## Audit Sampling Methods

There are two broad categories: Random (probability-based) and Non-random.

---

### A. Random Sampling

All items have an equal chance of selection — no auditor bias.

#### i. Simple Random Sampling

Every unit in the population has an equal and independent chance of selection.
No bias from auditor judgement.
Suitable for homogeneous populations.
Method: use a random number table or computerised tools (e.g., Excel RAND function).

#### ii. Stratified Sampling

Divide a heterogeneous population into homogeneous sub-groups (strata).
Allocate different weights (sample sizes) to each stratum based on risk or value.
Example: high-value debtors form Stratum 1; low-value debtors form Stratum 2.
Reduces overall variability → allows a smaller total sample while maintaining coverage.

#### iii. Systematic (Interval) Sampling

Sampling Interval = Population Size ÷ Sample Size (e.g., every 50th item).
Select a random starting point within the first interval, then select every nth item thereafter.
To avoid predictability, auditor may use multiple starting points.
Risk: if the population has a periodic pattern matching the interval, bias can occur.

#### iv. Monetary Unit Sampling (MUS)

A type of value-weighted selection.
Each individual rupee (monetary unit) in the population is a sampling unit.
Higher-value items have a proportionally higher chance of selection.
Sample size, selection, and evaluation are all expressed in monetary amounts.
Particularly effective for detecting overstatements.

---

### B. Non-Random Sampling

#### v. Haphazard Sampling

Items chosen without any structured technique — no random number tables.
Aim: avoid bias and predictability through conscious effort.
Not suitable for statistical sampling because it cannot be mathematically evaluated.
Acceptable only for non-statistical sampling.

#### vi. Block Sampling

Selecting a contiguous sequence of items (a 'block') from the population (e.g., all invoices from April).
Not suitable for audit sampling because items within a block share similar characteristics — they are not representative of the whole population.
Items in a block may differ significantly from items in other parts of the population.

Worked example

### Example 1

Systematic Sampling — Worked Calculation:

Population = 1,000 purchase invoices. Required sample size = 50.

Sampling interval = 1,000 ÷ 50 = 20.

Auditor picks a random start within 1–20, say item #7.

Selected items: 7, 27, 47, 67, 87 … continuing every 20th item up to item #987.

### Example 2

MUS — Why high-value items get selected more often:

Imagine a population of 3 debtors: D1 = ₹1,00,000; D2 = ₹5,000; D3 = ₹500.

Total monetary units = 1,05,500.

D1 occupies 95% of the monetary units → has a 95% probability of being selected in MUS.

This naturally focuses audit effort on material balances.

### Example 3

Block Sampling — Why it fails:

Auditor selects all 50 invoices from January. In January, the company ran a year-end promotion with unusual discounting terms. The sample is 100% January invoices and captures none of the normal pricing patterns from the rest of the year — the sample is not representative of the population.

⚠️ Common exam mistakes

Confusing haphazard sampling (no structure, acceptable for non-statistical work) with random sampling (structured, uses random number tables — suitable for statistical sampling).
Using block sampling in audit and assuming it provides representative coverage — it does not because sequential items share common characteristics.
Forgetting that in MUS, the sampling unit is an individual monetary unit (₹1), not an account or transaction — this is why high-value items are automatically over-represented.
Not using multiple starting points in systematic sampling — a single start point on a cyclically patterned population introduces bias.

Reference:

Now that you've read this — what's next?

Move from understanding → mastery in 3 clicks. Each option below picks up from this lesson's topic.

🔥 Drill 5 Qs on this → ✍️ Adaptive practice →