In businesses that have hundreds of thousands of sales or purchases, it is simply not feasible for an auditor to examine each sale or purchase invoice. Instead, some kind of sample must be chosen.
Ideally, the sample or portion of invoices to be examined will be representative of the entire invoice population. But there is a risk that the sample will not be representative.
There are two different approaches to handling this.: non-statistical sampling and statistical sampling.
In statistical sampling the risk is quantified. In non-statistical sampling, the risk is not quantified, but is left to the professional judgment of the auditors performing the examination.
To compensate for the uncertainty in non-statistical samples, most states select substantially more transactions than achieving an accurate statistical sample would require.
In non-statistical sampling it is impossible to know whether the sample is too large, too small, or how accurately it reflects the records as a whole.
Expanding the sample size frequently results in a more homogenous sample. But it does so at the expense of the taxpayer’s and auditor’s time.
Block sampling is the most commonly used example of non-statistical sampling. In block sampling, the auditor chooses a block of time (weeks, months, days) of checks, invoices or vouchers to examine. The auditor assumes that the block of time is representative of the entire population. An error rate is determined from the sample chosen and then extrapolated over the entire audit period. There is no way of measuring the accuracy of such a method.
According to our research, the following 26 states and one district use statistical sampling: