Sampling- an Audit Tool

Presented by
Mr. Preman Dinaraj,
Prin. Director, RTI, Mumbai.
 Need
• Audit resources- man power or time are very limited,
scarce and costly.
• Huge volume of entity records makes it not possible
or even advisable to examine each and every
transaction in audit.
• It optimise the available resources without
compromising on the quality of its results.
• Hence audit working procedures must prescribe and
rely on an effective system of sampling.
• Sampling tests less than 100% items in a given
population for specified characteristics & imposes
test results on the entire population which can be a
class of transactions or records.
Advantages
Saves valuable time, money and effort.
The results are obtained on the basis of objective
criteria and hence defensible.
Since sampling size is predetermined, the time
and money spent on individual audits can be
justified.
Sampling provides more accurate results than
examination of complete population.
The results can be objectively evaluated and
projected to known limits of reliability.
Types of Sampling Techniques
• Can be statistical or non-statistical but the governing
principles are the same.
• Require professional judgement in planning, testing and
evaluation stages.

Non Statistical

Techniques

Statistical
Non-Statistical Sampling
• User does not apply statistical measurements.
• Forgoes the advantage of quantifying and measuring the sampling
results and risks.
• Technique relies heavily upon the user’s experience and judgment
for conclusions.
• One disadvantage- the element of subjectivity and less precision in
the results.
Statistical Sampling
• Relies on the Theory of Probability & provide a
measurable degree of sampling risk.
• Has a measurable relationship between the size of
the sample and the degree of risk.
• Decide the sample size selection method.
• Based on testing & evaluation, the results are
presented mathematically.
• Statistical sampling methods control the auditor’s
exposure to sampling risk
• Drawback- to convert a statistical conclusion into a
value say 5%.
Advantages of Statistical Sampling
• Ensures that the results are measurable in terms of the Theory of
Probability.
• The conclusions arrived at will apply to the population as a whole.
• It provides greater objectivity in sample selection and audit
conclusions.
Categories of Statistical Sampling

Attribute

Sampling
Categories

Variable
Attribute Sampling
• Requires practitioner to identify the relevant
characteristics or attributes in a population.
• Find out what proportion of the population ‘has’
or ‘does not have’ the selected attribute(s).
• Then quantify the results & apply it to the
population as a whole, for e.g. 5 times out of 100.
• Remember that by merely knowing the proportion
in the population is not enough but also required
to know the value of these ‘deviations’/ ‘errors’.
• Widely used in financial audit.
Variable Sampling

• Examines ‘how much’ deviations or ‘errors’ have occurred within

the selected sample.
• It provides an estimate based on the measurement of the average
or total value of a population.
• Due to complications in converting the averages into values, variable
sampling is not generally resorted to in audit.
Sampling Risk
• Risk that the decision made based on the sample
differs from the decision that would have been
made by examining the population
• Cause is a nonrepresentative sample
• Controlled by:
• Determining an appropriate sample size
• Ensuring that all items have an equal opportunity of selection
• Mathematically evaluating sample results
Summary of Sampling Risks

• Effectiveness Losses (Major Risk)

• Risk of assessing control risk too low (overreliance) - Attribute
• Risk of incorrect acceptance - Variables
• Efficiency Losses (Minor Risk)
• Risk of assessing control risk too high (underreliance) - Attribute
• Risk of incorrect rejection - Variables
 Systematic Error vs. Random Error
Systematic Error :
• It has a definite cause and is uni-directional (i.e. either (+) or (-)
which typically reduce the accuracy without necessarily
lowering the result precision.
• To reduce it, identify the source of the error and remove it.
Random Error :
• Arises from inherent limits of any measuring devices and is bi-
directional (being randomly positive or negative) and cannot
be corrected.
• May occur for each measurement in a data set i.e. every time a
data point is taken it could be off target.
• Occur due to a wide variety of largely unpredictable reasons.
Monetary-Unit or $ Sampling
• Used to assess the amount of monetary misstatement that may
exist in an account balance.
• Also known as dollar-unit or probability-proportional-to-size
sampling has been used for many years and is widely accepted
among auditors.
• MUS requires calculations that introduce the possibility of error
if made manually.
• Can be done on a Microsoft Excel spreadsheet that automates
and enhances the use of MUS and reduces the chance of
errors.
• Using MUS involves three key steps:
1) Determining the proper sample size.
2) Selecting the sample and performing the audit procedures.
3) Evaluating the results and arriving at a conclusion about the
recorded population value.
Acceptance- Rejection Sampling

• Using random sampling methods draw a sample of

a given size.
• Examine it with reference to pre determined error/
deviation criteria & if they are less, accepted.
• Otherwise the sample is rejected.
• This type of sample allows for only an accept or
reject decision. The various types of acceptance
sampling plans are discussed in detail in the Arkin
text.
Discovery Sampling
• Also referred to as exploratory sampling.
• Used where evidence of a single error or instance of irregularity
would call for intensive investigation.
• Discovery sampling is frequently of value when fraud, avoidance of
internal controls, evasion of regulation or other critical performance
and quality control measures are in question.
Major Steps in Sampling

1. Determine the objective

2. Define characteristic of interest
Planning 3. Define the population
4. Determine sample size
5. Select the sample
6. Measure sample items
Performing 7. Evaluate sample results

Evaluating
Selecting the Sampling Method

• Determine the ‘purpose’ of the sample and the

type of test it will be subjected to.
• Understand the Unit’s accounting system, data
capture & retrieval, volume of transaction/ data
reliability.
• The type of audit, its extent and intensity.
• Ensure that sampling plans & selection techniques
to be used are precisely documented.
• Since each sampling method has its practical use
and limitations, the auditor must judge & decide
the method best suited for his audit.
 Judgmental

Random

Cluster
Sampling
Methods Convenience

Stratified

Systematic
Test Checking
• Done when the number of transactions are large.
• Involves checking of a few selected transactions &
based on it forming an opinion.
• Must be applied only if satisfied that the entity’s
internal control system is robust & working.
Precautions include:
• Checking of entries of all description.
• Conduct random check of selected entries.
• Ensure that periods & entries selected for test
check are different for different audits.
Judgemental Sampling
• Most commonly used method in audit.
• Auditor relies largely on the previous experience
and knowledge of the Unit.
• Sampling done on the basis of ‘test check’ of
selected accounts and is a non-statistical approach.
• Involves checking a pre-determined proportion of
the population based on auditor’s judgement.
• Useful for testing small samples and can be
effective, if it is properly designed.
• Drawback - inability to measure sampling risk &
prone to allegations of auditor’s biases. Since the
sample size is generally small, the conclusions may
not be credible and difficult to extrapolate on the
entire population.
Simple Random Sampling
• Requires the entire population to be known before
the sample for testing is selected.
• Each and every member of a given population has
an equal chance of being selected and hence
completely free of any bias.
• The sample can be easily designed and interpreted.
• Leads to reliable estimates in the population and
reduces the chances of sampling errors.
• To ensure reliability, the population listed must be
complete and accurate.
Cluster Sampling
• Selection of subjects are done in groups or clusters.
• Overcomes costs and time constraints associated
with highly dispersed populations & simplifies
selections from the entire multistage population,
e.g. boys and girls studying in schools in a region.
• Population assigned to groups or clusters.
• Define the clusters, take a random sample &
examine all its components.
• Disadvantage- possibility of large sampling errors.
Stratified Sampling
• Identify the population based on shared
homogeneity or similar characteristics into
different stratums e.g. Students and non-students.
• Since this method requires samples from each
identified main group to be included, it is more
reliable than random sampling due to the
reduction in sampling errors.
• Main difficulty- sample selection is more complex
requiring reliable population information and
estimates involve complex calculations.
Systematic Sampling
• Also called the ‘n’th name selection technique.
• Begin by selecting a starting point randomly,
between 1 and ‘n’ and thereafter select every ‘n’th
unit, where ‘n’ is derived by dividing the
population size by the sample size.
• Facilitates easy extraction of sample and ensures
cases are spread across the population.
• Disadvantage- can be a costly process both in
terms of time and costs & cannot be used where
there is periodicity in population.
Convenience Sampling
• Relies on data or volunteers readily available.
• Useful for preliminary audit examination to get a general idea of the
issue being tested.
• The larger the group examined, the more the information gathered.
• Main disadvantage - likelihood of bias & the sample results cannot be
extrapolated to the entire population.
• Hence recommended only during preliminary audit examination.
Designing a Sample

Define Sample
Population
Determine
Sample Size
Decide Sample
Sample Selection Method
Designing
Perform Audit
Tests
Project Results
to
Entire Population
Designing a Sample- (1) Defining the Sample
Population
• Ensure all elements within the given population are
represented.
• Identify the specification of the target population
clearly and completely.
• Above step can be on the basis of the Unit’s data
already available with the auditor. E.g. Monthly
expenditure statements, pay rolls etc.
• Identify the sample population & the target
population depending upon the sampling
technique.
Designing a Sample- (2) Determining the
Sample Size
• Requires consideration of the 5 key factors mentioned
below. All these factors are considered together to
achieve the right balance & objectivity.
(1) Margin of Error (Tolerable error):
• Recognises that sample estimates will not be exact and its
inference to the population will have a margin of error.
• Possible to reduce this margin with better designing and
larger sample sizes.
• The tolerable error is the maximum acceptable error for
certifying that the Unit is following the procedures
properly.
• Requires the auditor to prescribe an ‘upper error limit’
(UEL) which is equal to the rate of control failure that he is
willing to accept.
(1) Margin of Error (Tolerable error) Contd.:
• At the evaluation stage, the auditor compares
the actual results with the ‘UEL’ to arrive at the
estimate of ‘true error’ for the population.
• The acceptability of the sample results depends
upon the extent of variation between the UEL
and the ‘true error’ and the level of precision
desired in arriving at defensible conclusions i.e.
the higher the precision requirement, the lower
must this variation be and vice versa.
Components of UEL:
• Projected misstatement
• Incremental allowance for sampling risk
• Basic allowance for sampling risk
Designing a Sample-
(2) Amount of variability: This factor explains the range of
values or opinions in the population. This in turn affects the
accuracy of the results and consequently the size of sample
required when estimating a value. Hence, as the amount of
variability increases, the accuracy of the estimate decreases
requiring larger sample size to be used.
(3) Confidence level: Confidence level states how certain the
auditor is, that the actual population measure is within the
sample estimate and its associated precision level. In other
words, confidence level indicates the likelihood that the
sample results are within the defined precision levels.
Further, confidence level is directly proportional to the
sample size i.e. higher the required confidence level, the
larger the size of the sample required.
Designing a Sample-
(4) Population size: This provides an indication of the total number of
items in the population. Normally, the sample size is not influenced by
the population size, but by practical limitations such as time, easy
availability of data, competent software etc. Generally, a sample
between 50 and 100 will ensure that the results are sufficiently reliable
though the upper limit may be restricted to 5 percent, especially while
dealing with large volume data. Though case study samples by its very
nature tend to be smaller in size, they still form a defensible and
reliable audit methodology.
(5) Proportion of population: This is useful for attribute sampling. Here,
the proportion of items in the population displaying the pre-
determined attributes is determined rather than the calculation of their
average values. This can be estimated from the information already
known about a population e.g. the proportion of infrastructural
projects having time and (or) cost overruns.
Designing a Sample- (3) Method for Sample
Selection
• Selection of sample will depend upon the sampling method proposed to
be applied. If the intention is to do non-statistical sampling, due care
must be taken by the auditor to ensure that the representative sample
selected is unbiased by the auditors perception of the organization or
activity being audited. Generally, sample selection is closely dependent
upon the circumstances.
• The first step here is to determine the extent of precision required in the
sample results. If tight precision is required, then simple random,
systematic or cluster sampling can be adopted. If tight precision is not
required, apart from simple random sampling, judgment and
convenience sampling can be resorted to. Audit software like IDEA is an
ideal tool for selection of the sample.
Designing a Sample- (4) Perform Substantive
Audit Tests
• Once the sampling method has been decided,
identified audit tests will have to be done on the
sample to determine the selected sample
characteristics. These audit tests have to be
designed keeping the sampling objective and
population characteristics in mind. He must also
consider the specific objectives of these tests and
define what will constitute an error, before the
examination commences.
Designing a Sample- (5) Projection of Results

• Interpretation and projection of the test results on the population is the last
critical step in sampling. To arrive at reliable results:
•  Define the conditions that constitute an error (with reference to the audit
objectives) at the sample design stage itself and what the UEL is.
•  Evaluate the test results for each characteristic independently. Also
consider the qualitative aspects of the error like its nature, cause and its likely
effect on other phases of audit.
•  Superimpose the error results on the population as a whole to determine
the total possible error in the population and to determine whether the test
results are acceptable or otherwise.
• 
Projection of results
• Then assess the significance of such errors for the entire population. If it
exceeds the UEL, do not derive assurance from it. Instead, subject a larger
population of the relevant control system to substantive testing and based on
the results obtained, arrive at audit findings.
• In conclusion, it is important to mention that the reporting procedure must
identify the key factors considered during the evaluation process. These include
the sample size, the sample selection methodology, the substantive tests
performed and their results along with such parameters like the precision
levels, reliability factor and confidence levels.