Professional Skepticism by keralaguest


									                Professional Skepticism and Auditors' Workpaper Review

                                        Kathy Hurtt
                                   University of Wisconsin

                                      Martha Eining
                                      David Plumlee
                                     University of Utah

                                        Draft version
                                       February, 2002

Please do not quote without permission of the authors. Comments are welcome.

We thank the workshop participants at Arizona State University, University of Utah, University
of Wisconsin and Bentley College for their comments on earlier versions of this paper.
Professional Skepticism and Auditors' Workpaper Review

                 Professional Skepticism and Auditors' Workpaper Review

       Review of subordinates' workpapers by more experienced auditors consumes a

significant portion of the effort on an audit engagement (Bamber and Bylinski 1987). Research

into the behavior of workpaper reviewers shows that experience affects their ability to detect

certain kinds of errors (e.g., Ramsay 1994, Bamber and Ramsay 1997, 2000, Harding and

Trotman 1999). Audit risk and information importance are situational factors positively

associated with the accuracy of an auditor's memory for evidence contained in workpapers

(Sprinkle and Tubbs 1998). Beyond experience and the situational factors investigated by

Sprinkle and Tubbs (1998), our understanding of the workpaper review process remains

remarkably incomplete. One unexamined potential influence on the review process can be

found in the audit standards, which state that, "(s)ince evidence is gathered and evaluated

throughout the audit, professional skepticism should be exercised throughout the audit process"

(AU ' 230.7-9). Thus, the due care requirement that auditors exercise professional skepticism is

a critical mandate that should extend to auditors' workpaper review. This paper examines the

relationship between the degree of skepticism possessed by an auditor and four specific

behaviors exhibited during workpaper review.

       The audit profession's institutionalization of workpaper reviewing arises from the

requirement that assistants be adequately supervised (AU ' 150.02). However, from a practical

perspective, the review process is time consuming and expensive (Bamber and Bylinski 1987,

Asare and McDaniel 1996), and finding ways of producing efficiency gains in workpaper

reviewing are likely be welcomed by the auditing profession. Early studies of the review

process (Trotman and Yetton 1985, Trotman 1985) find no performance difference between

interacting groups comprised of two senior auditors and groups comprised of a senior and a

manager. Ramsay (1994) finds that seniors detect mechanical errors better than managers do,

while the reverse is true for conceptual errors. A similar relationship exists between staff and

senior auditors, where staff auditors are better at detecting mechanical errors (Harding and

Trotman 1999). Other studies show that combining the reviews of managers and seniors results

in more effective reviews than either seniors or managers alone (Bamber and Ramsey 1997),

and composite teams consisting of a senior and a staff auditor out perform pairs of either staff

or senior auditors (Harding and Trotman 1999). While this research generally focuses on the

efficiency and effectiveness gains that could result from specialization within the review

process, significant other influences, remain unexamined. This paper addresses one such

influence: the reviewer’s professional skepticism.

       In addition to the due care requirement for auditors to be skeptical, Statement on

Auditing Standards (SAS) No. 53 (AICPA 1988c), No. 57 (AICPA 1988b), No. 67 (AICPA

1992), and No. 82 (AICPA 1997a) reinforce the necessity for auditors to exhibit professional

skepticism in a variety of situations and on different types of engagements. However, there is

anecdotal evidence that auditors either do not consistently behave skeptically or there is some

variation in the degree of skepticism possessed by auditors. Former chief accountants of the

Security and Exchange Commission's (SEC) enforcement division enforcement division have

indicated that a lack of independence and a lack of professional skepticism were among the

primary causes of SEC actions against accountants (The CPA Journal 1996). Research on SEC

Enforcement Actions (1987-1997) by Beasley, Carcello and Hermanson concurred with this

assessment, indicating that 60% of the enforcement actions were related to a lack of

professional skepticism (AICPA 2000). In response to the SEC's concern about the quality of

financial audits, the Public Oversight Board established a panel that recommended to audit

firms that they provide guidance to their audit personnel about the concept of professional

skepticism (POB 2000). Thus, understanding the role of professional skepticism in workpaper

review may provide important insight into a recognized problem within the audit profession.

       A critical step in conducting research involving professional skepticism is a means of

identifying individuals who can be characterized as skeptical. Recent research by Hurtt (2001)

has resulted in a 30-item psychological scale that measures the degree of skepticism possessed

by an individual. In addition, a model linking an auditor's degree of skepticism and certain

behaviors has been proposed (Hurtt, Eining and Plumlee 2001). The model of skepticism,

developed from surveys of professional accountants as well as literatures from philosophy and

psychology, predicts that four behaviors will be associated positively with increasing levels of

skepticism: (1) expanded information search, (2) increased identification of contradictions, (3)

increased generation of alternatives and (4) increased examination of information from and

about people. Our study tests the assertion that auditors with high levels of skepticism will

exhibit these behaviors when reviewing workpapers.

       The AICPA accepts the position that professional skepticism results in behavioral

differences. SAS 53 (AICPA 1988c), one of the pronouncements intended to reduce the

"expectation gap" between the general public and public accountants, re-emphasized the

necessity for an auditor to have an attitude of professional skepticism and described actions that

a skeptical auditor would take. An even stronger position on professional skepticism, and one

with more behavioral implications, is found in SAS 82 which states that, "Due professional

care requires the auditor to exercise professional skepticism - that is, an attitude that includes a

questioning mind and a critical assessment of audit evidence" (emphasis added). SAS 82 also

modifies SAS 1 to more clearly define and explain professional skepticism, and lists examples

of skeptical behavior, including sensitivity in the selection and extent of documentation and

increased corroboration of management representations (AICPA 1992b).

        We test whether more skeptical auditors exhibit these four behaviors when reviewing

 workpapers as well as the impact of higher audit risk on less skeptical auditors in an on-line

 workpaper review task administered to auditors from a major international accounting firm.

 The experiment involved two phases. In the first phase, participants accessed a web site and

 completed the Skepticism Scale and demographic questions. The second phase consisted of a

 workpaper review task, which was also performed on-line, where the reviewer was ask to

 create review notes based on the final review of a subset (debt and inventory) of audit

 workpapers. Embedded in the workpapers were two risk treatment conditions (high and

 normal audit risk indicators). Using the Skepticism Scale score to differentiate between those

 who were more and less skeptical, we find that more skeptical auditors detect more

 contradictions and demonstrate a stable search pattern across risk situations. In contrast, a

 high risk audit situation induced less skeptical auditors, to generate more information search

 queries, spending more time evaluating the information and look at more information from

 and about people, than they did in the low risk situation. However, this additional work by less

 skeptical auditors does not result in more contradictions being detected. This suggests that

 although less skeptical auditors recognize the risk present in a high risk situation, they are not

 effective in translating that awareness of risk into better detection of contradictions.

       The remainder of this paper is organized as follows. The next section reviews prior

research into both workpaper review and professional skepticism along with the behavioral

predictions. The fourth section describes the experimental methods and data collections. The

data analysis is in the fifth section with the conclusions and implications in the last section.


Research on the workpaper review process

       Bamber and Bylinski (1987) document the substantial portion of total audit hours that

audit firms devote to the review process. Certain situational factors have been shown to affect

the review process. Asare and McDaniel (1996) show that when reviewer's are familiar with

preparers they reperform less of the original work. They also find that reviewers who are more

familiar with preparers detect more conclusion errors in a complex task relative to routine task,

while reviewers were not familiar with preparers are more effective in a the routine task.

       Other experimental studies of the review process compare judgments between

sequential reviewing and interacting groups, and show that, whether performance is measured

by consensus or accuracy, the sequential review process was not found to be better than the

group performance (Trotman and Yetton 1985, Trotman 1985). Research involving actual

review of simulated workpapers explicitly recognizes that the nature of the review task differs

between managers and seniors. Ramsey (1994) finds that managers were better at detecting

conceptual errors, while seniors were more accurate at detecting mechanical errors. Harding

and Trotman (1999) generalize Ramsey's (1994) finding by showing that composite groups

with a staff and a senior auditor outperform composite groups with either two seniors or two

staff auditors. Extending this research to the question of whether specialization improves

reviewer performance, Bamber and Ramsey (1997) compared "all-encompassing" reviews with

specialized reviews that focused exclusively on either conceptual or mechanical errors.

Consistent with Ramsey they find that managers were more effective at identifying conceptual

errors, while seniors were more effective at identifying mechanical errors. However, when ex

post teams comprised of both a senior and a manager engage an all-encompassing review, they

are more effective than teams performing specialized reviews. Examining review efficiency

and reviewer confidence, Bamber and Ramsey (2000) find that specialized reviews are less

efficient in terms of time, and seniors performing specialized reviews were more confident

when performing specialized reviews, while managers were not.

       In studies involving reviewer's memory for audit evidence, Moeckel (1990) finds that

inexperienced auditors fail to connect related pieces of audit evidence more often than

experienced auditors were more likely to mentally reconstruct inconsistent pieces of evidence

to form a coherent memory of the evidence. Moeckel and Plumlee (1989) correlate confidence

in memory with recognition accuracy and find that, when the mental representations resulted

from inference, recognition accuracy and auditor’s confidence in their memories were inversely

related, implying that inferred (i.e., internally generated memories) were seen as valid audit

evidence. Looking for situational factors that might mitigate misplaced confidence, Sprinkle

and Tubbs (1998) show that audit risk and information importance mitigate misplaced

confidence. Libby and Trotman (1993) demonstrate that the review process can serve as a

countervailing force to the potentially biased recollections of subordinates. Bamber, Bamber

and Bylinski (1988) surveyed audit managers using an instrument that focused on four audit

areas. They created a taxonomy of eight workpaper review activities. For each activity,

participants indicated whether they would engage in that activity and estimated their anticipated

time spent reviewing. Bamber et al. find that managers engage in directed search, that is,

pursuing related audit evidence rather than analyzing information sequentially. Bamber et al.

describe the review process as a stable search process where auditors are sensitive to account

materiality and risk, but that does not change significantly their review strategies. They also

find that over three-quarters of their subjects read all working papers for each account and

perform active inquiry such as thinking of competing explanations performing analytical tests.

       In summary, research into the workpaper review process finds that it consumes a large

portion a firm's total audit hours, and situational factors effect performance. Auditors' level in

the firm corresponds to how they mentally view the review task, which, in turn, affects their

performance in error detection. Auditors' search through working papers is directed and,

typically, thorough. Research into audit workpaper reviewing has not examined the impact of

professionally mandated factors such as skepticism.

Research on skepticism

       Much of the skepticism research in accounting (McMillan and White 1993, Shaub and

Lawrence 1996, Shaub 1996, Choo and Tan 1998) attempts to measure an individual's

skepticism level then examine behavioral differences among individuals with different levels of

skepticism. Most of these studies build upon the premise that suspicion is the opposite of trust,

and skepticism is synonymous with suspicion. However, it is difficult to find either theoretical

or anecdotal support for equating these constructs. Suspicion and skepticism are never equated

in the philosophical literature on skepticism, rather skepticism is portrayed as a complex

construct, rather than being unidimensional. Also, Hurtt (2001) indicates that practicing

accountants did not equate skepticism with suspicion.

       Several other studies employ a manipulation intended to induce skepticism for a

treatment group, then examined the behavioral differences between groups (Peecher 1996,

Turner 1997, Shaub and Lawrence 1997). Both Peecher (1996) and Turner (1997) find

behavioral differences between the auditors induced to behave skeptically and those in a

control condition. Turner finds that auditors in the skepticism-induced condition examined

more total items and examined more disconfirmatory items (items that would tend to disprove

or discount the client's position) than did the other auditors. Peecher finds that auditors in a

skepticism-induced situation were less likely to rely on client-provided explanations. Shaub

and Lawrence (1997) manipulate situational variables in eight different scenarios so that part of

the information in the scenario either signaled or did not signal a risk of fraud. They ask

participants to predict what they would do in each scenario, and find that new hires generally

exhibit greater skepticism (defined as increased audit testwork and increased questioning of

management and other personnel) than do seniors, managers and partners. These studies

demonstrate that auditors behave differently when induced to do so. It appears that auditors

respond to either situational cues indicating increased risk of fraud or to explicit instructions to

be more skeptical.

        In contrast to the situational view of previous studies, Hurtt (2001) adopts the view that

 skepticism is a psychological characteristic of individuals. She develops an understanding of

 philosophical skepticism as a basis for her multidimensional definition of skepticism, from

 which she develops an instrument that can be used to measure individuals' level of skepticism.

 She identifies six subconstructs of skepticism from questioning professional accountants and

 from the philosophical literature on skepticism: a quest for knowledge (curiosity), a desire to

 understand people, a questioning nature, a slowness in forming judgments, a reluctance to

 accept others' claims, and self-confidence. These constructs were used to develop a 30-item

 scale which demonstrated strong test-retest reliability, indicating that skepticism is a relatively

 stable individual characteristic. This skepticism scale provides useful measure of individual’s

 skepticism level.

       As Hurtt et al. (2001) explain, research into the writings of early philosophers (Annas

and Barnes 1985, Bunge 1991, Kurtz 1992, McGinn 1989, Popkin 1979) suggests that

skepticism results in four behaviors: increased information search, increased detection of

contradictions, increased generation of alternative explanations, expanded scrutiny of

information from and about people. Citing Kurtz (1992) they explain that skeptics have a

'questioning nature' characterized by searching and examining, and skeptics are people who

persist in their investigations (Annas and Barnes 1985). Hurtt et al. also argue that skeptics are

motivated to answer questions that are raised through an expanded information search and are

comfortable with uncertainty because of their willingness to suspend judgment. Thus, skeptics

exhibit more extensive information searches than less skeptical individuals. The second

behavior of skeptics asserted by Hurtt et al. is that skeptics are more adept at discovering

contradictions. They attribute this ability of skeptics to their willingness to suspend judgment,

making it comfortable cognitively for them to delay reaching conclusions and search more

information. The third behavior Hurtt et al. contend skeptics exhibit is constructing alternative

hypotheses regarding statements or claims (McGinn 1989). Support for this comes for the

following, "Skeptics wish to examine all sides of a question; and for every argument in favor of

a thesis, they usually can find one or more arguments opposed to it" (Kurtz 1992, p 22). So,

skeptical individuals can be expected to construct alternate interpretations for the information

they observe. The fourth behavior Hurtt et al. attribute to skeptics is an interest in information

about people. They claim that skeptics want to understand underlying assumptions behind

claims and beliefs and that skeptics try to understand people in order to understand the

assumptions that were made (Popkin 1979). This reluctance 'to accept others' claims' results in

an expanded investigation of information provided by people, since by understanding a person

skeptics can understand the perceptions and assumptions made by that person.

          These philosophical behavioral expectations are also consistent with the audit

procedures required by auditing standards. The expanded information search expected of

skeptics is consistent with the general standard on fieldwork (SAS 1) that requires an auditor to

obtain sufficient competent evidential matter before opining on the financial statements. This

mandates that an auditor perform testwork until she or he believes that sufficient evidential

matter exists. In addition, both contradiction detection and alternative generation are expected

of auditors. For example, SAS 56 (AICPA 1988a) requires auditors to develop expectations

about a company's financial statements prior to performing any analytical review procedures.

As part of the procedure, auditors are expected to identify and explain (generate alternative

explanations for) any unexpected differences. It is only after differences have been identified

and alternatives have been generated by the auditor that any inquiry as to management's

explanations should be made. Evaluating the explanations provided by management again

allows an auditor to detect contradictions in explanations or among analytic review results. In

addition, SAS 56 (AICPA 1988a) indicates that an auditor must obtain corroborating evidence

for management's assertions. SAS 82 requires that auditors specifically assess management's

characteristics by obtaining information about management's abilities, pressures, style and

attitude toward internal control. This is consistent with the suggestion from philosophical

skepticism that a skeptic will exercise increased scrutiny of information both from and about


Hypothesis Development

Hypothesized effect of skepticism

While Peecher (1996) and Turner (1997) find that participants instructed to exercise

professional skepticism exhibit skeptical behaviors, we believe that skepticism is also a

personality trait. That is, skeptical individuals will exhibit skeptical behaviors across

situations. Specifically, we expect that skeptical auditors will exhibit the four behaviors found

in the model of skeptical behavior developed by Hurtt et al. (2001): greater information search,

better detection of contradictions, greater generation of alternative explanations, and expanded

scrutiny of information from and about people. This leads to the following research hypothesis:

        H1: During work paper review, auditors with higher levels of skepticism will display a
        greater levels of information search, contradiction detection, alternative generation,
        and scrutiny of information from and about people than those with lower levels of

Hypothesized effect of risk on individual’s with lower skepticism levels

        SAS 82 instructs auditors to be aware of situational cues that indicate a need for a

heightened sense of professional skepticism. Shaub and Lawrence (1997) find that auditors

responded to situational risk cues embedded in several experimental scenarios. GAAS

specifies the need for skeptical behavior in high risk situations, and research indicates that

auditors do respond to situational cues. What is more uncertain is whether there will be

behavioral differences in the manner that more and less skeptical auditors behave in a "normal"

audit situation. Johnson (1978), indicates that although all people are "skeptical" about certain

things or in certain situations, not all people are skeptics, nor are skeptics skeptical in all

situations. We assume that skeptical auditors maintain a higher level of skeptical behavior at all

times, and less skeptical auditors will exhibit skeptical behavior in higher audit risk situations,

but they will not exhibit skeptical behaviors when the audit situation is not higher risk. This

leads to the following hypotheses:

       H2: During work paper review, less skeptical auditors in the low risk condition will
       exhibit lower level of information search, contradiction detection, alternative
       generation, and scrutiny of information from and about people than either more
       skeptical auditors in either condition or than less skeptical auditors in the high risk

Experimental Methods

Experimental design

       This research employs a randomized block experimental design where participants were

categorized as scoring higher or lower on the Skepticism Scale, then assigned to one of two risk

treatment groups. The risk treatments result from manipulation of situational cues that indicate

either a "normal" or a "high risk" audit condition. The dependent variables are information

examined and time spent (to measure information search behavior and scrutiny of information

from and about people), contradictions detected (to measure contradiction detection), and

alternatives generated (to measure alternative generation).

Experimental Task

       The experiment involved two distinct phases. In the first phase, participants completed

the Skepticism Scale and answered demographic questions. The scale and demographic

questions were coded into HTML and JavaScript and placed on a web-server where

participants answered by accessing a URL for the site. The participants' responses were

collected online using a CGI Script. After scoring each participant's responses to the

Skepticism Scale, we assigned the participants alternately to either the high or low risk

experimental condition based on their ranked skepticism score to ensure balanced cell sizes

between the experimental conditions.

       The second phase of the experiment was administered an average of 13 days after

completion of the pretest, and consisted of a workpaper review task with two treatment

conditions (high and normal audit risk indicators) embedded in the workpapers. Similar to

Shaub and Lawrence (1997), one set of workpapers contains several cues that are consistent

with a heightened audit risk level and the other did not have any specific high risk cues.

Appendix B lists the differences between the instructions in the two treatment conditions.

       The participants conducted general information search through "Workpaper" screens.

The auditors were expected to sign-off on these screens (these were 'normal' audit workpaper

debt and inventory screens) and the auditors had the ability to write review notes for them. In

contrast, information from and about people was contained within "People" screens. The

"People" screens contained brief biographies on the employees, and there was no ability to

sign-off or write review notes on these screens.

       The experimental task was based on the one used by Moeckel (1990). The written task

instructions inform the participants that he or she will be performing the only review of a subset

(debt and inventory) of the workpapers, and they ask the reviewer to document any review and

their reasons for making the note.

       For this study, the original workpapers were modified slightly and were coded into

HTML and JavaScript and placed on a web-server. The modifications included: 1)expanding

the workpapers to include "people" information consisting of an organizational chart and brief

biographies of company employees; 2) a "review notes" textbox frame located on the top

portion of every screen replacing the review note tablet in the M&P study; 3) tickmarks

modified to ones supported by a keyboard; and 4) the addition of separate screens for the

"people information." Dates were also changed to make the workpapers current. The original

workpapers were designed to contain a number of contradictions and errors and these remained

in the online version. Appendix A contains an example of both a contradiction and an error that

were in the workpapers.

       As a result of pilot testing, several changes were made to the workpapers: 1) inclusion

of a screen indicating specific inventory testwork that was performed, which replaced physical

copies of eight invoices that were included in the original workpapers; 2) modification of the

instructions from one long screen to three short screens which are easier to read online; and 3)

dividing the permanent file screen from one long to three shorter screens for the same reason.

A second pilot test indicated that these changes were effective and gave preliminary indication

that participants in the treatment conditions were responding as predicted.

       Participants completed the task by accessing the online site, examining the workpapers

and writing review notes which were captured using a CGI Script. An analysis of the

participant's workpaper review notes provides evidence regarding detection of contradictions,

detection of errors, and information search queries. In addition, unknown to the participants,

the program captured the screens examined, and the amount of time spent on each screen. This

information allowed analysis of the information search behavior and the scrutiny of information

from and about people.

       To examine the participant's generation of alternatives, three brief scenarios were

presented to the participants at the conclusion of the workpaper review task. Each participant

was asked to generate written alternative explanations for each scenario (Table 8), and an

analysis of the participant's answers to the scenarios provides evidence regarding alternative


Data Analysis and Results

              Prior to beginning statistical analysis, the alternative generation, identification of

    contradictions and information search variables were coded by two independent raters who

    were blind to the treatment conditions and skepticism levels of the participants. Inter-rater

    reliability was greater than .90 for the coding on contradiction detection, and greater than .85

    for alternative generation coding and information search coding. Differences were resolved by

    the researchers. Both the amount of time spent1 and the number of screens the auditor

    examined are used to measure general information search and the scrutiny of information

    from and about people.

              To obtain time spent and number of screens information, the raw data captured by the

    CGI Script was summarized by screens and time. An initial review of the time spent indicated

    that, although participants were instructed to complete the entire review without interruptions,

    in some cases it appeared that participants had been interrupted. The data was winsorized and

    all analyses use this winsorized sample data.2 Descriptive statistics for all variables are

    presented in Table 1.

                                                  Insert Table 1 about here
    Effect of skepticism

              Our initial test was a MANOVA …

               Participants were not aware that time or search pattern was being captured.
 For example, participants spent (on average) approximately two minutes on workpaper APF-3 (a single text page
containing information on client personnel). However, one participant (who wrote no review notes while this
workpaper was on his or her screen) spent over 115 minutes before exiting this screen. Congruent with the
procedure described by Winer (1971) on how to handle influential outliers, the sample was winsorized at g = 1,
where the highest and lowest time values were replaced by the second highest and second lowest time values.

           with all Table 13 presents the results of a series of ANOVA procedures used to test for

    the effect of skepticism. Consistent with philosophical predictions, skepticism level was a

    significant (p<.05, n = 75) predictor of the auditor's detection of contradictions.

    Contradictions are the type of errors that are most critical for an auditor to identify and include

    elements such as contradictory or inconsistent client explanations. Increased identification of

    this type of error should result in more effective audits. Skepticism level was not a significant

    predictor of increased information search or increased scrutiny of information about people.

    Neither was it significant in the number or type of alternatives generated.3 4

                                               Insert Table 1 about here

    Planned comparison with low risk and low skepticism level

           Table 15 presents the results of a series of ANOVA procedures performed prior to

    completing the planned comparison between the low skepticism, low risk condition group and

    the remaining three groups. This analysis examines whether differences exist among the

    groups. Detection of contradictions remains significant as does the scrutiny of information

    from and about people, which was measured by the number of people screens examined.

                                               Insert Table 1 about here

  Additional analyses were completed using regressions run with the participant's raw skepticism score. This did
not result in any meaningful changes in the results. The number of contradictions detected remained significant (t
= 2.666, sig t < .009; adj R2 = .076).
  Four auditors never examined the trial balance or subsequent event workpapers. This is significant because of
the nine contradictions present in the workpaper, six of them have part of the contradictory information on either
the trial balance or subsequent event workpapers. Participants who did not review these workpapers did not have
the information to detect the majority of the contradictions. Three of the four auditors who did not examine these
workpapers were in the high skepticism category. We reanalyzed the data after eliminating these four individuals.
The new analysis resulted in only one significant change: the number of mechanical errors detected was different
between the two treatment conditions.


       The results of the planned comparison are shown in Table 16. The number of

information search queries written in the review notes is significantly different between the

two groups, as is the number of people screens examined and the time spent on these.

However, the detection of contradictions or generation of alternatives is not significantly

different between the groups.

                                          Insert Table 1 about here


       This research examines the important issue of whether professional skepticism affects

four behaviors in workpaper review. a term widely used in auditing literature but not clearly

defined or well understood. The purpose of this study was test the theoretical predictions of

skeptical behaviors utilizing the instrument to differentiate between more and less skeptical

individuals. Following is a brief discussion of these three areas that highlights overall

contributions, areas of potential limitations, and future research.

       The model predicts that skeptics will exhibit four behaviors: expanded information

search, detection of contradictions, generation of alternatives, and scrutiny of information

from and about people. Using the Skepticism Scale score obtained from an on-line pretest to

differentiate between those who were more and less skeptical, the auditors in this study

completed an on-line workpaper review task and then responded to scenarios asking them to

generate alternatives.

        The most significant finding from the experiment is that more skeptical auditors find

more contradictions. More skeptical auditors also demonstrate a stable search pattern across

risk situations. This is consistent with my expectations that more skeptical auditors maintain a

level of skeptical behavior in all situations and do not react to changing risk conditions with

radical changes in behavior.

        In contrast, less skeptical auditors, when placed in a high risk audit situation, respond

by generating more information search queries, spending more time evaluating the

information and looking at more information from and about people; however, this additional

work does not result in more contradictions detected. This suggests that although less

skeptical auditors recognize the risk present in a high risk situation, they are not effective in

translating that awareness of risk into better results.

        The expanded information search behavior that was predicted for more skeptical

auditors was not evident in the number of screens examined or the time spent on those screens.

Upon reflection, the nature of the experimental task, a workpaper review which requires that

the participants "sign-off" on each workpaper, created a bias against finding significant

differences in search behaviors. Auditors are trained to review and sign each of these

workpapers which limited the availability of discretional additional information for expanded

information search.

        The behavioral model predicts that more skeptical auditors will generate more

alternative explanations and a greater number of different types of alternative explanations.

Neither of these behaviors was evident from the experiment. Once again, the nature of the

experimental task, where the participants were answering brief scenarios after completion of a

workpaper review, might have biased against finding significant differences. Additionally, the

mean time spent on this experiment was over seventy-five minutes, and at the end of that

amount of time, many participants might have been too exhausted to give complete attention

and provide alternatives to these scenarios.

       In retrospect, a workpaper review task is not one where a large number of alternatives

are generated. On the other hand, auditors performing analytic review procedures are required

to generate explanations for variances. Measuring an auditor's skepticism level before he or

she performs an analytic review task and using performance on that task to test the model's

prediction of alternative generation might be a more appropriate task.

       This study also contributes by using a new method of delivering the experimental

materials to the participants. The experiment conducted for this study was administered

entirely on-line using World Wide Web-based technology. The on-line nature of this

experiment allowed auditors to complete the workpaper review task at a time and place of

their choosing, recreating the most realistic audit environment. This type of experimental

design offers one alternative to expensive and time-consuming on-site experiments.


Bamber, E. M. and R. J. Ramsay. 2000. The effects of specialization in audit workpaper review
      on review efficiency and reviewers' confidence. Auditing: A Journal of Practice and
      Theory 19 (2):147-157.

Bamber, E. M. and R. J. Ramsay. 1997. An investigation of the effects of specialization in audit
      workpaper review. Contemporary Accounting Research 14 (Fall): 501-513.

Harding, N. and K. T. Trotman. 1999. Hierarchical differences in audit workpaper review
      performance. Contemporary Accounting Research 16 (Winter): 671-684.

Libby, R. and K. T. Trotman. 1993. The review process as a control for differential recall of
       evidence in auditor judgments. Accounting Organizations and Society 18 (August):

Messier, W. F., and R. M. Tubbs 1994. Recency effects in belief revision: the impact of audit
       experience and the review process. Auditing: A Journal of Practice and Theory 13

Moeckel, C. L. 1990. The effect of experience on auditors' memory errors. Journal of
      Accounting Research 28 (Autumn): 368-387.

Ramsay, R. J.. 1994. Senior/manager differences in audit workpaper review performance.
      Journal of Accounting Research 32 (Spring): 127-135.

Sprinkle, G. B. and R. M. Tubbs. 1998. Effects of audit risk and information importance on
       auditor memory during working paper review. The Accounting Review 73 (October):

Trotman, K. T. 1985. The review process and the accuracy of auditor judgments. Journal of
      Accounting Research 23 (Autumn): 740-752.

____, and P. Yetton. 1985. The effect of the review process on auditor judgments. Journal of
       Accounting Research 23 (Spring): 256-267.

                                                             Table 1

                                                  Subject Demographic Statistics
                         (Mean values are presented first, followed by median values (italicized)
                                         with standard deviation in parentheses)

                                          Risk Condition Low                       Risk Condition High

                                        Low               High                 Low           High Skepticism
        n=83                         Skepticism         Skepticism          Skepticism          (Group 4)
                                      (Group1)           (Group 2)          (Group 3)                n=20
                                        n=19                n=22               n=22

        Age                             28.8a               27.8a              26.6a                 28.8a
                                         28.0               28.0                26.0                 28.5
                                        (2.9)               (2.0)              (2.4)                 (3.1)

        Experience in                    62.6               62.8                48.6                 59.7
        Months                           52.0               61.5                47.5                 56.0
                                        (20.7)             (23.5)              (12.9)               (34.1)

        Skepticism Score               128.6**            149.6**             128.2**               147.7**
                                        129.0              148.5               129.0                 146.0
                                        (6.4)               (7.6)              (7.2)                 (7.8)

        Gender                          10 M                13 M               10 M                  12 M
        (M = Male;                       9F                  9F                12 F                   7F
        F = Female)                                                                            1 no response

     Significant difference exist between group 3 and groups 1 and 4, p < .05 (t-test)
**Significant difference between the high and low skepticism cell means (as planned), p < .001 (t-test)

                                                Table 1

                                         Descriptive Statistics

                  (Mean values are presented with standard deviation in parentheses.)

                                        Risk Condition Low                Risk Condition High

                                        Low               High            Low             High
                                    Skepticism        Skepticism      Skepticism        Skepticism

Information search measures            n=19               n=22           n=22             n=20

Workpaper Screens                       51.9              52.2            53.6             54.6
   Count                               (21.5)             (24.6)         (25.7)           (27.1)

Workpaper Screens                       76.6              75.6            74.7             79.3
   Time in minutes                     (34.5)             (37.9)         (22.1)           (25.6)

Information search queries              .47                1.9            1.6              2.1
                                        (1.0)             (3.1)           (2.1)           (3.1)

Contradiction detection
measures                               n=19               n=18           n=19             n=19

Contradictions                          1.2                1.6            1.3              2.5
                                        (1.4)             (1.1)           (1.2)           (1.8)

Errors                                  .74                .61            1.1              1.2
                                        (.65)             (.98)           (.85)           (1.07)

Alternative generation
measures                               n=19               n=18           n=19             n=19

Total number of alternatives            2.4                3.1            3.4              2.7
                                        (2.5)             (1.5)           (2.3)           (1.6)

Different types of alternatives          1.8               2.6            2.5              2.3
                                        (1.8)             (1.1)           (1.5)           (1.3)

People search measures                 n=19               n=22           n=22             n=22

People Screens                           .8                2.8            4.8              2.5
   Count                                (1.1)             (4.1)           (5.7)           (3.4)

People Screens                           .6                2.2            3.2              1.8
   Time in minutes                      (1.4)             (3.5)           (4.5)           (3.2)

                                                    Table 3

                                   ANOVA Results: Effect of Skepticism Level

Panel A: Information Search (Hypothesis 1a)

    Effect                                                     SS            F-Value      Sig. of F

    Workpaper screens examined                                       5.271         .009           .926

    Workpaper time (in minutes)                                     63.446         .069           .793

    Number of information search queries (n = 74)                   17.773        2.916           .092

Panel B: Detection of Contradictions (Hypothesis 1b)

    Effect                                                     SS            F-Value      Sig. of F

    Total logic errors detected                                     11.705        5.703           .020

    Total mechanical errors detected                                  .001         .000           .989

Panel C: Generation of Alternatives (Hypothesis 1c)

    Effect                                                     SS            F-Value      Sig. of F

    Total alternatives generated                                      .005         .001           .972

    Total different types of alternatives                             .917         .128           .510

Panel D: Scrutiny of Information from and about people (Hypothesis 1d)

    Effect                                                     SS            F-Value      Sig. of F

    People screens examined                                          2.297         .128           .722

    People time (in minutes)                                        .00006         .000           .998

                                                  Table 4

                                        ANOVA Among all Four Groups

Panel A: Information Search

    Effect                                                   SS             F-Value     Sig. of F

    Workpaper screens examined                                     97.457        .053               .984

    Workpaper time (in minutes)                                   249.351        .088               .966

    Number of information search queries (n=74)                    29.486       1.611               .195

Panel B: Detection of Contradictions

    Effect                                                  SS              F-Value     Sig. of F

    Total logic errors detected                                    19.523       3.253               .027

    Total mechanical errors detected                                3.711       1.529               .214

Panel C: Generation of Alternatives

    Effect                                                   SS             F-Value     Sig. of F

    Total alternatives generated                                    9.973        .773               .513

    Total different types of alternatives                           6.155        .988               .403

Panel D: Scrutiny of Information from and about people

    Effect                                                   SS             F-Value     Sig. of F

    People screens examined                                       172.944       3.547               .018

    People time (in minutes)                                       70.252       2.022               .118

                                                    Table 16

                                   Planned Comparison between Low Skepticism,
                                     Low Risk Condition and the Other Groups

Panel A: Information Search

    Effect                                                         t            df   Significance

    Workpaper screens examined                                         .243     79              .809

    Workpaper time (in minutes)                                        .014     79              .989

    Number of information search queries                            2.112       71              .038

Panel B: Detection of Contradictions

    Effect                                                         t            df   Significance

    Total logic errors detected                                     1.474       71              .145

    Total mechanical errors detected                                   .853     71              .397

Panel C: Generation of Alternatives

    Effect                                                         t            df   Significance

    Total alternatives generated                                    1.159       68              .250

    Total different types of alternatives                           1.585       71              .117

Panel D: Scrutiny of Information from and about people

    Effect                                                         t            df   Significance

    People screens examined                                         2.456       79              .016

    People time (in minutes)                                        2.019       79              .047

                                          Appendix A

Contradiction or                     Statements in Workpapers                             Inference or
 Error example                                                                              Problem

Contradiction      There were strong markets at the       Sales were down                 Were prices
                   end of the prior year. Philips Co.     5.5 %. There are two            increased or
                   had forecast that strong markets       related reasons for this. . .   decreased?
                   would bear both price and volume       . Second, Philips
                   increases. Management decided to       decreased its prices
                   be very aggressive and to do           across the board in both
                   whatever was necessary to take         March and June in order
                   full advantage of the situation. . .   to reduce its inventory
                   Second, to get in on the price         buildup (per schedule A-
                   increases it had predicted would       1, inventory: $205,500
                   result from the stronger market at     and $164,040 in 1998 and
                   the end of FY 1997, Philips Co.        1997), and to shore up
                   changed its pricing policies across    slackening sales.
                   the board.

      Error        Philips Co. has plants located in      This note was retired with      Which plant
                   Stillwater, New Mexico (also its       the proceeds from the           closed?
                   central office) and Coldton,           sale of land and the
                   Minnesota. Philips closed a plant      closing of the Coldton
                   in Sunnyvale, California in            plant.
                   December 1997 when its lease on
                   the building expired.

                                                  Appendix B

     ANormal" Audit Risk Instructions             AHigh@ Audit Risk Instructions

         GENERAL BACKGROUND                                  GENERAL BACKGROUND
             INFORMATION                                            INFORMATION
   Terry Brandon, who prepared the
     workpapers you have received, has          Terry Brandon, who prepared the workpapers you have
     about one year of experience, including      received, has about one year of experience with
     a limited amount of work on the prior        Henderson and Franks, including a limited amount of
     year's audit of Philips Co.                  work on the prior year's audit of Philips Co.

                                                You have no reason to question Terry's basic
                                                  competence, but since Terry was trained by another firm,
   You have no reason to question Terry's       there will probably be minor stylistic differences between
     basic competence.                            your approaches to the same work.

       CLIENT INFORMATION:                                    CLIENT INFORMATION:
   The client is Philips Co., a designer      The client, which is new to you, is Philips Co., a designer
     and manufacturer of water treatment          and manufacturer of water treatment systems.
                                                Your firm acquired Philips Co. as a client on 3/31/98
                                                  when your firm merged with Henderson & Franks, a
   Your firm has audited Philips Co. for        well respected regional CPA firm. Henderson & Franks
     the past seven years, and rendered           has audited Philips Co. for the past seven years, and
     unqualified opinions in all cases.           rendered unqualified opinions in all cases.

To top