Magazine article The CPA Journal

A Second Look at an Old Tool: Analytical Procedures

Magazine article The CPA Journal

A Second Look at an Old Tool: Analytical Procedures

Article excerpt

Advanced microcomputer technology now permits the use of traditional statistical techniques to assist in analytical procedures for use by auditors and management. A case study of one application shows how it works.

The mathematical technique known as regression analysis has been used for many years in the social and physical sciences to identify and quantify relationships among groups of variables. Regression analysis is also well established in the curriculum of many business schools. However, it has been used sporadically by management and auditors as a planning, testing, and control tool. One reason is the substantial computing power required by sophisticated regression-analysis software that until recently was available only on mainframes or minicomputers. Busy managers and auditors were unlikely to find it practicable to interact with a company's central computing resource either directly or through the information systems group. Second, many statistical packages were littered with terminology more suited to professional statisticians than to those who wanted to use the technique occasionally. The learning curve for generalists was just too steep.

Things have changed. Today's fast, power, portable, and affordable personal computers can handle sophisticated regression analysis packages. User interfaces have improved significantly. Turnaround time has improved so users can experiment with model building and see the results quickly. It is time to reconsider the possibility of regression analysis as a management and audit tool.


Regression analysis uses a mathematical technique called the "method of least squares" to create an equation that relates one variable to one or more other variables. A simple regression equation involves only one independent variable and takes the form y = a + bx, which provides an estimate of what the variable y should be given a known value of the variable x. Multiple regression analysis involves the use of two or more independent variables to predict the value of the dependent variable.

The regression equation is developed from a data set involving multiple observations of the variables. In a time-series regression analysis, the data set will represent observations taken at regular intervals over time, such as monthly or weekly. For example, a natural gas utility might build a model to capture the historical relationship between monthly sales to residential customers and such drivers as the volume of gas sold, the number of residential customers, temperature statistics in the markets served, and typical residential unit sales prices. In crossectional analysis--for example, retailing or branch banking--the data set will represent observations taken from different locations at the same point or period in time.

An example of cross-sectional analysis will show how regression analysis can be used in the accounting profession. It is based on an actual experience where regression analysis evolved from an auditing tool to a manement tool. To preserve the anonymity of the company, the actual data has been altered.


The company is a small regional retailer with approximately 90 locations. Virtually every store has the following principal departments: men and women's clothing, jewelry, sporting goods, housewares, and miscellaneous.

The auditors were interested in improving their ability to identify stores with operating results that were out of line with normal expectations. Management used the results to determine if the stores should be subjected to more detailed review and inquiry, including a possible visit by the audit staff. Historically, a rotational plan of store visits was followed. However, since only 15 stores per year were visited, it took at least six years to visit all stores at least once, assuming no new stores were added. The rotation had been augmented by desk reviews of each store's operating results. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.