LO 10.

5 Selecting the sample

Since the purpose of sampling is to draw conclusions about the entire population or stratum
from which the sample has been drawn, the sample needs to be typical of the
characteristics of that population or stratum. To achieve this, the sample needs to be
selected without bias, so that all sampling units in the population or stratum have a chance
of selection.

Appendix 4 of ASA 530 (ISA 530) outlines selection procedures. Sample items should be
selected in such a way that the sample is expected to be representative of the population.
As described in ASA 530.A13 (ISA 530.A13) there are three principal sampling Page 432
methods that achieve this: random selection, systematic selection and haphazard
selection.

Random selection
In random selection , each sampling unit making up the account balance or class of
transactions has a chance, often an equal chance, of selection. The concept requires that the
person selecting the sample will not influence or bias the selection either consciously or
unconsciously, so some form of impartial selection process is used to make the sample
truly random. Auditors commonly use computerised random number generators or random
number tables to generate numbers which can be linked to pre-numbered sampling units in
the population. If selecting from pre-numbered transactions the auditor can identify the
first and last transaction numbers for the period under audit, between which the random
numbers are generated. If sampling from components of account balances, such as
individual debtors or inventory items, the software can randomly select from the population
of valid debtor numbers or inventory item numbers.
Copyright © 2018. McGraw-Hill Australia. All rights reserved.

This approach is appropriate for both non-statistical and statistical sampling, since the
sample is selected on a basis that allows the auditor to measure the probability of selecting
the combination of sampling units actually chosen.

Systematic selection
Systematic selection involves selecting every nth item in the population, the sample
interval being determined by dividing the number of items in the population by the sample
size, and selecting a random starting point. For example, if the auditor wants a sample of

Gay, GE, & Simnett, R 2018, Auditing and Assurance Services in Australia, McGraw-Hill Australia, Sydney. Available from: ProQuest Ebook Central. [30 October 2020].
Created from usc on 2020-10-30 19:38:23.
 20 from a population of 20 000 items, the sample interval is every 1000th item. The auditor
randomly selects a starting point (falling between the first item in the population and the
sample interval) and selects every 1000th item from that starting point.

This method produces a random sample only if the population is randomly arranged. For
example, if large-value sales invoices are always assigned a number in a particular group, a
systematic sampling process may result in a biased sample of primarily large or small
invoices. Again, software can be used to select systematic samples from computer-based
files. This method is illustrated by Global example 10.2 .

GLOBAL EXAMPLE 10.2 Sampling by systematic selection

Assume the sample size has been determined as 20 and the number of
items in the population is 10 000.
Step 1: Calculate the sample interval:
 
No. of items in population 10 000
= = 500
Sample size 20

Step 2: Give every item in the population an equal chance of selection by

choosing a random number (random start) within the range of 1 and the
sampling interval (in this example, 500), for example, 217.
Step 3: Continue to add the sampling interval to the random start to
identify sample items, e.g. 217, 717, 1217, 1717 . . . 9217, 9717.
This gives an even spread of items throughout the population.
Copyright © 2018. McGraw-Hill Australia. All rights reserved.

Random selection and systematic selection are the most common sampling techniques. In
comparing these techniques, systematic selection may be advantageous in ensuring an even
distribution of selection throughout the population (for example, in testing the Page 433
continuity objective of internal control). However, if there are potential patterns in
the population (for example, control fails for every 100th transaction), care must be taken
in using systematic sampling: if the pattern corresponds to the sample interval an
unrepresentative sample will result.

Haphazard selection
Gay, GE, & Simnett, R 2018, Auditing and Assurance Services in Australia, McGraw-Hill Australia, Sydney. Available from: ProQuest Ebook Central. [30 October 2020].
Created from usc on 2020-10-30 19:38:23.
 Haphazard selection is permitted by the auditing standards, although it cannot
normally be used for statistical sampling applications because it does not allow the auditor
to measure the probability of selecting the combination of sampling units and it cannot
guarantee that the auditor will select without bias. In this method the auditor selects
sampling units without any conscious bias, and in a manner that can be expected to be
representative of the population. This technique may be useful when selecting a sample
from a population which is physically stored in an unsystematic manner. For example,
where all sales invoices are filed in a filing cabinet drawer, the auditor selects invoices
from the drawer irrespective of any distinctive feature of any of the individual documents.
Humans are, however, subject to subconscious biases which may affect the randomness of
this approach. These include an affinity with the number seven, and also a bias to select
towards the middle of a sequence rather than at the extremes. Both of these biases have
been demonstrated a number of times by research, with the affinity for the number seven
illustrated in Auditing in the global news 10.2 .
Copyright © 2018. McGraw-Hill Australia. All rights reserved.

Gay, GE, & Simnett, R 2018, Auditing and Assurance Services in Australia, McGraw-Hill Australia, Sydney. Available from: ProQuest Ebook Central. [30 October 2020].
Created from usc on 2020-10-30 19:38:23.
 10.2 Auditing in the global news ...

Biases from haphazard sampling

This note describes a technique that we have used for several years to
illustrate the problems inherent in judgment sampling techniques.

The question (asked) . . . select a number from 1 to 10 and write it in the

space below . . . The results reported below are from a group of
approximately 140 students.

A truly random selection would have led to each number being selected by
approximately one-tenth of the class. This question is, however, intended to
highlight the human affinity for the number seven (our weeks are seven days
long, we split our oceans into seven seas, we are tempted to indulge in
seven deadly sins, Snow White had seven dwarfs, we drink 7-Up™ and so on).
Whether the students were aware of this or not, their selection of a ‘random’
number was hopelessly biased by this innate affinity.
Copyright © 2018. McGraw-Hill Australia. All rights reserved.

Source: Stewart and Dunn, 2000.

Page 434

Unacceptable sample selection methods

In the past, auditors have used some selection methods that cannot be expected to produce
representative samples. While such methods of identifying items to test may be acceptable

Gay, GE, & Simnett, R 2018, Auditing and Assurance Services in Australia, McGraw-Hill Australia, Sydney. Available from: ProQuest Ebook Central. [30 October 2020].
Created from usc on 2020-10-30 19:38:23.
 evidence-gathering techniques, they are rarely suitable as an approach to selecting sample
items when the intention is to extrapolate the results from testing the sample to the
population. Examples are as follows:

Block selection With this method the auditor selects a block of items from a
population. For example, the auditor might examine all cash disbursement transactions in
the first week of June and the last week of December. The problem with this selection
method is that the sampling unit is a period of time rather than an individual transaction.
In the example given, the auditor has selected two units in a population of 52 units. If the
blocks or periods were selected randomly the sample might be representative, but a valid
sample size would normally be impractically large. For example, to have reasonable
assurance the auditor might need to select 30 out of 52 weeks.
Judgmental selection based on sample item characteristics With this method the
auditor selects large or unusual items from the population or uses some other judgmental
criterion for selection, with the result that there is bias towards high-value or key items.
This method is an example of ‘selecting specific items’, which was discussed earlier in
this chapter under Various means of gathering audit evidence , and is not a sampling
technique. Obviously, this method has a conscious bias and the items selected cannot be
considered representative of the population. Because these judgmental selection methods
have traditionally been used extensively in practice as a method of selecting sample
items, it is important to recognise that they are unacceptable today even for non-statistical
sampling.

The fact that judgmental selection with bias towards high-value or key items is not
appropriate for audit sampling does not mean that auditors should stop using judgment in
selecting items to examine. The point is that the items selected using judgmental criteria
are not necessarily representative of the population, and conclusions based on items
selected judgmentally should not be extended to the rest of the population.

In establishing the extent of an audit test, an auditor commonly first identifies a stratum of
large or unusual items and plans to examine all of them. If the remaining items are
sampled, this approach usually allows smaller sample sizes and is effectively equivalent to
stratification, which was discussed under Planning and designing the sample .
Copyright © 2018. McGraw-Hill Australia. All rights reserved.

However, while the large and unusual items are being tested, they are not being sampled,
they are being 100 per cent examined. Audit sampling is being used only for the remaining
items. ASA 500.A56 (ISA 500.A56) explains the point as follows:

While selective examination of specific items from a class of

transactions or account balance will often be an efficient means of
gathering audit evidence, it does not constitute audit sampling.
The results of procedures applied to items selected in this way
cannot be projected to the entire population; accordingly, selective
examination of specific items does not provide audit evidence
concerning the remainder of the population.

Gay, GE, & Simnett, R 2018, Auditing and Assurance Services in Australia, McGraw-Hill Australia, Sydney. Available from: ProQuest Ebook Central. [30 October 2020].
Created from usc on 2020-10-30 19:38:23.
 QUICK REVIEW
1. The auditor should select sample items so as to have a reasonable
expectation that all sampling units in the population have a chance of
selection.
2. Three selection methods commonly used are:
random selection
systematic selection
haphazard selection.
3. Unacceptable sample selection methods include:
block selection
judgmental selection, where the auditor uses judgmental criteria, such as
large or unusual items, for selection.
4. While judgmental selection will often be an efficient means of gathering
audit evidence, it does not constitute audit sampling.
Copyright © 2018. McGraw-Hill Australia. All rights reserved.

Gay, GE, & Simnett, R 2018, Auditing and Assurance Services in Australia, McGraw-Hill Australia, Sydney. Available from: ProQuest Ebook Central. [30 October 2020].
Created from usc on 2020-10-30 19:38:23.