Using Computerized Audit Software to Learn Statistical Sampling: An Instructional Resource

Robert C. Richardson and Timothy J. Louwers

INTRODUCTION

Section 404 of the Sarbanes-Oxley Act (U.S. House of Representatives 2002) requires an assessment of internal control. To carry out this assessment, public accounting firms must use either non-statistical or statistical sampling. Furthermore, their sample selection will fall under the scrutiny of the Public Company Accounting Oversight Board?s (PCAOB) rigorous firm inspec- tions. Although statistical sampling requires some judgment, non-statistical sampling requires more judgment. Hitzig (2004) argues that using non-statistical sampling instead of statistical sampling could subject auditors to even greater professional and legal criticism. Nevertheless, many firms have been reluctant to embrace statistical sampling. Hitzig (2004) states that this reluctance had historically been due to three major factors: the cost of sample selection, the cost of sample evaluation, and the cost of training. The cost of sample selection and the cost of sample evaluation are virtually eliminated today by the widespread availability of software packages (Hitzig 2004). All of the Big 4 firms have at least one CAATs software package (e.g., ACL, IDEA, etc.) available. Many smaller firms often have software available (Lanza 1998; Warner 1998), although they sometimes fail to understand how to properly use the software due to a lack of training and an absence of technology ?champions.?

In addition to the argument for more objective measures found in statistical sampling, certain types like Monetary-Unit-Sampling (MUS) can result in more efficient (i.e., smaller) sample sizes. MUS has the additional feature that sampling can begin before the population is complete. Al-

though the cost of training is not insignificant, at least one of the Big 4 firms believes that the potential benefits outweigh the cost as they have increased their training emphasis on Monetary- Unit-Sampling. Furthermore, the firm is moving the training to an earlier stage in their staff?s career, indicating a desire by the firm for younger staffers to get educated in statistical sampling.

Many partners are reluctant to embrace technological changes, making college students ideal apprentices of CAATs. Students stand to have the biggest initial impact in this area. It could take a decade for a college student to catch up to the industry experience of a partner, but it could take one week to surpass some partners on CAATs knowledge. Although the focus of this tutorial is statistical sampling, CAATs can also be used in conducting fraud detection procedures, performing analytical procedures, conducting data queries, as well as performing other auditing applications. We are using ACL in this tutorial; other audit software packages are similar. In other words,

learning one software package will be helpful when using another (e.g., IDEA).

Robert C. Richardson is an Associate Professor and Timothy J. Louwers is a Professor, both at James

Madison University.

The authors gratefully acknowledge the comments of the editor, Greg Gerard, and two anonymous reviewers whose comments greatly strengthened the instructional resource.

STATISTICAL SAMPLING WITH ACL

The following tutorial is divided into three main parts. In the first part, you will be asked to load ACL and the data file that you will use. In the second part, you will be asked to determine sample sizes under a variety of assumptions using Monetary Unit Sampling (MUS). MUS is used primarily for substantive testing (i.e., testing account balances). In the third part, you will be asked to determine sample sizes under a variety of assumptions using Attribute Sampling. Attribute sampling is used primarily for controls testing. Both the MUS and Attribute Sampling techniques you will use are considered statistical sampling. This tutorial will illustrate how different factors affect sample size. In addition to becoming familiar with factors that affect sample size, you will also become familiar with ACL.

Understanding the impact that certain factors have on sample size will benefit you in three ways. First, you can plan and modify your audit approach more efficiently if you can anticipate how sample sizes might change. Second, an anticipated sample size change (e.g., an increase in sample size) might reveal an error (e.g., an input error) if the actual sample size outcome is different than anticipated. Last, sample size determinants have been tested on prior CPA exams and could be included in future exams as indicated by the sections labeled ?applications of audit sampling? and ?computer-assisted audit techniques? in the Uniform CPA Exam content specifica- tion outline for auditing.

Loading ACL and Importing the Data (Estimated Time 2 minutes)

Find the file named ?Accounts Receivable for ACL? in Blackboard and save it to your desktop.

Load ACL with the disk that accompanies your textbook.

After loading ACL, select ?File,? ?New,? and ?Project.? Type in your name under file name and click ?Save.? Click next on the Data Definition Wizard. Select Disk. Select Desktop (or wherever you saved ?Accounts Receivable for ACL?), and click on ?Accounts Receivable for ACL.? Select Open.

Click through the defaults in Wizard. They are ?PCs and all other types of computers,? ?Excel File,? and ?Sheet1.? Type in your name and click ?Save.? Click ?Finish.? If a box pops up saying ?Table ?untitled? changed, save as,? click OK.

Using Computerized Audit Software to Learn Statistical Sampling Using Computerized Audit Software to Learn Statistical Sampling: An Instructional Resource Robert C. Richardson and Timothy J. Louwers INTRODUCTION Section 404 of the Sarbanes-Oxley Act (U.S. House of Representatives 2002) requires an assessment of internal control. To carry out this assessment, public accounting firms must use either non-statistical or statistical sampling. Furthermore, their sample selection will fall under the scrutiny of the Public Company Accounting Oversight Board's (PCAOB) rigorous firm inspections. Although statistical sampling requires some judgment, non-statistical sampling requires more judgment. Hitzig (2004) argues that using non-statistical sampling instead of statistical sampling could subject auditors to even greater professional and legal criticism. Nevertheless, many firms have been reluctant to embrace statistical sampling. Hitzig (2004) states that this reluctance had historically been due to three major factors: the cost of sample selection, the cost of sample evaluation, and the cost of training. The cost of sample selection and the cost of sample evaluation are virtually eliminated today by the widespread availability of software packages (Hitzig 2004). All of the Big 4 firms have at least one CAATs software package (e.g., ACL, IDEA, etc.) available. Many smaller firms often have software available (Lanza 1998; Warner 1998), although they sometimes fail to understand how to properly use the software due to a lack of training and an absence of technology \"champions.\" In addition to the argument for more objective measures found in statistical sampling, certain types like Monetary-Unit-Sampling (MUS) can result in more efficient (i.e., smaller) sample sizes. MUS has the additional feature that sampling can begin before the population is complete. Although the cost of training is not insignificant, at least one of the Big 4 firms believes that the potential benefits outweigh the cost as they have increased their training emphasis on MonetaryUnit-Sampling. Furthermore, the firm is moving the training to an earlier stage in their staff's career, indicating a desire by the firm for younger staffers to get educated in statistical sampling. Many partners are reluctant to embrace technological changes, making college students ideal apprentices of CAATs. Students stand to have the biggest initial impact in this area. It could take a decade for a college student to catch up to the industry experience of a partner, but it could take one week to surpass some partners on CAATs knowledge. Although the focus of this tutorial is statistical sampling, CAATs can also be used in conducting fraud detection procedures, performing analytical procedures, conducting data queries, as well as performing other auditing applications. We are using ACL in this tutorial; other audit software packages are similar. In other words, learning one software package will be helpful when using another (e.g., IDEA). Robert C. Richardson is an Associate Professor and Timothy J. Louwers is a Professor, both at James Madison University. The authors gratefully acknowledge the comments of the editor, Greg Gerard, and two anonymous reviewers whose comments greatly strengthened the instructional resource. Richardson and Louwers STATISTICAL SAMPLING WITH ACL The following tutorial is divided into three main parts. In the first part, you will be asked to load ACL and the data file that you will use. In the second part, you will be asked to determine sample sizes under a variety of assumptions using Monetary Unit Sampling (MUS). MUS is used primarily for substantive testing (i.e., testing account balances). In the third part, you will be asked to determine sample sizes under a variety of assumptions using Attribute Sampling. Attribute sampling is used primarily for controls testing. Both the MUS and Attribute Sampling techniques you will use are considered statistical sampling. This tutorial will illustrate how different factors affect sample size. In addition to becoming familiar with factors that affect sample size, you will also become familiar with ACL. Understanding the impact that certain factors have on sample size will benefit you in three ways. First, you can plan and modify your audit approach more efficiently if you can anticipate how sample sizes might change. Second, an anticipated sample size change (e.g., an increase in sample size) might reveal an error (e.g., an input error) if the actual sample size outcome is different than anticipated. Last, sample size determinants have been tested on prior CPA exams and could be included in future exams as indicated by the sections labeled \"applications of audit sampling\" and \"computer-assisted audit techniques\" in the Uniform CPA Exam content specification outline for auditing. Loading ACL and Importing the Data (Estimated Time 2 minutes) Find the file named \"Accounts Receivable for ACL\" in Blackboard and save it to your desktop. Load ACL with the disk that accompanies your textbook. After loading ACL, select \"File,\" \"New,\" and \"Project.\" Type in your name under file name and click \"Save.\" Click next on the Data Definition Wizard. Select Disk. Select Desktop (or wherever you saved \"Accounts Receivable for ACL\"), and click on \"Accounts Receivable for ACL.\" Select Open. Click through the defaults in Wizard. They are \"PCs and all other types of computers,\" \"Excel File,\" and \"Sheet1.\" Type in your name and click \"Save.\" Click \"Finish.\" If a box pops up saying \"Table 'untitled' changed, save as,\" click OK. Using Computerized Audit Software to Learn Statistical Sampling MUS or PPS Sampling (for Substantive Testing) Choosing customer accounts receivable for confirmation is a common task in auditing. One of the first steps you should perform on a data set is to gather some basic statistics. Step 1: The data file should already be highlighted under \"Project Navigator.\" Within the data file, highlight the \"Amount\" column by clicking on the word \"Amount.\" It should cause the entire column to be highlighted in black as pictured below. Step 2: Select the \"Analyze\" application in ACL. From the drop-down choices, select \"Statistical\" and then \"Statistics.\" Step 3: With the output provided from Step 2 above, answer the following questions: What is the book value of the population? You should calculate the total on the Excel spreadsheet of the \"amount\" column and check the calculation against the book value calculated in ACL to ensure that all of the data was imported. (Additionally, you would want to compare this total to the General Ledger to ensure that all accounts were included.) How many customer accounts are there in the accounts receivable population? How many customer accounts have zero balances? What is the largest customer balance? Step 4: Your supervisor in the field would give you the following three factors to assist you in determining your sampling plan: Tolerable misstatement (i.e., materiality) Allowable risk of incorrect acceptance Expected misstatement (i.e., expected total errors) $1,500,000.00 0.10 $200,000.00 Click on the \"Sampling\" application in ACL. From the drop-down choices, select calculate sample size and a popup box should appear as pictured below. Step 5: Make sure that \"Monetary\" is checked. This will ensure that we are calculating a sample for MUS (i.e., PPS) and not Attribute Sampling. Step 6: Input Confidence. This should always be 1 minus the risk of incorrect acceptance. For our sample, it will be 1 - 0.10 = 0.90. You should input 90 in the confidence box. Step 7: Using the information obtained in Step 3 above, input population book value in the \"Population\" box. This should be 40175738.32 without commas. Step 8: Input Materiality. This should be our tolerable misstatement of 1500000.00 without commas Step 9: Input Expected Total Errors (i.e., Expected Misstatements). This should be 200000.00 without commas. Step 10: Calculate sample size and sampling interval. Click on "Calculate" (NOT on "OK") to see the sample size of 78 and the sampling interval of $512,554.11 as revealed in the screen shot below. If you accidentally click on "OK," you will have to input all four numbers again to make any modifications (i.e., start with step 1 again). . j .... File o Edit I Data 1 I Analyze Sampling pected Error R ate(%) R esult s --------------, Calculate Sample Size 80 Int er val 6.25 002 00 3 0 04 0 05 006 0 07 00 8 00 9 010 011 01 2 013 01 4 015 01 6 017 01 8 019 0 20 021 0 22 0 23 0 24 0 25 026 027 19217.11 812 0 3.59 13862.62 961 64.06 78 001 .53 33507.0 7 157 4 63.59 1 06728.4 6 1361 9.44 12 5 2 0.4 5 46361 .01 1 22 9 83.1362.01 5 48726.88 35650.20 6 9 32.6 0 79969.96 491 19.93 0.00 927 1 44.50 01207.16 1 19 946.01 1 16666.77 79987.13 812 00.72 4 45 69.01 1 26252.38 91 0 28 Size 91 001 1555 94.62 Number of Tolera ble E rrors OK Cancel Help Initial Sample Size= 80 Expected error rate - the percentage Expl ain w hy changes i n expected error rate error rate expected in the p_opulation Changes Changes Sample Si ze From To Changes To 2% 1% 3% 1% 4% 1% 0.5% 1% caused the observed changes in sample size. Risk of Assessing Control Risk Too Low Expl ai n w hy cha nges in the risk of assessi ng Note: Be sure to change expected error rate control risk too l ow caused the observed changes in back to I % sample size. Changes Changes Sample Si ze Changes To From To 5% 10% 5% 15% 5% 5% 2.5% 1% REFERENCES Hitzig, N. B. 2004. Statistical Sampling Revisited. The CPA Journal (May): 30-35. Lanza, R. B. 1998. Take my manual audit, please. Journal of Accountancy (June): 33-36. U.S. House of Representatives. 2002. The Sarbanes-Oxley Act of 2002. Public Law 107-204 [H. R. 3763]. Washington, D.C.: Government Printing Office. Warner, P. D. 1998. ACL for Windows. The CPA Journal (November): 40-44