Attribute sampling

Attribute sampling involves selecting a small number of transactions and making assumptions about how their characteristics represent the full population of which the selected items are a part. The concept is frequently used by auditors to test a population for certain characteristics, such as the presence of an authorizing signature or approval stamp on a document. The concept can be used to determine whether various accounting controls are functioning in a reliable manner.

The result of attribute sampling is binary - either a condition exists or it does not exist. Thus, there is no gray area in attribute sampling. Examples of typical attribute sampling tests are:

  • 50 out of 60 invoices were supported by a sales order
  • 38 out of 40 supplier invoices that were greater than $1,000 contained an approval signature
  • 19 of 20 fixed asset purchases had a supporting authorization document signed by the company president
  • 3 out of 80 invoices are overdue for payment
  • The early payment discount was not taken on 2 out of 11 supplier invoices
  • 13 out of 211 journal entries were posted to the wrong account

The results of an attribute sampling test are then compared to the tolerable error rate established for that test. If the test results are worse than the tolerable error rate, the control point related to the test has failed, and should be revised or replaced. For example, if the acceptable failure rate for supplier invoice approvals is 3% and the tested rate is 5%, it may be necessary to impose additional controls, retrain the staff, and/or alter the purchase approval procedure to reduce the failure rate indicated by attribute sampling.

When the tested sampling rate falls just outside the acceptable error rate, it is possible that conducting more tests with a larger sample size will result in an actual error rate that falls within the acceptable error rate. Thus, the first reaction by many people to a marginal attribute sampling result is to keep on testing with a larger sample group. This expansion of the sample size frequently does not yield a better result, as the original smaller sample size already provided the correct insight into the underlying error rate.

Attribute sampling is heavily used for the testing of internal controls. The results of these tests can then be used by a company's external auditors, who can choose to rely (or not) on the tested abilities of the accounting controls when developing their own procedures for how the company's financial statements will be audited.