Page 184 - COSO Guidance Book

Regression analysis
Regression analysis is the use of statistical models to quantify expectations; it incorporates risk and
precision. It is similar to reasonableness testing in that there is an explicit prediction using the manager’s
or auditor’s knowledge of factors that affect the account balance. For example, a model to predict the
gross sales amount for new tires might include fuel prices (higher prices lead to less leisure driving),
airfare wars (consumers might fly instead of drive), and interest rates (higher interest rates might
decrease demand for new cars and encourage consumers to maintain existing vehicles).

Regression analysis is the most precise of the four techniques discussed because it is the only method
that provides a measure of statistical precision. An explicit relationship is formed using two or more
predictors and both internal and external data can be included. Regression analysis is recommended for
forensic investigations because it provides the highest degree of precision, which would be required if the
forensic case went to court. In practice, regression analysis is typically used with the aid of specialized
software programs that calculate complex statistical formulas.

Identification, investigation, and evaluation

An entity manager or auditor compares his or her expectation with the recorded amount. If the difference
between these amounts is greater than the manager’s or auditor’s predetermined acceptable amount,
then an investigation should be launched to find the reasons for the error. One reason could be that the
difference results from the use of unreliable data to formulate the expectation. Another reason could be
related to factors that affect the account under analysis, such as the account being the result of
subjective estimates (for example, allowance for doubtful accounts estimates). Yet another reason for
the difference could be that there is an actual misstatement. It is important to note that the more precise
the expectation, the more likely that the difference is due to a misstatement.

If the manager or auditor believes that the difference more likely derives from factors related to the
precision of the expectation, he or she should determine whether a more precise expectation could be
cost-effectively developed. For example, should more disaggregated data be used?

If the manager or auditor believes that the difference is due to a misstatement, then he or she should
perform additional analysis and inquiry, using knowledge of the industry and entity to evaluate the most
likely causes and identify a plausible explanation.

Possible adjustments to unadjusted client accounts

When using analytical procedures to plan the audit and assess risk, the auditor typically uses client
7
financial data that has yet to be adjusted (by either the entity’s own post-closing adjustments or by audit

7
For example, audit planning analytics (performed as part of the auditor’s risk assessment procedures) may occur
prior to year-end; therefore, the client may not have yet made their own adjusting journal entries for accruals.

© 2020 Association of International Certified Professional Accountants. All rights reserved. 1-14