,000 × 0.98 = $7,840,000
The estimated overstatement is: $8,000,000 − $7,840,000 = $160,000
Again, this is a point estimate, not a complete statistical conclusion.
When is ratio estimation useful?
Ratio estimation may be useful when audited amounts are closely related to recorded amounts and the relationship is reasonably stable.
For example, it may be appropriate for inventory pricing tests when errors tend to vary proportionally with recorded values.
The method may be less effective when recorded amounts are zero, relationships are unstable, or the population contains unusual items.
7. Comparing the Three Classical Variables Methods
Method | Estimation basis | Potentially useful application |
Mean-per-unit | Average audited value per item | Independent estimate of a population's value |
Difference | Average recorded-versus-audited difference | Estimation of monetary misstatement |
Ratio | Relationship between audited and recorded amounts | Populations with proportional valuation relationships |
The methods do not necessarily produce identical estimates or precision.
The auditor should select a method based on the population's characteristics and the intended audit conclusion.
The examples above are simplified illustrations of point estimation. They do not constitute complete statistical sampling plans or demonstrate that a balance is fairly stated.
8. Monetary Unit Sampling: An Alternative to Classical Variables Sampling
Monetary unit sampling is another common statistical method for substantive audit testing.
Instead of selecting physical items with equal probability, MUS generally selects monetary units.
Larger recorded balances therefore have a greater probability of selection.
For example, consider an accounts receivable population of $5,000,000.
A customer balance of $100,000 is more likely to be selected than a customer balance of $1,000.
This makes MUS particularly useful when the auditor is concerned about overstatements in recorded balances.
However, MUS has limitations.
It may be less suitable when the primary concern is understatement or omitted items.
Zero and negative balances also require special consideration.
Classical variables sampling may be more appropriate when both overstatements and understatements are expected or when the auditor wants to estimate the total audited population value independently.
9. Planning a Variables Sampling Engagement
Before selecting any sample, the auditor should establish the audit objective.
Consider an inventory valuation audit.
The objective might be: “Determine whether the recorded inventory balance is materially misstated because of incorrect unit costs or quantities.”
The auditor should then define:
Population: The inventory records subject to the audit.
Sampling unit: The individual inventory item, record, or other defined unit.
Recorded value: The amount included in the company's accounting records.
Audited value: The amount determined through the auditor's procedures.
Misstatement: The difference between recorded and audited values.
Tolerable misstatement: The maximum population misstatement the auditor is willing to accept for the planned procedure.
Expected misstatement: The auditor's anticipated level of misstatement based on available information.
Acceptable sampling risk: The risk the auditor is willing to accept that the sample will lead to an incorrect conclusion.
The sampling plan should also identify the statistical method, selection procedure, sample size, and evaluation methodology.
10. Understanding Tolerable Misstatement
Tolerable misstatement is a critical input into substantive sampling.
It represents the maximum monetary misstatement the auditor is willing to accept in the population for the particular sampling application.
Suppose a company reports $10 million in inventory.
The auditor establishes tolerable misstatement for the inventory population at $250,000.
The sampling procedure must be designed to provide appropriate assurance concerning whether population misstatement exceeds that amount.
A point estimate of $150,000 does not automatically establish that the balance is acceptable.
The auditor must consider sampling risk and the resulting statistical bound.
If the calculated upper misstatement limit exceeds $250,000, the sample may not provide sufficient support for accepting the balance under the planned sampling approach.
The auditor may need additional procedures, adjustments, or a revised conclusion.
11. Understanding Sampling Risk
Sampling risk is the risk that a conclusion based on a sample differs from the conclusion the auditor would reach by examining the entire population using the same procedures.
Two important risks in substantive testing are:
Risk of incorrect acceptance
The auditor concludes that a materially misstated balance is acceptable when it is not.
This is particularly important because it can affect the audit opinion.
Risk of incorrect rejection
The auditor concludes that a balance is unacceptable when the population is not materially misstated.
This may lead to unnecessary additional audit work.
Statistical sampling provides a way to quantify sampling risk under the assumptions of the selected method.
Nonstatistical sampling also requires consideration of sampling risk, but it does not quantify that risk through a statistical probability model.
12. How Auditors Determine Sample Size
Sample size is affected by several factors.
Factor | General effect on required sample size |
Lower tolerable misstatement | Increases sample size |
Higher expected misstatement | Generally increases sample size |
Greater population variability | Generally increases sample size |
Lower acceptable risk of incorrect acceptance | Increases sample size |
More assurance from other relevant audit procedures | May reduce required sample size |
Effective stratification | May improve sampling efficiency |
There is no universal sample size that is appropriate for every variables sampling application.
Selecting 25, 40, or 60 transactions because that number has historically been used does not establish statistical sufficiency.
Sample size must reflect the audit objective, population characteristics, sampling methodology, and required assurance.
13. The Importance of Population Completeness
An effective sample cannot compensate for an incomplete population.
Suppose an auditor intends to test all inventory items at year-end.
If the inventory listing excludes an entire warehouse, selecting a statistically valid sample from that listing will not address the omitted warehouse.
The auditor must evaluate whether the population is complete and appropriate for the objective.
Relevant procedures may include:
Reconciling population totals to the general ledger.
Evaluating system-generated report parameters.
Identifying excluded accounts or locations.
Reviewing cutoff dates.
Evaluating data extraction procedures.
Investigating duplicate or unusual records.
Population reliability is a prerequisite for meaningful sampling conclusions.
14. Selecting Representative Sample Items
Classical statistical variables sampling generally requires a probability-based selection method consistent with the estimation technique.
Simple random sampling is one common approach.
Each eligible population item has a known chance of selection.
Other probability-based designs may be appropriate, including stratified random sampling.
The auditor should preserve the integrity of the selection process.
For example, replacing a selected transaction merely because it is difficult to examine may introduce bias.
When the planned procedure cannot be performed on an item, the auditor should evaluate the reason and follow the applicable sampling methodology.
The objective is to obtain a sample that supports valid conclusions about the defined population.
15. Stratification Can Improve Audit Efficiency
Stratification divides a population into groups with similar characteristics.
For example, inventory may be divided into:
Inventory category | Recorded value |
High-value items | $4,000,000 |
Medium-value items | $3,000,000 |
Low-value items | $1,000,000 |
Total | $8,000,000 |
The auditor may test selected high-value items individually and apply statistical sampling to the remaining population.
Alternatively, separate samples may be designed for appropriately defined strata.
Stratification can reduce variability within groups and improve the efficiency of estimation.
However, the auditor must correctly account for each stratum when combining results.
A conclusion about one stratum does not automatically apply to another.
16. Evaluating the Sample Results
After performing the planned procedures, the auditor must evaluate the results.
This generally includes:
Determining the audited amount for each selected item.
Calculating individual misstatements.
Investigating the nature and cause of identified differences.
Calculating the population point estimate.
Calculating the applicable allowance for sampling risk.
Comparing the statistical result with tolerable misstatement.
Considering qualitative factors and other audit evidence.
Determining whether additional procedures are required.
An important distinction is that a sample's projected misstatement and its sampling allowance are not the same thing.
The projection estimates the amount of misstatement.
The sampling allowance reflects uncertainty arising from examining only part of the population.
Both are important to the conclusion.
17. Practical Example: Inventory Pricing Audit
Assume a company reports an inventory balance of $8 million.
The auditor is concerned that unit costs may be overstated.
The auditor establishes a statistical sampling plan and uses ratio estimation.
The results are:
Item | Amount |
Recorded inventory population | $8,000,000 |
Recorded value of sampled items | $400,000 |
Audited value of sampled items | $392,000 |
Audited-to-recorded ratio | 98% |
Estimated audited population value | $7,840,000 |
Estimated population overstatement | $160,000 |
Tolerable misstatement | $250,000 |
The point estimate is $160,000.
That amount is below tolerable misstatement.
However, the auditor cannot yet conclude that the population is acceptable.
The auditor must determine the applicable statistical precision or upper misstatement bound.
For illustration only, suppose the completed statistical analysis produces an upper misstatement limit of $290,000.
Because $290,000 exceeds the $250,000 tolerable misstatement, the result would not support accepting the population under that sampling plan.
The auditor would need to consider additional procedures or other appropriate responses.
The $290,000 figure is hypothetical; it is not calculated from the summarized information in this example.
This illustrates a fundamental point:
A projected misstatement below tolerable misstatement does not automatically mean the audit sample supports acceptance of the population.
18. Common Variables Sampling Mistakes
Using the wrong sampling method
Selecting attribute sampling when the objective is to estimate monetary misstatement.
Selecting an arbitrary sample size
Using historical sample sizes without considering risk, variability, or tolerable misstatement.
Failing to evaluate the population
Sampling from incomplete or inaccurate accounting records.
Ignoring unusual items
Failing to evaluate significant, unusual, or individually high-risk transactions.
Using the wrong estimation method
Applying ratio estimation when the relationship between recorded and audited amounts is unstable.
Ignoring sampling precision
Concluding based solely on projected misstatement.
Treating net misstatements as the complete evaluation
Allowing offsetting overstatements and understatements to obscure significant errors or risk patterns.
Failing to investigate the causes of errors
Calculating estimates without understanding whether errors are isolated, systematic, or potentially fraudulent.
Overstating the conclusion
Assuming a statistically acceptable sample establishes that every assertion concerning the population is properly stated.
19. Variables Sampling in Internal Auditing
Variables sampling is not limited to external financial statement audits.
Internal auditors can use these techniques when engagements require monetary estimates.
Examples include:
Accounts payable: Estimating duplicate payments or invoice overpayments.
Payroll: Estimating incorrect compensation amounts.
Inventory: Estimating inventory valuation differences.
Procurement: Estimating pricing discrepancies against contract terms.
Healthcare: Estimating reimbursement overpayments.
Banking: Estimating monetary errors in selected loan or transaction populations.
Government auditing: Estimating questioned costs or other monetary exceptions.
Internal auditors must ensure that the selected methodology supports the engagement objective.
For example, estimating the amount of overpayments requires a different approach from determining the percentage of transactions that failed an approval control.
20. How AI and Data Analytics Can Support Variables Sampling
Artificial intelligence and data analytics can assist auditors with the mechanics of sampling.
Potential applications include:
Preparing and validating population data.
Identifying unusual monetary values.
Supporting stratification.
Generating random selections.
Comparing recorded and audited amounts.
Calculating point estimates.
Identifying patterns in misstatements.
Preparing preliminary sampling documentation.
However, AI-generated calculations should not be accepted without validation.
Auditors should independently verify statistical formulas, selection methods, population totals, and calculated results.
They should also distinguish between exploratory anomaly detection and probability sampling.
Selecting transactions because an AI model considers them unusual may be an effective targeted audit procedure.
But that selection does not automatically support a statistically valid projection to the entire population.
AI can improve sampling efficiency, but statistical validity depends on the sampling design—not on the sophistication of the software.
21. Variables Sampling Workpaper Checklist
A well-documented variables sampling workpaper should address:
Audit objective and relevant assertion.
Population definition and reconciliation.
Sampling unit.
Sampling methodology.
Tolerable and expected misstatement.
Acceptable sampling risk.
Population variability and stratification.
Sample size determination.
Selection method and selected items.
Procedures performed.
Recorded and audited values.
Individual misstatements.
Investigation of unusual errors.
Population point estimate.
Sampling precision or appropriate misstatement limit.
Comparison with tolerable misstatement.
Additional audit procedures, if necessary.
Final conclusion and supervisory review.
Documentation should allow an experienced auditor to understand the procedures performed, evidence obtained, statistical evaluation, and resulting conclusion.
22. Applicable Auditing Standards and Professional Guidance
For PCAOB engagements, AS 2315, Audit Sampling, establishes requirements relevant to planning, performing, and evaluating audit samples.
The standard addresses statistical and nonstatistical approaches, sampling risk, sample selection, and evaluation.
PCAOB AS 1105, Audit Evidence, also remains relevant because sampling procedures must produce sufficient appropriate evidence.
For audits conducted under AICPA standards, AU-C section 530, Audit Sampling, addresses the use of audit sampling.
The AICPA's Audit Sampling guide provides more detailed discussion of classical variables sampling, monetary unit sampling, and practical applications.
Internal auditors can use these methodologies where appropriate, while applying the IIA Global Internal Audit Standards governing engagement planning, evidence, analysis, and evaluation.
The Bottom Line: Variables Sampling Requires More Than Calculating an Average Error
Variables sampling is a valuable technique for auditors who need to estimate monetary values or misstatements without examining an entire population.
Classical variables sampling offers three principal approaches: mean-per-unit, difference, and ratio estimation.
Monetary unit sampling provides another important statistical approach, particularly for certain overstatement objectives.
Each technique has strengths and limitations.
Effective use requires a properly defined population, appropriate sample selection, reliable audit evidence, suitable statistical calculations, and evaluation of sampling risk.
Most importantly, auditors must understand the difference between a point estimate and a conclusion supported by an appropriate statistical bound.
Professional audit sampling is not simply about how many items were tested. It is about whether the sampling design and resulting evidence support the conclusion the auditor intends to reach.
That is the foundation of effective statistical audit tradecraft.
Develop Your Audit Sampling Skills With Corporate Compliance Seminars
Corporate Compliance Seminars (CCS) provides continuing professional education covering audit sampling, statistical and nonstatistical techniques, audit evidence, risk assessment, internal controls, and professional audit methodology.
Our programs help internal and external auditors design more effective sampling procedures, evaluate results, and document defensible audit conclusions.
Explore our auditing and assurance CPE programs at Corporate Compliance Seminars.
Comments