1Chapter 17 Audit Sampling for Tests of Details of Balances

Review Questions

17-1 The most important difference between (a) tests of controls and substantive tests of transactions and (b) tests of details of balances is in what the auditor wants to measure. In tests of controls and substantive tests of transactions, the primary concern is testing the effectiveness of internal controls and the rate of monetary misstatements. When an auditor performs tests of controls and substantive tests of transactions, the purpose is to determine if the exception rate in the population is sufficiently low to ustify reducing assessed control ris! to reduce substantive tests. When statistical sampling is used for tests of controls and substantive tests of transactions, attributes sampling is ideal because it measures the fre"uency of occurrence (exception rate). In tests of details of balances, the concern is determining whether the monetary amount of an account balance is materially misstated. #ttributes sampling, therefore, is seldom useful for tests of details of balances. 17$tratified sampling is a method of sampling in which all the elements in the total population are divided into two or more subpopulations. %ach subpopulation is then independently sampled, tested and the results pro ected to the population. #fter the results of the individual parts have been computed, they are combined into one overall population measurement. $tratified sampling is important in auditing in situations where the misstatements are li!ely to be either large or small. In order for an auditor to obtain a stratified sample of &' items from each of three strata in the confirmation of accounts receivable, he or she must first divide the population into three mutually exclusive strata. # random sample of &' items is then selected independently for each stratum. 17-! The point estimate is an estimate of the total amount of misstatement in the population as pro ected from the !nown misstatements found in the sample. The pro ection is based on either the average misstatement in the sample times the population si(e, or the net percent of misstatement in the sample times the population boo! value. The true value of misstatements in the population is the net sum of all misstatements in the population and can only be determined by a 1'') audit. 17-" The statement illustrates how the misuse of statistical estimation can impair the use of an otherwise valuable audit tool. The auditor*s mista!e is that he or she treats the point estimate as if it is the true population value, instead of but one possible value in a statistical distribution. +ather than udge whether the point estimate is material, the auditor should construct a statistical confidence

1,-1

17-" #continued$ interval around the point estimate, and consider whether the interval indicates a material misstatement. #mong other factors, the interval will reflect appropriate levels of ris! and sample si(e. 17-% .onetary unit sampling is a method whereby the population is defined as the individual dollars (or other currency) ma!ing up the account balance. # random sample is drawn of these individual monetary units and the physical audit units containing them are identified and audited. The results of auditing the physical audit units are applied, pro rata, to the random monetary units, and a statistical conclusion about all population monetary units is derived. .onetary unit sampling is now the most commonly used method of statistical sampling for tests of details of balances. This is because it uses the simplicity of attributes sampling yet still provides a statistical result expressed in dollars. It does this by using attribute tables to estimate the total proportion of population dollars misstated, based on the number of sample dollars misstated, and then modifies this amount by the amounts of misstatements found. This latter aspect gives monetary unit sampling its /variables/ dimension, although normal distribution theory is not used0 rather an arbitrary rule of thumb is applied to ma!e the ad ustment. 17-& $ampling ris! is the ris! that the characteristics in the sample are not representative of those in the population. The two types of sampling ris! faced by the auditor testing an account balance are1 a. b. The ris! of incorrect acceptance (#+I#)2this is the ris! that the sample supports the conclusion that the recorded account balance is not materially misstated when it is materially misstated. The ris! of incorrect re ection (#+I+)2this is the ris! that the sample supports the conclusion that the recorded account balance is materially misstated when it is not materially misstated.

$ampling ris! occurs whenever a sample is ta!en from a population and therefore applies to all sampling methods. While #+I# applies to all sampling methods, #+I+ is only used in variables sampling and difference estimation. 17-7 The steps in nonstatistical sampling for tests of details of balances and for tests of controls are almost identical, as illustrated in the text. The ma or differences are that sampling for tests of controls deals with exceptions and sampling for tests of details of balances concerns dollar amounts. This results in differences in the application of the two methods, but not the steps. 17-' The two methods of selecting a monetary unit sample are random sampling and systematic sampling. 3nder random sampling, in this situation, 4, random numbers would be obtained (the sample si(e in 1,-15) between 1 and 16,764,'''. These would be sorted into ascending se"uence. The physical audit units in the

1,-6

17-' #continued$ inventory listing containing the random monetary units would then be identified by cumulating amounts with an adding machine or spreadsheet if the data is in machine-readable form. #s the cumulative total exceeds a successive random number, the item causing this event is identified as containing the random dollar unit. When systematic sampling is used, the population total amount is divided by the sample si(e to obtain the sampling interval. # random number is chosen between 1 and the amount of the sampling interval to determine the starting point. The dollars to be selected are the starting point and then the starting point plus the interval amount applied successively to the population total. The items on the inventory listing containing the dollar units are identified using the cumulative method described previously. In applying the cumulative method under both random sampling and systematic sampling, the page totals can be used in lieu of adding the detailed items if the page totals are considered to be reliable. 17-( # uni"ue aspect of monetary unit sampling is the use of the preliminary udgment about materiality, as discussed in 8hapter 9, to directly determine the tolerable misstatement amount for the audit of each account. .ost sampling techni"ues re"uire the auditor to determine tolerable misstatement for each account by allocating the preliminary udgment about materiality. This is not re"uired when monetary unit sampling is used. The preliminary udgment about materiality is used. 17-1) #cceptable ris! of incorrect acceptance (#+I#) is the ris! the auditor is willing to ta!e of accepting a balance as correct when the true misstatement in the balance is greater than tolerable misstatement. #+I# is the e"uivalent term to acceptable ris! of assessing control ris! too low for audit sampling for tests of controls and substantive tests of transactions. The primary factor affecting the auditor*s decision about #+I# is control ris! in the audit ris! model, which is the extent to which the auditor relies on internal controls. When internal controls are effective, control ris! can be reduced, which permits the auditor to increase #+I#, which in turn reduces the re"uired sample si(e. :esides control ris!, #+I# is also affected directly by acceptable audit ris! and inversely by inherent ris! and other substantive tests already performed on the account balance, assuming effective results. ;or example, if acceptable audit ris! is reduced, #+I# must also be reduced. If analytical procedures were performed and there is no indication of problem areas, there is a lower li!elihood of misstatements in the account being tested, and #+I# can be increased. 17-11 The statement reflects a misunderstanding of the statistical inference process. The process is based on the long-run probability that the process will produce correct results in a predictable proportion of the times it is applied. Thus,

1,-&

17-11 #continued$ a random sampling process that produces a 9') confidence interval will produce intervals that do, in fact, contain the true population value 9') of the time. <owever, the confidence limits of each interval will not all be the same. 17-1 #+I# for tests of details of balances is the e"uivalent of #+#8+ for tests of controls and substantive tests of transactions. If internal controls are considered to be effective, control ris! can be reduced. # lower control ris! re"uires a lower #+#8+, which re"uires a larger sample si(e for testing. If controls are determined to be effective after testing, control ris! can remain low, which permits the auditor to increase #+I#. #n increased #+I# allows the auditor to reduce sample si(es for tests of details of balances. 17-1! In using the binomial distribution, monetary unit sampling estimates the proportion of all population dollars misstated by some amount. ;or the sample items actually misstated, the amounts of those misstatements are used. <owever, many items in the population have a statistical probability of being misstated by some other amount. #n assumption must be made as to what this amount is in order to compute the monetary unit sampling results. This is called the /percent of misstatement assumption./ $ince the purpose of monetary unit sampling is to estimate the most the misstatements in the population are li!ely to be, there is an inherent need for conservatism in the .3$ process. $ince account balance details if they are overstated, are unli!ely to be overstated by more than their recorded value, a 1'') assumption is a conservative choice. =n this basis it is easier to ustify the 1'') misstatement assumption than a less conservative amount, and thus it is commonly used. 17-1" The preliminary sample si(e is calculated as follows1 Tolerable misstatement #verage misstatement percent assumption +ecorded population value > Tolerable exception rate 4'',''' 1.'' 4'',''' 16,764,''' 5)

3sing the table for a 1') #+#8+ with an expected population exception rate of (ero and a tolerable exception rate of 5), the preliminary sample si(e is 4,.

1,-5

17-1%

*isstatement +ounds using the attri+utes ta+les

*,SSTAT-*-.T3 R-C/RD-D A*/2.T .&61 1.''' .'71

*,SSTAT-*-.T 1 6 &

R-C/RD-D 0A12?9,.17 5,.'6 1,761.7?

A2D,T-D 0A127'9.17 ' 1,466.7?

*,SSTAT-*-.T 6??.'' 5,.'6 99.''

3sing the attributes sampling table for a sample si(e of 1'', and an #+I# of 1'), the 83%+ is1
,.CR-AS- ,. B/2.D R-S21T,.6 5R/* A. ADD,T,/.A1 *,SSTAT-*-.T

./4 /5 *,SSTAT-*-.TS ' 1 6 &

C2-R .'6& .'&9 .'4& .'77

.'17 .'15 .'1&

In order to calculate the upper and lower misstatement bounds, it will be assumed that for a (ero misstatement rate the percent of misstatement is 1''). The upper misstatement bound1
*,SSTAT-*-.T B/2.D 8/RT,/. 69',&,4 6'6,''' 47,,&, 1','16 449,165

./4 /5 *,SSTAT-*-.TS ' 1 6 &

R-C/RD-D 0A1216,764,''' 16,764,''' 16,764,''' 16,764,'''

2.,T C2-R *,SSTAT7 8/RT,/. 7 -*-.T 9 .'6& .'17 .'15 .'1& 1.''' 1.''' .&61 .'71

3pper .isstatement :ound

1,-4

17-1% #continued$ The lower misstatement bound1 :efore ad ustment1

*,SSTAT-*-.T B/2.D 8/RT,/. 69',&,4

./4 /5 *,SSTAT-*-.TS '

R-C/RD-D 0A1216,764,'''

2.,T C2-R *,SSTAT7 8/RT,/. 7 -*-.T 9 .'6& 1.'''

#d ustment1 @oint estimate for overstatements > sum of misstatement percents x recorded value A sample si(e > > > (.&61 B 1.''' B .'71) x (16,764,''' A 1'') 1.&?6 x 167,64' 1,5,5,?

#d usted lower misstatement bound > initial bound - point estimate for overstatements > > 69',&,4 - 1,5,5,? 114,?9,

:ased on this calculation method, the population is not acceptable as stated since the upper misstatement bound exceeds the C4'',''' materiality limit. 17-1& The difficulty in determining sample si(e lies in estimating the number and amount of misstatements that may be found in the sample. The upper bound of a monetary unit sample is sensitive to these factors. Thus, sample si(e varies a great deal with differing assumptions about them. Denerally, the auditor will determine sample si(e by ma!ing reasonable but conservative assumptions about the sample exception rate and average misstatement amount. In the absence of information about misstatement amount, which is most difficult to anticipate, a 1'') assumption is often used.

1,-7

17-17

The decision rule for difference estimation is1 If the two-sided confidence interval for the misstatements is completely within plus or minus tolerable misstatements, accept the hypothesis that the boo! value is not misstated by a material amount. =therwise, accept the hypothesis that the boo! value is misstated by a material amount. ;or example, assume the E8E is -1',''', the 38E is 5',''' and tolerable misstatement is C54,'''. The following illustrates the decision rule1 - T. - 54,''' - 1',''' E8E ' B T. B 54,''' B 5',''' 38E

The auditor can conclude that the population is not materially misstated since both E8E and 38E are within the tolerable misstatement limits. 17-1' When a population is not considered acceptable, there are several possible courses of action1 1. @erform expanded audit tests in specific areas. If an analysis of the misstatements indicates that most of the misstatements are of a specific type, it may be desirable to restrict the additional audit effort to the problem area. Increase the sample si(e. When the auditor increases the sample si(e, sampling error is reduced if the rate of misstatements in the expanded sample, their dollar amount, and their direction are similar to those in the original sample. Increasing the sample si(e, therefore, may satisfy the auditor*s tolerable misstatement re"uirements. Increasing the sample si(e enough to satisfy the auditor*s tolerable misstatement standards is often costly, especially when the difference between tolerable misstatement and pro ected misstatement is small. #d ust the account balance. When the auditor concludes that an account balance is materially misstated, the client may be willing to ad ust the boo! value. +e"uest the client to correct the population. In some cases the client*s records are so inade"uate that a correction of the entire population is re"uired before the audit can be completed. +efuse to give an un"ualified opinion. If the auditor believes the recorded amount in accounts receivable or any other account is not fairly stated, it is necessary to follow at least one of the above alternatives or to "ualify the audit opinion in an appropriate manner.

6.

&. 5. 4.

1,-,

17-1( The population standard deviation is a measure of the difference between the individual values and the mean of the population. It is calculated for all variables sampling methods but not for monetary unit sampling. ;or the auditor, it is usually estimated before determining the re"uired sample si(e, based on the previous year*s results or on a preliminary sample. The population standard deviation is needed to calculate the sample si(e necessary for an acceptable precision interval when variable sampling methods are used. #fter the sample is selected and audited, the population standard deviation is estimated from the standard deviation calculated from the values in the sample. The re"uired sample si(e is directly proportional to the s"uare of the population standard deviation. 17- ) This practice is improper for a number of reasons1 1. Fo determination was made as to whether a random sample of 1'' inventory items would be sufficient to generate an acceptable precision interval for a given confidence level. In fact, a confidence limit was not even calculated. The combined net amount of the sample misstatement may be immaterial because large overstatement amounts may be offsetting large understatement amounts resulting in a relatively small combined net amount. #lthough no misstatement by itself may be material, other material misstatements might not have exhibited themselves if too small of a sample was ta!en. +egardless of the si(e of individual or net amounts of misstatements in a sample, the effect on the overall population cannot be determined unless the results are evaluated using a statistically valid method.

6.

&. 5.

17- 1 Gifference estimation is a method for estimating the total misstatement in a population by multiplying the average misstatement (the audited value minus the recorded value) in a random sample by the number of items in the entire population. +atio estimation is "uite similar to difference estimation. <owever, instead of basing the estimate of total misstatement on the difference between audited and recorded values, it uses the ratio of misstatement amounts to recorded amounts. This ratio for the sample is multiplied times the total population recorded amount to estimate total misstatement. .ean-per-unit estimation is a method of estimating the total audited value of the population by multiplying the arithmetic average, or mean, audited value of the sample times the number of items in the population. $tratified mean-per-unit estimation is similar to mean-per-unit estimation except that the population is divided into groups of homogeneous items, called strata, for purposes of sample design. # separate random sample is selected from each stratum and the estimate of the total population audited amount is computed by determining an estimate for each stratum and adding the results.

1,-?

17- 1 #continued$ The following are examples where each method could be used1 a. Gifference estimation can be used in computing the balance in accounts receivable by using the misstatements discovered during the confirmation process, where a significant number of misstatements are found. +atio estimation can be used to determine the amount of the EI;= reserve where internal inventory records are maintained on a ;I;= basis but reporting is on EI;=. .ean-per-unit estimation can be used to determine total inventory value where the periodic inventory method is employed. $tratified mean-per-unit estimation can be used to determine total inventory value where there are several locations and each is sampled separately.

b. c. d.

.onetary unit sampling would generally be preferable to any of these where few or no misstatements are expected. Gifference and ratio estimation are not reliable where the exception rate is low, and mean-per-unit is generally not as efficient. <owever, in item HcI above, mean-per-unit must be used because there is only one value per sample item. 17Tolerable misstatement (8hapter 9) represents the portion of overall materiality allocated to each individual account. It is the amount of misstatement the auditor believes can be present in an account and the account balance still be acceptable for audit purposes. $ince hypothesis testing re"uires a decision rule based on materiality, that amount should be tolerable misstatement for an individual account balance. If test results provide a confidence limit greater than tolerable misstatement, the auditor would conclude the account is misstated. This would result in one or more of several actions1 1. 6. &. 5. 4. @erform expanded audit tests in specific areas. Increase the sample si(e. #d ust the account balance. +e"uest the client to correct the population. +efuse to give an un"ualified opinion.

In addition, it may be possible to ad ust tolerable misstatement (upward) and rema!e the decision. The basis for this would be a reconsideration of the original udgment concerning determining overall materiality and allocation to the accounts. ;or example, audit wor! completed on another account may indicate that a much lower tolerable misstatement exists for that account then originally planned. This would allow a reallocation providing a larger tolerable misstatement to the sub ect account.

1,-9

17- ! Gifference estimation can be very effective and very efficient where (1) an audited value and a boo! value is available for each population item, (6) a relatively high fre"uency of misstatements is expected, and (&) a result in the form of a confidence interval is desired. In those circumstances, difference estimation far outperforms both .3$ and mean-per-unit estimation. It may or may not outperform ratio estimation, depending on the relationship of misstatement amounts to recorded amounts, but it does re"uire less computational effort than ratio estimation in any case. If focus on large dollar value items is re"uired, difference estimation can be used with stratification. 17- " %xamples of audit conclusions resulting from the use of attributes, monetary unit, and variables sampling are as follows1 3se of attributes sampling in a test of sales transactions for internal verification1 We have examined a random sample of 1'' sales invoices for indication of internal verification0 two exceptions were noted. :ased on our sample, we conclude, with a 4) ris!, that the proportion of sales invoices to which internal verification has not been applied does not exceed 7.6). 3se of monetary unit sampling in a test of sales transactions for existence1 We have examined a random sample of 1'' dollar units of sales transactions for existence. #ll were supported by properly prepared sales orders and shipping documents. :ased on our sample, we conclude, with a 6') ris!, that invalid sales do not exceed C5','''. 3se of variables sampling in confirmation of accounts receivable (in the form of an interval estimate and a hypothesis test)1 We have confirmed a random sample of 1'' accounts receivable. We obtained replies or examined satisfactory other evidence for all sample items. # listing of exceptions is attached. :ased on our sample, we estimate, with 1') ris!, that the true population misstatement is between C6',''' understatement and C5',''' overstatement. $ince tolerable misstatement for accounts receivable is udged to be C4',''', we conclude, with a ris! of 4), that accounts receivable are not materially misstated. *ultiple Choice Questions from C8A -7aminations a. a. (5) (5) b. b. (&) (6) c. c. (&) (6)

17- % 17- &

1,-1'

Discussion Questions and 8ro+lems a. b. 96 (:oo! valueAtolerable misstatement) x assurance factor > (7,9'','''A14',''') x 6

17- 7

If poor results were obtained for tests of controls and substantive tests of transactions for sales, sales returns and allowances, and cash receipts, the re"uired sample si(e for tests of details of balances would need to be increased. 3sing the formula in the problem, the auditor would increase sample si(e by increasing the assurance factor. This has the same effect has specifying a lower acceptable ris! of incorrect acceptance (#+I#). # systematic sample can be selected based on the number of accounts, or the dollar value of the population. To select a systematic sample based on the number of accounts, the total number of accounts in the population is divided by the re"uired sample si(e to determine the interval. # random number is then selected between one and the interval as the starting point. :ecause each account has an e"ual li!elihood of selection, this method is appropriate if all the accounts are similar in si(e, or if the population is stratified into two or more samples. To select a systematic sample based on the dollar value of the population, the population value is divided by the re"uired sample si(e to obtain the appropriate interval. # random number is then selected between one and the interval as the starting point. The interval is added to the starting point to determine the dollar units selected. #ccounts are selected for testing where the cumulative total of accounts receivable includes the random number. This method of selection is similar to monetary unit selection, and accounts greater than the amount of the interval are automatically selected using this method. The direct pro ection of error for the sample can be computed as follows1 (%rrors in sampleAsample boo! value) x population boo! value > (1,4''A6&',''') x 7,9'',''' > C54,''' overstatement The pro ected error of C54,''' is well below tolerable misstatement of C14',''' and provides an allowance for sampling ris! of C1'4,'''. #ccordingly, the population is deemed to be fairly stated.

c.

d.

1,-11

17- '

a.

The following summari(es the confirmation responses1

Recorded 0alue Confirmation Response C1,&,619 17,9&, ,,?7, ' 7,?&6 ' 6&,759

*isstatement ' 7,46' 4,6 1,,55& 76' ' 1',9&5 C&7,'?9 Timing difference 8utoff error %rror in "uantity shipped 8utoff error @ricing error Timing difference 8utoff error

#cct. 11& #cct. 619 #cct. 67, #cct. 5,7 #cct. 4,& #cct. 7?9 #cct. ?5,

C1?&,619 6&,54, ?,5&9 1,,55& ,,546 5,&?1 &5,4?&

Total misstatement

b.

%stimate of total misstatement

Sample 0alue Sample *isstatements Boo: 0alue 8ro;ected *isstatement

$tratum 1 $tratum 6 $tratum & Totals

C 9&9,19, 1,1,5,471 ,1,6&9 C6,1?5,99,

C ' &5,?9, 1,196 C&7,'?9

C 9&9,19, 5,7?,,??7 ?96,461 C7,419,7'5

C ' 1&9,6?' 15,9&5 C145,615

c.

The population is not acceptable since the pro ected misstatement of C145,615 exceeds tolerable misstatement of C1'','''. The auditor is li!ely to propose an ad ustment andAor increase testing. In this situation, many of the errors involved cutoff, so the auditor could expand testing in this area and propose an ad ustment for the errors found. :ecause the cutoff errors were isolated and testing expanded in this area, the cutoff errors would not be included in the pro ection of error for each stratum. If random selection is performed using %xcel (@1,69.xls), the command to select numbers randomly from the population is1 >+#FG:%TW%%F(1,6',694) The 1' random numbers selected using this approach will vary for each student. The command for selecting the random numbers can be entered directly onto the spreadsheet, or can be selected from the function menu (math J trig) functions. It may be necessary to add the analysis tool pac! to access the +#FG:%TW%%F function. =nce the formula is entered, it can be copied down to select additional random numbers.

17- (

a.

1,-16

17- ( #continued$ ./T-< +andom dollar items are matched with population item numbers where the cumulative boo! value of the population includes the random dollar selected. b. Interval > > > @opulation total Fumber of items selected 6',,694 1' 6',,69 Interval

3sing 1?4, as a starting point, we have1

S=ST-*AT,C D/11AR 1 6 & 5 4 7 , ? 9 1' 1,?4, 66,4?7 5&,&14 75,'55 ?5,,,& 1'4,4'6 167,6&1 157,97' 17,,7?9 1??,51? 8/821AT,/. ,T-* ./4 6 7 ? ? 14 6' 67 &' &' &4

./T-< $ystematic dollar items are related to population item numbers in the same manner as for part a above. c. #ll items larger than the interval will be automatically included. If the interval is 6',,69 item &' will be included at least once, and item ? at least twice. The same is not necessarily true for random number selection, but the probability is high. Fote that for item ?, there is a probability of approximately 66) (55,11'A6',,694) of its being included in a given sample draw. It was included twice in a sample of 1'. There is no significant difference in ease of selection between computer generation of random numbers and systematic selection. $ome auditors prefer the use of random numbers because they believe this helps ensure an unbiased sample. .onetary unit sampling would be used because (1) it is efficient and (6) it focuses on large dollar items.

d.

e.

1,-1&

17-!)

a.

The differences that were uncovered include only five misstatements rather than seven. Items 6 and , are not misstatements, but only timing differences. Therefore, only the five misstatements are summari(ed in order to compute the upper and lower misstatement bounds. These misstatements are summari(ed below.
*,SSTAT-*-.T3 R-C/RD-D 0A12.'?5 .795 (.'&') (.?96) (.'65)

,T-* 1 & 5 4 7

R-C/RD-D 0A12C6,,6?.'' &,?9'.'' ,91.'' 45?.'' &,114.''

A2D,T-D 0A12C6,59?.'' 1,19'.'' ?14.'' 1,'&,.'' &,19'.''

*,SSTAT-*-.T C 6&'.'' 6,,''.'' (65.'') (5?9.'') (,4.'')

2pper misstatement +ound +efore ad;ustment<

*,SSTAT-*-.T B/2.D C54,564 61,9&' 6,&6& C79,7,?

./4 /5 *,SSTAT-*,SSTAT-- R-C/RD-D C2-R *-.T > *-.TS 0A127 8/RT,/. 7 ASS2*8T,/. 9 ' 1 6 C1,9,4,''' 1,9,4,''' 1,9,4,''' .'6& .'17 .'15 .'4& 1.''' .795 .'?5

1ower misstatement +ound +efore ad;ustment<

*,SSTAT-*-.T B/2.D C54,564 6?,1?, ?&' 717 C,4,'4?

./4 /5 *,SSTAT-*,SSTAT-- R-C/RD-D C2-R *-.T > *-.TS 0A127 8/RT,/. 7 ASS2*8T,/. 9 ' 1 6 & C1,9,4,''' 1,9,4,''' 1,9,4,''' 1,9,4,''' .'6& .'17 .'15 .'1& .'77 1.''' .?96 .'&' .'65

1,-15

17-!) #continued$ Ad;ustment of upper misstatement +ound< @oint estimate for understatement amounts > sum of misstatement percents x recorded value A sample si(e > (.?96 B .'&' B .'65) x (1,9,4,''' A 1'') > .957 x 19,,4' > 1?,7?5 #d usted bound > initial bound - point estimate for understatement amounts > 79,7,? - 1?,7?5 > 4',995 Ad;ustment of lower misstatement +ound< @oint estimate for overstatement amounts > sum of misstatement percents x recorded valueAsample si(e > (.795 B .'?5) x (1,9,4,''' A 1'') > .,,? x 19,,4' > 14,&77 #d usted bound > initial bound - point estimate for overstatements > ,4,'4? - 14,&77 > 49,796 b. The population is not acceptable as stated because both the lower misstatement bound and upper misstatement bound exceed materiality. In this situation, the auditor has the following options1 1. $egregate a specific type of misstatement and test it separately (for the entire population). The sample would then not include the specified type of misstatement since it is being tested separately. Increase the sample si(e.

6.

1,-14

17-!) #continued$ &. 5. 4. 7. #d ust the account balance (i.e., propose an ad ustment). +e"uest the client to review and correct the population. 8onsider "ualifying the opinion is the client refuses to correct the problem. 8onsider the criteria used in the test, possibly in connection with additional audit wor! in areas outside of accounts receivable.

=f these options, segregating a specific type of misstatement may prove to be the most beneficial. In this problem, items & and 4 are cutoff misstatements. $egregating these items, testing cutoff more extensively, and eliminating them from the sample would result in the following bounds1 2pper misstatement +ound<
*,SSTAT-*-.T B/2.D C54,564 6,745 C5?,',9 (1,'7,) C5,,'16

./4 /5 *,SSTAT-*,SSTAT-- R-C/RD-D C2-R *-.T > *-.TS 0A127 8/RT,/. 7 ASS2*8T,/. 9 ' 1 C1,9,4,''' 1,9,4,''' .'6& .'17 .'&9 Eess ad ustment K(.'&' B .'65) (19,,4')L 1.''' .'?5

1ower misstatement +ound<

*,SSTAT-*-.T B/2.D C54,564 95? 775 C5,,'&, (1,749) C54,&,?

./4 /5 *,SSTAT-*,SSTAT-- R-C/RD-D C2-R *-.T > *-.TS 0A127 8/RT,/. 7 ASS2*8T,/. 9 ' 1 6 C1,9,4,''' 1,9,4,''' 1,9,4,''' .'6& .'17 .'15 .'4& Eess ad ustment K(.'?5) (19,,4')L 1.''' .'&' .'65

1,-17

17-!) #continued$ It can be seen that both misstatement bounds are now within materiality after cutoff misstatements were segregated. These misstatements were significant in two ways. Their existence increased the overall estimated population exception rate, and their magnitude contributed to the amount of estimated misstatements in the portion of the population represented by the misstatements in the sample. 17-!1 a. The audit approach of testing all three account balances is acceptable. This approach is also desirable when the following conditions are present1 1. 6. &. The auditor can obtain valid, reliable information to perform the re"uired tests in all of the areas. The internal controls for each of the three areas are comparable. .isstatements are expected to occur evenly over the entire population. ;or instance, the auditor does not expect a large number of misstatements in accounts receivable and few, if any, in inventory.

b.

The re"uired sample si(e for all three accounts is1 Tolerable misstatement +ecorded population value 1'','''M 1',''',''' > .'1

;rom the attributes table sample si(e n cannot be determined, but using interpolation it is approximately 115 B (115 - ,7)>146. (This is not an appropriate method to determine sample si(e in practice.) c. The re"uired sample si(es if each account is tested separately are1 A88R/?4 SA*81S,@> .'& > .'6 > .'7 ,7 115 &?

ACC/2.T #ccounts receivable Inventory .ar!etable securities n> n> n>

5ACT/R 1'','''M &,7'',''' 1'','''M 5,?'',''' 1'','''M 1,7'','''

The important point is that sample si(e under b is much smaller than for the combined samples in c.

1,-1,

17-!1 #continued$ d. The population would be arranged so that all accounts receivable would be first, followed by inventory and mar!etable securities. The items would be identified by the cumulative totals. In the example, the number 5,76,,?,1 would relate to an inventory item since it is between the cumulative totals of C&,7'',''' and C?,5'','''. #ccordingly, for this number the inventory audit procedures would be performed. The misstatement data are as follows1
R-C/RD-D A*/2.T C9?,.16 A2D,T-D A*/2.T C??,.16 *,SSTAT-*-.T3 R-C/RD-D A*/2.T 1'.1)

e.

D,55-R-.CC1''.''

#ssuming a 1'') average misstatement in the population when there are no misstatements found and an #+I# of 1'), the misstatement bounds are1 3pper misstatement bound1 C1',''',''' x .'16 x 1.' > C16',''' ?,'?' C16?,'?'

C1',''',''' x .''? x .1'1 > Eower misstatement bound1 :efore ad ustment1 C1',''',''' x .'16 x 1.' #d ustment1 .1'1 x C1',''',''' 6'' >

> C16','''

(4,'4') C115,94'

:ased on the sample results and the stated combined acceptable misstatement of C1'',''', the population (i.e., accounts receivable, inventory, and mar!etable securities combined) should not be accepted as stated without further testing. 17-! 1. (a) 6. (c) &. (a) 5. (d) 4. (d)

1,-1?

17-!! Computer Solution. This is an excellent problem to demonstrate the use of the computer in auditing, as it re"uires a great deal of computational wor!. # solution prepared using %xcel is included on the 8ompanion Website (;ilename @1,&&.xls). Important points to stress are1 1. 6. The spreadsheet program is set up in two sections1 one for data entry and one for computations. 8ells are set up for variables by name, and the values for the variables are then entered in those cells (e.g., sample si(e > ). 8omputations are then done by reference to the cells rather than by entering values in the formulas. This allows the wor!sheet to be used as a general program for similar problems. #lthough the program assures computational accuracy, the formulas must be correct. They should always be reviewed and double chec!ed, and test data should be processed to assure accuracy. a. 8alculating the point estimate1

&.

:efore computing the computed precision interval, we must compute the standard deviation1

e C(,6.'') 74.,' 51.1' &7.1' 41.?' (.16) &'.'' 61.11 C1,&.79

(e )6 4,1?5.'' 5,&17.59 1,7?9.61 1,&'&.&1 6,7?&.65 .'1 9''.'' 554.7& 17,461.,9

1,-19

17-!! #continued$ 8omputed precision interval1

The confidence interval is expressed as &,995.?, B 5,,1?.57. To compute the confidence limits, 38E E8E b. > > N B 8@I > &,995.?, B 5,,1?.57 > ?,,1&.&& N - 8@I > &,995.?, - 5,,1?.57 > -,6&.49

The auditor should not accept the boo! value of the population since the maximum misstatement in the population that she was willing to accept, C7,''', at a ris! level of 4), is less than the possible amount of true misstatement indicated by the 38E of C?,,1&.&&. The options available to the auditor at this point are1 1. 6. &. 5. 4. @erform expanded audit tests in specific areas. Increase the sample si(e. #d ust the account balance. +e"uest the client to correct the population. +efuse to give an un"ualified opinion.

c.

17-!"

a.

It would be desirable to use unstratified difference estimation when the auditor believes that there is not a small number of misstatements in the population that are in total material, and the population has a large number of small misstatements that in total could be material. 3nstratified difference estimation would not be appropriate when either of the above characteristics is not present. ;or example, if the auditor believes that certain large accounts payable may contain large misstatements that are material, they should be tested separately. # significant consideration in this situation is whether the auditor can identify the entire population. This consideration applies whether using stratified or unstratified difference estimation. The auditor in this instance is identifying the population based upon an

1,-6'

17-!" #continued$ accounts payable list. If this list includes only those accounts with an outstanding balance, the sample is ignoring those accounts that have a recorded balance of (ero. Thus, many accounts could be understated but not considered in the sample or the statistical inferences drawn from the sample. b. Ignoring the #+I+, the re"uired sample si(e may be computed as follows1

where T. - %O > 54,''' - 6',''' > C64,'''

c.

In order to determine whether the population is fairly stated, the computed precision interval must be calculated.

8I 8I 38E E8E

> > > >

N B 8@I 61,''' B 66,&,5 5&,&,5 -1,&,5

$ince both 38E and E8E are less than tolerable misstatement, the auditor can conclude that the population is fairly stated. The primary reasons the population is acceptable is that (1) the actual point estimate is reasonably close to the expected misstatement, and (6) the actual sample standard deviation is less than the estimated standard deviation.

1,-61

17-!" #continued$ d. 8onsidering the #+I+, the sample si(e may be computed from the following formula1

e.

The sample si(e increases significantly with the inclusion of the #+I+ because by including it the auditor is establishing the ris! he or she will ta!e of re ecting an acceptable population, as well as considering the ris! of accepting an unacceptable population. It ta!es more effort (sample items) to control two ris!s, rather than ust one. The effect can be seen from reviewing the formula for calculating the sample si(e. The approach described will only result in an appropriate sample si(e by chance. This would occur when the 64) increment is e"ual to the increase in the sample si(e re"uired when the #+I+ is considered. This is not a li!ely occurrence. This approach is not desirable because it is inefficient in terms of time and cost. 3nless by chance the sample si(e is approximately e"ual to the sample si(e re"uired by considering #+I+, the sample si(e will be either too small or too large. Too small a sample will re"uire the sample to be increased. This may be both time consuming and expensive, if it is even possible. 8onversely, too large a sample results in the auditor performing more wor! than is re"uired.

f.

Cases a. Determination of ARIA - Fote that there are many ways to estimate #+I#. =ne method is as follows1 #+I# > > > > ##+ A (I+ x 8+ x #@+) .'4 A (.? x .4 x 1.') .'4 A .5 .1& rounded to .1' (to be conservative)

17-!%

1,-66

17-!% #continued$ Tolerable misstatement as a percent1 T%+ > > > T. A @opulation ?'',''' A 16,''',''' .'7, rounded to .'7 (to be conservative)

$ample si(e determined using Table 14-? (assumes an expected misstatement of (ero and a misstatement percent of 1''))1 n b. > &?

Determination of ARIA - Fote that there are many ways to estimate #+I#. =ne method is as follows1 #+I# > > > > ##+ A (I+ x 8+ x #@+) .'4 A K1.' x .? x (1 - .7)L .'4 A .&6 .17 rounded to .14

Tolerable misstatement as a percent1 T%+ > > > T. A @opulation ?'',''' A 6&,''',''' .'&4 rounded to .'& (to be conservative)

There is no table available for an #+I# of 14). Inherent ris! and control ris! for inventory are greater than for accounts receivable. <owever, due to the inclusion of a component for analytical procedures ris!, #+I# for inventory is not significantly greater than #+I# for accounts receivable. :ecause the boo! value of the population for inventory is much larger, the tolerable misstatement as a percent is much lower for inventory. #s a result, the sample si(e for inventory should be larger than the sample si(e for accounts receivable in re"uirement a. c. The same #+I# must be used for the entire combined test. It would be most prudent to use the lower of the #+I#s calculated for the separate tests, (i.e. 1') from the examples shown in re"uirements a and b). Tolerable misstatement as a percent1 T%+ > > > > T. A @opulation ?'',''' A (16,'''.''' B 6&,''',''') ?'',''' A &4,''',''' .'6& (rounded to .'6)

1,-6&

17-!% #continued$ $ample si(e computed using Table 14-? (allows for a .''4 exception ratePan average of the expected misstatements for accounts receivable and inventory2and assumes misstatement percent of 1''))1 n d. > 195

The generation of random numbers using %xcel (@1,&4.xls) to obtain the sample of &? accounts receivable for confirmation would be obtained as follows1 @opulation boo! value > C16,''',''' 8ommand to obtain each random number1 >+#FG:%TW%%F(1,16'''''') =nce the formula is entered, it can be copied down to select additional random numbers. To obtain a sorted list, the list of random numbers should be copied to a separate column, and pasted as a value (use the H@aste $pecialI command and select HvalueI). Then use the HGata $ortI command to obtain a sorted list. The command for selecting the random numbers can be entered directly onto the spreadsheet, or can be selected from the function menu (math J trig) functions. It may be necessary to add the analysis tool pac! to access the +#FG:%TW%%F function. #n example prepared using %xcel is included on the 8ompanion Website (filename @1,&4.xls).

17-!&

a.

This nonstatistical (i.e., nonprobabilistic or udgmental) sample is a stratified sample. #ll 6& items over C1',''' were examined 1''). The remaining ,,69, items were tested with a sample of ,, items. #lthough this was not a probabilistic sample, auditing standards re"uire that in the auditor*s udgment, it is a representative one. #ccordingly, the results must be pro ected to the population and a udgment made about sampling ris!, although sampling ris! and precision cannot be measured. @ro ection of the total population misstatement would be as follows1 Items over C1','''1 @ro ected .isstatement > #udited value - +ecorded value > 5&6,''' - 574,''' > (&&,''') overstatement

1,-65

17-!& #continued$ Items under C1',''' - average misstatement amount method1 @ro ected .isstatement > #verage sample misstatement x population si(e > K(5,&4') A ,,L x (,,&6' - 6&) > (47.59) x ,69, > (516,6',) overstatement Items under C1',''' - proportional amount method1 @ro ected .isstatement > $ample misstatement ratio x population boo! value > K(5,&4') A ?1,4''L x (6,,7',''' 574,''') > (.'4&) x 6,694,''' > (161,7&4) overstatement Where sample misstatements are1
,T-* 16 19 && &4 41 49 ,5 Totals A2D,T-D 0A125,?6' &?4 64' &,?,4 1,?64 &,,?' ' 15,9&4 R-C/RD-D 0A124,16' 5?4 1,64' &,9,4 1,?4' 5,6'' 6,5'4 19,6?4 *,SSTAT-*-.T (&'') (1'') (1,''') (1'') (64) (56') (6,5'4) (5,&4')

Fote that the sample misstatements are divided by the sample boo! value of C?1,4'' to calculate the sample misstatement ratio. The pro ected misstatement is significantly lower using the proportional amount method because the average account si(e in the sample is large than the average account si(e in the population. Total misstatement is either1 (&&,''') B (516,6',) > (554,6',) overstatement or (&&,''') B (161,7&4) > (145,7&4) overstatement In either case, the following can be said1 There are a significant number of misstated items in the sample, and the amount is "uite large. $ince the sample is representative, it is clear that there is a

1,-64

17-!& #continued$ material misstatement of the population. The amount of misstatement is not easily estimable from the sample. It could be significantly higher or lower than either point estimate. #t this point, the best course of action would be to as! the client to ma!e a study of their records for all population items to identify more accurately the misstatements that exist and correct them. b. If this were a @@$ sample, the sampled portion would be evaluated as follows1 .isstatement taintings1
,T-* 16 19 && &4 41 49 ,5 A2D,T-D 0A125,?6' &?4 64' &,?,4 1,?64 &,,?' ' R-C/RD-D 0A124,16' 5?4 1,64' &,9,4 1,?4' 5,6'' 6,5'4 *,SSTAT-*-.T (&'') (1'') (1,''') (1'') (64) (56') (6,5'4) 8-RC-.T (.'49) (.6'7) (.?'') (.'64) (.'15) (.1'') (1.''')

8alculation of overstatement bound1

2.,T *,SSTAT-*-.T ASS2*8T,/. 1.' 1.' .?'' .6'7 .1'' .'49 .'64 .'15 *,SSTAT-*-.T B/2.D 8/RT,/. 91,?'' 4',59' &7,,6' ?,9?& &,9'6 6,5&, 91? 457

/0-RSTAT-*-.T ' (1) 1 6 & 5 4 7 ,

(1)

281 .'5' .'66 .'6' .'19 .'1, .'1? .'17 .'1,

R-C/RD-D 0A126,694,''' 6,694,''' 6,694,''' 6,694,''' 6,694,''' 6,694,''' 6,694,''' 6,694,'''

;rom Table 14-9 using an #+I# of 4 percent and a sample si(e of ,4. =verstatement bound from sample .isstatement of 1'') items Total overstatement bound 1,-67 194,,97 &&,''' 66?,,97

17-!& #continued$ #n ad usted understatement bound is calculated as follows1 Initial understatement bound > .'5' x 6,694,''' > 91,?'' @oint estimate for overstatements > sum of unit misstatement assumptions A sample si(e x recorded population amount > 6.6'5 A ,, x 6,694,''' > 74,791 #d usted understatement bound > initial bound - point estimate for overstatements > 91,?'' - 74,791 > 67,1'9 #s would be expected, this is very small. $ince all misstatements were overstatements, one wouldn*t expect a net understatement to occur. The results of a @@$ sample indicate that the accounts receivable balance is overstated by as much as C66?,,97. This is about ? percent of the recorded boo! amount. It is significantly greater than tolerable misstatement, indicating that the population is unacceptable and must be sub ect to more scrutiny either by the client andAor the auditor. c. # template for the @@$ portion of the problem is prepared using %xcel on the 8ompanion Website (;ilename @1,&4.xls). This template is a complete wor!sheet for .3$, including appropriate tables for various exception rates and ris! levels. Qou will note that the results are very similar to those computed manually, the differences being due to rounding.

,nternet 8ro+lem Solution< *onetarA 2nit Sampling Considerations

17-1 .onetary unit sampling (.3$) is the most commonly used statistical method of sampling for tests of details because of its simplicity and its ability to provide statistical results in dollars. #n article about using .3$ appeared in the .ay 6''4 issue of The CPA Journal. $ee the following1 Khttp1AAwww.nysscpa.orgAcpa ournalA6''4A4'4AessentialsAp&7.htmL.

1,-6,

,nternet 8ro+lem 17-1 #continued$ 1. The authors suggest that there are three critical steps in applying .3$. What are these stepsR Answer< Getermining the proper sample si(e0 $electing the sample and performing the audit procedures0 and %valuating the results and arriving at a conclusion about the recorded population value. 6. <ow do the authors indicate that a .3$ sample si(e is determinedR Answer< #ccording to the paperSs authors H:ecause .3$ is based on attribute sampling, the sample si(e may be determined by the same basic procedures as for a statistical sample si(e for tests of controls.I &. What two factors must be considered when evaluating resultsR Answer< The authors state that two factors must be considered when evaluating results. These factors are1 the tApe of exception meaning whether it is an understatement or an overstatement and the e7tent of the exception must be measured and considered in estimating the misstatement.
(.ote1 Internet problems address current issues using Internet sources. :ecause Internet sites are sub ect to change, Internet problems and solutions may change. 8urrent information on Internet problems is available at www.pearsonhighered.comAarens.)

1,-6?