Probability-proportional-to-size sampling

In some cases the sample designer has access to an "auxiliary variable" or "size measure",
believed to be correlated to the variable of interest, for each element in the population. These
data can be used to improve accuracy in sample design. One option is to use the auxiliary
variable as a basis for stratification, as discussed above.
Another option is probability proportional to size ('PPS') sampling, in which the selection
probability for each element is set to be proportional to its size measure, up to a maximum of 1.
In a simple PPS design, these selection probabilities can then be used as the basis for Poisson
sampling. However, this has the drawback of variable sample size, and different portions of the
population may still be over- or under-represented due to chance variation in selections.
Systematic sampling theory can be used to create a probability proportionate to size sample. This
is done by treating each count within the size variable as a single sampling unit. Samples are then
identified by selecting at even intervals among these counts within the size variable. This method
is sometimes called PPS-sequential or monetary unit sampling in the case of audits or forensic
sampling.
Example: Suppose we have six schools with populations of 150, 180, 200, 220, 260, and 490
students respectively (total 1500 students), and we want to use student population as the basis
for a PPS sample of size three. To do this, we could allocate the first school numbers 1 to 150,
the second school 151 to 330 (= 150 + 180), the third school 331 to 530, and so on to the last
school (1011 to 1500). We then generate a random start between 1 and 500 (equal to 1500/3)
and count through the school populations by multiples of 500. If our random start was 137, we
would select the schools which have been allocated numbers 137, 637, and 1137, i.e. the first,
fourth, and sixth schools.
The PPS approach can improve accuracy for a given sample size by concentrating sample on
large elements that have the greatest impact on population estimates. PPS sampling is commonly
used for surveys of businesses, where element size varies greatly and auxiliary information is
often availablefor instance, a survey attempting to measure the number of guest-nights spent in
hotels might use each hotel's number of rooms as an auxiliary variable. In some cases, an older
measurement of the variable of interest can be used as an auxiliary variable when attempting to
produce more current estimates.

AU Section 350 - Audit Sampling

Sampling and Nonsampling Risk

AU Section 350 of the AICPA Professional Standards indicates that audit risk arises from
sampling and nonsampling risk. Sampling risk results from performing an audit procedure on
less than one hundred percent of the population. It represents the risk that the audit sample is not
representative of the population. In other words, that the auditor's evaluation of a population
based on an audit sample is different from what it would be if the entire population was tested.
Nonsampling risk results from human error. It represents the risk that the selected audit
procedure is not appropriate for the intended purpose or the evidence from an audit procedure is
misinterpreted. Nonsampling risk exists regardless of the number of items selected from a
population for testing.

Sampling risk should be considered when an auditor performs an audit procedure on less than
one hundred percent of a clearly definable population for the purpose of evaluating the
population. There are two aspects to sampling risk when performing tests of controls:

The risk of assessing control risk too low represents the risk that an audit sample supports
the conclusion that the design and operation of an internal control is effective when in
fact it is not.
The risk of assessing control risk too high represents the risk that an audit sample
supports the conclusion that the design and operation of an internal control is not
effective when in fact it is effective.

Similarly there are two aspects to sampling risk when performing substantive tests:

The risk of incorrect acceptance represents the risk that an audit sample supports the
conclusion that a material misstatement does not exist when in fact a material
misstatement does exist. This risk is similar to the risk of assessing control risk too low.
The risk of incorrect rejection represents the risk that an audit sample supports the
conclusion that a material misstatement exists when in fact a material misstatement does
not exist. This risk is similar to the risk of assessing control risk too high.

The risk of assessing control risk too low and the risk of incorrect acceptance are concerned with
the effectiveness of audit tests while the risk of assessing control risk too high and the risk of
incorrect acceptance are concerned with the efficiency of audit tests.

Sampling risk can be considered using a non-statistical or statistical approach. Both approaches
require professional judgement. The main difference between the two approaches is that
statistical approaches allow auditors to quantify sampling risk.

Sample Selection Methods

Five methods are commonly used to select population items for audit testing:

Random Sampling involves selecting items from the population so that each item has an
equal chance of being selected. Random selection requires the use of random number
tables or computer programs to guarantee that each population item has an equal chance
of selection.
Systematic Sampling involves selecting every kth item from the population after a random
start.
Haphazard Sampling involves selecting items from the population without consideration
to known characteristics of the items in the population (i.e. any conscious bias in the
selection of population items).
Block Sampling involves selecting items from the population in contiguous groups (or
blocks).
 Purposive or Directed Sampling involves selecting items from the population using some
prespecified criteria (i.e., selecting accounts receivable for confirmation based on the size
of the outstanding balance).

The first two sample selection methods are referred to as probability (statistical) sample selection
methods since every population item has a known probability of selection. The last three
methods are referred to as non-probability (non-statistical) sample selection methods since every
population item does not have a known probability of selection. This distinction is important
because sample results from probability selection methods can be assessed using statistical
theory whereas sample results from non-probability samples cannot. The subjectivity of non-
probability selection precludes the development of a theoretical framework for evaluating sample
results. Non-probability samples can be assessed only by subjective evaluation, not by
assumption-free statistical methods. Nevertheless both probability and non-probability selection
methods are considered acceptable and used in practice.

Attribute Sampling

Attribute sampling is a statistical approach used with tests of controls. It requires the use of a
probabilistic sample selection method (random or systematic sampling). Attribute sampling
allows the auditor to estimate the proportion of population items containing a specified
characteristic. The characteristic auditors are concerned with for tests of controls is deviations
from internal controls.

Sample size for attribute sampling can be determined by reference to attribute sampling tables.
These sample determination tables require the auditor to establish three factors:

Risk of assessing control risk too low represents the risk that the auditor concludes that
the design and operation of an internal control is effective when in fact it is not. The level
used for this risk is based on the auditor's desired control risk assessment. The lower the
desired control risk assessment the lower the needed risk of assessing control risk too
low. This risk is inversely related to sample size.
Expected Population Deviation Rate represents the auditor's best estimate of the
population deviation rate. This rate is normally based on prior experience with the client.
This rate is directly related to sample size.
Tolerable Deviation Rate represents the highest deviation rate the auditor could accept
and still conclude that the design and operation of an internal control is effective. This
rate is based on the tolerable misstatement relative to the number and dollar size of
traansactions included in the population. Tolerable misstatement represents the maximum
misstatement that could occur before the population would be considered materially
misstated. The lower the required tolerable misstatement relative to the number and
dollar size of transactions the lower the needed tolerable deviation rate. This rate is
inversely related to sample size.

Sample results are evaluated by comparing the computed maximum population deviation rate to
the tolerable deviation rate. The computed maximum population deviation rate equals the sample
deviation rate plus an allowance for sampling risk. If the maximum population deviation rate is
larger than the tolerable deviation rate the auditor will conclude that the design and operation of
the internal control is not effective. If the computed maximum population deviation rate is less
than or equal to the tolerable deviation rate the auditor will conclude that the design and
operation of the internal control is effective.

Attribute sampling tables have been constructed to determine the computed maximum
population deviation rate. These attribute evaluation tables require the auditor to know three
factors:

Risk of assessing control risk too low (see above).

Sample Size represents the number of population items tested.
Sample Deviations represents the number of sample items that deviated from internal
controls.

Probability-proportionate-to-size Sampling (Monetary-unit Sampling)

Probability-proportionate-to-size Sampling (PPS) is used with substantive tests. It requires the

use of a probabilistic sample selection method (random or systematic sampling). PPS allows the
auditor to estimate the total misstatement of a population. It is different from other sampling
approaches used by auditors in that each dollar in the population is treated as a separate sampling
unit instead of each customer, invoice, check, vendor, etc.

Sample size for a PPS application can be determined by reference to a PPS sampling table. The
factors considered when determining sample size for substantive tests are:

Risk of incorrect acceptance represents the risk that the auditor concludes that a material
misstatement does not exist when in fact a material misstatement does exist. The level
used for this risk is based on the auditor's planned detection risk and other planned
substantive tests. A higher risk of incorrect acceptance is used with a higher planned
detection risk and/or other planned substantive tests. This risk is inversely related to
sample size.
Expected Number of Misstated Dollars represents the auditor's best estimate of the
number of sample dollars that will be misstated. This estimate is based on prior
experience with the client and is normally expected to be zero when PPS is used. This
estimate is directly related to sample size.
Tolerable Misstatement represents the highest misstatement that could occur before the
population would be considered materially misstated. This amount has an inverse
relationship with sample size.
Reported $ Balance represents the total recorded dollar balance in the population.

The risk of incorrect acceptance and expected number of misstated dollars are used to determine
the PPS Factor from the PPS table. The sample size is then determined using the following
formula:
Obviously, substantive tests are not performed on individual dollars but rather are performed on
logical units (invoice, customer, check, vender, etc,) containing individual dollars. Therefore,
evaluation of sample results requires restatement of the logical unit results in terms of the
sampling unit (individual population dollars). Sample results are evaluated by comparing the
adjusted upper and lower limits on misstatements (error bounds) to tolerable misstatement. If the
adjusted upper and lower limits on misstatements are smaller than or equal to tolerable
misstatement the auditor will conclude that a material misstatement does not exist. If either the
upper or lower limit on misstatement is greater than tolerable misstatement the auditor will
conclude that a material misstatement does exist.

The adjusted upper and lower limits on misstatements can be calculated using the PPS table. The
first step in calculating the adjusted upper and lower limits on misstatement is to divide misstated
dollars into over and understated dollars and then to rank them from the highest to lowest percent
misstatement. The percent misstatement for each misstated dollar is call tainting and is calculated
using the following formula:

The second step is to calculate an unadjusted upper and lower limit on misstatement using the
PPS table. These calculations require the auditor to know the risk of incorrect acceptance, sample
size, recorded dollars in the population (reported balance), sample dollars misstated, and tainting
of misstated dollars. The unadjusted upper limit on misstatement is based on the observed
overstated dollars and the lower limit on misstatement is based on the observed understated
dollars. The unadjusted upper and lower limits on misstatement are calculated using the
following formula:

The taintings are used in order of highest to lowest.

The third step is to calculate the estimated dollar overstatement and understatement in the
population. The estimated dollar overstatement is calculated based on the number of observed
overstated dollars. The estimated dollar understatement is calculated based on the number of
observed understated dollars. These amounts are calculated using the following formula:

The final step is to calculate the adjusted upper and lower limits on misstatement. The adjusted
upper limit on misstatement is calculated by subtracting from the unadjusted upper limit on
misstatement the estimated dollar understatement in the population. The adjusted lower limit on
misstatement is calculated by subtracting from the unadjusted lower limit on misstatement the
estimated dollar overstatement in the population.

Non-statistical Sampling

Non-statistical sampling can be used with tests of controls or substantive tests. Non-statistical
sampling does not require the use of a probabilistic selection method. The main advantage of
non-statistical sampling is that it is less complex and less time consuming than statistical
sampling. The main disadvantage is that sampling theory cannot be used to quantify sampling
risk.

Sample size for non-statistical sampling is left entirely to the auditors professional judgement.
The factors considered when determining sample size for tests of controls using a non-statistical
approach are the same as those considered for attributes sampling. The factors considered when
determining sample size for substantive tests using a non-statistical approach are:

Risk of incorrect acceptance represents the risk that the auditor concludes that a material
misstatement does not exist when in fact a material misstatement does exist. The level
used for this risk is based on the auditor's planned detection risk and other planned
substantive tests. A higher risk of incorrect acceptance is used with a higher planned
detection risk and/or other planned substantive tests. This risk is inversely related to
sample size.
Risk of incorrect rejection represents the risk that the auditor concludes that a material
misstatement exists when in fact a material misstatement does not exist. The level used
for this risk is based on the cost and difficulty of obtaining additional evidence. A lower
risk of incorrect acceptance is used when more costly or difficult evidence will be
required if expanded testing is needed. This risk is inversely related to sample size.
Expected Misstatement represents the auditor's best estimate of the population
misstatement. This estimate is normally based on prior experience with the client. This
estimate is directly related to sample size.
Tolerable Misstatement represents the highest misstatement that could occur before the
population would be considered materially misstated. This amount has an inverse
relationship with sample size.
Sample results for tests of controls are evaluated by comparing the sample deviation rate to the
tolerable deviation rate and calculating an allowance for sampling risk. The sample deviation rate
is calculated by dividing the number of sample deviations by the sample size. The allowance for
sampling risk is calculated by subtracting the sample deviation rate from the tolerable deviation
rate. If the allowance for sampling risk is large and positive the auditor would most likely
conclude that the design and operation of an internal control is effective. However, if the
allowance for sampling risk is small or negative the auditor would conclude that the design and
operation of an internal control is not effective. What constitutes a large enough difference is a
matter for professional judgement. Generally, smaller allowances for sampling risk are tolerated
with higher risks of assessing control risk too low and larger sample sizes.

Sample results for substantive tests are evaluated in a similar manner. A projected population
misstatement is calculated based on the sample results and compared to the tolerable
misstatement. The projected population misstatement is computed by dividing the sample
misstatement by the dollar value of the sample and multiplying this amount by the dollar value of
the population. The difference between the projected population misstatement and tolerable
misstatement is called the allowance for sampling risk. If the allowance for sampling risk is large
and positive the auditor would most likely conclude that a material misstatement does not exist.
However, if the allowance for sampling risk is small or negative the auditor would conclude that
a material misstatement does exist. What constitutes a large enough difference is a matter for
professional judgement. Generally, smaller allowances for sampling risk are tolerated with
higher risks of incorrect acceptance and larger sample sizes.

Analyze Exceptions/Misstatements

Regardless of whether a statistical or non-statistical approach is used to evaluate sample results,

the auditor needs to consider the nature and cause of every observed exception or misstatement.
Exceptions or misstatements that are intentional will normally affect the audit tests performed
even when the quantitative analysis supports effective controls or no material misstatement.
Additionally, the auditor may need to revise the original inherent risk and control risk
assessments based on subsequently performed substantive tests. The original assessments may
need to be revised upward if misstatements are observed.