The dual role of audit evidence.
The expression of an audit opinion is an implicit reflection of two conclusions based on evidence collected and processed during the audit: (1) that the risk of undetected material misstatement in the client's financial statements is appropriately low, (i.e. the opinion is accurate), and (2) that the evidence collected provides sufficient evidential support for the opinion rendered (i.e. the opinion is justified). Audit evidence may thus be viewed as having a dual role in the opinion formulation process--an "opinion-forming" role and an "opinion-justifying" role (Waller and Felix 1984).
This paper analyzes the distinction between the opinion-forming and opinion-justifying roles of audit evidence. Specifically, the paper proposes that standards of evidence usefulness in forming and revising beliefs about the risk of misstatement in a client's financial statements are, as a practical matter, less stringent than standards of evidence usefulness in providing justification for those beliefs. Applying this argument to an internal control evaluation setting, the paper proposes that this difference in standards may lead auditors to form beliefs that control risk is below the maximum level before they have sufficient evidential matter to justify a control risk assessment at that level.
This proposition and related propositions were tested in a field experiment involving practicing auditors. Subjects were presented with information which was suggestive of a relatively strong internal control structure design, but which provided only limited evidence on operating effectiveness. Consistent with the theory, subjects expressed beliefs that control risk was significantly below the maximum level but indicated that the evidence provided was not sufficient to justify an assessment at that level.
Waller and Felix (1984, 27) describe the dual role of audit evidence as follows:
(T)he dual role of audit evidence is to reduce the auditor's uncertainty in two of its facets: (1) a "first-order" uncertainty about whether the client's financial statements contain a material error and (2) a "second-order" uncertainty about whether a set of audit evidence provides adequate support for a conclusion about the "first-order" uncertainty. Reducing the "first-order" uncertainty may be seen as the "opinion-forming" role of audit evidence, while reducing the "second-order" uncertainty may be seen as the "opinion-justifying" role of audit evidence.
This characterization highlights a distinction between the "opinion-forming" and "opinion-justifying" roles of audit evidence. While this distinction is important, it has received little attention in the auditing literature to date. The goal of this paper is to further our understanding of the distinction between the two roles of audit evidence.
A descriptive model of the opinion formulation process is presented which incorporates opinion-forming and opinion-justifying as distinct, though not independent processes. In this model, opinion-forming is described as a sequential belief revision process which guides and directs the search for evidential matter. The paper further proposes that opinion-justifying can also be represented as a sequential process which drives decisions to continue or terminate evidence collection activities.
Based on the descriptive model of the audit process, a theory is presented of how audit evidence contributes to the auditor's opinion-forming and opinion-justifying processes, and additionally, how these two roles differ. Specifically, the paper proposes that standards for evaluating the usefulness of audit evidence differ across the two roles in predictable ways. While there is some theoretical support for this proposition (e.g. Cohen 1977; Goldman 1986; Smith et al. 1991), no study has attempted to examine this issue empirically.
In an effort to test the propositions about the dual role of audit evidence, a field experiment was conducted involving practicing auditors. Subjects were presented with information regarding a hypothetical audit client's internal control structure and asked to express (1) their beliefs about internal control risk, and (2) the degree to which they felt an assessment of control risk at that level would be justified.
It is important to note that, in attempting to examine the dual role of audit evidence empirically, the paper is venturing into largely unexplored territory. Practicing auditors are assumed to be familiar with making explicit judgments about internal control risk, since they are required by professional standards to do so in practice. However, auditors are not as likely to be familiar with making explicit judgments about degrees of justification. Indeed, opinion-justifying is an implicit (and perhaps even unconscious) part of the audit process.
This raises a number of issues with respect to the design and interpretation of the empirical tests. First, since justification judgments are not explicit in practice, it is not clear that the choice of measurement scales in this paper is appropriate. Furthermore, auditors' lack of familiarity with making these judgments explicit is a potentially confounding issue regardless of the measurement scale used. In light of these issues, the empirical part of this study should be viewed as a first attempt at examining the dual role of audit evidence in a field setting and consequently, the empirical results should be interpreted with caution. Nevertheless, this is an important first step in contributing to our understanding of how auditors view the two roles of audit evidence.
Also examined, as a secondary issue, is the extent to which auditors agree on the amount of evidential support that individual items of evidence provide. While prior research has examined the degree of consensus among auditors' assessments of the relative assurance provided by different tests of controls procedures (e.g. Spires 1991), little is known about how much "credit" is given to individual items of evidence in reaching the required degree of evidential support. Again, while the tests should be viewed as a first attempt at empirical examination of the justification process, such an examination is important in light of the increasing need for auditors to defend their actions in the event of litigation. The results, at a minimum, highlight the need to explore the justification process further.
Waller and Felix (1984, 28) characterize the dual role of audit evidence in the opinion formulation process as follows:
(T)he auditor collects and processes evidence E in order to reduce uncertainty in two of its facets: R = Bel(M|E) and S = Bel(R|E), where M is a material misstatement in the client's financial reports and Bel (.|.) is an evidential support or belief support function. The auditor's goal is to collect and process E such that R <= [R.sup.*], S >= [S.sup.*], and C <= [C.sup.*], where [R.sup.*] is the maximum allowable audit risk, [S.sup.*] is the minimum allowable justification for the conclusion about audit risk, C is the cost of E, and [C.sup.*] is the maximum allowable cost of E.
In addition, Waller and Felix (1984) define audit evidence as useful if it affects R and/or S (ignoring C). That is, audit evidence is useful if it has an impact on (1) the auditor's belief about the likelihood of misstatement in the client's financial statements (its "opinion-forming" role) and/or (2) the degree to which a conclusion that R <= [R.sup.*] is justified (its "opinion-justifying" role). A key implication of this characterization is that auditors' opinion-forming and opinion-justifying processes are conceptually distinct, though not independent. That is, E drives both R and S, although a given piece of evidence may not be equally useful to each.
The following section outlines a descriptive model of the audit process which incorporates opinion-forming and opinion-justifying as distinct, though interrelated, processes.
The audit process involves the sequential collection and processing of audit evidence (Felix and Kinney 1982; Gibbins 1984). Early in the audit, the auditor's primary activities involve learning about the client firm and its operations. Initial learning activities normally include procedures such as inquiry of client personnel, gathering information about the client's industry, reading the minutes of board and stockholder meetings, review of prior period financial statements, etc. This information, along with the auditor's own experience-based knowledge of the audit process, forms the basis for the auditor's initial beliefs about the likelihood of misstatement in the client's financial statements.
At this point in the audit, two possibilities for R exist. First, it is conceivable that R may be less than [R.sup.*]. One can envision situations (e.g. the audit is a recurring engagement for an audit client with whom past experience has been extremely favorable) when the auditor might believe that audit risk is below the maximum allowable level before any current evidence is collected. However, there would not be sufficient evidential matter at this point to justify a conclusion that R <= [R.sup.*]. Thus, the auditor would make the determination (either implicitly or explicitly) that S < [S.sup.*], which in turn would lead to a decision to collect additional audit evidence.
Alternatively, suppose that after initial learning activities, R is greater than [R.sup.*]. In this case, S must necessarily be less than [S.sup.*]. Simply put, it would not be prudent for the auditor to conclude that sufficient evidential matter exists to support a level of audit risk less than the maximum allowable when she does not believe that audit risk is less than the maximum allowable level. As in the earlier scenario, the determination that S < [S.sup.*] would lead to a decision to collect additional evidence.
The audit process continues in a sequential fashion, with R and S being revised accordingly in response to new evidence (e.g. from results of analytical procedures, tests of internal controls, or tests of details). As long as S < [S.sup.*], the auditor will continue to collect additional evidence. When S >= [S.sup.*], the auditor terminates her information search. Note that S >= [S.sup.*] implies that R <= [R.sup.*] since a conclusion that sufficient evidential matter exists to support a level of audit risk at or below the maximum allowable level implies that the auditor believes that audit risk is appropriately low.
The model described above incorporates a number of conjectures that are worth noting. First, R may reach a point less than (or equal to) [R.sup.*] before there is sufficient evidential matter to support a conclusion that R <= [R.sup.*] (i.e. before S >= [S.sup.*]). Note, however, that the reverse is not true. That is, S >= [S.sup.*] cannot occur before R <= [R.sup.*] for the reason described earlier.
Second, the model focuses on the process by which the auditor's "ultimate belief" about audit risk comes to be justified. In other words, while the auditor is primarily concerned that the conclusion about audit risk be justified, the model conjectures that the opinion-justifying process is a sequential process whereby the auditor continually evaluates the degree of evidential support for a conclusion about audit risk throughout the audit process as a basis for deciding whether or not to continue evidence collection activities. The termination of evidence collection implicitly signals a conclusion that the expression of an opinion is justified.
Third, the model implies that the auditor's intermediate beliefs about the risk of misstatement in the client's financial statements are not subject to the same rigorous standards of justification as the "ultimate" belief. That is, while the conclusion about audit risk must be justified, the auditor's intermediate, or "pre-conclusion" beliefs need not, and as argued below, cannot be justified (at least not to the same degree as the ultimate belief). As a practical matter, the auditor must make decisions about where and how the search for audit evidence should be directed without the benefit of complete information. The direction of this information search is based primarily on the auditor's beliefs about the risk of error in various components of the client's financial statements (e.g. account balances or assertions) at the time the decision is made to collect additional evidence.
Unarguably, it is desirable that the auditor's intermediate beliefs be an unbiased reflection of the evidence processed up to that point and in this sense there should be some degree of "justifiedness" in auditors' intermediate beliefs. It is less crucial, however, that they be justified in the same sense that the audit conclusion must be justified. This is primarily due to the fact that, unlike the audit opinion, evidence collection strategies can be easily modified when results occur which are inconsistent with expectations.
Furthermore, requiring intermediate beliefs to be "fully justified" in order to direct evidence collection activities would necessarily result in a "Catch-22" situation. Auditor's beliefs about the risk of error at intermediate points in the audit are based on limited information. Justification of an intermediate belief would involve establishing that sufficient evidential matter exists to support a conclusion that the belief is reasonably accurate. However, because the intermediate belief is based on incomplete information, determining that the belief is reasonably accurate necessarily depends on whether or not subsequent audit evidence is consistent with the belief. Thus, as a practical matter, intermediate beliefs cannot ordinarily be "fully justified" prior to the collection of additional audit evidence.
Finally, the model proposes a distinction between the roles of the opinion-forming and opinion-justifying processes. Specifically, the model suggests that the opinion-justifying process drives the decision to collect additional audit evidence while the opinion-forming process guides and directs that information search.
The Dual Role of Audit Evidence
The foregoing discussion suggests that auditors' beliefs at intermediate points in the audit process play an important role in guiding the search for additional information. Audit evidence is useful in this process if it provides the auditor with new information regarding the presence or absence of potential areas of weakness in the client's financial statements. This differs from the role of audit evidence in providing justification for the expression of an audit opinion. Audit evidence is useful in this process if it provides the auditor with some degree of competent evidential support for the expression of an audit opinion. These two roles can be viewed as distinct since audit evidence is used in different ways under each role.
As discussed above, the auditor's intermediate beliefs which form the basis for directing the search for additional audit evidence are not subject to the same rigorous standards of justification as the ultimate belief about audit risk. Indeed, they cannot be, since decisions about where and how to search for error must necessarily be made without the benefit of complete information. Based on this argument, auditors will employ greater latitude in their use of audit evidence in forming and revising beliefs about the risk of misstatement than in assessing the usefulness of audit evidence in providing the required degree of evidential support for the audit opinion. In other words, the standards for evaluating the usefulness of audit evidence are more stringent for the justification process than for the belief-forming process.
This proposition is consistent with observations made in a number of studies by epistemologists and philosophers. For example, Cohen (1977, 352) states:
"... it may be rational to believe something yourself on evidence which does not make it completely certain, ... but it is sensible to require better evidence than this when you are making assertions which, because they tend to generate beliefs in others, may rebound to affect your own credit."
Similarly, Goldman (1986, 21) asserts:
"Evidential standards for a public claim may differ from those for a private belief."
Consistent with these views, Smith, Benson, and Curley (1991, 294) state:
"Because knowledge ... is widely shared, rigorous standards of justification must be applied in authenticating claims. Standards for personal belief need not be so demanding. Indeed, they cannot afford to be since practical affairs require action. For practical purposes, beliefs are knowledge surrogates that allow decisions to be made when relevant propositions cannot be justified as true."
With this proposition in mind, consider now the circumstances under which audit evidence affects the belief-forming process but not the justification process and vice versa. As discussed earlier, the auditor's primary evidence-gathering activities early in the audit involve information which is somewhat general in nature. This information is useful in forming beliefs about the risk of misstatement in the client's financial statements, but of limited usefulness in providing evidential support for the expression of an audit opinion. Indeed, auditors' initial expectations about misstatement in the financial statements may start forming even prior to the collection of current audit evidence as a result of prior experience with the audit client and/or the auditor's experience-based knowledge of the audit process. This experience, however, would not be expected to contribute significantly to the justification process as useful audit evidence.
Thus, to the extent that early audit evidence is more general in nature, it will be more useful in forming beliefs than in providing evidential support. Furthermore, when this information is "positive" in nature, this difference in the usefulness of evidence across these two roles may result in the auditor reaching a belief that audit risk is at or below the maximum allowable level before there is sufficient evidential matter to justify such a conclusion.
When this occurs, subsequent audit evidence (which is generally more specific in nature; e.g. results of substantive testing procedures) would not be expected to further affect beliefs unless it is inconsistent with beliefs. That is, as long as additional evidential matter suggests that the financial statements are free of material misstatement, one would not expect beliefs to change. However, one would expect this information to affect the degree of evidential support for beliefs. Thus, later audit evidence which is consistent with beliefs will be more useful to the justification process than the belief-forming process. On the other hand, later audit evidence which is not consistent with beliefs (e.g. when an unexpected error is discovered) would be expected to affect both beliefs about the risk of misstatement and the degree of evidential support.
The model presented in the previous section describes the process by which the auditor reaches and justifies the audit opinion. While the audit opinion may be viewed as the ultimate "judgment" which must be justified in the audit process, the logic of the model may also be applied to specific phases of the audit in which a justified conclusion must be drawn. One such phase is the evaluation of the audit client's internal control structure.
In this section, the logic of the model is applied to the internal control evaluation process in order to develop testable hypotheses based on the propositions about the dual role of audit evidence. While the nature of internal control judgments differ somewhat from those described earlier, the basis logic of the model is the same. In particular, the propositions with respect to the dual role of audit evidence remain essentially unchanged.
The internal control evaluation process differs from the opinion formulation process in several respects. First, while an audit opinion implicitly reflects the belief that audit risk is appropriately low, control risk assessments need not reflect the belief that control risk is appropriately low. In other words, a control risk assessment may be made at any level that the auditor deems appropriate. Second, control risk assessments need only be justified to the extent that they are below the maximum level (see SAS No. 55, AICPA 1988). That is, when an auditor assesses control risk below the maximum level, she should have sufficient evidential matter to support an assessment at that level. With these points in mind, a brief description of the role of evidence in the belief-forming and justification processes of internal control evaluation is now presented.
In the context of internal control evaluation, initial evidence-gathering activities comprise procedures performed to obtain an understanding of the client's internal control structure. These might include, for example, documentation of the client's accounting system, identification of existing control procedures, and inspection of the client's policy and procedures manuals. Consistent with the discussion in the previous section, these procedures will be most useful in forming beliefs about the likelihood that significant misstatement will get through the client's internal control structure undetected.
Furthermore, when this information is suggestive of a relatively strong internal control design, it is conceivable that auditors will form beliefs that control risk is below the maximum level. However, the information collected in obtaining an understanding of the client's internal control structure would normally not be sufficient to justify a control risk assessment below the maximum level. As discussed earlier, this situation can occur as a result of differences in the standards of evidence usefulness across its two roles. This proposition is the focus of the first hypothesis regarding the dual role of audit evidence.
H1: When information from procedures performed to obtain an understanding of the client's internal control structure are suggestive of a relatively strong internal control design, auditors will form "unjustified" beliefs that control risk is below the maximum level.
If initial control risk beliefs are below the maximum level, the auditor uses these beliefs in order to plan the nature and extent of additional evidence collection activities. Note that if initial beliefs are at or near the maximum level, this implies a weak internal control structure design. In this case, the auditor would not continue collecting additional evidential matter, but would assess control risk at the maximum level. (Recall that an assessment at the maximum level requires no justification). Additionally, even if the auditor believes control risk is less than the maximum level, she may choose to assess control risk at the maximum level if gathering additional evidence to support a control risk assessment below the maximum level would be inefficient.
Assume, however, that the auditor decides to continue collecting evidential matter. This decision implies that initial control risk beliefs are below the maximum level (Morton and Felix 1991). When additional evidence collected is also "positive" in direction (i.e. it is consistent with beliefs), although it may lead to a further decrease in control risk beliefs, its primary role will be to increase the degree of evidential support for control risk beliefs below the maximum level. This proposition is the focus of the second hypothesis regarding the dual role of audit evidence.
H2: When subsequent evidence is "positive" in direction, its primary role will be to increase the degree of evidential support for control risk beliefs below the maximum level.
In order to justify a control risk assessment at the level of control risk beliefs, the auditor would continue the evidence collection process until such time that sufficient evidential exists to support a control risk assessment at that level. However, the auditor may elect to terminate evidence collection activities at an earlier point (i.e. before control risk beliefs are justified). In this situation, the auditor would assess control risk at that level (above the level of control risk beliefs) that is sufficiently supported by the evidential matter collected. This proposition has been examined elsewhere (see Morton and Felix 1990) and is not addressed in this study. That is, the present focus is not on control risk assessments but on the distinction between the belief-forming and justification processes and the role of audit evidence with respect to each.
As noted in the introduction, a secondary issue which is examined empirically in the field experiment described below is the degree to which auditors agree on the amount of evidential support that specific items of evidence provide. However, because little is known about how much "credit" auditors do (or should) give to specific types of evidence and because this process is normally of an implicit, rather than explicit nature in practice, no predictions are made with respect to judgment consensus.
In order to test the hypotheses described above, a case was developed which provided information regarding the internal control structure of a hypothetical audit client. The information was presented in two forms: (1) general information regarding the audit client's background, accounting system, and personnel policies, and (2) four items of additional (subsequent) information regarding the design and/or operation of specific control procedures. This information is presented in Appendix A.
The background information was intended to provide a description of a relatively strong internal control design as a basis for testing H1. The four items of additional information, while more specific in nature, would not normally be sufficient to justify a control risk assessment significantly below the maximum level. These information items were chosen for two reasons.
First, while the inspection of a sample of client documents provides the most persuasive evidence to justify a control risk assessment below the maximum level (see Spires 1991), it is less clear what the contribution of items such as management inquiry and prior year test results should be to providing evidential support. Thus, the experiment provides initial evidence on how auditors view the contribution of "less persuasive" evidence.
Second, the choice of the additional information items provides the basis for a stronger test of H1. That is, not only are control risk beliefs based on the background information alone expected to be perceived as "unjustified", but control risk beliefs based on all the information should be similarly "unjustified" (although less so). In addition to providing an additional test of H1, the additional information items are intended to provide a basis for tests of H2 and for an examination of the degree of consensus on the amount of evidential support provided by specific items of evidence.
The experiment was designed as a repeated measures design; that is, subjects were presented with a series of eight "cases" containing different combinations of the information items and asked to consider each case independently. The background information was included in every case (see Table 1). This design was chosen primarily because it provides a means for calculating the impact of individual evidence items on judgments. This design is subject to some limitations, however, in that subjects' responses to one case may affect their judgments in other cases (i.e. they may not be treated independently). In order to reduce the impact of these effects on our results, the eight cases were randomized across subjects.
Each subject received a questionnaire which included an introduction, instructions, information about a hypothetical audit client's internal control structure, and a response booklet for recording judgments. In the introduction, subjects were informed that the purpose of the experiment was to examine auditor's use of information in internal control judgments. Specifically, they would be required to make two types of judgments: (1) control risk judgments (CR), and (2) evidence sufficiency judgments (ES).
In the instructions, subjects were told that their CR judgments should correspond to their beliefs about the likelihood that a material misstatement could occur and get through the client's internal control structure undetected. The instructions noted that professional standards require sufficient evidential support for a control risk assessment below the maximum level. Subjects were told, however, that the emphasis is on what they actually believe control risk to be. Evidential support considerations, they were told, was the purpose of the ES judgments. Subjects were told that their ES judgments should reflect the degree to which the specified information provided sufficient competent evidential matter to support their beliefs.
After reading the instructions, subjects were provided with general information regarding a hypothetical audit client. This information was presented as "excerpts from the permanent file" of the client and included a description of the firm, it's accounting system, and personnel policies. After reading this information, subjects were shown a list of four "additional information" items. Subjects were asked to make CR and ES judgments based on the permanent file excerpts and different combinations of the additional items. The additional information items were designed to have as little information overlap as possible in order to minimize the potential for significant interactions among the four items.
Different combinations of the additional items were presented in a series of eight cases. Each subject provided CR and ES judgments for all eight cases, the order of which was randomized across subjects. Table 1 lists the additional information items included in each case. The eight cases comprise a 1/2 replicate of a 24 orthogonal design. The combinations of additional information items were chosen to allow calculation of all two-way interactions between information items (Winer 1971). An ANOVA model was estimated in order to test whether any of these interactions could account for a significant portion of the variation in judgments. The results, not reported here, indicate that all two-way interactions are insignificant. Based on this result, it is assumed that all higher order interactions are also insignificant.
The "Y"s ("N"s) in the columns under each case indicate which items subjects were asked to consider (not consider) in addition to the permanent file excerpts when making their judgments. Subjects recorded their CR and ES judgments by marking an 'X' on the appropriate response line provided for each case. As Appendix B indicates, subjects' responses could range from zero percent to 100 percent. Actual judgments were measured by ruler to the nearest one percent. Appendix B contains a sample case.
After completing the cases, subjects were asked to describe their understanding of the meaning of each of the two judgments (CR and ES) without referring back to the instructions. Based on the responses to this task and discussions with subjects after completion of the experiment, it appears that subjects had a good understanding of the task and the meaning of the judgments.
The case materials were pre-tested using a group of 14 audit seniors. The pre-test group reviewed the case for readability, realism, and clarity of instructions. This process resulted in substantial revision of the format and content of the case materials. Several of the auditors in the pre-test group subsequently reviewed the revised case and had no further suggestions for improvement.
A total of 164 audit seniors participated in the experiment, representing each of (what was then) the Big-Six accounting firms. The subject pool included auditors from widely dispersed geographical locations, including Seattle, Boston, Los Angeles, New York, and Phoenix. In addition, the experiment was administered to two groups of auditors who were gathered for CPE from all parts of the United States.
The first hypothesis predicts that auditors' control risk beliefs based on initial evidence will be below the maximum level and that the corresponding evidence sufficiency judgments will also be below the maximum level. In this experiment, initial evidence is represented by excerpts from the permanent file of a hypothetical audit client (see Appendix A). Consequently, subjects' responses to Case 1 are examined in order to test the prediction of H1. In Case 1, subjects were asked to base their judgments on the permanent file excerpts alone (see Table 1).
Table 2 presents the Case 1 mean control risk (CR) and mean evidence sufficiency (ES) judgments for each firm and for the total sample. H1 predicts that both mean CR judgments and mean ES judgments will be less than 1.00. As Table 2 indicates, the mean CR and mean ES judgments are all significantly less than 1.00, both at the firm level and for the sample as a whole. The mean CR judgment for the total sample is .522 with the highest firm mean CR being .615 (Firm 2). In addition, the mean ES judgment for the total sample is .213 with the highest firm mean ES being .337 (Firm 3). These results are consistent with the predictions of H1.
As discussed in the previous section, auditors' control risk beliefs based on all of the information are also predicted to be "unjustified". This is because the additional information items selected for this experiment are assumed to be less persuasive than other types of evidence relevant to the justification of internal control risk beliefs. To test this prediction, auditors' responses to Case 8 (i.e. which included all information) were examined. Table 3 presents the Case 8 mean control risk (CR) and evidence sufficiency (ES) judgments for each firm and for the total sample.
Again, both the mean CR judgments and mean ES judgments are predicted to be less than 1.00. As Table 3 indicates, the mean CR and mean ES judgments for Case 8 are all significantly less than 1.00, both at the firm level and for the sample as a whole. The mean CR judgment for the total sample is .237 with the highest firm mean CR being .284 (Firm 2). In addition, the mean ES judgment for the total sample is .746 with the highest firm mean ES being .827 (Firm 1). These results provide additional support for H1.
The second hypothesis predicts that the primary role of subsequent ("positive") evidence will be to increase evidential support for control risks beliefs below the maximum level. In this experiment, subsequent evidence is represented by the four "additional information" items (see Appendix A). Note that when this information is "positive" in direction, it should reduce control risk beliefs (CR) and increase evidence sufficiency judgments (ES). Consequently, this hypothesis suggests that the increase in ES judgments will exceed the (absolute) decrease in CR judgments. To test this hypothesis, subjects' responses to Case 1 and Case 8 were compared. As noted above, subjects were asked to base their judgments for Case 1 on the permanent file excerpts alone. For Case 8, subjects were asked to consider all of the additional information items along with the permanent file excerpts in making their judgments (see Table 1). Table 4 presents the mean CR and ES judgments for Cases 1 and 8 for each firm and for the overall sample. For each subject, the change in judgments was calculated by subtracting the Case 1 judgment from the Case 8 judgment. The mean change in judgments is presented for each firm and for the overall sample. The direction of the change is negative for CR judgments and positive for ES judgments, as expected.
Since CR and ES judgments are correlated, a new variable was formed by adding the change in CR and ES judgments for each subject in order to test whether the information had a greater impact on ES judgments than on CR judgments as H2 predicts. If this new variable is not significantly different from zero, this would indicate that the information had an equal (but opposite) impact on each of the two judgments. As Table 4 indicates, however, the means of this variable's distributions are significantly greater than zero for all firms except firm 2 and for the sample as a whole. These results indicate that the additional information had a greater impact on ES judgments, as predicted.
Test of Auditor Consensus
Current professional standards for auditors do not provide a great deal of guidance for determining the amount of evidential support that different types of information provide. However, the need for justification is becoming increasingly important in light of recent criticism of and litigation against auditors.
In order to examine the extent to which auditors agree on the amount of evidential support provided by each of the additional items of information, the following regression was estimated for each subject:
ES = a + [b.sub.i]*[ITEM.sub.i], i = 1 to 4
In this equation, a and b are least-squares regression coefficients and ITEMi refers to a particular additional information item. The variable ITEM takes on a value of '1' for Cases in which the information item is taken into consideration in making judgments and '0' if it is excluded. The estimated coefficients represent the change in ES judgments resulting from the inclusion of a particular item.
Next, changes in ES judgments were correlated (i.e. the regression coefficients from equation 1) for each subject with those of every other subject both within the same firm and for the overall sample. The mean values for all between-subjects pairwise correlations are reported in Table 5.
As Table 5 indicates, the mean correlations for the six firms range from a low of .0729 for Firm 1 to a high of .2384 for Firm 5. The mean between-subjects pairwise correlation for the sample as a whole is .1486. While these results suggest that auditors have widely varying perceptions of the degree of evidential support that the additional information items provide, these results may also be the consequence of using an unfamiliar scale for the explicit assessment of an implicit process, as discussed in the paper's introduction. Thus, although conclusions cannot be drawn with regard to auditor consensus, the results may indicate the need for more guidance in assessing evidential support.
This study's goal is to further our understanding of the dual role of audit evidence. A descriptive model of the audit process was outlined which characterizes opinion-forming and opinion-justifying as distinct, though not independent, processes. Based on the model, differences in the role of audit evidence between the opinion-forming and opinion-justifying processes were analyzed. The analysis suggests that standards for evaluating the usefulness of evidence in forming beliefs are less stringent than standards for evaluating the usefulness of evidence in providing evidential support for beliefs.
In an internal control evaluation setting, the difference in standards of evidence usefulness across the two roles of audit evidence can in some situations (i.e. when internal control design is strong), lead auditors to form beliefs that control risk is below the maximum level before they have sufficient evidential matter to support a control risk assessment at that level. In addition, when this occurs, the primary role of additional evidence is to provide the evidential support necessary to support a control risk assessment below the maximum level.
The results are consistent with both of these propositions. It should be emphasized, however, that due to potential measurement problems with "evidence sufficiency" judgments, the empirical tests should be viewed as a first attempt at examining the dual role of audit evidence in a field setting. However, to the extent that beliefs do "lead" justification as the results suggest, this raises a potentially important issue with respect to subsequent testing strategies. In particular, prior research suggests that in some cases, auditors may be prone to a "confirmation bias" in their search for and evaluation of audit evidence (Church 1990).
In addition, the extent to which auditors agree on the amount of evidential support provided by individual items of evidence was also examined. Again, while the results should be interpreted with caution, they suggest a potential need for more guidance in determining the contribution of different types of evidence in providing evidential support for beliefs. Boritz and Wensley (1990) take an important first step in this direction by proposing and developing an expert systems to the assessment of evidence sufficiency.
In this study, the empirical analysis of the dual role of audit evidence is restricted to the situation in which all information is "positive" in order to test specific propositions related uniquely to this situation. However, our understanding of the dual role of audit evidence would be enhanced by replication and extension of this analysis to other phases of the audit, or to situations where information is conflicting or negative in direction.
In conclusion, while the exploratory nature of this study is emphasized, it takes an important first step in furthering our understanding of an issue which has potentially far-reaching implications, but which has received little attention to date--the dual role of audit evidence.
APPENDIX A EXCERPTS FROM THE PERMANENT FILE
Description of Client
Midwest Lumber Company, Inc. is a medium-sized firm which supplies lumber to local builders in a large midwest community. The company was founded twenty years ago by James R. Smith and was operated as a sole proprietorship for ten years. The company was incorporated ten years ago and its shares have been publicly traded since that time. Mr. Smith continues to have a controlling interest in the firm with ownership of 52% of the shares outstanding. In addition, Mr. Smith serves as president and CEO of Midwest Lumber Company, Inc.
There are three managers reporting to the company president; a Chief Financial Officer (CFO) who oversees the operation of the Accounting Department, a Sales Manager who oversees the staff of sales clerks, and a Product Manager who oversees the company's buying, storing, and delivery of merchandise to customers. Each of these managers have been with the company since its incorporation.
The Board of Directors is comprised of seven prominent members of the community. Each of the Board members owns less than five percent of the shares outstanding.
Description of Accounting System
All sales of lumber to local builders are made on credit. Sales originate at a central order desk where customers may place their orders in person or over the phone. When orders come in, sales clerks enter the customer number, item numbers, and quantities on one of the several computer terminals located at the central order desk. The computer automatically assigns sequential sales order numbers to each order and enters the date, the customer's name and address, unit prices, and total. This information is then printed out by inserting a four-part carbonated sales order form into the printer. The Sales Manager reviews all sales orders and initials the top copy. One copy of the sales order is filed at the central order desk.
Completed sales orders are batched and delivered to the office of the controller at regular intervals. The controller reviews the customer's credit history and, if satisfactory, approves the sales order for further processing. Approved sales orders are signed by the controller. Those which are not approved are sent back to the sales clerks who enter a "cancelled" flag on the computer terminal and notify the customer that their order cannot be processed. These sales orders are kept on file at the central order desk.
The approved sales orders are forwarded to the loading dock where orders are loaded onto flatbed trucks for delivery. One copy of the sales order is delivered to the customer with the order. The loading dock retains a second copy and forwards the original sales order to the accounting department.
When sales orders are received in the accounting department, accounts receivable clerks enter the sales order number on the computer which automatically records the sale and the receivable and updates the customer's account. The original sales orders are filed in the accounting department by customer number. At the end of every month, billing statements are generated by the computer for each customer with outstanding sales orders. These three-part statements provide an itemized list of sales order numbers and corresponding amounts due, as well as the total due for the month. The original statement and a remittance copy are sent to the customer and the final copy is retained along with the original sales orders by customer number. These are kept in a pending file until payment is received.
The firm's mail is opened on a daily basis by the firm's receptionist, who keeps a log of all incoming correspondence. Any checks received are separated from the remittance copies of billing statements. Remittance copies are sent to the accounting department and the checks are given to the firm's secretary who prepares a three part bank deposit form. One copy of this form is sent to the accounting department where it is compared to the remittance copies and payments are posted to customers' accounts.
The company maintains a policy manual which describes the process for the hiring of staff personnel. This process includes the preparation by applicants of a detailed employment application and a rigorous interviewing process. During interviews with prospective employees, the interviewer uses a preprinted form to record the applicant's responses to a series of questions regarding their background, personality, work ethic, etc.
The company also maintains a manual of complete job descriptions for each position within the company. These job descriptions provide for the appropriate segregation of duties and also provide the basis for regular employee evaluations. The company uses preprinted evaluation forms corresponding to the employee's position which are completed by the employee's immediate supervisor on a biannual basis. In addition, the company maintains a manual which describes the company's goals, operations, code of conduct for employees, and grievance process for the resolution of conflicts among employees. Employee turnover for the company has been quite low.
 Observation of Loading Dock Supervisor
During the walkthrough of the client's accounting system, a loading dock supervisor was observed verifying an order to be delivered. After checking off items on the loading dock's copy of the sales order, the supervisor initialed both his copy and the original sales order to be forwarded to the accounting department. During a cursory examination of sales orders filed in the loading dock, no sales order copies were encountered that were missing checkmarks or initials.
 Sales Manager Statement Regarding Pricing Policy
According to the Sales Manager, upper management meets on a monthly basis to discuss pricing policy. Any revisions in prices agreed upon at these meetings are entered into the computer by the Sales Manager. This process requires the use of a password known only to the Sales Manager. After the required revisions are entered, computer-generated dated copies of the revised price list are forwarded to the CFO and Product Manager for review. Approved price lists are then distributed to accounting department personnel for price verification of sales orders.
 Test Results from Prior Year
In the prior year's audit, a sample of recorded sales orders were examined for verification of approval by the Sales Manager and controller. No exceptions were noted.
 CFO Statement Regarding Management Style
According to the CFO, the firm has adopted an active management style which includes regular interaction with staff personnel. He indicated that supervisors inspect subordinates' work at regular intervals and make unannounced visits to employees' work stations on a daily basis. The CFO feels that this approach has improved communications between supervisors and their subordinates and provides a basis for swift resolution of any problems which are encountered by staff personnel.
APPENDIX B Sample Case INFORMATION AVAILABLE YES NO Permanent file excerpts X Additional information:  Observation of loading dock supervisor X  Sales manager statement regarding pricing policy X  Test results from prior year X  CFO statement regarding management style X JUDGMENT #1 Estimate the likelihood that a material misstatement (in the valuation assertion for gross Sales) will occur and get through the client's internal control structure: 0% 50% 100% Lowest 50/50 Highest Likelihood Chance Likelihood JUDGMENT #2 Assess the extent to which the information provides sufficient competent evidential matter to justify reliance on the above likelihood estimate: 0% 50% 100% No Competent Moderate Sufficient Evidential Amount of Competent Matter Competent Evidential Evidential Matter Matter
AICPA. 1988. Statement on Auditing Standards No. 55. New York, NY: AICPA.
Boritz, J.E., and A.K.P. Wensley. 1990. Structuring the Assessment of Audit Evidence - An Expert Systems Approach. Auditing: A Journal of Practice and Theory 9 (Supplement): 49-87.
Church, B.K. 1990. Auditors' Use of Confirmatory Processes. Journal of Accounting Literature 9: 81-112.
Cohen, L.J. 1977. The Probable and the Provable. Oxford: Clarendon Press.
Gibbins, M. 1984. Propositions about the Psychology of Professional Judgment in Public Accounting. Journal of Accounting Research (Spring): 103-125.
Goldman, A.I. 1986. Epistemology and Cognition. Cambridge, MA: Harvard University Press.
Morton, J.E., and W.L. Felix, Jr. 1990. Assessing Control Risk: Effects of Procedural Differences on Auditor Consensus. Auditing Symposium X. R.P. Srivastava, ed. The University of Kansas: 109-131.
Morton, J.E., and W.L. Felix, Jr. 1991. A Critique of Statement on Auditing Standards No. 55. Accounting Horizons 5 (March): 1-10.
Smith, G.F., P.G. Benson, and S.P. Curley. 1991. Belief, Knowledge, and Uncertainty: A Cognitive Perspective on Subjective Probability. Organizational Behavior and Human Decision Processes 48: 291-321.
Waller, W.S., and W.L. Felix, Jr. 1984. Cognition and the Auditor's Opinion Formulation Process: A Scematic Model of Interactions Between Memory and Current Audit Evidence. Decision Making and Accounting: Current Research: The University of Oklahoma.
Winer, B.J. 1971. Statistical Principles in Experimental Design. McGraw-Hill.
Jane E. Morton, Merrimack College
TABLE 1: Combinations of Additional Information Items Used for Each Case Additional Information Items Case (a) 1 2 3 4 5 6 7 8 Item 1: Observation of Loading Dock N Y Y Y N N N Y Supervisor Item 2: Sales Manager Statement N Y N N Y Y N Y Regarding Pricing Policy Item 3: Test Results From Prior Year N N Y N Y N Y Y Item 4: CFO Statement Regarding N N N Y N Y Y Y Management Style (a) 'Y' ('N') indicates that an item should (should not) be considered for that case. Note: The permanent file excerpts were included in every case. TABLE 2: Control Risk and Evidence Sufficiency Judgments Based on Permanent File Excerpts Only (a) CR t-value ES t-value N Mean for CR<1 Mean for ES<1 (stdev) (Prob>t) (stdev) (Prob>t) Firm 1 39 0.473 11.40 0.240 20.82 (0.288) (0.0001) (0.228) (0.0001) Firm 2 18 0.615 5.58 0.375 8.40 (0.293) (0.0001) (0.316) (0.0001) Firm 3 19 0.449 9.70 0.214 19.74 (0.248) (0.0001) (0.174) (0.0001) Firm 4 38 0.489 12.72 0.268 19.53 (0.248) (0.0001) (0.231) (0.0001) Firm 5 12 0.507 6.49 0.222 12.19 (0.263) (0.0001) (0.221) (0.0001) Firm 6 38 0.602 9.73 0.275 16.42 (0.252) (0.0001) (0.272) (0.0001) All Subjects 164 0.522 22.81 0.265 38.42 (0.268) (0.0001) (0.245) (0.0001) (a) The data in this table are based on Case 1 judgments only. TABLE 3: Control Risk and Evidence Sufficiency Judgments Based on All Information (a) CR t-value ES t-value N Mean for CR<1 Mean for ES<1 (stdev) (Prob>t) (stdev) (Prob>t) Firm 1 39 0.221 22.63 0.827 6.17 (0.215) (0.0001) (0.176) (0.0001) Firm 2 18 0.284 14.07 0.712 4.56 (0.216) (0.0001) (0.268) (0.0003) Firm 3 19 0.200 14.12 0.730 5.75 (0.247) (0.0001) (0.204) (0.0001) Firm 4 38 0.197 28.65 0.745 8.26 (0.173) (0.0001) (0.190) (0.0001) Firm 5 12 0.277 8.72 0.683 4.11 (0.287) (0.0001) (0.268) (0.0017) Firm 6 38 0.275 18.83 0.711 7.78 (0.237) (0.0001) (0.229) (0.0001) All Subjects 164 0.237 44.26 0.746 15.04 (0.221) (0.0001) (0.216) (0.0001) (a) The data in this table are based on Case 8 judgments only. TABLE 4: Comparison of Case 1 and Case 8 Control Risk and Evidence Sufficiency Judgments Prob> N CR ES CR Change + T [absolute Mean Mean ES Change value of T] (b) Firm 1 39 Case 1 0.473 0.240 Case 8 0.221 0.827 Change (a) -0.252 0.587 0.335 5.08 0.0001 Firm 2 18 Case 1 0.615 0.375 Case 8 0.284 0.712 Change -0.331 0.337 0.006 0.06 0.9510 Firm 3 19 Case 1 0.449 0.214 Case 8 0.200 0.730 Change -0.249 0.516 0.267 4.69 0.0002 Firm 4 38 Case 1 0.489 0.268 Case 8 0.197 0.745 Change -0.292 0.477 0.185 3.29 0.0022 Firm 5 12 Case 1 0.507 0.222 Case 8 0.277 0.683 Change -0.230 0.461 0.231 2.31 0.0413 Firm 6 38 Case 1 0.602 0.275 Case 8 0.275 0.711 Change -0.327 0.436 0.109 2.29 0.0277 All Subjects 164 Case 1 0.522 0.265 Case 8 0.237 0.746 Change -0.285 0.481 0.196 7.05 0.0001 (a) Change = Case 8 Judgment - Case 1 Judgment. (b) Two-sided t-test for: CR Change + ES Change = 0. TABLE 5: Between-Subjects Pairwise Correlations Between Changes in Evidence Sufficiency Judgments Number Number of of Pairwise Mean Subjects Correlations Correlation Firm 1 37 666 0.0729 Firm 2 18 153 0.1275 Firm 3 19 171 0.0809 Firm 4 38 703 0.1934 Firm 5 12 66 0.2384 Firm 6 37 666 0.1575 All Subjects 161 12,880 0.1486
|Printer friendly Cite/link Email Feedback|
|Author:||Morton, Jane E.|
|Publication:||Academy of Accounting and Financial Studies Journal|
|Date:||Sep 1, 2002|
|Previous Article:||Earnings response to auditor switches using a multi-tiered auditor classification.|
|Next Article:||Accounting diversity and the value relevance of accounting earnings and book value in four countries-the United States, the United Kingdom, Canada...|