9.1 Classical Variables Sampling Methods (MPU, Ratio, Difference)

9.1 Classical Variables Sampling Methods (MPU, Ratio, Difference)

Core Principle: Classical variables sampling is a statistical sampling approach used in substantive tests of details to estimate the true monetary value of an account balance or class of transactions, or to estimate the monetary amount of misstatement therein. Unlike attribute sampling (which measures deviation rates in tests of controls), classical variables sampling evaluates monetary amounts and tests financial statement assertions such as existence, accuracy, and valuation or allocation under AU-C 530 (Audit Sampling).

---

1. Statistical Foundations: Normal Distribution Theory & The Central Limit Theorem

Classical variables sampling relies upon classical statistical theory, specifically the properties of the normal distribution and the Central Limit Theorem.

                                  THE NORMAL DISTRIBUTION OF SAMPLE MEANS
                                                     |
                                                   *****
                                                 *   |   *
                                               *     |     *
                                             *       |       *
                                           *         |         *
                                         *           |           *
                                      *              |              *
                                  *                  |                  *
                             *                       |                       *
      -------------------+---------------------------+---------------------------+-------------------
                        -2σ                         Mean                        +2σ
                         |                           (μ)                         |
                         |<---------------- 95.4% of Sample Means ------------->|

The Central Limit Theorem

The Central Limit Theorem (CLT) states that regardless of the underlying distribution of the population (even if the individual account balances are heavily skewed, bimodal, or non-normal), the distribution of sample means drawn from that population approaches a normal bell-shaped curve as the sample size increases (generally n ≥ 30 to $50$).

This principle provides two critical mathematical advantages to the auditor:

1. Predictable Sampling Distribution: The mean of the sample means equals the true population mean (μ).
2. Standard Error of the Mean: The dispersion of sample means around the true population mean is measured by the standard error (SE = σ / √n), where σ is the population standard deviation and n is the sample size.

Confidence Intervals & Precision in Audit Testing

By applying normal distribution theory, the auditor can construct a confidence interval around the estimated population value:

$\text{Confidence Interval} = \text{Point Estimate} \pm (Z \times SE)$

Where:

• Point Estimate: The projected total balance calculated from the sample.
• Z (Reliability Coefficient): The standard normal factor for the desired confidence. For a two-sided interval, Z = 1.645 gives 90% confidence and Z = 1.96 gives 95%. Audit sample-size work treats the risk of incorrect acceptance as one-sided, so a 5% risk uses about 1.645 and a 10% risk about 1.28.
• Precision (Allowance for Sampling Risk): The monetary interval (Z × SE) within which the true population value is expected to fall.

---

2. Population Variability & The Power of Stratified Sampling

In classical variables sampling, the required sample size is directly proportional to the square of the population standard deviation (σ²):

$n = \left( \frac{N \times Z \times \sigma}{\text{Tolerable Misstatement} - \text{Expected Misstatement}} \right)^2$

This simplified form folds the confidence factors into one Z; full AICPA formulas use separate factors for the risks of incorrect acceptance and incorrect rejection and may apply a finite population correction. Because population variability (σ) is in the numerator, an unstratified population with high variability (items ranging from $10 to $500,000) produces an enormous standard deviation, necessitating an impractically large sample size.

+-------------------------------------------------------------------------------------------------------+
|                                 THE MECHANISM OF STRATIFIED SAMPLING                                  |
|                                                                                                       |
|   UNSTRATIFIED POPULATION (N = 5,000)              STRATIFIED DESIGN (Subgroups by Dollar Magnitude)  |
|   Wide dispersion: $5 to $500,000                  - Stratum 1: > $100,000 (100% audit / 20 items)    |
|   Standard Deviation (σ) = $12,400                 - Stratum 2: $10,000 - $100,000 (σ = $3,200)       |
|   Required Sample Size: n = 420 items              - Stratum 3: < $10,000 (σ = $850)                  |
|                                                    Combined Required Sample Size: n = 110 items       |
|                                                    --> RESULT: 74% Sample Size Reduction!             |
+-------------------------------------------------------------------------------------------------------+

Why Stratification is Essential

1. Homogeneity Within Strata: Grouping items into relatively homogeneous dollar categories drastically reduces the standard deviation within each stratum.
2. 100% Audit Stratum: The largest, individually significant items (any balance exceeding performance materiality or tolerable misstatement) are isolated into a separate stratum and audited 100%. These items contain zero sampling risk because no sampling is performed on them.
3. Radical Sample Size Reduction: Because the total required sample size is the sum of the sample sizes of the individual strata, reducing the intra-stratum standard deviation dramatically reduces the overall sample size required to achieve the desired audit precision.

Exam Trap: Stratification is used exclusively when the auditor desires to reduce sample size by controlling population variability, or when the population contains highly skewed items. Stratification does not change the inherent risk of the account, nor does it convert a substantive test into a test of controls.

---

3. Method 1: Mean-Per-Unit (MPU) Estimation

Definition & Core Concept

Mean-Per-Unit (MPU) estimation establishes a sample mean from the audited values of sample items and projects this average across the total number of physical units in the population.

Mathematical Formulas

$\text{Sample Mean Audited Value (}\bar{x}\text{)} = \frac{\sum \text{Audited Values of Sample Items}}{n}$

$\text{Estimated Audited Population Total (Point Estimate)} = \bar{x} \times N$

$\text{Projected Misstatement} = \text{Recorded Book Value of Population} - \text{Estimated Audited Population Total}$

Where:

• n = Number of items in the sample
• N = Total number of physical units in the population

Step-by-Step Numerical Walkthrough

An auditor tests an accounts receivable subsidiary ledger consisting of 4,000 customer accounts with a total recorded book value of $2,400,000.

• The auditor randomly selects a sample of 200 accounts (n = 200).
• The sum of the audited values of the 200 sample accounts is $114,000.

Step 1: Compute the mean audited value per unit:
        x̄ = $114,000 / 200 = $570 per account

Step 2: Project the mean across the total population units:
        Point Estimate = $570 × 4,000 = $2,280,000

Step 3: Calculate the projected misstatement:
        Projected Misstatement = $2,400,000 (Book Value) - $2,280,000 (Audited Value)
                               = $120,000 (Overstatement)

Preconditions & When to Use MPU

• No Individual Book Values Required: MPU is the only classical variables method that does not require the recorded book value of individual sample items. It requires only the audited value of each sample item and the total physical count (N) of items in the population.
• Best Use Case: When individual recorded book values are unavailable, incomplete, or fundamentally unreliable (e.g., perpetual inventory records have failed, cost ledger is corrupted, or client only tracked physical counts).
• Limitation: MPU typically produces the largest standard deviation and therefore requires the largest sample size among classical variables methods unless the population is heavily stratified.

---

4. Method 2: Difference Estimation

Definition & Core Concept

Difference estimation calculates the difference between the audited value and the recorded book value for each item in the sample, computes the average difference per sample item, and projects this average difference across the total physical units in the population.

Mathematical Formulas

$\text{Sample Difference per Item (}d_i\text{)} = \text{Audited Value}_i - \text{Book Value}_i$

$\text{Average Difference per Unit (}\bar{d}\text{)} = \frac{\sum d_i}{n}$

$\text{Projected Total Difference} = \bar{d} \times N$

$\text{Estimated Audited Population Total} = \text{Total Recorded Book Value} + \text{Projected Total Difference}$

Step-by-Step Numerical Walkthrough

Using the same accounts receivable population (N = 4,000; recorded book value $2,400,000; n = 200):

• The total recorded book value of the 200 sample items is $122,000.
• The total audited value of the 200 sample items is $114,000.

Step 1: Compute the net sample difference:
        Total Sample Difference = $114,000 (Audited) - $122,000 (Book) = -$8,000
        (Sample is overstated by $8,000)

Step 2: Calculate average difference per unit:
        d̄ = -$8,000 / 200 = -$40 per account

Step 3: Project the difference across the entire population:
        Projected Total Difference = -$40 × 4,000 = -$160,000
        (Population is projected to be overstated by $160,000)

Step 4: Determine estimated audited population total:
        Point Estimate = $2,400,000 + (-$160,000) = $2,240,000

Preconditions & When to Use Difference Estimation

• Preconditions:
  1. Every sampled item must have both a recorded book value and an audited value.
  2. Differences between book and audited values must not be too rare; if zero differences exist in the sample, the standard deviation of differences becomes zero, invalidating the statistical calculation.
  3. Misstatements must be independent of dollar amounts (i.e., errors are constant and do not grow proportionally with account size).
• Advantage: Requires a significantly smaller sample size than MPU because the standard deviation of differences is generally much smaller than the standard deviation of individual account balances.

---

5. Method 3: Ratio Estimation

Definition & Core Concept

Ratio estimation computes the ratio of the total audited value of the sample to the total recorded book value of the sample, and multiplies this ratio by the total recorded book value of the population to arrive at the estimated audited population total.

Mathematical Formulas

$\text{Sample Ratio (}R\text{)} = \frac{\sum \text{Audited Values of Sample Items}}{\sum \text{Book Values of Sample Items}}$

$\text{Estimated Audited Population Total (Point Estimate)} = R \times \text{Total Recorded Book Value}$

$\text{Projected Misstatement} = \text{Total Recorded Book Value} - \text{Estimated Audited Population Total}$

Step-by-Step Numerical Walkthrough

Using the same population (N = 4,000; recorded book value $2,400,000; n = 200) but a different sample of 200 items:

• Total recorded book value of sample = $120,000.
• Total audited value of sample = $114,000.

Step 1: Calculate the sample ratio:
        R = $114,000 / $120,000 = 0.95 (Audited value is 95% of recorded book value)

Step 2: Project ratio to total population book value:
        Point Estimate = 0.95 × $2,400,000 = $2,280,000

Step 3: Calculate projected misstatement:
        Projected Misstatement = $2,400,000 - $2,280,000 = $120,000 (Overstatement)

Preconditions & When to Use Ratio Estimation

• Preconditions:
  1. Every sampled item must have both a recorded book value and an audited value.
  2. Differences between book and audited values must exist in reasonable frequency.
  3. Misstatements must be approximately proportional to recorded dollar amounts (e.g., a 5% pricing error or uniform percentage discount applied across varied account balances).
• Advantage: Highly efficient (yields smallest sample size) when errors are proportional to dollar magnitude.

---

6. Comprehensive Comparative Matrix: Classical Variables Methods

---

7. Classical Variables Sampling Risks: Alpha vs. Beta Risk

In substantive testing of account balances, the auditor faces two distinct sampling risks under AU-C 530:

+-------------------------------------------------------------------------------------------------------+
|                                 SUBSTANTIVE SAMPLING RISK MATRIX                                      |
|                                                                                                       |
|                                              ACTUAL POPULATION CONDITION                              |
|   AUDITOR'S CONCLUSION              Not Materially Misstated          Materially Misstated            |
|   --------------------              ------------------------          --------------------            |
|   Accept as Fairly Stated               CORRECT DECISION              BETA RISK (Type II Error)       |
|                                                                    - Risk of Incorrect Acceptance     |
|                                                                    - IMPAIRS AUDIT EFFECTIVENESS      |
|                                                                    - Ultimate Audit Risk / Lawsuits   |
|                                                                                                       |
|   Reject as Materially Misstated    ALPHA RISK (Type I Error)             CORRECT DECISION            |
|                                  - Risk of Incorrect Rejection                                        |
|                                  - IMPAIRS AUDIT EFFICIENCY                                           |
|                                  - Unnecessary Extra Audit Work                                       |
+-------------------------------------------------------------------------------------------------------+

The Two Critical Substantive Risks

1. Risk of Incorrect Acceptance (Beta Risk / Type II):

   • The Risk: The sample results support the conclusion that the recorded balance is not materially misstated when, in fact, it is materially misstated.
   • Consequence: Impairs audit effectiveness. The auditor issues an inappropriate unmodified opinion on materially misstated financial statements, creating legal liability and audit failure.
   • Auditor Response: Controlled by the auditor through setting the desired confidence level (reliability factor) and precision.

2. Risk of Incorrect Rejection (Alpha Risk / Type I):

   • The Risk: The sample results support the conclusion that the recorded balance is materially misstated when, in fact, it is not materially misstated.
   • Consequence: Impairs audit efficiency. The client will contest the adjustment, prompting the auditor to expand sample sizes, perform alternative procedures, or re-examine client reconciliations. The true facts ultimately emerge, so audit effectiveness is not compromised, but substantial time and cost are wasted.