TUUKKA JÄRVINEN 
 
Factors Influencing Auditors' 
Information Usage: Experience, Risk, 
Task Structure and Information 
Reliability 
 
 
 
 
 
 
ACTA WASAENSIA NO 272 
________________________________ 
 
BUSINESS ADMINISTRATION 110 
ACCOUNTING AND FINANCE 
 
 
 
 
 
 
 
 
 
 
 
 
 
UNIVERSITAS WASAENSIS 2012 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Reviewers Professor Iris Stuart 
Department of Accounting, Auditing and Law 
Norwegian School of Economics (NHH) 
NO–5045 Bergen 
Norway 
 
Professor Claus Holm 
Department of Economics and Business 
Aarhus University 
DK–8210 Aarhus V 
Denmark 
 
 
 
 III 
Julkaisija  Julkaisupäivämäärä  
Vaasan yliopisto  Marraskuu 2012 
Tekijä(t)  Julkaisun tyyppi  
Tuukka Järvinen Monografia 
Julkaisusarjan nimi, osan numero  
Acta Wasaensia, 272 
Yhteystiedot  ISBN  
Vaasan yliopisto 
Kauppatieteellinen tiedekunta 
Laskentatoimi ja rahoitus 
PL 700 
65101 VAASA 
978–952–476–420–9 
ISSN  
0355–2667, 1235–7871 
Sivumäärä Kieli  
179 Englanti 
Julkaisun nimike  
Tilintarkastajien informaation käyttöön vaikuttavat tekijät: Kokemus, riski,  
tehtävän rakenne ja informaation luotettavuus 
Tiivistelmä  
Tässä tutkimuksessa tarkastellaan tilintarkastajien informaation käyttöä ja tilin-
tarkastuksen tehokkuutta. Tutkimuksen päätavoitteena on tutkia, kuinka yksilöön, 
ympäristöön, tehtäviin ja informaatiopalasiin liittyvät tekijät vaikuttavat tilintar-
kastajien informaation käyttöön yksittäisessä tilintarkastustehtävässä. Tutkimuk-
sessa pyritään valottamaan tilintarkastajien päätöksentekoa pohjautuen näiden 
tekijöiden päivitettyyn ja laajennettuun luokitteluun. 
Empiirisiä analyyseja varten toteutettiin kokeellinen tutkimus internetissä. Tutki-
muksen otoksessa on mukana 271 havaintoa auktorisoiduilta ja auktorisoimatto-
milta tilintarkastajilta sekä tilintarkastuksen opiskelijoilta. Analyyseissa on muka-
na neljä tekijää: tilintarkastajan kokemus, riski, tehtävän rakenne ja informaation 
luotettavuus. Informaation käyttöä toimeksiannon jatkamis- ja hyväksymistehtä-
vissä on mitattu tehtävään käytetyllä ajalla ja informaation määrällä. 
Tutkimuksen päätulokset voidaan tiivistää seuraavasti: Ensiksikin kokeneet tilin-
tarkastajat käyttävät kaiken kaikkiaan informaatiota tehokkaammin kuin vähem-
män kokeneet koehenkilöt. Toiseksi tulokset osoittavat, että korkea riski lisää 
informaatiopalasten yhdistelyyn ja arviointiin käytettyä työmäärää. Kolmanneksi 
tarkastustehtävän rakenteen vaativuus lisää vähemmän kokeneiden koehenkilöi-
den käyttämää työmäärää informaatiopalasten yhdistelyssä ja arvioinnissa. Vii-
meiseksi kokeneiden tilintarkastajien tehokkuusedut informaation käytössä rajoit-
tuvat vain olosuhteisiin, joissa informaatio on luotettavaa. 
Kaiken kaikkiaan tutkimuksen tulokset tuovat uutta tietoa tilintarkastuksen tehok-
kuuteen vaikuttavista seikoista ja sitä voidaan käyttää myös käytäntöjen, kuten 
aikabudjetoinnin ja tilintarkastajien koulutuksen, kehittämiseen. 
Asiasanat  
tilintarkastajien päätöksenteko, informaation käyttö, kokemus, riski, tehtävän ra-
kenne, informaation luotettavuus 
 

 V 
Publisher  Date of publication  
Vaasan yliopisto  November 2012 
Author(s)  Type of publication  
Tuukka Järvinen Monograph 
Name and number of series  
Acta Wasaensia, 272 
Contact information ISBN  
University of Vaasa 
Faculty of Business Studies 
Accounting and Finance 
P.O. Box 700 
FI–65101 VAASA, Finland 
978–952–476–420–9 
ISSN  
0355–2667, 1235–7871 
Number 
of pages 
Language 
179 English 
Title of publication  
Factors Influencing Auditors' Information Usage: Experience, Risk, Task  
Structure and Information Reliability 
Abstract 
This thesis investigates auditors' information usage and audit efficiency. In par-
ticular, the main objective of the thesis is to examine how individual, environ-
mental, task-related and cue-related factors affect auditors’ information usage in 
a single audit task. Based on a revised and expanded taxonomy of these factors, 
this study sheds further light on auditors’ decision-making in this context.  
For the empirical analysis, a web-based experiment was conducted. The sample 
consists of 271 observations from certified and non-certified auditors and audit-
ing students. Four factors are included in the analysis: auditor experience, risk, 
task structure and information reliability. Information usage in client continuance 
and acceptance tasks is measured by the time spent on the task and number of 
used information cues.  
The main results can be briefly summarized as follows. First, experienced audi-
tors are, on the whole, more efficient in their information usage than less experi-
enced subjects. Second, the results indicate that high risk increases the effort 
used in combining and evaluating the information cues. Third, the unstructured-
ness of a task increases less experienced subjects' effort in the combination and 
evaluation of the information cues. Finally, the efficiency advantage of the more 
experienced auditors in the information usage is limited to circumstances where 
information is reliable. 
Overall, the findings of this thesis contribute to the audit literature by providing 
new evidence on audit efficiency. This information can be used to improve audit 
practices, such as time budgeting and auditor training. 
Keywords 
auditors' decision-making, information usage, experience, risk, task structure, 
information reliability 

 VII 
ACKNOWLEDGEMENTS  
At the end of this long project, the word that best describes my feelings is “final-
ly”. A very little is left from the original research plan, now that this thesis has 
evolved  to  its  final  form.  During  this  project,  a  number  of  people  have  encour-
aged and supported me in several ways. Now, I want to thank all those people 
who have influenced my doctoral studies the most and the institutions whose fi-
nancial support made this project possible. 
I  owe a  great  depth  of  gratitude  to  the  supervisor  of  this  thesis,  Professor  Teija  
Laitinen, for encouraging me to pursue doctoral studies in 2004. Since then, we 
have worked closely together and we have had many lively conversations about 
the research, studies and about almost everything. These sessions have encour-
aged me to carry on whenever I have felt confused about my work. I am also very 
grateful for her comments and suggestions that have contributed especially to my 
thesis’s theoretical part and my research instrument. 
I am also extremely grateful to my two pre-examiners, Professor Iris Stuart of 
Norwegian School of Economics NHH and Professor Claus Holm of Aarhus Uni-
versity for their valuable comments on the manuscript. I wish to express my grati-
tude to Professor Erkki K. Laitinen for his help on several occasions. His exten-
sive knowledge and creative thinking has generated valuable pieces of advice that 
have helped me to overcome many challenging issues that arose during this re-
search. I also want to thank Dr. Mikko Zerni for his excellent guidance on meth-
odological issues and those many comments that have improved my thesis signif-
icantly. I thank him for sharing his expertise and also his friendship. 
I wish to thank all the professors and colleagues in the Department of Accounting 
and Finance at the University of Vaasa. I feel privileged to have had the possibil-
ity to work with all these talented people. Seeing different views on doing scien-
tific research has taught me greatly.  I would like to thank Professor Emeritus 
Timo Salmi for many precious pieces of advice concerning the doctoral studies 
and the research in general.  I am also grateful to Professors Stefan Sundgren and 
Seppo Pynnönen for their comments and suggestions. I want to thank Professors 
Allen Blay and Gustav Lundberg for providing me with their outline of experi-
ments. I would like to thank my colleagues who have already graduated, Dr. Jo-
hanna Kurtti, Dr. Kim Ittonen and Dr. Annukka Jokipii, for helping me in my 
doctoral studies in many ways and also for their encouragement. I am also grate-
ful to Ms. Emma-Riikka Myllymäki for many useful comments. 
During the years, the thesis has been financially supported by several foundations 
and organizations. I am especially grateful to the South Ostrobothnia Regional 
VIII 
Fund and the Evald and Hilda Nissi Foundation, which enabled me to work as a 
full-time doctoral student for two years. I also gratefully acknowledge the gener-
ous financial support from the Jenny and Antti Wihuri Foundation, the Ostro-
bothnia Chamber of Commerce, the Finnish Foundation for Economic and Tech-
nology Sciences (KAUTE), Evald and Hilda Nissi Foundation, the Finnish Sav-
ings Banks Research Foundation, the Foundation for Economic Education, the 
Oscar Öflund Foundation, the Marcus Wallenberg Foundation, the Foundation for 
Promoting Equity Markets in Finland (SAE) and Academy of Finland (grant 
126630). 
Different versions of this thesis have been presented at several workshops and 
seminars. I have received valuable advice and comments that have helped me to 
improve my thesis. I would like to thank the participants of the Conference in 
Accounting and Performance Management Perspectives in Business and Public 
Sector Organizations (Tartu, September 2005), the USBE Workshop on Auditing 
and Financial Accounting Research (Umeå, January 2007), EARNet PhD work-
shop (Aarhus, October 2007), Doctoral Tutorial in Accounting (Vaasa, May 
2008), Workshop on Auditing and Financial Accounting Research (Vaasa, June 
2008), Workshop on Auditing and Financial Accounting Research (Vaasa, Febru-
ary 2010), and the Doctoral Tutorial in Accounting (Tampere, August 2011). 
They have provided me with many helpful comments. I am especially grateful to 
Professor  Aasmund  Eilifsen,  Professor  Mervi  Niskanen  and  Professor  Aila  Vir-
tanen for their valuable comments and suggestions. 
Finally, I would like to thank my family and friends for their strong support and 
encouragement. They have given me other things to think about when I have 
needed it  the most.  I  would like to give special  thanks to my father,  Matti,  who 
has inspired me to think abstractly ever since I was a little boy. Finally, I would 
like to express my deepest gratitude to my wife Marjaana. Without her uncondi-
tional love and support this project would have not been possible. 
 
 
Vaasa, October 2012 
Tuukka Järvinen 
 
 
 IX 
CONTENTS   
ACKNOWLEDGEMENTS ............................................................................. VII 
1 INTRODUCTION ........................................................................................ 1 
1.1 Research problem .............................................................................. 3 
1.2 Contribution of the study and main findings ....................................... 6 
1.3 Structure of the study ......................................................................... 9 
2 AN AUDITOR’S DECISION-MAKING AND  INFORMATION 
ACQUISITION AND USAGE ................................................................... 11 
2.1 An auditor’s information acquisition and usage in  the decision- 
making process ................................................................................ 11 
2.2 Process tracing approaches in previous studies ................................. 16 
3 AUDITORS’ INFORMATION USAGE  FRAMEWORK .......................... 18 
3.1 Individual factors ............................................................................. 19 
3.1.1 Auditor experience........................................................... 19 
3.1.2 Motivated reasoning ........................................................ 22 
3.1.3 Confirmation bias ............................................................ 24 
3.1.4 Confirmation bias and auditor experience ........................ 25 
3.2 Environmental factors ...................................................................... 25 
3.2.1 Accountability ................................................................. 25 
3.2.2 Time pressure .................................................................. 27 
3.2.3 Client risk ........................................................................ 29 
3.3 Task-related factors .......................................................................... 30 
3.3.1 Task complexity .............................................................. 31 
3.3.2 Task framing ................................................................... 32 
3.3.3 Task type ......................................................................... 33 
3.3.4 Task response mode ......................................................... 34 
3.4 Cue-related factors ........................................................................... 34 
3.4.1 Information order ............................................................. 35 
3.4.2 Information reliability ...................................................... 36 
3.4.3 Presentation mode............................................................ 38 
3.4.4 Irrelevant information ...................................................... 39 
3.4.5 Information types ............................................................. 40 
3.5 Interactions ...................................................................................... 42 
3.5.1 Interactions and individual factors ................................... 42 
3.5.2 Interactions and environmental factors ............................. 47 
3.5.3 Interactions and task-related and cue-related factors......... 49 
3.6 Conclusions about the studies .......................................................... 49 
4 FACTORS AND HYPOTHESES DEVELOPMENT.................................. 52 
4.1 Factors for the empirical analysis ..................................................... 52 
4.2 Hypotheses development.................................................................. 55 
4.2.1 Hypothesis 1 – Auditor experience .................................. 55 
4.2.2 Hypothesis 2 – RMM ....................................................... 57 
X 
4.2.3 Hypothesis 3 – Task structure ........................................... 59 
4.2.4 Hypothesis 4 – Information reliability .............................. 60 
4.2.5 Hypothesis 5 – RMM and auditor experience ................... 62 
4.2.6 Hypothesis 6 – Task structure and auditor experience ....... 63 
4.2.7 Hypothesis 7 – Information reliability and auditor  
experience ........................................................................ 64 
5 EXPERIMENT AND DATA ...................................................................... 66 
5.1 Client acceptance and continuance tasks ........................................... 66 
5.2 Information cues ............................................................................... 70 
5.3 Web-based experiment ..................................................................... 73 
5.4 Experimental design ......................................................................... 74 
5.5 Subjects and experimental procedures ............................................... 79 
5.6 Dependent variables.......................................................................... 80 
5.7 Independent variables ....................................................................... 82 
5.8 Data and exclusion of outliers ........................................................... 83 
5.8.1 Descriptive statistics of subjects ....................................... 85 
5.8.2 Descriptive statistics of the dependent variables ............... 87 
5.8.3 Descriptive statistics of cue usage ..................................... 88 
6 METHODOLOGY AND RESULTS ........................................................... 94 
6.1 Manipulation check .......................................................................... 94 
6.2 Univariate analyses ........................................................................... 94 
6.3 Multivariate analyses ...................................................................... 103 
6.3.1 Continuous dependent variables – ANOVA .................... 104 
6.3.2 Discrete dependent variables – ordered logistic  
regression ....................................................................... 112 
6.4 Supplementary analyses and robustness tests .................................. 118 
6.5 Analyses of independent variables effect on task-specific  
judgments ....................................................................................... 120 
6.5.1 Effect of information usage on task-specific judgments .. 124 
6.6 Summary and discussion of the results ............................................ 125 
7 CONCLUSIONS ....................................................................................... 130 
7.1 Summary of the study and practical implications ............................ 130 
7.2 Suggestions for future research ....................................................... 135 
7.3 Limitations ..................................................................................... 136 
REFERENCES ................................................................................................ 139 
APPENDICES ................................................................................................ 153 
 
  
 XI 
  
List of Figures 
Figure 1. Individual, environmental, task-related and cue-related factors 
influence task-specific information acquisition and usage as well as 
task-specific judgment (adopted from El-Masry & Hansen 2008) ... 4 
Figure 2. Independent and dependent variables of the study ........................... 6 
Figure 3. The diagnostic, sequential and iterative decision-making process 
(adopted from Moroney 2007) ...................................................... 13 
Figure 4. Graphical illustration of hypothesis 5 ............................................ 63 
Figure 5. Graphical illustration of hypothesis 6 ............................................ 64 
Figure 6. Graphical illustration of hypothesis 7 ............................................ 65 
Figure 7. Information classes of client continuance and acceptance tasks ..... 71 
Figure 8. Treatment groups of the study ....................................................... 75 
Figure 9. Overview of the research instrument ............................................. 78 
Figure 10. Distribution of observations between treatment groups ................. 85 
Figure 11. Interaction plot – estimated marginal means for experience and 
information reliability ................................................................. 108 
Figure 12. Interaction plots – estimated marginal means for experience, risk  
and structure ............................................................................... 109 
Figure 13. Interaction plots – estimated marginal means for experience and 
structure/information reliability .................................................. 110 
Figure 14. Interaction plots – estimated marginal means for risk, structure  
and information reliability .......................................................... 111 
Figure 15. Non-linear increase in judgment time depending on the number  
of uncertainty factors .................................................................. 128 
 
  
XII 
List of Tables  
Table 1. Cues of the information menu .................................................. 73 
Table 2. Descriptive statistics of subjects ............................................... 86 
Table 3. Descriptive statistics of the dependent variables ....................... 87 
Table 4. Descriptive statistics of cue usage ............................................ 89 
Table 5. Descriptive statistics of cue usage – auditor experience ............ 90 
Table 6. Descriptive statistics of cue usage – RMM ............................... 91 
Table 7. Descriptive statistics of cue usage – task structure .................... 92 
Table 8. Descriptive statistics of cue usage – information reliability ...... 93 
Table 9. Univariate tests of hypothesis 1 ................................................ 95 
Table 10. Univariate tests of hypothesis 2 ................................................ 97 
Table 11. Univariate tests of hypothesis 3 ................................................ 98 
Table 12. Univariate tests of hypothesis 4 ................................................ 99 
Table 13. Univariate tests of hypothesis 5 .............................................. 100 
Table 14. Univariate tests of hypothesis 6 .............................................. 101 
Table 15. Univariate tests of hypothesis 7 .............................................. 103 
Table 16. ANOVA tables for continuous variables ................................ 106 
Table 17. Ordered logistic regression analyses for discrete variables ..... 114 
Table 18. Coefficients of discrete variables with TREATMENT and  
INEXP as the independent variables....................................... 117 
Table 19. Ordinal regression models estimated to test how  
the independent variables affect task-specific judgments in  
semi-structured and unstructured tasks ................................... 122 
Table 20. Summarized results of the hypotheses testing ......................... 132 
 
  
 XIII 
  
List of abbreviations 
ANOVA Analysis of Variance 
CPA Certified Public Accountant 
HTM Certified auditor approved by the Auditing Committee of regional 
 Chamber of Commerce (in Finnish: Hyväksytty tilimies) 
JDM Judgment and decision-making 
JHTT Certified auditor approved by the Finnish Board for Chartered Public 
 Finance Auditors (in Finnish: Julkishallinnon ja -talouden tilin- 
tarkastaja) 
KHT Certified auditor approved by the Auditing Board of the Central 
Chamber of Commerce (in Finnish: Keskuskauppakamarin 
hyväksymä tilintarkastaja) 
RMM Risk of Material Misstatement 
 
 
  
 
  
1 INTRODUCTION  
Auditors are surrounded by an increasing amount of information in the current 
global environment. Modern audit approaches such as business risk auditing and 
strategic systems auditing emphasize the use of complex information from many 
different  sources  for  holistic  risk  assessments  (Bell,  Peecher  &  Solomon  2005;  
Peecher, Schwartz & Solomon 2007). Numerous databases and complex business 
networks provide multiple information sources to acquire information for deci-
sion-making (Trotman 2005). At the same time, increased competition between 
audit firms (see Dunn, Kohlbeck & Mayhew 2011) and cost pressures induce au-
ditors to emphasize efficiency in audit work. Consequently, information acquisi-
tion and usage in the decision-making should be performed in an efficient man-
ner, that is, in a short time, yet without jeopardizing the audit effectiveness. At the 
core of this issue is an individual auditor who performs information acquisition 
and usage among various audit tasks as the audit process progresses towards an 
audit report. 
The auditing standards (e.g. ISA 500) require an auditor to obtain sufficient 
amount of appropriate information (evidence) in order to draw reasonable conclu-
sions on which to base the final audit  opinion. Before an auditor is  able to give 
such a written opinion to shareholders, he/she must conduct numerous audit phas-
es and subphases, which include accomplishing a great number audit tasks.  An 
auditor’s decision-making process in accomplishing a single audit task includes 
three phases: information acquisition, information usage and judgment (Einhorn 
& Hogarth 1981). Before a task-specific judgment can be made, the information 
acquisition and usage phases consist of several elements that require professional 
judgment.  For  example,  an  auditor  must  consider  which  type  of  information  to  
acquire as well as the amount of information and its cost. These judgments are 
presumably routine procedures if important information is defined clearly in audit 
manuals and the circumstances surrounding the judgment are normal. However, 
the opposite setting may threaten both audit effectiveness and efficiency. 
When judgment is required for information acquisition and usage, audit effective-
ness may be threatened if an auditor does not acquire sufficient appropriate in-
formation. Underacquiring information does not help a decision-maker make an 
accurate  decision  (Connolly  &  Thorn  1987).  For  example,  if  an  auditor  fails  to  
gather all relevant information by overlooking important pieces during the acqui-
sition phase, it may lead to a suboptimal judgment/decision (Abdel-Khalik & El-
Sheshai 1980; Simnett & Trotman 1989). Ultimately, a failure in information ac-
quisition and usage when making going-concern judgments may cause a wrong 
audit report to be issued (Bonner 2008). 
2      Acta Wasaensia 
Audit efficiency might also be decreased when an auditor overacquires infor-
mation or uses an extensive amount of time for the acquisition or usage of infor-
mation. Overacquiring may occur when a decision-maker gathers too much of the 
available information and therefore incurs additional costs in gathering infor-
mation that adds little value to the final decision (Connolly & Thorn 1987). Thus, 
audit effectiveness might be decreased if, as a result of overacquiring, redundant 
and/or irrelevant information is used in the decision-making (see Joe 2003; 
Hackenbrack 1992). The use of an extensive amount of time can happen when an 
auditor does not believe that her/his client’s internal information is sufficient or 
appropriate for making an accurate decision. The acquisition of elusive infor-
mation from external sources may incur additional auditing costs. 
Auditors’ information usage is an important and timely topic. First, it is suggested 
that further research is warranted for investigating whether recent audit scandals 
are partly caused by auditors’ poor information usage (El-Masry & Hansen 2008). 
Second, in high information environments and increasingly larger auditees, indi-
vidual auditors’ information acquisition and processing skills are important in 
detecting misstatements since it is not possible to review all available information 
(Hammersley 2006). Despite the topic’s inherent importance, previous research 
has left many avenues unexplored. While previous studies (see El-Masry & Han-
sen 2008 for a review) have recognized several factors that affect auditors’ infor-
mation acquisition and usage, many factors have not been studied, especially at an 
audit task level. 
For instance, previous studies (e.g. Anderson, Koonce & Marchant 1994; Good-
win 1999) have found that auditors adjust their task-specific judgments according 
to information reliability and make more conservative judgments when the avail-
able information is less reliable. However, they have not addressed whether in-
formation reliability affects the number of information gathered or the processing 
efficiency of information (e.g. time used) when auditors make judgments. Fur-
thermore, it is not known whether some less studied factors interact with common 
factors (e.g. auditor experience), which have been previously found to be im-
portant determinants of information usage. Thus, there is a considerable research 
gap in this area, since these factors may have important implications on audit effi-
ciency. 
 Acta Wasaensia     3 
  
The present study investigates how different factors affect auditors’ information 
usage behavior in a single audit task1. This study investigates previously recog-
nized factors by applying empirically El-Masry and Hansen’s (2008) taxonomy of 
different factors. The ultimate aim of the study is to add knowledge on auditors’ 
information usage as well as help improve audit efficiency practices. 
1.1 Research problem 
The objective of the study is to examine how individual, environmental, task-
related and cue-related factors affect auditors’ information usage in a single audit 
task. In this study, information usage is measured by the time spent on the task 
and by the number of used cues in two different audit tasks, namely client contin-
uance and acceptance. This study focuses on investigating information usage in 
single audit tasks as opposed to many previous studies that have examined audi-
tors’ information or evidence acquisition in audit planning. 
The categorization of the factors is based on El-Masry and Hansen’s (2008) litera-
ture review of the major factors that influence auditors’ information acquisition 
and usage. In their study, they proposed a taxonomy that classifies factors into 
four categories according to their properties. As illustrated in Figure 1, these cate-
gories are individual factors, environmental factors, task-related factors and cue-
related factors. The theoretical framework suggests that these factors have direct 
effects and/or interaction effects on auditors’ information acquisition and usage.  
In the present study, one factor from each category is selected to represent that 
category in the empirical analyses. While it would be possible to study any com-
bination of the factors, this selection aims to capture the factors that have (i) gath-
ered much less attention at an audit task level and/or (ii) important and previously 
identified factors in the auditing context that have meaningful theoretical interac-
tions with the factors mentioned in section (i). 
From the individual factors, the category auditor experience is chosen for the 
empirical analysis. Experience is the most studied factor in this category and it 
has been found to significantly affect information usage. A vast number of previ-
                                               
 
1 This study specifically focuses on examining auditors’ information usage, namely the number 
of information or time that is used by subjects to read/assimilate and process the information. 
However, previous studies have also examined more directly the acquisition (selection, 
choice) of information, thus the term “information acquisition and usage” is used when refer-
ring to all previous studies in the research area. 
4      Acta Wasaensia 
ous studies (e.g. Biggs & Mock 1983; Bonner 1990; Bonner & Pennington 1991; 
Davis 1996; Moroney 2007) have indicated that experienced auditors’ better 
knowledge and advanced acquisition strategies help them perform information 
usage more efficiently compared with less experienced auditors. Previous studies 
(e.g. Kennedy 1993; Shelton 1999; Cianci & Bierstaker 2009) have also shown 
that experience interacts with other factors and that these interactions reduce cog-
nitive bias as well as improve judgment and decision-making (JDM) quality. 
Thus, these previous findings and the general cognitive differences between expe-
rienced and less experienced auditors suggest that it is important to study if the 
effects of experience affect other investigated factors. 
  
 
Figure 1. Individual, environmental, task-related and cue-related factors influ-
ence task-specific information acquisition and usage as well as task-
specific judgment (adopted from El-Masry & Hansen 2008) 
 
From the environmental factors, the category risk of material misstatement 
(RMM) is chosen. Previous research (e.g. Houston, Peters & Pratt 1999; Beaulieu 
2001; Johnstone & Bedard 2001) indicates that auditors adjust their audits de-
pending on the client’s risks (e.g. planned amount of substantive tests), but there 
exists no evidence in the literature on how auditors adjust their information usage 
in a single audit task when risk varies. For instance, high risk may lead auditors to 
acquire a greater number of available information or process the information 
more carefully when performing an audit task. 
 Acta Wasaensia     5 
  
Task structure is chosen from the task-related factors category. Task structure is a 
dimension of task complexity that has rarely been studied in auditing-related in-
formation acquisition and usage studies. Based on the definition of task structure 
(Bonner 1994), it is argued that when task structure is less structured (more struc-
tured), important task-specific information is less specified (more specified). This 
study investigates empirically whether auditors compensate for this increased 
uncertainty in less structured tasks by performing more extensive information 
usage compared with in more structured tasks. 
Previous studies (e.g. Hirst 1994; Anderson, Koonce & Marchant 1994; Goodwin 
1999; Glover, Jiambalvo & Kennedy 2000) have shown that auditors adjust their 
judgments depending on information reliability. However, they have not ad-
dressed whether information reliability affects auditors’ information usage, i.e. 
the number of information used or the processing of information (e.g. used time) 
when auditors form these judgments. This study addresses this gap in the litera-
ture by examining whether information reliability from the category of cue-
related factors influences auditors’ information usage. 
The main research question of the study is: How does (i) auditor experience, (ii) 
RMM, (iii) task structure and (iv) information reliability affect auditors’ infor-
mation usage in a single audit task? Furthermore, interactions are used to study if 
the effect of the factors depends on the levels of other factors. 
Regarding the main research question, this study attempts to answer whether the 
studied factors separately affect auditors’ information usage, and if they do, what 
is the magnitude and direction of this effect (direct effect). Furthermore, it is stud-
ied if the factors’ potential effects are conditional on the levels of other factors 
and the direction of this effect (indirect effect). 
Based on previous accounting and psychology studies, seven hypotheses are de-
veloped. To test these hypotheses, a web-based experiment is conducted. In the 
between-subjects experiment, two audit tasks are used: one semi-structured task 
and one unstructured task. Two of the studied factors, RMM and information reli-
ability, are operationalized by phrase manipulations within the experiment. Audi-
tors’ experience is collected from the post-experimental questionnaire. The sam-
ple consists of Finnish CPAs, non-certified auditors and Master’s level auditing 
students. 
In the main empirical analyses, the analysis of variance (ANOVA) and ordered 
logistic regression methods are applied. Information usage is measured by multi-
ple variables. One variable is used as a dependent variable at the time. There are 
two main variables and four alternative variables. The main dependent variables 
6      Acta Wasaensia 
are total time per task and number of used cues. The alternative variables are total 
cue time, judgment time and number of important cues (self-estimated and care-
fully read). The independent and dependent variables of the study are summarized 
in Figure 2. 
 
 
 
Figure 2. Independent and dependent variables of the study 
 
1.2 Contribution of the study and main findings 
The present study examines auditors' information usage and audit efficiency in a 
single  audit  task.  The  purpose  of  this  study  is  to  relax  the  assumption  that  indi-
vidual, environmental, task-related and cue-related factors have only direct effects 
on information usage. More specifically, four factors are investigated to examine 
whether their potential effects are conditional on the levels of other factors. This 
approach makes it possible to consider how the information usage process affects 
these factors in a broad manner. Furthermore, this approach enables us to gain 
insights into how different factors interact with each other in a mundane decision-
making environment, where auditors are constantly surrounded by a vast number 
of environmental, task-related and cue-related factors at the same time. 
This study makes several contributions to the auditing literature. First, it revises 
the taxonomy of El-Masry and Hansen (2008) of individual, environmental, task-
related and cue-related factors. More specifically, it contributes by expanding this 
 
 
- Total time per task (main) 
- Number of used cues (main) 
- Total cue time 
- Judgment time 
- Number of important (self-estimate) cues 
- Number of important (carefully read) cues 
 
 
 
INDEPENDENT VARIABLES  DEPENDENT VARIABLES 
- Individual factor:  
 Task-specific experience 
- Environmental factor:  
 RMM 
- Task-related factor:  
 Task structure 
- Cue-related factor:  
 Information reliability 
 
 Acta Wasaensia     7 
  
original taxonomy with interaction effects between factors from the different cat-
egories. 
Second, this study contributes empirically to the existing audit JDM literature by 
investigating the three factors from different categories (El-Masry & Hansen 
2008) that have rarely been studied in an auditors’ information usage context, 
namely RMM, task structure and information reliability. Previous studies have 
mainly investigated the effect of these factors on auditors’ task-specific judg-
ments (e.g. Simnett & Trotman 1989; Simnett 1996; Houston, Peters & Pratt 
1999; Beaulieu 2001; Hirst 1994; Goodwin 1999), but not on auditors’ infor-
mation usage per se. This study contributes specifically by examining whether 
and how auditors adjust their information usage behavior in a single audit task 
when the levels of these factors change. 
Third, this study also contributes empirically by examining how the factors from 
different categories (El-Masry & Hansen 2008) interact with each other. One fac-
tor from every category is brought into the simultaneous empirical investigation. 
The aim of this approach is to examine whether these factors’ potential effects on 
information usage are conditional on the levels of other factors. Some of these 
interaction effects are suggested from the auditing and psychology literature, but 
have not yet been studied in the auditing context. 
The main findings of the thesis can be summarized as follows. First, the audit 
expertise literature (e.g. Biggs & Mock 1983; Davis 1996; Moroney 2007) has 
found that experienced auditors generally use less information, apply directed 
information search strategies and are overall more efficient in their information 
usage than less experienced auditors. In this study, auditor experience is measured 
by an auditor’s task-specific experience. This study extends the audit expertise 
literature by finding that task-specific experienced auditors also use less time for 
tasks than less experienced subjects. This study further finds that experienced 
auditors spend less time on cue screens as well as outside of cue screens while 
processing information. These findings suggest that experienced auditors may 
have better reading or assimilation techniques of cues that require less effort than 
the techniques used by less experienced subjects. However, robustness tests sug-
gest that longer cue screen times are only realted to subjects having no or very 
little prior task-specific experience. The findings also suggest that experienced 
auditors are more efficient in evaluating and combining cues (i.e. processing) 
outside of cue screens than less experienced subjects. In summary, these results 
suggest that high task-specific experience leads to better overall efficiency in de-
cision-making. 
8      Acta Wasaensia 
Second, previous studies investigating the effect of risk on auditors’ decision-
making (e.g. Mock & Wright 1993; Houston, Peters & Pratt 1999; Beaulieu 2001) 
have found that auditors adjust their information acquisition plans at an audit en-
gagement level depending on client-related risks. This study extends this stream 
of research by finding that in a single audit task high RMM increases the total 
time for the task and the time used outside of cue screens, but not that used on cue 
screens. This finding suggests that auditors also seem to read and assimilate se-
lected cues in the conventional way in a risky setting, but demonstrate greater 
effort in information processing outside of cue screens when making judgments. 
Third, previous studies have rarely addressed how auditors’ information acquisi-
tion and usage behavior varies depending on the complexity of an audit task (e.g. 
Simnett & Trotman 1989; Simnett 1996). This study contributes to the task com-
plexity literature by finding that the unstructuredness of a task increases the time 
that is used outside of cue screens. This finding suggests that while the quantity of 
used information is not affected by task structure, acquired information is pro-
cessed with a greater effort in unstructured than in semi-structured tasks. Howev-
er, the results indicate significant interactions between task structure and auditor 
experience. These interactions show that only less experienced subjects are af-
fected by task structure, while experienced auditors are not. 
Fourth, previous studies investigating information reliability (e.g. Hirst 1994; 
Anderson, Koonce & Marchant; Goodwin 1999; Glover, Jiambalvo & Kennedy 
2000) have particularly found that under less reliable information auditors tend to 
be more skeptical towards their clients and increase their information gathering in 
subsequent tasks. However, they have not examined whether auditors need more 
information to make judgments or use the same number of information but just 
process it more carefully when it is less reliable. This study extends this stream of 
research by finding evidence that less reliable information is processed for longer 
outside of cue screens, but not on cue screens. This finding suggests that the pro-
cessing of less reliable information is more time-consuming, i.e. less efficient in 
auditors’ decision-making processes, implying that information reliability affects 
audit judgment as well as the efficiency of the preceding decision-making pro-
cess. 
This study further contributes to this literature by finding interaction effects in 
information usage between auditor experience and information reliability. Specif-
ically, it finds evidence that when information is less reliable, experienced audi-
tors do not use less total  time for a task or time outside of cue screens than less 
experienced subjects. This finding suggests that experienced auditors’ more effi-
cient cue processing exists only in conditions where information is reliable. 
 Acta Wasaensia     9 
  
Therefore, with less reliable information even experienced auditors may need to 
engage in more effortful information processing similar to that used by less expe-
rienced subjects. 
Finally, this study further contributes to the information usage literature by find-
ing a significant three-way interaction between information reliability, task struc-
ture and RMM. The results indicate that used time outside of cue screens increas-
es non-linearly as those factors with high values (=uncertainty factor2) increase. 
This shows that after the appearance of one uncertainty factor, used time does not 
increase significantly when the second uncertainty factor appears. However, the 
simultaneous appearance of three uncertainty factors increases used time signifi-
cantly. This finding suggests that auditors do not increase linearly their efforts for 
information processing when the number of uncertainty factors increases. Thus, 
auditors may have only a few different information processing styles when there 
is an indication of a problem in an audit. 
1.3 Structure of the study 
The present chapter starts with the introduction of the research area and motiva-
tion behind the research. After this, the research problem is defined more specifi-
cally. The following section presents the contributions of the study to the existing 
literature and the main findings of the empirical analyses. 
The second chapter presents an overview of auditors’ information acquisition and 
usage processes and explains how they are related to the JDM process. This sec-
tion also presents the two most common process tracing approaches used in this 
research area. 
The third chapter reviews previous studies of auditors’ information acquisition 
and usage. These studies are classified based on El-Masry and Hansen’s (2008) 
taxonomy into four categories. The fifth section presents the interaction studies of 
the different categories. In each section, the factors are ordered by their common-
ness in the literature. The chapter concludes with a discussion of auditors’ infor-
mation acquisition and usage based on the results of previous studies. 
The fourth chapter starts by presenting the factors included in the empirical anal-
yses. These choices are built mainly on the gaps in the audit literature. The specif-
                                               
 
2  The term “uncertainty factor” is used to refer to any of the factors that have a "high value" 
(i.e. less experienced auditor/ high RMM/unstructured task/ less reliable information). 
10      Acta Wasaensia 
ic rationale behind the selection of each factor is explained. In the last section, the 
hypotheses of the study are developed. 
The fifth chapter presents the experiment and the data of the study. In the first 
section, the studied tasks are outlined based on previous research. The second 
section presents the process used to form information cues for the experiment. 
The third section introduces the general advantages and disadvantages of web 
experiments, before the present experimental design, subjects and experimental 
procedures are introduced in detail. In the following section, the dependent and 
independent variables of the study are presented. After this, the sample is de-
scribed as well as the rationale and criteria for excluding some observations from 
the data set. The chapter concludes by presenting extensive descriptive statistics 
of the data. 
The sixth chapter presents the results of the study. This chapter also includes se-
lected non-hypothesized tests that are aimed to further shed light on the role of 
information usage in auditors’ JDM processes. 
The final chapter summarizes the study and reviews the main results from a theo-
retical perspective. In this chapter, the practical implications of the results as well 
as suggestions for future research are discussed. The study concludes with a dis-
cussion of the limitations of the study. 
 Acta Wasaensia     11 
  
2 AN AUDITOR’S DECISION-MAKING AND 
 INFORMATION ACQUISITION AND USAGE 
The purpose of this chapter is to review auditors’ decision-making processes, par-
ticularly from the information acquisition and usage perspective, and to present 
the two most common research approaches used in previous audit studies investi-
gating this research area. 
While there is no unified theory of an individual’s information acquisition and 
usage, in previous studies several decompositions of decision-making processes 
have highlighted the importance of the different phases of this process. The first 
section begins by presenting previous decompositions of decision-making pro-
cesses. The activities related to auditors’ information acquisition and usage are 
then sorted according to Moroney’s (2007) decomposition model, as this is the 
most universal of the presented decompositions. Specifically, the role of problem 
representation and the procedures related to information acquisition are discussed 
in details. The first section concludes by presenting some common biases that 
may distort information acquisition and usage processes and thus cause subopti-
mal judgments or decisions. 
The second section introduces the process tracing approach generally as well as 
the advantages and disadvantages of different process tracing approaches. Two 
common process tracing approaches, verbal protocol analysis and applied com-
puter-based tracking techniques, are introduced in this chapter. 
2.1 An auditor’s information acquisition and usage in 
 the decision-making process 
Over the years, several theories have been developed to understand how individu-
als make decisions. Researchers have decomposed the decision-making process 
into different phases to understand better this sequential process and ultimately to 
find out the determinants of high JDM quality. Several decompositions (e.g. Ein-
horn & Hogarth 1981; Bonner & Pennington 1991; Koonce 1993) have recog-
nized the importance of the information acquisition and usage phases in a deci-
sion-making process and their impact on overall JDM quality. These phases are 
especially suggested to require a considerable number of effort in decision-
making. For instance, Biggs and Mock (1983) found that information usage-
related activities formed a significant proportion (53.4%) of all decision-making 
activities in a complex audit task. 
12      Acta Wasaensia 
Pioneering work by Einhorn and Hogarth (1981) described that the decision-
making process consists of three phases: information acquisition, information 
usage (evaluation) and judgment (action/choice). They especially stressed that 
information acquisition and evaluation (i.e. usage) are interdependent phases that 
should be examined together. Bonner and Pennington (1991) looked closely at 
auditors’ decision-making processes. Their decomposition contains eight phases:  
 
1. Retrieving knowledge from memory 
2. External information search 
3. Comprehension 
4. Hypotheses generation  
5. Evaluation of hypotheses  
6. Design 
7. Estimation  
8. Choice  
 
However, they noted that not all these phases are needed in every audit task, but 
rather in each audit phase (e.g. planning or completing the audit). 
Koonce (1993) outlined the decision-making process of auditors’ analytical re-
views to be a diagnostic, sequential and iterative process. Diagnostic refers to the 
mental representation of the problem, hypothesis generation, information search 
and hypothesis evaluation. Sequential refers to the order of the process and itera-
tive means that some of the components may be re-performed. Moroney (2007) 
proposed a similar model as Koonce (1993) but extended it to cover other audit 
tasks than analytical reviews. In her classification, the decision-making process 
consists of three phases: pre-information search, information acquisition and deci-
sion (Figure 3). 
Pre-information search refers to the mental representation of the task, which de-
fines how the rest of the task will be performed. Information acquisition encom-
passes many judgments that define the extent of information usage. For instance, 
an auditor needs to consider different information sources, type of acquired in-
formation, number of information and the order of search. Finally, when the audi-
tor has gathered enough information, she/he is  ready to make a decision. As the 
focus  of  the  present  study  is  on  an  auditor’s  information  usage,  the  pre-
information search and information acquisition phases are discussed in details in 
the following paragraphs. 
 
 Acta Wasaensia     13 
  
 
Figure 3. The diagnostic, sequential and iterative decision-making process 
(adopted from Moroney 2007) 
 
The main purpose of information acquisition in the decision-making process is 
uncertainty reduction to an acceptable level (Koonce 1993). Before an auditor is 
able to begin information acquisition, the first step is to gain an understanding and 
interpretation of the problem at hand (Moroney 2007). This problem representa-
tion is based on an auditor’s previous knowledge about the task and it guides what 
information will be searched in actual information acquisition (Christ 1993). Con-
sequently, a lack of domain-specific knowledge in problem representation may 
constrain information acquisition (Bédard & Mock 1992). It is argued that initial 
problem representations are persistent and resist modification unless significant 
contradictory information emerges during information acquisition (Schultz, Bier-
staker & O'Donnell 2010).  
Depending on the audit environment and other circumstances, an auditor may 
choose different problem representations. For instance, auditors may choose be-
tween normal-audit and problem-audit representation, which may reflect on the 
subsequent information acquisition phase (Waller & Felix 1984). Further, it is 
suggested that switching to a problem-audit schema increases information acqui-
sition (Asare & Knechel 1995). 
In Bonner and Pennington’s (1991) decomposition of a decision-making process, 
the first three phases particularly concern information acquisition and usage activ-
ities for the problem representation. First, an auditor may retrieve from memory 
information that concerns the task at hand. Second, many tasks require obtaining 
information from external sources. All sources other than an auditor’s memory 
are considered to be external sources. Finally, the auditor comprehends infor-
mation from both sources to form a mental representation of the problem (Bonner 
& Pennington 1991). 
14      Acta Wasaensia 
Highly developed domain-specific knowledge and the complete knowledge struc-
ture of the decision-maker are suggested to assist in making more efficient and 
effective problem representations (Koonce 1993; Biggs, Mock & Watkins 1988). 
Thus, advanced problem representation may sharpen finding and interpreting im-
portant information more efficiently in the information acquisition phase than 
undefined problem representation (Moroney 2007). 
The information acquisition phase can be divided into two dimensions, depth of 
search and order of search. Depth of search refers to the total number of infor-
mation searched, while order of search means the order in which information is 
acquired by the decision-maker (Ford et al. 1989). The order of search can be 
either directed or sequential. In a directed search, an auditor looks for specific 
information from the available information, but in a sequential search an auditor 
goes through information in the order in which it is presented (Bonner 2008). 
Previous research has widely documented that decision-makers’ knowledge struc-
tures affect both these dimensions. Several audit studies (e.g. Simnett 1996; Bé-
dard & Mock 1992; Davis 1996) suggest that experienced auditors use less infor-
mation than less experienced auditors and that experienced auditors perform a 
goal-oriented, directed acquisition of audit evidence instead of a sequential search 
(Hoffman, Joe & Moser 2003). 
An auditor must usually also choose which information to acquire and process, as 
it is not possible to examine all available information (Knechel & Messier 1991). 
As mentioned before, there are two main sources of information (Bonner & Pen-
nington 1991). First, an auditor can use her/his internal memory to retrieve infor-
mation about the client, client industry and earlier experiences that she/he thinks 
is relevant for the task. The second source of information is external sources, such 
as discussions with the client or with audit team members, work papers and indus-
try databases. An auditor may also consider choosing between financial and non-
financial information. In many tasks, financial information is easily available 
from the client or public sources, while the acquisition of complex nonfinancial 
information from external sources might be more time-consuming and incur addi-
tional costs. 
Upon reaching a sufficient level of confidence for judgment or decision, the audi-
tor will cease information acquisition. The information acquisition process can 
thus be described as a sequential process3 where an auditor selects one piece of 
                                               
 
3 It is important to note that “sequential search” and “sequential processing” are different con-
cepts that should not be mixed. 
 Acta Wasaensia     15 
  
information, evaluates it and then decides whether to acquire additional infor-
mation or stop the acquisition phase (Gibbins 1984). Sequential information pro-
cessing assumes that an auditor evaluates each piece of information with respect 
to the earlier collected information (Knechel & Messier 1991). Auditors tend to 
initially search information that they deem the most diagnostic for the task 
(Knechel & Messier 1991). If the acquired information is sufficient to make a 
decision, the acquisition will be terminated, otherwise an auditor will continue 
acquisition until a decision can be made (Knechel & Messier 1991). 
An auditor may use so-called stopping rules to determine when information ac-
quisition should be terminated. Stopping rules can be simple, such as stopping 
after collecting a predetermined number of evidence items or the completion of a 
standard checklist. More complex stopping rules include mental models, where 
the current situation is compared with experiences and when there are no incon-
sistencies between the mental model and these experiences, information acquisi-
tion will be terminated (White & Harding 2011). 
Research has shown that in some circumstances information acquisition might be 
biased. When auditors have directional goals, they might perform biased acquisi-
tions and emphasize the importance of those cues that support their desired goals 
(Cianci & Bierstaker 2009). One form of this phenomenon is confirmation bias, 
where an auditor directs his/her information search towards information that con-
firms a favored hypothesis rather than that related to all generated hypotheses 
(Bonner 2008). For instance, susceptibility to confirmation bias can weaken inde-
pendence from the client, because the favored hypothesis is often the same as the 
client-preferred position (Kadous, Magro & Spilker 2008). 
Further, biased or unbalanced information acquisition may cause important in-
formation to be overlooked in the decision-making process. Some accounting 
studies (Andersson 2004; Thayer 2011) have shown that the judgments of the task 
outcome made in the initial decision-making phase are persistent and guide fur-
ther information acquisition. These studies suggest that once a preliminary judg-
ment has been formed, it is rarely challenged and information acquisition is di-
rected to justify it. For instance, in analytical procedures once a hypothesis for the 
unexpected fluctuation has been explicitly or implicitly formulated, this reason 
will dominate subsequent information acquisition behavior (Libby 1985). 
16      Acta Wasaensia 
2.2 Process tracing approaches in previous studies 
This study employs process tracing approaches to study auditors’ information 
usage. Ford et al. (1989) states that process tracing approaches “observe the pre-
decisional behavior in analyzing a situation by tracing the steps leading to a de-
cision”. Thus, one of their aims is to observe the “steps” in an individual’s deci-
sion-making process to gain insights into how the process is evolving to the deci-
sion and find the underlying cognitive processes that drive information acquisi-
tion behavior. More specifically, process tracing studies trace the decision-
maker’s information acquisition and usage behavior in real-time when performing 
a task. Real-time tracing is particularly intended to prevent the subsequent ration-
alization of decision behavior (Andersson 2004). The general advantage of real-
time tracing approaches is that they allow getting a rich data set of a decision-
maker’s behavior. 
Previous process tracing studies (e.g. Biggs & Mock 1983; Bédard & Mock 1992; 
Moroney 2007) investigating auditors’ information acquisition and usage have 
employed either verbal protocol analysis or computer-based tracking techniques. 
In verbal protocol analysis, a decision-maker thinks aloud in a supervised space 
when making a decision. By thinking aloud, she/he is presumed to express her/his 
thoughts and actions that reflect the cognitive processes used in the decision-
making process (Ford et al. 1989; Andersson 2004). In particular, in information 
acquisition studies, a cue that is mentioned is considered to be used for the deci-
sion (Bédard & Mock 1992). For the analysis of research, obtained data from the 
research sessions are transcribed and coded. 
The advantages of verbal protocol analysis studies are that studied tasks can be 
complex and realistic, as there are few instrumental restrictions other than the 
presence of the researcher. Verbal protocol analysis also allows for observing 
directly the decision-maker’s behavior in a semi-natural environment. Thus, it is 
possible to incorporate a large number of information available for decision-
making and present the information in its natural form (e.g. the format of audit 
work papers). A general disadvantage is that cues can be acquired without verbal-
izing acquisition. This threat is obvious when decision-makers are highly experi-
enced subjects whose decision processes are automatic (Bédard & Mock 1992). 
Other disadvantages are that a large amount of verbal protocol data is laborious 
and time-consuming to analyze, as the observations are usually coded by two in-
dependent coders to decrease subjective interpretations and to avoid coding errors 
(Andersson 2004; Biggs & Mock 1983; Wright & Bedard 2000). Furthermore, 
because of this tediousness, the data usually contain observations from only a few 
 Acta Wasaensia     17 
  
subjects that are not necessarily a representative sample of the whole subject pop-
ulation (Biggs, Mock & Watkins 1988). 
Computer-based tracking techniques are based on software that accurately record 
information usage patterns when decision-makers are actively searching for in-
formation and making decisions in a computerized environment. This technique 
can handle large amounts of data in a reliable way, as recording happens automat-
ically to the database (Andersson 2004). For instance, the number of used infor-
mation, order of the search, time spent on the task, and each decision-making 
phase are followed effortlessly by this technique. The computer environment also 
minimizes researcher intrusion in the actual decision-making situation (Rosman 
& O’Neill 1993).  
As a main limitation, the computer environment is usually much more simplified 
and artificial than a natural decision-making environment (Andersson 2004). For 
instance, tracking information usage accurately may require that information be 
separated into multiple cues or that cues be shortened. Information must also be 
located under menus or search engines, which can be considered to be artificial 
elements. Ultimately, in more complex tasks an information menu can even act as 
unwanted decision help that simplifies tasks compared with the natural environ-
ment (Boritz 1992). 
In summary, both verbal protocol analysis and computer-based tracking tech-
niques have advantages and disadvantages. In general, these approaches should be 
seen as complementary methods and the choice should be made based on the re-
search problem of the study. 
  
18      Acta Wasaensia 
3 AUDITORS’ INFORMATION USAGE 
 FRAMEWORK 
A large body of psychological research indicates that information acquisition and 
usage are highly dependent on the demands and characteristics of the task (Payne 
1982; Ford et al. 1989). In other words, the nature of each task largely determines 
what information is needed to perform that task. Consequently, the determination 
of a client’s going-concern status requires the acquisition and usage of vastly dif-
ferent information than the decision to accept or to reject a prospective client. 
However, audit tasks do not exist in a vacuum. Thus, performing a task is contin-
gent upon the properties of the individual decision-maker and the context in 
which decision-making occurs (Andersson 2004). For example, individual auditor 
attributes affect an audit task, as auditors bring to bear these individual character-
istics to the task (Nelson & Tan 2005). 
As stated before, this study adopts El-Masry and Hansen’s (2008) taxonomy, 
which classifies factors concerning auditors’ information acquisition and usage 
into four categories. The aim of this chapter is to review audit studies that have 
investigated:  
 
1. Individual factors 
2. Environmental factors  
3. Task-related factors 
4. Cue-related factors (i.e. factors related to the nature of cues), and  
5. Interactions between these factors 
 
Compared with El-Masry and Hansen’s (2008) study, the present study includes 
studies that have used external auditors or tax auditors4 as subjects. Another pre-
requisite for including a particular study is that a study is suggested to have impli-
cations on auditors’ information usage in a single audit task. Further, studies in-
vestigating interactions between the factors representing these four categories are 
central to this study and these studies are presented in a separate section. The 
chapter concludes with a discussion and conclusion of the results of the presented 
studies. 
                                               
 
4 Those studies using tax auditors as subjects are also included as their information acquisition 
and usage process is expected to be similar to that of external auditors. Specifically, in most of 
these studies tax auditors were working in one of the Big-4/6 accounting firms, suggesting 
that their in-house training might also be similar to that of external auditors. 
 Acta Wasaensia     19 
  
3.1 Individual factors 
Individual factors are factors that relate to single auditor characteristics that vary 
between auditors (El-Masry & Hansen 2008). They also encompass the cognitive 
processes that an auditor uses while performing information acquisition and usage 
(Bonner 2008). The following factors have been studied previously in the auditing 
context: auditor experience, motivated reasoning and confirmation bias. From 
these three, auditor experience has received most attention in the previous studies.   
3.1.1 Auditor experience 
Almost all published audit studies have found that general audit experience (i.e. 
experience in years) decreases the acquisition of information for various audit 
tasks. In audit program planning, Biggs, Mock and Watkins (1988) found that 
audit seniors upon finding a problem in accounts receivables increased substan-
tially their information acquisition compared with audit managers. Similarly, Da-
vis (1996) found in control risk assessment that experienced auditors selected 
fewer cues and weighted cues’ importance more unequally than less experienced 
auditors. In going-concern assessment, Simnett (1996) provided evidence that 
experienced auditors selected fewer ratios than less experienced auditors. 
Contrary to the above studies, Moeckel (1991) in supervisory review task showed 
that experienced auditors used more information than less experienced auditors 
when information cues’ forms deviated significantly from each other. The author 
concluded that experienced auditors were better able to utilize information than 
less experienced auditors, because they were better at making links between in-
formation cues when the information was not in “close proximity” or “semanti-
cally similar”. 
Research has also found that in less common audit tasks, i.e. tasks that are not 
prepared by every auditor on a regular basis, the amount of task-specific experi-
ence may have an important role in information acquisition and usage. Bédard 
and Mock (1992) found that computer audit specialist auditors utilize less infor-
mation in internal control system evaluation task than non-specialists. Bonner 
(1990) specifically investigated the role of task-specific experience in two differ-
ent audit tasks, namely analytical risk assessment and control risk assessment. It 
was hypothesized that a much larger experience-related effect would occur in 
analytical risk assessment than in control risk assessment, as the former requires 
knowledge that is acquired in the later years of an auditor's career. The results 
showed some support that task-specific experience helped auditors in analytical 
risk task recognize the relevant cues better and weight cues more accurately. As 
20      Acta Wasaensia 
expected, the study did not find any significant experience-related effect in con-
trol risk task. 
Some experimental studies have measured the time that auditors spend acquiring 
and processing information in a task. The amount of time spent in the task is usu-
ally defined as a proxy for audit efficiency in these studies. Thus, a shorter time 
for a task is deemed to signal more efficient information processing behavior. 
Davis (1996) found that in control risk assessment, experienced auditors used less 
total time than less experienced auditors, but the study did not conclude unambig-
uously whether this difference stemmed from the fact that experienced auditors 
acquired less information or processed the acquired information faster. Moroney 
(2007) showed that industry experience increased audit information acquisition 
efficiency in two industries when the auditor was an industry specialist. The result 
also held when the number of used information was controlled. Thus, the finding 
can be interpreted that industry-experienced auditors process single information 
cues more efficiently than non-industry-experienced auditors5. 
The literature has offered some explanations why experienced auditors acquire 
less information than less experienced auditors. The first explanation is that they 
have better knowledge of the relevant cues for specific tasks (Bonner 1990). The 
second explanation is that they are able to recognize better relevant information 
patterns by recalling similar situations that they have encountered earlier (Leh-
mann & Norman 2006). Thus, experienced auditors’ more accurate problem rep-
resentations help them better ignore irrelevant information in their information 
acquisition (Glover 1997; Shelton 1999). The third explanation is that experi-
enced auditors already have more knowledge or information in their memories, 
which they rely on for their decision-making (Bonner 2008). For example, expe-
rienced auditors may have benchmark data or in-depth knowledge of accounting 
standards ready in their memories, which can be applied directly to the present 
task without external information acquisition. 
The literature has developed theories to explain how experienced auditors’ infor-
mation acquisition behavior differs from that of less experienced auditors. Biggs 
and Mock (1983) recognized that auditors use either sequential or directed infor-
mation acquisition strategies in internal control evaluation and audit scope deci-
sions. In the sequential strategy, an auditor acquires extensively available infor-
                                               
 
5  Both Davis (1996) and Moroney (2007) also studied whether more efficient subjects are also 
more effective (i.e. accurate) in their judgments. The results did not support these expectations 
and showed even some support for opposite findings (i.e. taking more time leads to the better 
performance).  
 Acta Wasaensia     21 
  
mation before making any subjudgment in a task. By contrast, in the directed 
strategy information on each subjudgment is collected separately. According to 
Bonner and Pennington (1991), experienced auditors apply directed strategies in 
information acquisition and use internal “checklists” as a guideline in acquisition 
processes. Hoffman, Joe and Moser (2003) described experienced auditors’ in-
formation acquisition as a goal-oriented and directed process. Davis (1996) theo-
rized that experienced auditors use a “top-down approach” in their information 
acquisition, i.e. they holistically recognize the general features of a given situation 
and use these features to select the relevant information for the task at hand. Less 
experienced auditors instead use a “bottom-up” approach, where they look for 
potential cues and then select these one after the other (Davis 1996). Knowledge 
of typical accounting and control systems guides experienced auditors in infor-
mation acquisition compared with less experienced auditors, who usually acquire 
information in the order in which it is presented (Davis 1996). Experienced audi-
tors are also suggested to use more rules of thumb, whereas less experienced audi-
tors may only use simple sequential searches (Davis 1996). 
While theories suggest that experienced auditors are more selective in their in-
formation acquisition, some studies have documented that experience increases 
balanced information usage. Specifically, these studies have found that because of 
experienced auditors’ well-developed knowledge of tasks, they consider more 
effectively information that is inconsistent with their initial expectations or fram-
ing of the problem (Waller & Felix 1984; Choo & Trotman 1991). For instance, 
in a going-concern task experienced auditors are better at evaluating information 
that mitigated the threat to continued existence (Choo & Trotman 1991). Further, 
in an analytical procedures task experienced auditors recognize better whether the 
acquired information was sufficient for decision-making, while less experienced 
auditors were more likely to rely on weak information in their decision-making 
(Glover et al. 2005). 
Vast experience may not always lead to better information usage because of over-
confidence6. In a real estate valuation task, Earley (2002) expected the type of 
initially received information to influence how additional information was pro-
cessed. It was hypothesized that whether client-provided data were consistent or 
inconsistent with industry average data impacted the usage of additional infor-
mation. The results showed that experienced auditors performed generally better, 
                                               
 
6 Generally, overconfidence occurs when someone’s average rated confidence is higher than 
average (Van Swol & Sniezek 2005). However, in this study the term is used in a more gener-
ic sense. 
22      Acta Wasaensia 
i.e. made more accurate judgments than less experienced auditors. However, 
when client-provided information was consistent with the industry data, there 
were no differences in the judgment performance between experience levels. This 
finding was explained by the fact that experienced auditors overlooked client-
specific information of a sufficient depth and relied too much to their previous 
expectations, when the data were found to be consistent with industry data. Thus, 
when experienced auditors deem a task to be unchallenging they may not engage 
fully their cognitive processing activities that would need to consider both indus-
try and client-specific information. 
Another related finding shows that when experienced auditors are not allowed to 
process information in their natural ways, they will perform more like less experi-
enced auditors (Hoffman, Joe & Moser 2003). In other words, when experienced 
auditors apply their usual information acquisition strategies, their judgments re-
flect their “expert processing” of the information. Hoffman, Joe and Moser (2003) 
showed that the possibility of performing unconstrained processing, compared 
with the fixed order of information processing, increased experienced auditors’ 
attention to mitigating information (i.e., the cues that decrease the threat of con-
tinued existence), which lead to more optimistic survival judgments of going-
concern firms. 
Taken together, the theories derived from the psychology literature as well as the 
empirical evidence from audit studies generally suggest that because of better 
developed knowledge content and structure and advanced cognitive processes, 
experienced auditors are more effective and efficient in their information acquisi-
tion and usage than less experienced auditors. 
3.1.2 Motivated reasoning 
Motivated reasoning theory posits that individuals with preferred conclusions use 
biased mechanisms to reach a desired goal (Kunda 1990). One of these mecha-
nisms in the decision-making process is biased cue acquisition and usage (Cianci 
& Bierstaker 2009). If an auditor is susceptible to motivated reasoning, this may 
turn up in unbalanced information acquisition or the over-/underweighting of cues 
in information evaluation. Thus, an auditor who attempts to maximize audit effi-
ciency or favor client-preferred judgments may do so by under-auditing or ignor-
ing important information either in information acquisition or in processing 
(Cianci & Bierstaker 2009). 
Motivated reasoning may also have positive effects on information acquisition 
and usage if it increases the motivation to provide objectivism and skepticism 
 Acta Wasaensia     23 
  
(Cianci & Bierstaker 2009). Specifically, if auditors have these so-called accuracy 
goals in their decision-making, they may use more cognitive efforts, and therefore 
perform more comprehensive and balanced information acquisition and pro-
cessing than would otherwise occur. 
Blay (2005) investigated specifically how one reason for motivated reasoning, 
economic incentives, affected information evaluation in a going-concern report-
ing task. This experimental study manipulated the fear of losing the client (de-
fined as independence threat) and litigation risk7. It was hypothesized that when 
the fear of losing the client is high, auditors evaluate acquired information con-
sistently  with  a  client-preferred  conclusion.  By contrast,  high  litigation  risk  was  
expected to result in the opposite direction of evidence evaluation. The results 
indicated that auditors with a high fear of losing the client evaluated those clients 
more likely to survive after the evaluation of information, whereas auditors who 
faced high litigation risk evaluated clients less likely to survive after information 
evaluation. 
Kadous, Magro and Spilker (2008) examined how the level of practice risk ef-
fected the motivated reasoning of tax auditors. They theorized that when a prac-
tice risk of engagement is high, biased judgments are more likely to lead to nega-
tive consequences,  such as monetary penalties or the loss of reputation, as these 
threats become more salient to the decision-maker. Therefore, auditors may want 
to protect themselves from such losses by emphasizing accuracy goals in those 
circumstances. As expected, the study found that when practice risk was high, 
auditors performed comprehensive and balanced information acquisition, while 
auditors with low risk clients mainly acquired information that was consistent 
with client preferences. 
Closely related to motivated reasoning theory, auditors’ attitudes towards the con-
sequences of certain judgment outcomes may influence information usage. 
Guiral,  Ruiz  and  Rodgers  (2011)  hypothesized  that  the  fear  of  the  self-fulfilling  
prophecy of issuing going-concern reports affects information evaluation. The 
study investigated how auditors’ expectations of the existence of the self-fulfilling 
prophecy affected confirming (business termination) and disconfirming (business 
continuance) information evaluation. It was theorized that auditors who are afraid 
of the self-fulfilling prophecy are concerned that a going-concern opinion leads to 
                                               
 
7 Independence threat and litigation risk can also be considered to be environmental factors. 
However, they are presented in this category because Blay (2005) theorized these factors con-
vincingly as economic incentives that influence auditors’ decision-making via motivated rea-
soning. 
24      Acta Wasaensia 
the loss of future economic rents, as the client relationship would terminate be-
cause of business failure. The results confirmed these expectations, namely audi-
tors who had higher expectations of the self-fulfilling prophecy displayed greater 
sensitivity to disconfirming cues and simultaneously a lower tendency to confirm-
ing cues. 
3.1.3 Confirmation bias 
Confirmation bias is defined as a proneness to direct information acquisition to 
information that confirms the favored or initial hypothesis generated by an audi-
tor8 (Bonner 2008). An example of confirmation bias is a situation when an audi-
tor finds a ratio that indicates unexpected fluctuation, assumes a reason (hypothe-
sis) for the cause of fluctuation and searches only this hypothesis. Thus, an audi-
tor’s concentration on one hypothesis leads to biased information acquisition. 
Confirmation bias is also closely related to motivated reasoning in circumstances 
where an auditor favors one hypothesis over others. 
Audit studies have found mixed evidence about the existence of confirmation 
bias. McMillan and White (1993) found that auditors who believed the error hy-
pothesis when they encountered fluctuations in financial statement ratios used 
more confirming and disconfirming information than those who believed there 
was  no  error.  Further,  their  results  showed  that  all  auditors,  regardless  of  their  
initial hypothesis, were prone to using error-related evidence in the later stages of 
the task. This finding was interpreted to be consistent with Smith and Kida’s 
(1991) view that auditors are educated to design information acquisition so that 
potential material errors are uncovered. However, a later study by Anderson and 
Maletta (1994) did not find evidence that initial beliefs affect auditors’ attendance 
to positive or negative information. 
Contrary  to  the  findings  of  McMillan  and  White  (1993),  Bamber,  Ramsay  and  
Tubbs (1997) found that auditors are more sensitive to information that confirms 
their initial hypotheses in both fraud (material employee fraud in inventory) and 
non-fraud (collectability of a material accounts receivable) tasks. Brown, Peecher 
and Solomon (1999) found that whether auditors have a particular incentive to be 
efficient or effective in their decision-making influences their confirmation bias 
                                               
 
8 This initial hypothesis may also stem from the task, namely depending on how the decision 
problem is described for the decision-maker. Studies investigating this effect (i.e. framing) are 
presented in Chapter 3.3. 
 Acta Wasaensia     25 
  
proneness. Their results showed that unless auditors are encouraged to be effec-
tive in their hypothesis testing, they became confirmation bias prone. 
3.1.4 Confirmation bias and auditor experience 
Two studies have investigated whether increased experience mitigates proneness 
towards confirmation bias. Kaplan and Reckers (1989) found evidence that only 
less experienced auditors were confirmation bias prone in explaining ratio fluc-
tuations. Thus, their initial hypothesis of the fluctuation cause correlated positive-
ly with the type of information they subsequently searched. By contrast, experi-
enced auditors performed balanced information search strategies that did not cor-
relate with their initial hypothesis. 
While Bamber, Ramsay and Tubbs (1997) found that auditors are more sensitive 
to information that confirms their initial hypothesis, the low experience of a sub-
ject did not exacerbate confirmation bias proneness. However, the study provided 
evidence that less experienced auditors are more prone to rate more highly infor-
mation that confirmed their initial hypothesis than experienced auditors, but this 
proneness did not have a significant effect on final judgment. 
3.2 Environmental factors 
Environmental factors are defined as factors that surround auditors while they 
perform audit tasks (El-Masry & Hansen 2008). They are not related to specific 
people or tasks, but to the particular environment in which the audit occurs (Bon-
ner 2008). Several environmental factors have been found to affect auditors’ in-
formation acquisition and usage. These factors are accountability, time pressure 
and client risk.  
3.2.1 Accountability 
Accountability means that there is pressure to justify one’s actions or judgments 
to other people (Tetlock 1983). In auditing, the amount of accountability may 
vary from firm to firm depending on how much reviews, documentations and 
justifications are required by an audit firm. Generally, it is posited that accounta-
bility has positive effects on information acquisition and processing, as it moti-
vates decision-makers to engage in effortful information processing (Lee et al. 
1999). Thus, accountability information processing may become more complex, 
integrative and careful (Glover 1997; DeZoort, Harrison & Taylor 2006). 
26      Acta Wasaensia 
It must be noted that accountability may not always improve information acquisi-
tion and usage performance. Increased effort may increase the use of irrelevant 
information if an auditor lacks the necessary expertise to define task-relevant in-
formation (Bonner 2008). Thus, the additional effort might be used to process all 
available information without a consideration of its relevance. 
Still, several studies have shown that accountability increases auditors’ efforts in 
information acquisition and usage. Cloyd (1997) found that the existence of ac-
countability increased tax auditors’ efforts in information acquisition as evi-
denced by increased time spent on the task. Asare, Trompter and Wright (2000) 
found that auditors who were accountable when investigating the cause of unex-
pected fluctuations acquired more information on all the possible reasons for the 
fluctuations compared with auditors who were not held accountable. It was ar-
gued that auditors who worked under accountability did not want to eliminate any 
judgment options. Therefore, the overall acquiring of information usually increas-
es as each judgment option needs its own information search. However, the study 
did not find that accountability influenced the depth of information search on au-
ditors’ primary hypothesis.  
DeZoort, Harrison and Taylor (2006) found that depending on the amount of ac-
countability pressure, auditors’ usage of different information types varied in ma-
teriality judgment-related task. The amount of accountability pressure was ma-
nipulated via four forms of accountability pressure (none, review, justification 
and feedback). The results indicated that when pressure increased, auditors shifted 
to using more qualitative information than quantitative information in their expla-
nations of the judgment. Further, the time spent on the task increased along with 
this, suggesting that accountability induced the more complex and careful analy-
sis of available information. 
Some studies have evidenced the negative effects of accountability, which have 
caused auditors to perform biased information processing. Specifically, when 
accountability is to a supervisor whose view of the matter is known before the 
decision-making process begins, there is a risk that information acquisition and 
usage is biased to support the desired conclusion if a decision-maker wants to 
please her/his supervisor (Bonner 2008). Thus, there is a risk that the decision-
maker engages in motivated reasoning to reach the same conclusion as her/his 
supervisor. 
 Acta Wasaensia     27 
  
Few studies have examined how the preferences of reviewers are affected when 
auditors are asked to justify9 their decisions. Peecher (1996) found evidence that a 
supervisor’s preference of the justification type (credence, objectivity or skepti-
cism-inducing) in analytical procedures affects how extensively auditors search 
information for alternative explanations of fluctuations. Specifically, when the 
supervisor’s preference was “client-provided information credence-inducing” and 
the client possessed high integrity, auditors terminated searching additional in-
formation earlier than auditors in other experimental conditions. 
Turner (2001) asked auditors to justify their decisions in an accounts receivable 
collectability task. The study examined accountability’s effect on audit efforts by 
measuring the number of information and the time that auditors spent reading 
each cue. Reviewer preference was manipulated in three levels: unknown prefer-
ence, credence to client preference and skepticism preference. The study found 
that auditors with unknown reviewer preferences generally made similar infor-
mation choices to auditors who had a reviewer with skepticism preference. The 
main finding was that auditors with credence preference acquired relatively more 
client-preferred information, but overall less information than auditors in other 
manipulation groups.  However,  the time spent reading each cue was not signifi-
cantly different.  
Wilks (2002) studied how the supervisor’s view of the judgment affected audi-
tors’ information evaluation in a going-concern task. It was found that those audi-
tors who knew their supervisors’ views before information evaluation evaluated 
information more consistently with those compared with subjects who did not 
know their supervisors views beforehand. The former group also made more go-
ing-concern judgments that were consistent with the view of the supervisor than 
the latter group. However, the study emphasized that this so-called pre-decisional 
distortion of information seems to happen unknowingly and unintentionally by 
decision-makers. 
3.2.2 Time pressure 
Time pressure may affect auditors’ information acquisition and usage in several 
ways.  First,  it  is  generally suggested that a low (but existent)  level of time pres-
sure accelerates the speed of information processing and thereby has a positive 
                                               
 
9 Justification is the process of providing explanations to support one’s beliefs (Bonner 2007). 
For example, staff auditors normally prepare a memo of audit procedures they have performed 
in certain audit task. 
28      Acta Wasaensia 
impact on audit efficiency. Hence, auditors tend to increase effort intensity and 
process information more efficiently in a given task. When the time pressure 
starts to increase from low to modest or high, it is suggested that auditors start to 
filter out information from their processing in order to perform a task in a given 
time. At the beginning, auditors might be able to filter out only information that 
they perceive less relevant but as the time pressure increases relevant information 
might also be ignored (Bonner 2008). 
Second, time pressure may influence information processing strategies. It is sug-
gested that increased time pressure might change processing strategy from com-
pensatory to noncompensatory. In a compensatory strategy, different information 
cues can compensate each other, i.e. the low value of one cue versus another 
cue’s high value. Noncompensatory strategies meanwhile do not permit such 
tradeoffs between the positive and negative features of the information. Compen-
satory strategies stress the comprehensive use of all available information, while 
noncompensatory strategies encourage using simplified rules and thus ignoring 
relevant information. As the time pressure increases to modest or high, auditors 
may believe that shifting to noncompensatory strategies (i.e. processing infor-
mation  more  superficially)  allows  them  to  work  more  efficiently  (Ford  et  al.  
1989; Choo 1995; Bonner 2008). 
The empirical results in the auditing domain have been consistent with the predic-
tions from theory. McDaniel’s (1990) results indicated that in an inventory audit 
task increased time pressure decreased the amount of collected information meas-
ured by the amount of sampling adequacy10, whereas increased time pressure sped 
up subjects’ information processing. Specifically, while the overall effectiveness 
decreased (total number of found errors) as time pressure increased to the high 
level it improved audit efficiency, as auditors in the most extreme time pressure 
group found seeded errors from the experimental material more efficiently (errors 
per used time) than those in other treatments. 
Two studies (Braun 2000; Cianci & Bierstaker 2009) have indicated that in-
creased time pressure leads auditors to consider the overall efficiency of the audit. 
Both have suggested that auditors favor cues that support “no-problem audit” 
under high time pressure11. Braun (2000) examined whether time pressure affect-
                                               
 
10 Sampling adequacy refers to the value that indicates whether the examined sample size by the 
subject was adequate compared with the desired level of certainty. 
11 Whether this happens because of motivated reasoning, is not obvious. Therefore, these studies 
are presented together with time pressure instead of motivated reasoning, as time pressure is 
the common denominator in these studies and is a common environmental factor in auditing. 
 Acta Wasaensia     29 
  
ed subsidiary fraud detection when auditors’ focus was directed to an inventory 
audit task. The results showed that auditors under low time pressure were able to 
process effectively a broader set of cues and recognize fraud signals more often 
than auditors under high time pressure. It was indicated that the low time pressure 
group recognized the fraud signals in the same period of time (i.e. during the first 
45 minutes) as the high time pressure group.  It was concluded that low time pres-
sure, not the available time itself, helped an auditor be more attentive to all avail-
able cues. 
Cianci and Bierstaker (2009) investigated how time (budget) pressure effects in-
formation evaluation in an internal control evaluation task. It was hypothesized 
that when time pressure increases, auditors may alter their information processing 
strategies to meet the efficiency demands of a task. Thus, in order to improve ef-
ficiency, auditors may discount the effect of negative information and emphasize 
the positive information of internal controls to progress towards an unqualified 
audit opinion. The empirical results strongly supported these expectations of bi-
ased information processing. 
3.2.3 Client risk 
The audit literature has mainly investigated indirectly how the perceived client 
risk affects auditors’ information acquisition and usage. These studies have par-
ticularly examined how audit program plans and evidence collection plans are 
adjusted in the presence of different risk factors. The results of these studies gen-
erally suggest how auditors change their information acquisition behavior at an 
audit engagement level depending on client-related risks. However, they also pro-
vide suggestions on how information acquisition and usage adapts to the micro 
level of an audit, i.e. a single audit task. 
Mock and Wright (1993) examined evidential planning decisions on actual audits 
with archival data. They found that account-specific risk factors, but not engage-
ment-wide risks, particularly the number of previous errors, influenced the 
amount of evidence acquisition. Mock and Wright (1999) extended their earlier 
work by examining auditors’ risk assessment and audit planning in accounts re-
ceivable tasks in two industries. They found that the type of the needed audit evi-
dence was responsive to changes in a number of risk factors such as client’s li-
quidity and profitability. Likewise, Pratt and Stice (1994) found that an audit cli-
ent’s poor financial condition, high value of receivables and inventory, high 
growth in sales and high market value of equity increased auditors’ recommenda-
tions to obtain more audit evidence for the engagement. 
30      Acta Wasaensia 
Other indirect findings of the risk to information acquisition suggest that changes 
in client-related risks affect the extent of testing according to the number of staff 
level work hours. It is thus reasonable to assume that increased budgeted audit 
hours are also related to increased information acquisition, as audit staff are re-
sponsible for collecting audit evidence (Hackenbrack & Knechel 1997). Specifi-
cally, O’Keefe, Simunic and Stein (1994) found that the greater inherent risk of a 
client  resulted  in  significantly  greater  audit  hours  of  staff  and  senior  auditors.  
Hackenbrack and Knechel (1997) measured client risk by type of firm (public vs. 
private) and found that this affected the number of auditors’ labor hours. Bedard, 
Mock and Wright (1999) summarized the findings of previous audit planning 
studies and concluded that client-related nonfinancial risk factors and manage-
ment explanations for unusual discrepancies are related to the extent of testing. 
Several studies have investigated more directly how the riskiness of the client 
affects the amount of needed audit evidence. Houston, Peters and Pratt (1999) 
found that when client risk was manipulated by the discovery of accounting 
treatment that misspecified inventory, auditors increased the planned amount of 
audit evidence in order to complete the audit. Beaulieu (2001) found that a pro-
spective client’s management integrity was linked to auditors’ recommendations 
of needed audit evidence. More specifically, when a CFO was characterized as 
possessing negative integrity, auditors planned to acquire additional evidence for 
the engagement. Johnstone and Bedard (2001) found that in audit engagement 
planning that the presence of risk factors affected subsequent audit evidence col-
lection strategies. 
3.3 Task-related factors 
Task-related factors refer to the characteristics of different audit tasks, which vary 
from task to task (Bonner 2008). As it may be unreasonable to compare audit 
tasks with each other, because they have specific purposes in the audit process, 
task-related factors aim to recognize and classify the common denominators of 
different tasks to disentangle those effects on auditors’ information acquisition 
and usage. However, task-related factors can also vary within tasks (Bonner 
2008) – for example, the same going-concern task can be initially framed differ-
ently – which may affect subsequent information acquisition and usage. The task-
related factors that have been studied previously are task complexity, task fram-
ing, task type and task response mode.  
 Acta Wasaensia     31 
  
3.3.1 Task complexity 
Bonner (1994) suggested that the definition of task complexity12 encompasses 
task difficulty and task structure. The difficulty element of task complexity relates 
to the number of processing steps of the task, the number of information cues 
available, the number of possible hypotheses/decision alternatives and the correla-
tion between information cues (Bonner 2008). Generally, an increase in the num-
ber of these factors increases task complexity. However, a strong correlation be-
tween cues might mitigate such an increase in task complexity, as the information 
load will not increase linearly with the number of information (Bonner 2008). 
The task structure element refers to the clarity of cognitive processing in a task 
(Bonner 1994). This structure can vary depending on how well task-specific in-
formation is defined in terms of the degree of cue measurement, familiarity of the 
information form to the decision-maker and relations between information cues 
and task outcomes (Bonner 2008). When the clarity of task processing is not well-
specified (e.g. information or its relation to the outcome), the task is classified as 
semi- or unstructured. 
While task complexity is frequently studied in auditing (e.g. Tan & Kao 1999; 
Tan, Ng & Mak 2002), few studies have particularly studied task complexity’s 
influence on information acquisition and usage. Simnett and Trotman (1989) 
studied the effect of information choice and information processing on auditors’ 
judgments in business failure prediction task. They manipulated one element of 
task complexity, namely task structure, with two levels by giving one or two year 
before-the-event financial ratio information to auditors. Auditors then had to pre-
dict  whether  firms  were  going  to  go  bankrupt  within  one  or  two  years.  Infor-
mation choice, namely by selecting four ratios out of 10, was carried out by sub-
jects or statistical models. Information processing was performed by an auditor or 
statistical model depending on the treatment. 
The results indicated that information choice was a limiting factor in predictive 
accuracy, i.e. model-selected information outperformed human-selected infor-
mation. By contrast, in the treatments where auditors selected information, there 
were no differences in judgment accuracy between subjects and model pro-
cessing. In addition, while the results showed that predictive accuracy decreased 
steadily in all treatments when task structure shifted to less structured, the study 
                                               
 
12 There are numerous definitions of task complexity (see e.g. Wood 1986, Byström 2002) in the 
literature. For the sake of brevity, only Bonner’s (1994) is discussed in this section. 
32      Acta Wasaensia 
did not find that task structure affected the types of ratios chosen by auditors. In 
other words, selected ratios were similar in both task structure levels. 
Simnett (1996) examined another element of task complexity, namely task diffi-
culty, in a corporate failure prediction task. Information load was manipulated 
using two levels (low/high) depending on the number of financial ratios given for 
the task. Similar to Simnett and Trotman (1989), information selection and pro-
cessing varied so that phases were performed by auditors or statistical models (or 
a combination). Thus, there were 10 treatment groups. The results generally 
showed that auditors did not perform well in information acquisition and pro-
cessing when task difficulty increased. Specifically, when information was select-
ed by the model, it led to a better predictive accuracy of failure than when ratios 
were picked by an auditor. This finding was consistent with both low and high 
information load conditions. In addition, when information load was high, audi-
tors’ information processing performances were significantly lower compared 
with the statistical models. Thus, as in Simnett and Trotman (1989), information 
processing under high task complexity was deemed to be a limiting factor in 
judgment performance. Finally, the study found systematic differences between 
the types of selected ratios depending on information load. While these differ-
ences were not examined in detail, this result suggests that task difficulty influ-
ences the type of information that auditors select. 
These two studies show that suboptimal information selection by auditors limits 
their judgment performances. They also indicate that deficiencies in auditors’ 
information processing limit performance when the number of available infor-
mation is high. Finally, these studies are inconclusive about whether task com-
plexity affects the type of information that an auditor acquires for her/his deci-
sion-making. 
3.3.2 Task framing 
Framing refers to a change in a task that does not alter its true substance, but 
changes decision-makers’ perceptions about it (Mueller & Anderson 2002). A 
framing effect is said to occur when a change in the description of the task (i.e. a 
frame) changes the decision that is made (Jamal, Johnson & Berryman 1995). In 
addition, task framing may also alter the decision-making process itself. Specifi-
cally, this happens when an auditor applies the initial frame as an anchor that di-
rects the subsequent acquisition process. Thus, framing may activate confirmation 
bias if the auditor is not able to transform problems into a standard representation 
(Jamal, Johnson & Berryman 1995). 
 Acta Wasaensia     33 
  
Overall, auditing studies have not found task framing to seriously direct auditors’ 
information acquisition. Kida’s (1984) experimental study divided subjects into 
two groups, namely failure and viability. In the failure (viability) group, subjects 
were asked to list information that they consider to be relevant to decide whether 
a firm would fail (remain viable) within two years. The study reported that the 
initial task framing influenced the information that auditors considered to be rele-
vant. The results indicated that auditors in the viability group selected more via-
bility cues than auditors in the failure group, but overall auditors in both groups 
listed more failures than viability cues. 
Trotman and Sng (1989) extended Kida’s (1984) research setting by taking into 
account that auditors may process information sequentially in their decision-
making. The study found evidence that when an initial situation was set to posi-
tive (strong ratios) and the problem was framed to be positive (viability), auditors 
acquired more positive cues than when the initial situation was negative (weak 
ratios) and the problem was framed to be negative (failure).  Consistent with the 
findings of Kida (1984), the study reported that regardless of problem framing, 
auditors overall acquired more negative than positive cues, meaning that there 
was no clear evidence of auditors’ applying confirmation bias strategies depend-
ing on the initial problem frame. 
In an analytical procedure task, Ayers and Kaplan (1993) found evidence that 
hypothesis framing affected auditors’ information usage. It was hypothesized that 
problem framing influences subsequent information choice when either a mis-
statement or a non-misstatement frame is presented to auditors to explain unusual 
financial statement fluctuation. The results showed that a (non-)misstatement 
frame increased auditors’ usage of (non-)misstatement cues. However, the study 
did not find any evidence that framing affected the actual judgments about the 
reason for the unusual fluctuation, i.e. only cue usage was affected. 
3.3.3 Task type 
While task type in information usage has only been examined in two audit stud-
ies, this factor might be an important determinant of how different cues are inte-
grated into auditors’ decision-making. Generally, tasks can be classified as either 
estimation or evaluation types (Hogarth & Einhorn 1992). A well-known belief 
adjustment model (Hogarth & Einhorn 1992) suggests that cues are integrated 
differently depending on task type. When an auditor estimates an amount of mon-
ey,  the  task  can  be  considered  to  be  an  estimation  task  (Kerr  &  Ward  1994),  
whereas when an auditor evaluates whether an account balance has been fairly 
34      Acta Wasaensia 
stated, i.e. hypothesis testing, the task can be considered to be an evaluation task 
(Kerr & Ward 1994). 
The belief adjustment model predicts that in estimation tasks every new piece of 
information is evaluated with respect to the current belief. Thus, this belief is ad-
justed through an averaging process of new pieces of information. By contrast, in 
evaluation tasks pieces of information are evaluated irrespective of the current 
belief by summing their values. Kerr and Ward’s (1994) empirical results of audit 
planning (estimation type) and internal control evaluation (evaluation type) task 
support the above predictions, suggesting that task type may have implications on 
how the order of cues affects final judgments (El-Masry & Hansen 2008). 
3.3.4 Task response mode 
Turner (2001) investigated how different response modes in otherwise identical 
tasks affect information usage in an accounts receivable collectability task. In one 
set of treatments (belief), subjects were asked to assess the likelihood of each ac-
count balances’ collectability. In the other set of treatments (action), subjects had 
to classify account balances according to whether they were collectible or uncol-
lectible. It was expected that auditors in belief treatments would be judged (by 
their superiors) more by the quality of their decision-making processes compared 
with auditors in action treatments and thus the former use more effort for infor-
mation usage. Consistently with this expectation, the study found that auditors in 
action treatments acquired less information cues and spent less time per infor-
mation cue than auditors in belief treatments. 
3.4 Cue-related factors 
Several characteristics of cues per se may influence how they are acquired or pro-
cessed in auditors’ decision-making (El-Masry & Hansen 2008). For example, the 
(perceived) reliability of a cue may influence how much effort an auditor uses to 
process the cue to find out its information value. In this study, cue-related factors 
refer to the nature and characteristics of cues. In previous audit studies, cue-
related factors13 such as information order, information reliability, irrelevant in-
formation, presentation mode and information types have been studied. 
                                               
 
13 While some studies in this area have focused on cue effects only, others have investigated 
information effects in general. As there is no theoretical argument to distinguish between 
 
 Acta Wasaensia     35 
  
3.4.1 Information order 
Several studies have investigated how information order affects audit judgments, 
as depending on the presentation order of information in audit tasks different 
judgments or decisions will be made (Arnold et al. 2000). Ideally, judgments and 
decisions should only be influenced by the substance of information (Kennedy 
1993). However, a significant amount of evidence (e.g. Knechel & Messier 1990; 
Messier 1992; Asare 1992; Messier & Tubbs 1994) suggests that auditors display 
information order effects known as recency. In recency effects, an auditor’s be-
liefs are revised more negatively when negative information follows positive than 
if positive information follows negative (Anderson & Maletta 1999). In addition, 
few studies have investigated the opposite effect of recency, known as primacy. A 
primacy effect occurs when an auditor’s belief is revised more negatively when 
negative information precedes positive information than if positive information 
precedes negative (Anderson & Maletta 1999). 
Ashton and Ashton (1988) tested whether recency effects existed in an internal 
control task. They conducted two experiments, where some subject groups were 
given four pieces of positive or negative information (consistent information) and 
others were given mixed evidence, i.e. both positive and negative information. 
The results of these experiments were consistent with their predictions; no order 
effects were found for consistent information group, but recency effects were 
found in mixed evidence groups. Similarly, Butt and Campbell (1989) investigat-
ed the existence of recency effects in an internal control task. They observed re-
cency effects only when subjects had negative beliefs about the internal control 
system. When subjects started with high positive beliefs about the internal control 
system, no recency effect was found. In a going-concern context, Asare (1992) 
found that auditors displayed recency effects in their probability estimates of the 
continued existence of a firm. Recency effects also affected their subsequent re-
porting behavior. Auditors who evaluated first negative (indicating failure) and 
then positive (indicating viability) information issued more unqualified audit re-
ports than those who evaluated the same information but in the reverse order. 
By contrast, some studies have evidenced that auditors are not always susceptible 
to the order effect. In an inherent risk assessment task, Monroe and Ng (2000) 
found that auditors always focused on factors that indicated high inherent risk 
regardless  of  the  order  in  which  they  were  presented.  They  explained  this  con-
                                                                                                                                
 
these two terms in this context, all cue-related factors in this study are named “information” 
instead of “cue”. 
36      Acta Wasaensia 
servative behavior by stating that the existence of high inherent risk factors domi-
nates the order effect. Further, Favere-Marchesi (2006) documented that an 
awareness of the temporal order of information decreases the recency effect in a 
going-concern judgment. He argued that the temporal order of information domi-
nates presentation order, because the chronological order of information empha-
sizes cause-and-effect relationships and attaches greater weight to more recent 
information. 
3.4.2 Information reliability 
Information reliability is suggested to be an important factor in auditors’ JDM, as 
less reliable information is expected to be weighted less in the decision-making 
process compared with more reliable information (Kizirian, Mayhew & Sneathen 
2005). Information reliability is often estimated as the credibility of its source 
(Hirst 1994; Goodwin 1999). Thus, information that comes from a more credible 
source is considered to be more reliable than information from a less credible 
source. 
In previous audit studies, source credibility has been suggested to consist of sev-
eral components: independence, competence and integrity (Rebele, Heintz & 
Briden 1988; Hirst 1994; Goodwin 1999). For example, an auditor may evaluate 
differently information provided by an internal auditing function depending on its 
competency and position in the organizational hierarchy. However, even if the 
source is competent, it is not necessarily objective if it has an incentive not to be 
truthful. Thus, as the source’s incentive not to be truthful increases, its credibility 
decreases. Thus, a biased source can be considered to be less credible and less 
persuasive than an unbiased one (Robertson 2010). 
The seminal work by Joyce and Biddle (1981) investigated how source independ-
ence affected auditors’ judgments. They experimentally manipulated a source in 
which the description of the client’s financial difficulties had been obtained. Their 
presumption was that when the description was from an independent third party, 
the information should be considered to be more reliable compared with the same 
description obtained from the client’s credit manager, as the latter has more in-
centive to give a biased description. Their empirical results did not support the 
premise that auditors were sensitive to source independence when carrying out a 
task. However, in a another experiment that employed a within-subjects design, 
i.e. where auditors were performing two tasks successively with different levels 
of source reliability, auditors were sensitive to source reliability as hypothesized. 
 Acta Wasaensia     37 
  
Subsequent studies have found more support for auditors’ sensitivity to source 
credibility. Rebele, Heintz and Briden (1988) found that when estimating the 
amount of uncollectible accounts, auditors placed significantly more reliance on 
information received from a more competent source than on information received 
from a less competent source. Further, Hirst’s (1994) results in an inventory mis-
statement task evidenced that both source competence (specialist/non-specialist) 
and independence (other auditor/CFO) influenced auditors’ misstatement esti-
mates. Similarly, Goodwin and Trotman (1996) found in an asset revaluation task 
that when information came from a less competent source, auditors planned to use 
more audit hours for audit evidence gathering than if the source was perceived as 
more competent. 
Anderson, Koonce and Marchant (1994) investigated whether the competence of 
management (low or high) influenced auditors’ judgments about the reliability of 
management explanations of unexpected fluctuations. It was also studied if the 
timing of competence information (before or after the explanation) affected 
judgments. Their results indicated that auditors were sensitive to the competence 
of information but that the timing of the information did not affect judgments. 
Interestingly, the answer of the control group (i.e. those who did not receive in-
formation about the competence of management) was significantly different from 
that of the low competence group, but not significantly different from that of the 
high competence group. This finding suggests that auditors may generally deem 
management as competent if there are no opposite indicators of this. 
An auditor’s assessment of the integrity of the information source may particular-
ly affect how information is processed. Goodwin (1999) studied whether audit 
evidence was evaluated differently if the source was internal (management) or 
external. The consistency of other information with the integrity of the infor-
mation source was also studied simultaneously. In two tasks (probability of law-
suit and overstatement of inventory), the integrity of the information source was 
manipulated by two levels (low or high). The results indicated that auditors were 
sensitive to the integrity of the information source regardless of whether it was 
internal or external. The study also provided evidence that auditors were even 
more concerned about source integrity when the other information was incon-
sistent with the information from the actual source. 
Further, Krhisnamoorthy, Mock and Washington (1999) used four belief revision 
models to show that auditors were sensitive in their judgments to differences in 
the reliability of internal control system. Glover, Jiambalvo and Kennedy (2000) 
found that when management did not justify extensively unexpected fluctuation 
and had an explicit incentive to misstate the reason of the fluctuation, the majority 
38      Acta Wasaensia 
of auditors planned to acquire additional information compared with the other 
three experimental conditions where the level of justification and presence of in-
centive varied. 
In addition to the independence, competence and integrity of the source, the de-
gree of personal involvement in the information processing stage might also af-
fect how the credibility of the information source is perceived. Reimers and Fen-
nema (1999) expected that less involved auditors (reviewers of work papers) pay 
less attention to the details of information compared with more involved auditors 
(preparers of work papers), but that reviewers are more sensitive to general per-
suasiveness, i.e. the credibility of the information source. Their experimental re-
sults in an accounts receivable task supported the expectation that reviewers were 
more sensitive to the credibility of the information source. 
3.4.3 Presentation mode 
Presentation mode refers to the form in which the information is presented to the 
decision-maker (Bonner 2008). For example, in many cases the piece of infor-
mation can be presented in either a graphical or a tabular format. Research has 
investigated whether presenting the same piece of information in different formats 
affects how information is acquired or how task-specific judgments are made. 
Presenting data graphically instead of in a tabular format may result in better 
judgments in tasks that require information to be compared for two reasons. First, 
it may help information acquisition by reducing cognitive efforts when infor-
mation is presented more concisely and in meaningful patterns. Thus, information 
can be more easily and quickly interpreted, which enables an auditor to direct 
more cognitive resources towards advanced information processing. Second, 
graphical data may thus reduce the cognitive efforts needed to correctly weight 
and combine information into a judgment (Wright 1995). 
Few studies have investigated how information presentation affects auditors’ 
JDM. While no studies have investigated how presentation mode affects auditors’ 
information acquisition, some have examined how different information formats 
affect task-specific judgments. Thus, these results give suggestions for how 
presentation mode affects information usage. 
Studies examining presentation mode have found generally mixed results, alt-
hough they have evidenced some benefits of a graphical over a tabular format. An 
early study by Ricchiute (1984) evidenced that auditors are most likely to adjust 
audits when information is presented in a written format instead of discussed oral-
 Acta Wasaensia     39 
  
ly. Kaplan (1988) examined whether the presentation mode of the previous year’s 
accounts affected auditors when they were asked to estimate the same account’s 
next year’s value. Presentation mode was either graphical or tabular. The results 
showed no evidence that presentation format significantly affected subjects’ ex-
pected value judgments. 
Anderson and Reckers (1992) extended Kaplan’s (1988) study by creating a more 
realistic estimation task. In their study, expected sales value was evaluated with 
ratio analysis instead of simple trend analysis. To make a ratio analysis, subjects 
had first to establish a historical relationship of sales with five other variables, i.e. 
find the best variable that correlates with sales. The results showed that the graph-
ical display of data led to better judgment performance and confidence compared 
with tabular data. 
Schulz and Booth (1995) provided more variables and a longer data period for a 
similar estimation task. Their results provided evidence that graphically presented 
data leads to the more accurate prediction of sales accounts than tabularly pre-
sented data. Inconsistent with the findings of Anderson and Reckers (1992), they 
found that auditors with graphically presented data were no more confident about 
their predictions compared with subjects in the tabular group. However, graph-
ically presented data led to greater information processing efficiency, as the time 
spent on the task decreased. 
In an inherent risk assessment task, Dilla and Stone (1997) showed that represent-
ing cues in quantitative (financial) instead of qualitative (nonfinancial) form in-
creased consensus in auditors’ information processing and decreased the time to 
process cues. In an internal control evaluation task, Bierstaker and Brody (2001) 
found that additional flowcharts over the narrative description of internal controls 
did not increase judgment accuracy, i.e. presentation format had no effect on au-
ditors’ judgments. 
3.4.4 Irrelevant information 
The audit environment usually contains a lot of information, and some infor-
mation is relevant for a specific task, whereas other information is not. One 
stream of research has examined how the presence of irrelevant and relevant in-
formation affects auditors’ judgments and their decision-making processes. Spe-
cifically, these studies have examined whether auditors are susceptible to the so-
called dilution effect. The dilution effect occurs when the presence of irrelevant 
information leads to less extreme decisions when presented together with relevant 
information (Hackenbrack 1992). The theory behind the dilution effect assumes 
40      Acta Wasaensia 
that the presence of irrelevant information cues (with relevant information cues) 
diminishes the similarity of the situation compared with decision-makers’ mental 
models of a problem (Bonner 2008). Thus, increased dissimilarity dilutes the 
judgment, as the current situation is not identical to the mental image of a typical 
situation. 
The pioneering study by Hackenbrack (1992) was the first to assess the dilution 
effect in auditing. In this study, auditors were given two fraud-risk assessment 
tasks; the first task included both relevant and irrelevant information and the se-
cond included only relevant information. The results indicated strong support for 
the dilution effect, as auditors’ fraud-risk assessments became less extreme when 
irrelevant information was present. By contrast, irrelevant information has mainly 
been studied with other variables, i.e. interaction effects. These studies are pre-
sented in the next chapter.  
A closely related factor to irrelevant information is redundant information. For 
example, an information cue might be repeated several times because of media 
coverage (Joe 2003). Thus, the repetition of information can be seen as irrelevant 
information, as it is has no informative value to the decision-maker after it has 
been presented once (Bonner 2008). Typically, an auditor may be exposed to in-
formation redundancy when there are multiple information sources that the audi-
tor  considers  to  be  relevant.  Joe  (2003)  investigated  how  the  press  coverage  of  
client debt default was associated with the increased likelihood of modified audit 
opinions, even though this information was already known by the auditor. The 
study had two competing hypothesis for the reason for this. The first explanation 
was that auditors might react to this overlapping information to mitigate their per-
ceived litigation risk (strategic hypothesis). The alternative explanation was the 
cognitive reason that the high salience of debt default leads auditors to conclude 
that the firm has a higher probability of failure (cognitive hypothesis). The exper-
imental results showed that auditors modified their audit opinions because of the 
cognitive hypothesis, meaning that they were prone to redundant information be-
cause of biased information processing. 
3.4.5 Information types 
Some studies have considered how information types affect information acquisi-
tion and processing. Information types have been categorized in these studies into 
two classes: financial and nonfinancial information. Financial information refers 
to quantitative information such as financial statements, cash flow charts and rati-
os. This information is usually readily available for an auditor as it is produced in 
a client firm. Nonfinancial information refers to qualitative information outside 
 Acta Wasaensia     41 
  
financial statements and footnotes (Cohen, Krishnamoorthy & Wright 2000; Bra-
zel, Jones & Zimbelman 2009). 
It has been suggested that financial information acts as a primary information 
source in many audit tasks, e.g. materiality judgments, risk assessments, analyti-
cal procedures and going-concern decisions (Krogstad, Ettenson & Shanteau 
1984; Mutchler 1986; Rosman, Seol & Biggs 1999; Cohen, Krishnamoorthy & 
Wright 2000). This is expected to arise because auditors’ education and training 
places emphasize financial information (Cohen, Krishnamoorthy & Wright 2000). 
However, some studies have found that nonfinancial information is also important 
in certain audit tasks. 
First, nonfinancial information has been suggested to act as corroborating evi-
dence for financial information in some audit tasks. Cohen, Krishnamoorthy and 
Wright (2000) investigated how financial and nonfinancial information are used 
in two analytical review tasks: establishing the level of audit scope and generating 
hypotheses for unusual fluctuation. In the former task, their results indicated that 
auditors place more reliance on financial trends than on nonfinancial trends. The 
study found that auditors mainly use nonfinancial information as corroborating 
evidence in this task. Thus, by using nonfinancial information, an auditor can be 
more convinced about the financial information’s accuracy and consistency. 
However, in the hypotheses generation task, auditors generated an equal number 
of hypotheses when either financial or nonfinancial information indicated a de-
cline. This finding suggests that auditors value both information types equally 
important in this task. 
Second, nonfinancial information can also supplement financial information. In 
going-concern tasks, forward-looking nonfinancial information can be used to 
predict future financial information (Arnold & Edwards 1993; Behn & Riley 
1999). For example, a general prediction of poor economic conditions is visible 
much later in the firm’s financial statements than in the surrounding environment, 
where an auditor can acquire this cue first. 
Changes in a task’s problem representation may lead to different acquisitions of 
financial and nonfinancial information. Rosman, Seol and Biggs (1999) hypothe-
sized that different task settings, stages of organizational development (start-up or 
mature) and financial health (bankrupt or non-bankrupt) affect auditors’ infor-
mation acquisition and judgment in going-concern tasks. The results indicated 
that information acquisition was sensitive to different task settings. Specifically, 
in start-up (mature) company treatments, auditors acquired more nonfinancial 
(financial) cues when they needed additional information for the judgment. How-
ever, in both organizational development levels, auditors acquired more mitigat-
42      Acta Wasaensia 
ing (i.e., cues that decrease the threat of continued existence) financial cues than 
mitigating nonfinancial cues. Thus, financial information dominated nonfinancial 
information in this part of the task. 
3.5 Interactions 
Considerable studies of auditors’ information acquisition and usage have investi-
gated interactions between factors from the four categories discussed above. In 
general, the majority of these studies has examined whether the presence of miti-
gating factors (e.g. auditor experience, accountability) decreases information ac-
quisition and usage-related judgmental biases that are caused by the presence of a 
known negative phenomenon (e.g. recency, high time pressure). From these miti-
gating factors, auditor experience has received most attention in the previous lit-
erature. Consequently, in the section 3.5.1 all interaction studies with individual 
factors except one study have examined whether the effects of other factors are 
conditional on auditor experience. In the section 3.5.2, interactions with environ-
mental factors are presented. Those studies have mainly investigated the effect of 
accountability in mitigating or exacerbating cognitive biases related to infor-
mation  acquisition  and  usage.  Rest  of  the  studies  in  this  section  have  examined  
time pressure  and  client  risk  with  cue-related  factors.  Finally,  section  3.5.3  pre-
sents a miscellaneous interaction between cue-related and task-related factors. 
3.5.1 Interactions and individual factors 
Studies of auditor experience have the general premise that experienced auditors 
are less susceptible to different biases in their information acquisition and usage, 
because they have superior domain- or task-specific knowledge and better prob-
lem-solving schemas than less experienced auditors. In many studies, these prop-
erties are expected to lead to better information acquisition and processing, and 
this is expected to be one of the determinants that explain the better overall JDM 
quality of the experienced auditors. However, not all studies have found these 
hypothesized interactions. 
 
Time pressure (environmental factor) and auditor experience 
The studies of time pressure and auditor experience have expected that experi-
enced auditors suffer less from moderate or high time pressure than less experi-
enced auditors. Cianci and Bierstaker (2009) expected that less experienced audi-
tors are more susceptible to time (budget) pressure than experienced ones. It was 
 Acta Wasaensia     43 
  
reasoned that experienced auditors have greater cognitive resources, better mental 
representations of the task and more sensitivity to the legal environment, which 
helps them mitigate time pressure effects. The results brought some evidence 
supporting this expectation, as negative internal control information was evaluat-
ed differently by less experienced auditors under high time pressure. However, 
the evaluation of positive internal control information and final judgments were 
not affected by experience. 
Two studies where tax auditors were the subjects have suggested that when audi-
tors are experienced the presence of time pressure increases the usage of more 
important14  information. In both these studies, it was expected that experienced 
tax auditors’ advanced knowledge structures allow them to selectively attend to 
more important information under high time pressure. By contrast, because less 
experienced tax auditors lack this proper knowledge, they were expected to be 
less able to change their search behavior under high time pressure. 
Spilker (1995) measured how tax professionals’ type of knowledge (declara-
tive/procedural) and time pressure affected information acquisition. Declarative 
knowledge was defined as knowledge of facts and concepts and procedural 
knowledge15 as knowledge about how to perform a task. The results indicated that 
when subjects had significant procedural knowledge they acquired more crucial 
information than subjects who had only declarative knowledge. Similarly, sub-
jects with declarative knowledge acquired more crucial information than subjects 
who had little or none of either type of knowledge. When examined jointly with 
time pressure, the procedural knowledge group acquired more crucial information 
than the declarative knowledge group under moderate time pressure than under 
low time pressure. By contrast, subjects who had little or no declarative or proce-
dural knowledge acquired less crucial information under moderate time pressure 
than under low time pressure. 
Spilker and Prawitt (1997) extended Spilker’s (1995) study by measuring the time 
utilization of tax auditors under time pressure. They hypothesized that experi-
enced tax auditors adapt better to time pressure than less experienced ones. Spe-
cifically, it was expected that experienced tax auditors use less time in each phase 
of the task, i.e. problem representation, re-reviewing initial problem and infor-
mation search than less experienced tax auditors. However, the empirical results 
                                               
 
14 Spilker (1995) used the term “relevant” for this information in his study, but he did not im-
plicitly state that the rest of the information was totally irrelevant. 
15 It is reasonable to assume that procedural knowledge is acquired through task-specific experi-
ence. 
44      Acta Wasaensia 
indicated that experienced subjects spent less time only information searching, as 
they focused on more important information under high time pressure. 
 
Information order (cue-related factor) and auditor experience 
The findings of several auditing studies have indicated that experience may miti-
gate the recency effect. These studies have assumed that experienced subjects are 
generally more familiar with various tasks, and therefore their increased confi-
dence of the initial impression of a problem makes them react less strongly to the 
contrasting cues they encounter (Messier & Tubbs 1994; Trotman & Wright 
1996). Less experienced auditors’ cognitive strain may also mean that they cannot 
process large amounts of information simultaneously, which leads them to em-
ploy information processing strategies that cause more easily recency bias (Ken-
nedy 1993; Trotman & Wright 1996). 
In a going-concern task, Kennedy (1993) did not find the recency effect for audit 
managers, whereas students displayed it strongly. Thus, students who received 
information in a positive/negative order judged the likelihood of failure to be 
greater than subjects who received information in the opposite order. In an ac-
counts receivable task, Messier and Tubbs (1994) used audit seniors and manag-
ers as subjects and found that managers were less exposed to the recency effect. 
By contrast, Krull, Reckers and Wong-On-Wing (1993) found that when subjects 
faced fraud signals in the writedown of an inventory task, increased experience 
raised exposure to the recency effect. Their experiment provided very little back-
ground information to subjects and expected experienced auditors to better recog-
nize the inadequacy of the initial information and therefore emphasize subsequent 
information compared with less experienced auditors. Thus, the increased usage 
of subsequent information may have caused greater order effects. The second 
explanation for the results was that between-treatment subjects varied only slight-
ly in their experience (managers vs. senior managers). 
 
Information order, task complexity (task-related factors) and auditor experience 
One study has examined the three-way interaction between information order, 
task complexity and auditor experience. The belief adjustment model (Hogarth & 
Einhorn 1992) predicts that the recency effect continues to bias experienced audi-
tors’ decisions, as experience only mitigates the degree to which experience re-
duces task complexity (Arnold et al. 2000). Trotman and Wright (1996) studied 
this issue in two tasks, namely internal control evaluation and going-concern 
 Acta Wasaensia     45 
  
tasks, which were defined as less and more complex tasks, respectively. Their 
subjects consisted of students, audit seniors and audit managers. The results indi-
cated that students displayed a significant recency effect in both tasks, while sen-
iors displayed the recency effect only in the going-concern task. However, contra-
ry to the predictions of the belief adjustment model, managers did not display the 
recency effect in either task. 
 
Irrelevant information (cue-related factor) and auditor experience 
Experienced auditors may also be able to better ignore irrelevant information than 
less experienced auditors. As suggested in the experience literature generally (see 
e.g. Choo 1989), experienced auditors employ directed information acquisition 
strategies and have advanced knowledge structures, which may help them ignore 
irrelevant information in the audit environment. Shelton (1999) specifically hy-
pothesized that the better knowledge of task-specific relevant cues may help pre-
vent the influence of irrelevant cues in decision-making. In her study, experienced 
subjects were audit partners or managers and less experienced subjects were audit 
seniors. The results indicated that less experienced subjects’ going-concern judg-
ments were significantly influenced by irrelevant information, while experienced 
subjects’ judgments were unaffected by the presence of irrelevant information. 
 
Presentation mode (cue-related factor) and auditor experience 
Compared with the above studies examining experience’s interaction effects, 
studies of presentation mode have not focused on any specific judgmental bias. 
Instead, they have investigated whether presenting information in one particular 
format diminishes judgment quality-related differences between more and less 
experienced auditors. However, there exists little empirical evidence showing that 
presentation mode and experience interact in auditors’ information processing. In 
Bierstaker and Brody (2001), subjects needed to evaluate the quality of the inter-
nal control system, where one group had a narrative description of internal con-
trols  and  the  other  group  was  given  an  additional  flowchart  of  these  controls.  
While the flowchart was considered to be overlapping information with the narra-
tive description, it was expected to interact with experience (no direction was hy-
pothesized). The study failed to find any interaction effect. However, experienced 
auditors outperformed less experienced auditors regardless of the used documen-
tation format. 
Anderson and Mueller (2005) examined the interaction between experience and 
presentation mode in two analytical review tasks. Subjects were audit seniors and 
46      Acta Wasaensia 
accounting students. The presentation mode of given information varied between 
tabular and graphical formats. In the first task, subjects were asked to predict cor-
relations between income statement accounts and related factors that were corre-
lated with the accounts. The results indicated that students benefited more when 
information was presented in a graphical format instead of in a tabular format. 
While both subject groups were more accurate with graphical data, auditors out-
performed students only in the tabular information group. The finding was ex-
plained by the argument that auditors had more previous experience with tables 
only, but not with graphs. 
In a second task, subjects were asked to predict the future value of a sales ac-
count. The results showed again that both subject groups were more accurate with 
graphical data, but that experience had no effect on performance. It was conclud-
ed that the second task was less complex than the first task and that it did not re-
quire the deep information processing of tabular information as the previous task. 
 
Information types (cue-related factor) and auditor experience 
Krogstad, Ettenson and Shanteau (1984) investigated how auditors and students 
utilized financial and nonfinancial information in materiality judgments. The re-
sults indicated that auditors consistently focused on the effect on net income (one 
of the financial cues), but also used various nonfinancial information to fine-tune 
their judgments. By contrast, students did not concentrate on any single cue and 
they used considerably more nonfinancial cues than financial cues. However, the 
overall mean number of cues did not vary between auditors and students, indicat-
ing that less experience does not always increase the number of used cues. 
 
Presentation mode (cue-related factor) and confirmation bias 
A recent study by Ricchiute (2010) argued that whether auditors search confirm-
ing or disconfirming information is conditional on the details of the information. 
The results from financial restatement-related tasks suggested that when infor-
mation is presented in a detailed format instead of in a summary format (as used 
in many previous studies), auditors tend to start their information searches by 
looking for disconfirming information (i.e. information that contradicts the ac-
counting in previous financial statements) from the given material. 
 Acta Wasaensia     47 
  
3.5.2 Interactions and environmental factors 
In this section, the interactions related to environmental factors that have no indi-
vidual factors are presented. The second most common factor in interaction stud-
ies, albeit much less studied than auditor experience, is accountability. These 
studies have hypothesized that accountability either mitigates or exacerbates 
judgmental bias, depending on the interacting factor. However, the empirical re-
sults have mainly evidenced only the positive effects of accountability as an inter-
acting variable. Another factor that has been studied more than in one interaction 
study is time pressure. Specifically, these studies have evidenced that the exist-
ence of time pressure has positive effects in decreasing irrelevant information. 
 
Information order (cue-related factor) and accountability 
Two studies have found strong evidence that the presence of accountability acts 
as a debiasing mechanism for the recency effect. Kennedy (1993) hypothesized 
that those decision-makers who are accountable for their judgments exert more 
effortful information processing that may overcome the recency effect. The study 
had three experimental groups: pre-accountability, post-accountability and no-
accountability. Subjects in the pre-accountability group knew before information 
evaluation that their judgments would be reviewed, while subjects in post-
accountability were told this after information evaluation, but before the likeli-
hood judgment of firm failure. The results indicated that the judgments of sub-
jects in the pre-accountability group showed no exposure to the recency effect, 
while the other groups were affected. However, the difference in the magnitude of 
the recency effect between the post-accountability and no-accountability groups 
was only marginal. 
Cushing and Ahlawat (1996) examined how the documentation requirement af-
fected the recency effect in a going-concern judgment. The documentation re-
quirement was expected to increase overall effort in decision-making, especially 
increasing attention, comprehension and the recall of important information. In 
the experiment, some subjects had to write a memorandum to the senior partner to 
provide reasons to support their report choices. The results showed that subjects 
who had a documentation requirement did not display the recency effect at all, 
while the no-documentation requirement subjects were susceptible to recency 
bias. 
 
 
 
48      Acta Wasaensia 
Irrelevant information (cue-related factor) and accountability 
Few studies have suggested that additional effort caused by the presence of ac-
countability increases proneness to process all available information without con-
sidering its relevance. Hoffman and Patton (1997) hypothesized that accountabil-
ity exacerbates the dilution effect in fraud-risk assessments. Their empirical re-
sults did not confirm the expected exacerbation effect but evidenced that auditors 
were susceptible to the dilution effect regardless of the presence of accountability. 
Likewise, Glover (1997) hypothesized that increased effort and attention encour-
age auditors to employ more integrative and complex thinking in information 
processing that would cause a greater dilution effect. However, the study did not 
find that accountability increased the use of irrelevant information in RMM as-
sessment tasks. By contrast, Cloyd (1997) found that accountability improved tax 
auditors’ abilities to filter out irrelevant information. However, the positive effect 
was only limited to subjects who possessed high knowledge of the relevant tax 
rules. 
 
Irrelevant information (cue-related factor) and time pressure 
Few studies have suggested that moderate time pressure has positive effects on 
audit efficiency and judgment quality. Glover (1997) found in an RMM assess-
ment task that when time pressure increased from a low to a moderate level, audi-
tors’ utilization of irrelevant audit information decreased. This finding suggests 
that moderate time pressure sharpens auditors’ focus on a smaller set of infor-
mation that reduces the acquisition of irrelevant information. However, Choo’s 
(1995) results in an audit confirmation-related task evidenced that when time 
pressure shifted to a high level, the usage of relevant cues also decreased along 
with irrelevant cues. 
 
Information reliability (cue-related factor) and client risk 
Alexander (2003) found that the perceived risk of engagement could interact with 
source credibility in certain circumstances. In this study, the perceived risk of the 
engagement variable was included as a control variable. It used a tax consulting 
domain to investigate simultaneously how the competence of the information 
source and the source’s incentives to be honest affect tax auditors’ information 
acquisition. The results showed that tax auditors responded to a more competent 
source by performing less detailed information acquisition, but this tendency de-
creased when the source had incentives to be less honest. However, the perceived 
risk of engagement was found to be an intervening variable in the experiment. 
 Acta Wasaensia     49 
  
When engagement risk was considered to be low by a subject, auditors relied on 
more competent sources. When the risk was perceived to be high, less competent 
sources were trusted more. This mixed finding was explained by the fact that the 
incentives of the information source dominated the competence of the source. 
3.5.3 Interactions and task-related and cue-related factors 
 
Task complexity (task-related factor) and presentation mode (cue-related factor) 
Wright (1995) suggested that how information presentation affects judgment ac-
curacy depends on task complexity. It was hypothesized that an additional graph-
ical summarization of the information leads to more accurate judgments than bare 
tabular information. However, graphical data were expected to improve perfor-
mance only when the task was highly complex. This is because highly complex 
tasks requiring additional effort and graphical data help free up cognitive efforts 
for information acquisition and processing. The experiment consisted of three 
loan collectability-related judgments. Each judgment was from a different task 
complexity level. The results confirmed the expectations that additional graphical 
information was beneficial only in the most complex task by allowing for unbi-
ased estimates to be made. 
3.6 Conclusions about the studies 
The studies presented in the above sections show that many factors affect both 
consciously and unconsciously auditors’ information acquisition and usage in 
varying audit tasks. The results of these studies suggest at least the following four 
conclusions. 
First, some studies of environmental and cue-related factors suggest that the 
amount and direction of auditors’ cognitive efforts in single audit tasks depend on 
whether particular factors are present in a decision-making situation. For instance, 
studies (e.g. McDaniel 1990; Asare, Trompeter & Wright 2000) have shown that 
a moderate time pressure or the presence of accountability leads auditors to per-
form more efficient or effective information acquisition and usage than would 
occur otherwise. Furthermore, studies of the effect of presentation mode (e.g. 
Anderson & Reckers 1992; Schulz & Booth 1995) suggest that graphical data 
allow users to direct cognitive efforts from more simple information processing 
activities to more critical task activities. 
50      Acta Wasaensia 
Second, several results of the studies of individual and cue-related factors indicate 
that auditors’ cognitive limitations affect information processing. For instance, 
studies of confirmation bias (e.g. McMillan & White 1993) and the order of in-
formation (e.g. Asare 1992) show that auditors are susceptible to judgmental bias. 
Furthermore, studies of task complexity (Simnett & Trotman 1989; Simnett 1996) 
show that the information processing limitations of an auditor constrain perfor-
mance in complex audit tasks. Overall, these studies have found various factors 
that intervene negatively in auditors’ information processing, indicating that the 
substance of information is not the only determinant of auditors’ JDM. Those 
factors are also likely to decrease JDM quality (e.g. Earley 2002). 
Third, studies of individual and environmental factors indicate that auditors’ in-
formation acquisition and usage are responsive to different directional goals in the 
audit environment.  For instance, the studies by Blay (2005) and Guiral, Ruiz and 
Rodgers (2011) indicated that auditors’ information acquisition and processing is 
adjusted according to the pressure to favor client preferences or maximize future 
audit rents. In addition, while individual factors, like motivated reasoning and 
confirmation bias, decrease audit effectiveness (i.e. decision accuracy), they can 
also increase audit efficiency if biased information acquisition leads to the de-
creased usage of available information. Furthermore, increased time pressure may 
lead to the client-preferred evaluation of the information, enhancing audit effi-
ciency, but at the same time, it may compromise audit effectiveness in some de-
gree (Braun 2000; Cianci & Bierstaker 2009). In addition, studies have shown 
that the presence of accountability leads to less objective information evaluation 
when the reviewer’s preference is known before the judgment, as subordinates 
seem to adjust their judgments towards this preference (Peecher 1996; Turner 
2001). Thus, the results of these studies suggest that various directional goals and 
surrounding preferences play an important role, not only in making final decisions 
but also in the preceding information acquisition and usage process. However, 
biased information processing, similar to motivated reasoning, can happen uncon-
sciously by the auditor (Wilks 2002). 
Finally, the results of interaction studies indicate that a relatively large number of 
factors from different categories interact with each other. In particular, these stud-
ies have produced a considerable amount of evidence that auditor experience and 
the presence of accountability mitigate or even reduce judgmental biases related 
to information acquisition and usage. For instance, these results suggest that the 
influence of irrelevant information (Shelton 1999) and the recency effect (Kenne-
dy 1993) on audit judgment is eliminated when a task is performed by an experi-
enced auditor or accountability is present, respectively. 
 Acta Wasaensia     51 
  
Furthermore, interaction studies have documented that initially unseemingly re-
lated factors interact with each other in information acquisition and usage. For 
instance, studies have found evidence that the impact of irrelevant information on 
audit judgments is revoked under moderate time pressure (Glover 1997) and that 
the graphical format of information is beneficial only in complex audit tasks 
(Wright 1995). Overall, the results of interaction studies give valuable insights 
when estimating whether and to what extent judgmental bias is likely to exist in a 
mundane audit environment. 
52      Acta Wasaensia 
4 FACTORS AND HYPOTHESES DEVELOPMENT 
As concluded in Chapter 3, previous research has recognized a considerable num-
ber of factors that affect auditors’ information acquisition and usage. More im-
portantly, these studies have shown that different types (i.e. factors from different 
categories) of factors interact with each other and that the observed interactions 
have explained auditors’ information acquisition and usage behavior in the deci-
sion-making process (see Chapter 3.6). Despite the evident importance of the top-
ic, several factors and their interactions have been studied only unilaterally in an 
audit context. Many previous studies have only focused on examining whether the 
existence of a particular factor changes the processing of information and conse-
quently task-specific judgments. For instance, previous studies (e.g. Anderson, 
Koonce & Marchant 1994; Goodwin 1999) have shown that auditors adjust their 
task-specific judgments depending on the reliability of client-related information. 
Anyhow, these studies have not examined whether an auditor needs more infor-
mation to make a judgment or process it  in a more effortful  way when it  is  less 
reliable. Thus, the notion that many factors influence auditors’ information usage 
in a single audit task has been insufficiently addressed in previous studies. 
4.1 Factors for the empirical analysis 
As noted before,  this study aims to shed light on the gaps in the auditors’ infor-
mation usage literature. To select relevant factors they were first classified into 
four categories in previous chapters. This categorization makes it possible to con-
sider the decision-making process-affecting factors in a broad manner i.e. so that 
all the significant aspects of these factors are included into one analysis. This is 
done by selecting one factor from each category to represent the category in ques-
tion. This approach enables us to gain insights into how different types of factors 
interact with each other in a mundane decision-making environment, where audi-
tors are constantly and simultaneously surrounded by a vast number of environ-
mental, task-related and cue-related factors. To further analyze information usage 
in auditors’ decision-making processes the main interactions between the selected 
factors are also included in the analysis. In the following paragraphs, the selected 
factors and rationale for their selection are presented. 
 
Individual factor: Auditor experience 
Auditor experience is clearly the most studied single individual factor (see Chap-
ter 3.1). A vast amount of auditing literature indicates that experienced auditors’ 
 Acta Wasaensia     53 
  
information acquisition and usage differs from less experienced auditors in sever-
al ways. For example, previous studies have suggested that experienced auditors 
acquire less information and process information more efficiently (e.g. Davis 
1996; Moroney 2007) than less experienced auditors. These observed differences 
have been suggested to stem from experienced auditors’ advanced information 
acquisition strategies (Biggs & Mock 1983) and well-developed knowledge of 
tasks (Waller & Felix 1984; Choo & Trotman 1991). Furthermore, as concluded 
in Chapter 3.6, the effect of many previously investigated factors on information 
usage is conditional on the level of auditor experience. Thus, owing to the strong 
theories from the psychology literature and previous empirical findings, auditor 
experience has been selected to represent individual factors in the empirical anal-
ysis. 
 
Environmental factor: Client risk 
Professional auditing standards e.g. ISA 330 (IFAC 2009a) emphasize that audi-
tors should carry out procedures designed to reduce client risk to an acceptably 
low level. According to theoretical predictions, as client risk increases, the level 
of audit effort should also increase (Sharma, Boo & Sharma 2008). Previous stud-
ies (e.g. Mock & Wright 1993; Houston, Peters & Pratt 1999; Beaulieu 2001) 
have generally found that auditors adjust their information acquisition plans at an 
audit engagement level depending on client-related risks and have found that the 
presence of risk factors increases demand for additional audit evidence. Despite 
the importance of risk in the auditing context, previous research has rarely ad-
dressed the influence of risk on auditors’ information usage behavior in a single 
audit task. In this study, risk has been selected to represent environmental factors 
in the empirical analysis. The present study employs one measure of client risk, 
namely RMM, which is an integrated measure of the different types of audit client 
risks. It can be formed based on source-based risk factors, i.e. inherent and con-
trol risks, or type-based risk factors, i.e. error and fraud risks (Popova 2008). 
Thus, RMM is a combined risk measure that should affect how auditors plan their 
audit procedures, e.g. the nature, timing and extent of audit evidence acquisition 
(Messier 2003). When RMM increases, auditors are expected to adjust these pro-
cedures, especially increasing their evidence acquisition. However, it is unknown 
how auditors adjust their information usage behavior in a single audit task when 
RMM is high. For example, auditors can acquire additional information, use more 
effort to find information or do both to mitigate increased uncertainty. 
  
54      Acta Wasaensia 
Task-related factor: Task structure 
In this study, task structure has been selected to represent task-related factors in 
the empirical analysis. According to Bonner (1994) task structure is the other di-
mension of task complexity,  while the task difficulty is  the other one.  While the 
task complexity overall has been recognized in many studies as an important fac-
tor that affects auditors’ JDM quality (Abdolmohammadi & Wright 1987; Simnett 
& Trotman 1989; Simnett 1996; Tan & Kao 1999; Tan, Ng & Mak 2002), apart 
from the studies by Simnett and Trotman (1989) and Simnett (1996), the effect of 
different task complexity dimensions (i.e. structure or difficulty) on information 
acquisition and usage has been exiguously studied. This study focuses on a task 
complexity dimension, namely task structure, while trying to keep the task diffi-
culty dimension constant between the studied tasks. It is important to distinct 
these  dimensions  of  task  complexity  from each  other  as  changes  in  them might  
have opposite consequences to the information usage16. 
Task structure is an important task dimension, as less structured tasks are more 
demanding for a decision-maker than more structured tasks and because the level 
of structure is likely to affect audit judgment quality (Abdolmohammadi 1999). 
By definition, when a task structure changes from more structured to less struc-
tured, important task-specific information becomes less specified or less clear 
(Abdolmohammadi & Wright 1987). Consequently, a decision-maker may com-
pensate for increased information uncertainty in less structured tasks through 
more extensive information usage than in more structured tasks. This study ad-
dresses empirically this issue in the auditing context. 
 
Cue-related factor: Information reliability  
An auditor must take into account the various properties of information when 
acquiring it. For instance, auditors must consider information cost, relevance and 
reliability. In this study, information reliability has been selected to represent cue-
related factors in the empirical analysis. 
Previous studies (e.g. Hirst 1994; Anderson, Koonce & Marchant 1994; Goodwin 
1999; Glover, Jiambalvo & Kennedy 2000) have widely demonstrated that audi-
                                               
 
16 For example, Ford et al. (1989) suggests that increase in task difficulty increases the use of 
simplifying information search strategies to make tasks more manageable, i.e. decreases in-
formation usage, while the present study argues that increase in a task structure has opposite 
effect.  See also Wood (1986) for other definitions of the task complexity. 
 Acta Wasaensia     55 
  
tors adjust their task-specific judgments depending on information reliability. 
Some (Goodwin & Trotman 1996; Glover, Jiambalvo & Kennedy 2000) have 
found that when management-provided information is less reliable, auditors tend 
to  be  more  skeptical  towards  their  clients  and  plan  to  increase  their  information  
gathering in subsequent tasks. However, these studies have not examined whether 
an auditor needs more information for making judgments or whether they process 
information in a more effortful way when it is less reliable. 
In summary, three factors – RMM, task structure and information reliability – 
have been chosen for this study’s empirical investigation, as how they affect audi-
tors’ information usage behavior in a single audit task has been virtually unstud-
ied. In addition, it is important to study their effects on information usage as they 
are common factors in a mundane audit environment and their levels are likely to 
vary between audit engagements. Auditor experience has also been chosen as it 
has been found to be an influential factor alone and an interacting factor in previ-
ous studies (see e.g. Chapters 3.1 and 3.6). 
4.2 Hypotheses development 
In this section, the hypotheses of the study are developed. All selected factors are 
expected to affect auditors’ information usage. Thus, the first four hypotheses of 
the study concern these main effects. As discussed in previous sections, it is sug-
gested that many influencing factors on information usage might be conditional 
on the level of auditor experience. Therefore, this study builds three interaction 
hypotheses where the effects of auditor experience on the factors’ main effects are 
examined. 
4.2.1 Hypothesis 1 – Auditor experience 
The majority of previous audit studies have found that experienced auditors use 
less information than less experienced auditors. Bédard and Mock (1992) found 
that experts utilized less information in control evaluations than novice auditors. 
Simnett (1996) found that experienced auditors selected fewer ratios in going-
concern assessment than less experienced auditors. Similarly, Davis (1996) found 
that experienced auditors based on their previous knowledge of decision-making 
situations weighted cues more unequally and selected fewer information cues than 
less experienced auditors. 
These observed differences may occur for several reasons. First, the more accu-
rate problem representations of experienced auditors may sharpen finding and 
56      Acta Wasaensia 
interpreting important information more efficiently than the undefined problem 
representations of less experienced auditors (Moroney 2007). 
Second, the research has demonstrated that experienced auditors are able to ig-
nore irrelevant information better than less experienced auditors (Glover 1997; 
Shelton 1999). Experienced auditors may also focus on one type of information 
that decreases overall information usage. Krogstad, Ettenson and Shanteau (1984) 
found that students used more nonfinancial information than auditors. 
Third, experienced auditors may already have more knowledge in their memories, 
which  they  can  rely  on  (Bonner  2008).  Thus,  they  are  exempted  from  external  
information acquisition. Further, experienced auditors may have useful bench-
mark data in their memories for interpreting the problem at hand. For example, 
Butt (1988) showed that experienced auditors had better frequency information 
about the existence of errors in financial statements than less experienced ones. 
Fourth, experienced auditors’ knowledge of correlations between information 
cues in a task may also decrease information usage. If information cues are ex-
ante known to be highly correlated, then additional information may not be used, 
because it is presumed not to have a positive effect on judgment quality (Bonner 
2008). By contrast, when an auditor lacks knowledge on important task-specific 
information, then the correlations might not be properly recognized before con-
ducting the task and all available information will thus be acquired. 
Finally, general information acquisition and usage strategies may vary by experi-
ence level. Experienced auditors are suggested to apply a goal-oriented, directed 
evaluation of audit evidence in their information search processes (Hoffman, Joe 
& Moser 2003).  They also use more rules of thumb and structured checklists in 
information search processes, whereas less experienced auditors use only simple 
sequential searches (Davis 1996). Experienced auditors may even have an internal 
“checklist” as a guideline for the search process (Bonner & Pennington 1991). 
Further, it is suggested that highly experienced professionals are able to recognize 
relevant information patterns by recalling similar situations encountered earlier 
(Lehmann & Norman 2006). Knowledge of typical accounting and control sys-
tems drives experienced auditors in information search, instead of less experi-
enced auditors who usually acquire information in the order in which it is pre-
sented (Davis 1996).  
 Acta Wasaensia     57 
  
Thus, it is expected that in experimental decision-making task less experienced 
subjects use more extensively information17 than experienced auditors. The above 
discussion leads to the first empirically testable hypothesis: 
H1: Less experienced auditors use the available information in deci-
sion-making more extensively than experienced auditors. 
4.2.2 Hypothesis 2 – RMM  
Generally, an auditor has several options when encountering an audit client with 
high RMM. The auditor may try to get rid of this client, compensate for it by car-
rying out additional audit procedures or bill fee premiums to cover the possible 
consequences of loss of reputation or litigation (Johnstone 2000). If the auditor 
decides to retain the high risk client, performing additional audit procedures is 
usually necessary. 
Previous research has investigated how auditors conduct audits depending on cli-
ent risk factors. One stream of research has investigated how audit resources are 
allocated when there exist risk factors or the assessment of overall risk has 
changed. The focus of these studies has been on how the perceived level of inher-
ent and/or control risk affects subsequent audit planning decisions. 
Some archival studies (Mock & Wright 1993; 1999) have found that changes in 
client risk factors have only a minor impact on general audit planning decisions. 
Other such studies (O’Keefe, Simunic & Stein 1994; Hackenbrack & Knechel 
1997) have found that the level of inherent risk (but not control risk) affects staff 
work hours. Audit staff work usually includes collecting audit evidence 
(Hackenbrack & Knechel 1997). Johnstone and Bedard (2001) found that in the 
initial engagement planning stage, clients with error risk factors were more likely 
to be tested more intensively than other clients. In behavioral studies, the strong 
link between an auditor’s risk judgments and audit planning has been demonstrat-
ed. Gaumnitz et al. (1982) provided evidence that auditors’ evaluations of internal 
control riskiness influenced their judgments of the audit hours used to accomplish 
the task. Kaplan (1985) reported that high inherent risk and high control risk 
jointly  (i.e.  RMM)  increased  significantly  the  number  of  planned  audit  hours  
compared with when both risks were low. 
                                               
 
17 In the present study term ”the more extensive use of information” refers to spending more 
time for reading, combining and evaluating of cues and/or acquiring greater number of infor-
mation cues 
58      Acta Wasaensia 
Another stream of research (Houston, Peters & Pratt 1999; Beaulieu 2001) has 
investigated specifically how client riskiness affects audit evidence. Houston, 
Peters and Pratt (1999) found that when there was unintentional, inconsistent ac-
counting treatment, auditors planned to acquire more audit evidence. Beaulieu 
(2001) found that auditors’ judgments of a prospective client’s CFO integrity 
were linked to the recommendations of audit evidence amount. The results specif-
ically indicated that when the CFO was perceived as dishonest, auditors planned 
to collect more audit evidence and vice versa (i.e. evidence collection judgments 
were adjusted in response to client integrity). Blay, Sneathen and Kizirian (2007) 
used archival data from audit engagements to show that auditors’ assessments of 
RMM were associated with their evidence searches. 
There are at least four arguments for high risk increasing information usage in a 
single audit task. First, high RMM may prompt auditors to act more conservative-
ly18 in a decision-making process, which ultimately aims to minimize economic 
losses (Smith & Kida 1991; Mueller & Anderson 2002). Alternatively, high 
RMM may cause auditors to demonstrate greater professional skepticism than 
they do usually (Shaub & Lawrence 1996; Quadackers, Groot & Wright 2009). 
Consequently, auditors who show increased professional skepticism have been 
shown to acquire more information (Hurtt, Eining & Plumlee 2008). Increased 
conservatism and/or professional skepticism can also trigger auditors to increase 
effort for the task, particularly leading to more extensive information processing 
in order to find errors or inconsistencies between different cues. 
Second,  high  RMM  may  change  auditors’  problem  representations  of  a  task.  It  
has been found that different problem representations affect subsequent infor-
mation usage within the same audit task (Rosman, Seol & Biggs 1999). Specifi-
cally,  because of high RMM auditors may switch to a problem audit  representa-
tion from a normal audit representation, leading to increased information usage 
(Waller & Felix 1984; Asare & Knechel 1995). 
Third, initially encountered negative characteristics (i.e. high RMM in this case) 
in audit tasks may cause auditors to display confirmation bias more easily than 
when neutral/positive characteristics are encountered (Trotman & Sng 1989). If 
an auditor is susceptible to confirmation bias, she/he may perform extensive in-
formation acquisition to find any negative information to support this initial view 
(assuming that negative information is not instantly found) (McMillan & White 
                                               
 
18 An auditor’s conservatism is defined by Smith and Kida (1991) as follows: “A tendency to 
give more attention to, and to be more influenced by, negative information or outcomes”. 
 Acta Wasaensia     59 
  
1993). Furthermore, there is also contrary evidence that high client risk may miti-
gate proneness towards confirmation bias if the risk of litigation is more salient 
(Kadous,  Magro  & Spilker  2008).  This  has  also  been  shown to  lead  to  effortful  
and balanced information usage (Kadous, Magro & Spilker 2008). Finally, the 
presence of high RMM may increase the uncertainty of the task outcome and con-
sequently lead to more extensive information usage to reduce this increased un-
certainty (Blay, Kadous & Sawers 2012). 
To summarize, at the audit planning level earlier research indicates that auditors 
adjust their audits and plans to increase audit effort depending on client risk (in-
cluding RMM). In addition, in single audit tasks auditors’ cognitive processes 
related to conservatism, professional skepticism, problem representation, confir-
mation bias and increased uncertainty suggest that auditors use available infor-
mation more extensively when client-related risks are high. The above discussion 
leads to the second empirically testable hypothesis: 
H2: Information is used more extensively in decision-making when 
RMM is high compared with when RMM is low. 
4.2.3 Hypothesis 3 – Task structure 
Task structure is  a dimension of task complexity that defines how well  a task is 
specified a priori for a decision-maker (Bonner 1994). Performing a task usually 
consists of three phases of general information processing: input, processing and 
output.  In  this  connection,  task  structure  specifically  refers  to  a  task’s  different  
elements in each of these information processing phases, i.e. defines the task’s 
overall structure. These elements are related to information clarity, e.g. 
cue/procedure/goal specification, cue measurement, magnitude of input/output 
relations and cue consistency (Bonner 1994). 
Previous research in this area has been scarce. Simnett and Trotman (1989) ma-
nipulated task structure by giving one- or two-year before-the-event financial ra-
tio information to auditors. They found no evidence that task structure affected 
the types of ratios chosen by auditors in failure prediction task. However, it is not 
clear if the finding is related only to that particular task, as the environmental pre-
dictability manipulation (i.e. will failure occur in one or two years) of bankruptcy 
is task-specific.  
Following the model of audit task complexity by Bonner (1994), when task struc-
ture shifts from structured to unstructured and as the cue, procedure or goal speci-
fication becomes less clear, a decision-maker may choose more information to 
60      Acta Wasaensia 
alleviate the uncertainty of less structured tasks. Similarly, when the magnitude of 
the input/output relations of information is not clear for the decision-maker, 
she/he may feel that more information is needed to make a judgment. 
Abdolmohammadi and Wright (1987) suggested three types of task structures: 
structured, semi-structured and unstructured. Specifically, in structured and semi-
structured tasks there is less uncertainty about important information than in un-
structured tasks (Abdolmohammadi & Wright 1987). In addition, the lack of 
specified procedures in unstructured tasks might cause information usage to be 
exaggerated if a decision-maker perceives that additional procedures can compen-
sate for this uncertainty. Taken together, when an auditor is performing unstruc-
tured tasks, she/he is expected to use information more extensively than in more 
structured tasks in order to alleviate the anxiety that stems from the increased 
uncertainty of the important information and the less specified procedures of a 
task. The above discussion leads to the third empirically testable hypothesis: 
H3: Information is used more extensively in decision-making when a 
task is less structured compared with when a task is more structured. 
4.2.4 Hypothesis 4 – Information reliability 
When information is acquired or processed, it is necessary to consider infor-
mation reliability, as this affects information persuasiveness (Pornpitakpan 2004). 
In audit research, information reliability has been estimated by source credibility 
(see Chapter 3.4.2). This is based on the assumption that a highly credible source 
is more persuasive in decision-making (Pornpitakpan 2004). Thus, it is assumed 
that information from a more credible source is more reliable than information 
from a less credible source. 
Several studies have examined the effects of source credibility components on 
auditors’ judgments. First, studies of source competence have found that auditors 
place more weight in their decision-making on information received from more 
competent information sources compared with when the same information is re-
ceived from less competent information sources (Bamber 1983; Rebele, Heintz & 
Briden 1988; Anderson, Koonce & Marchant 1994). In addition, Goodwin and 
Trotman (1996) found that when information was obtained from the less compe-
tent source, auditors planned to use more audit hours for subsequent audit evi-
dence gathering. 
Second, the studies of the independence component of source credibility have 
documented mixed results. Joyce and Biddle (1981) found that auditors were not 
 Acta Wasaensia     61 
  
always sensitive to the information source in an accounts receivable task, while 
Hirst (1994) found that both source competence and source objectivity influenced 
auditors’ misstatement estimates. Finally, studies (Goodwin 1999; Glover, Jiam-
balvo & Kennedy 2000) of the integrity component of source credibility have 
found that auditors are sensitive to the perceived integrity of management. Good-
win (1999) indicated in two separate audit tasks that auditors are always sensitive 
to the integrity of the information source regardless of whether it from inside or 
outside of the client. Glover, Jiambalvo and Kennedy (2000) found that when 
management had an incentive to misstate the reason for an unexpected fluctua-
tion, auditors planned to acquire more additional information than when an incen-
tive was not present. 
To summarize, the results of these studies indicate that auditors adjust their judg-
ments depending on information reliability, which stems from varying compo-
nents of source credibility. From this study’s standpoint, the results of Goodwin 
and Trotman (1996) and Glover, Jiambalvo and Kennedy (2000) are especially 
important, as they indicate that information from a less competent or low integrity 
source causes auditors to deem that information to be less reliable and conse-
quently increase plans for additional information acquisition. 
From a single audit task perspective, it is suggested that less reliable information 
is weighted less than more reliable information in decision-making (Kizirian, 
Mayhew & Sneathen 2005). Specifically, less reliable information has a smaller 
correspondence between information signal and the resulting outcome compared 
with more reliable information (Knechel & Messier 1991). Therefore, auditors 
might compensate for the smaller persuasiveness of less reliable information by 
acquiring more information (i.e. increasing the depth of search) or processing it in 
a more effortful way to assess its usefulness (i.e. finding out its information val-
ue). 
Furthermore, it is important to note that the whole information acquisition and 
usage process aims at reducing uncertainty in decision-making (Koonce 1993). 
Thus, if an auditor perceives information as less reliable, she/he may need to per-
form more extensive information usage to reach a sufficiency threshold in uncer-
tainty reduction than when information is more reliable. The above discussion 
leads to the fourth empirically testable hypothesis: 
H4: Information is used more extensively in decision-making when 
information is less reliable compared with when it is more reliable. 
62      Acta Wasaensia 
4.2.5 Hypothesis 5 – RMM and auditor experience 
Auditor experience is expected to interact with RMM for three reasons. First, pre-
vious audit studies (Choo & Trotman 1991; Moeckel 1990) have hinted that pro-
fessional skepticism about the existence of errors decreases when experience in-
creases. This may stem from the fact that experienced auditors possess more ac-
curate knowledge of the errors in financial statements than less experienced ones 
(Nelson 2009). Experienced auditors are also suggested to be able to explain non-
errors  in  their  audit  findings  more  easily  than  they  can  errors  (Nelson  2009).  
Thus, even under high RMM, experienced auditors may perceive many unusual 
audit findings as non-errors and are less likely to alter their usual information 
usage. By contrast, high RMM may exaggerate less experienced auditors’ reac-
tions to deem any unusual finding as an indication of an error and consequently 
cause more extensive information acquisition. 
Although experienced auditors react to high RMM by increasing their information 
usage, they may not engage in extensive information search if their expectations 
of the error are not supported by the first pieces of acquired information (Earley 
2002). 
Second, degree of auditor conservatism might be reflected in the extent of infor-
mation usage. Previous research has suggested that less experienced auditors are 
generally more conservative and less optimistic than experienced ones (Libby 
1995). For example, Libby (1995) summarized that experienced auditors have 
been found to give more positive estimates of probabilities of continuation as a 
going-concern than less experienced ones. Similarly, Abdolmohammadi and 
Wright (1987) found that less experienced auditors were more likely to make au-
dit adjustments than experienced ones. These findings imply that the conservative 
judgments of less experienced auditors may also affect information usage. They 
might especially use more effort to find negative information or even errors from 
the available information when RMM is high. Consequently, this could lead less 
experienced auditors to perform extensive information usage. 
Finally, less experienced auditors may also be more susceptible to confirmation 
bias than experienced auditors. Kaplan and Reckers (1989) found that in explain-
ing ratio fluctuations only less experienced auditors were sensitive to confirma-
tion bias. When RMM is high, less experienced auditors might be more willing to 
find confirming information of initially negative indicators than experienced ones. 
This may lead to more extensive information usage than under circumstances in 
which disconfirming information is considered to be diagnostic information. The 
above discussion leads to the fifth hypothesis: 
 Acta Wasaensia     63 
  
H5: In the context of high RMM, a less experienced auditor uses in-
formation more extensively than an experienced one. 
 
 
Figure 4. Graphical illustration of hypothesis 5 (“+” indicates increase in in-
formation usage, “0” is a baseline) 
 
4.2.6 Hypothesis 6 – Task structure and auditor experience 
Differences between experienced and less experienced auditors in information 
usage behavior might exist only when a task is unstructured (Bonner 1990). In 
particular, when a task structure changes from structured to unstructured and 
when the clarity of important information becomes more unspecified, the better 
knowledge of experienced auditors may help them perform their usual infor-
mation usage process. 
Different task structures demand different cognitive processes from an auditor. 
These cognitive processes are construction and reduction processes. Construction 
processes are demanding functions usually needed only in unstructured tasks, 
while reduction processes are required in all structure levels. When a task requires 
construction processes, e.g. information comprehension, hypothesis generation 
and hypothesis design, the benefits of experience may have positive impacts for 
decision-making. If a task requires only reduction processes, e.g. hypothesis eval-
uation, estimation and choice, knowledge differences may have only a small or no 
influence at all on decision-making (Bonner & Pennington 1991). 
High risk
Low risk
Risk of material 
misstatement
Experience
Low High
+ + + + +
+ 0
H5
64      Acta Wasaensia 
Further, despite a task being an unstructured one, experienced auditors may still 
be able to apply their usual directed information acquisition approach if they pos-
sess the necessary knowledge of the task. Experienced auditors may have a better 
developed understanding of task-specific diagnostic information and an ability to 
recognize information input/output relations better than less experienced ones. 
In summary, experienced auditors’ better cognitive processes and highly devel-
oped knowledge of unstructured tasks may compensate for a task’s complexity 
and help perform information usage in a similar way compared with more struc-
tured tasks. The above discussion leads to the sixth hypothesis: 
H6: The level of task structure does not affect the information usage 
of experienced auditors. 
 
 
Figure 5. Graphical illustration of hypothesis 6 (“+” indicates increase in in-
formation usage, “0” is a baseline) 
 
4.2.7 Hypothesis 7 – Information reliability and auditor experience 
Information reliability is expected to interact with auditor experience. A general 
theory from psychology suggests that a decision-maker who trusts an information 
source and perceives that information to be highly reliable may only focus on the 
content of information instead of questioning its relevance and pertinence (Sacchi 
& Burigo 2008). Thus, when experienced decision-makers rely on the reliability 
of the information source, they tend to use less cognitive efforts and apply se-
Semi-struct.
Unstruct.
Task structure
Experience
Low High
+ 0
+ + 0
H6
 Acta Wasaensia     65 
  
quential information search strategies that are similar to those usually used by less 
experienced decision-makers (Sacchi & Burigo 2008). 
The above theory suggests that experienced and less experienced auditors may 
acquire information in the order of its presentation when they trust that infor-
mation is from a competent source. Thus, experienced auditors may reject their 
typical goal-directed information usage strategies (Hoffman, Joe & Moser 2003), 
which they normally apply in the search of diagnostic information, and use more 
information in their decision-making. Therefore, the degree of the sequential in-
formation usage of experienced auditors is expected to increase when information 
is more reliable. By contrast, when information is less reliable, differences in in-
formation usage strategies across different experience levels are expected to 
emerge. The above discussion leads to the seventh and final hypothesis of the 
thesis: 
H7: In the context of less reliable information, a less experienced au-
ditor uses information more extensively than an experienced one. 
 
 
Figure 6. Graphical illustration of hypothesis 7 (“+” indicates increase in in-
formation usage, “0” is a baseline) 
 
Less-reliable
More-reliable
Reliability of 
information
0
Experience
Low High
+ + + + +
+
H7
66      Acta Wasaensia 
5 EXPERIMENT AND DATA 
This chapter presents the experiment and data of this study. First, the two tasks 
that are applied in this study’s experiment, client acceptance and client continu-
ance, are introduced by presenting previous research on them. Next, based on 
previous studies in the client acceptance literature, four information classes are 
formed that are used for the experiment. This is followed by the presentation of 
12 cues in the experiment’s information menu. 
In the next section, the general properties of web experiments are briefly dis-
cussed. Then, the experimental design, subjects and procedures for this study are 
introduced. The next two sections define the dependent and independent variables 
of the study, respectively.  Then, the handling of the data is  described as well  as 
the rationale to exclude observations from the data set. Finally, the descriptive 
statistics of subjects’ demographic information and cue usage are presented. 
5.1 Client acceptance and continuance tasks 
Client acceptance and continuance tasks are chosen for this study because alt-
hough they have many similarities, their task structure level varies. As argued in 
the following paragraphs, client acceptance can be classified as an unstructured 
task and client continuance as a semi-structured task, while both consist of many 
similar judgmental elements. Furthermore, because of their high similarities, the 
available information for a decision-maker can be presented in almost identical 
form in both tasks, which allows for the direct comparison of information usage 
between them. 
Before an auditor can accept a prospective client, s/he must usually become fa-
miliar with its background and business environment. Thus, the auditor needs to 
exercise a considerable amount of consideration and effort for information acqui-
sition and usage. This can be especially emphasized in the audit environment 
where an auditor is worried about a prospective client’s high risk indicators or 
other unusual circumstances, and therefore wants to find confirming or discon-
firming information to justify the rejection or acceptance of a client. 
In contrast to client acceptance decisions, client continuance decisions are usually 
less demanding, as the suitability of a client has already been assessed in the ini-
tial acceptance decision. In continuance decisions, auditor may only focus on 
monitoring the material changes in factors and conditions that have happened 
recently to the client (IFAC 2010). Thus, in a client continuance task less effort 
from the auditor is expected (Waller & Felix 1984). Specifically, it is proposed in 
 Acta Wasaensia     67 
  
this study that when a new auditor represents the same audit firm, s/he is mainly 
interested in the success of the previous audit and the most important material 
changes that have occurred recently in the client entity and/or business environ-
ment. Therefore, it is argued that a new auditor is generally able to make a con-
tinuance decision with the smaller usage of available information than a new au-
ditor in an initial acceptance decision. 
In addition, following the argumentation of Bonner’s (1994) framework of task 
structure, it is posited that client continuance tasks are more structured than client 
acceptance tasks. In particular, information clarity may be greater in client con-
tinuance tasks than in new client acceptance tasks because the input/output rela-
tion of the information is better specified, i.e. the firm has been already accepted 
by the auditor to be the client. In other words, the client’s input information has 
led to a certain output (acceptance). Thus, continuing with the existing client can 
be considered to be the default option for the decision. The procedures are better 
specified, because an auditor can focus on the major changes that have occurred 
in the client entity and environment recently. By contrast, in acceptance tasks an 
auditor is required to perform an in-depth analysis of the potential client’s back-
ground and environment and use more overall consideration, as there are no pre-
vious judgments on which to rely. 
The other reason for choosing client acceptance and continuance tasks for this 
study is that pre-engagement decisions for evaluating risky clients are highly im-
portant in audit firms’ overall risk management strategies (Huss & Jacobs 1991). 
However, little is known about how auditors acquire and use information in client 
acceptance and continuance judgments. Asare and Knechel (1995) investigated 
how the information type (positive/negative) of a prospective client affected the 
number of acquired information. Positive information favored acceptance, while 
negative information recommended rejection. They found that on average audi-
tors terminated information acquisition after receiving seven negative information 
cues and then made the rejection decision. When subjects received only positive 
cues, they tended to acquire all available information and then accept the client. 
The findings confirmed their expectations that auditors focus only on finding 
negative information that would indicate that the rejection of a prospective client 
is necessary. 
However, the study by Asare and Knechel (1995) has some limitations that 
should be taken into account when generalizing the results. First, subjects re-
ceived all cues in a predetermined order and they could not re-read previously 
seen information. In addition, subjects did not know the information pool that was 
available to them before they acquired the maximum number of information. Se-
68      Acta Wasaensia 
cond, the study was conducted in the US where exposure to legal liability is much 
higher than in Scandinavia (La Porta, Lopez-de-Silanes & Shieifer 1998). Thus, 
the generalizability of the findings in Asare and Knechel (1995) to other envi-
ronments is potentially hindered by the environment that emphasizes the total 
avoidance of risky clients. Therefore, it is proposed that in less litigious environ-
ments auditors may perform more balanced information acquisition, where neu-
tral or positive information is also considered to be diagnostic. 
Client acceptance is described as a multidimensional decision that consists of two 
phases (Johnstone 2000). The first phase includes the evaluation of relevant risks. 
Engagement risk is the overall risk associated with accepting a new client and it 
comprises three components: 1) client’s business risk, 2) audit risk and 3) audi-
tor’s business risk. A client’s business risk consists of factors that affect the prof-
itability and continued existence of the client. Audit risk is defined as the risk that 
the  auditor  fails  to  modify  his/her  audit  opinion  when  financial  statements  are  
materially misstated. An auditor’s business risk concerns the possibility that an 
auditor is sued or that s/he receives adverse publicity because of being connected 
to a particular client (Johnstone 2000). 
The second phase is a risk-adaption phase where an auditor considers ways to 
respond to evaluated risks. These screening strategies are based on clients’ risk 
levels, the interaction of risks (considering the mediation effect) and proactive 
measures for controlling risks (Johnstone 2000). First, an auditor may make a 
screening decision only based on levels of risk, i.e. by rejecting clients that have a 
high level of audit risk or client business risk. Second, some audit firms may con-
sider  the  mediating  effect  of  auditor  business  risk  on  high  audit  risk  or  client  
business risk. For example, an auditor can decrease the loss on engagement risk 
by using highly experienced or industry-specialist auditors in a risky audit. Third, 
an auditor may use proactive steps to adjust  heightened audit  risk or client busi-
ness risk by increasing the audit fee or acquiring more information during the 
acceptance process for the decision (Johnstone 2000). 
In an experimental study, Johnstone (2000) reported evidence that auditors pre-
ferred to avoid risky clients rather than proactively adapt to the increased risk. 
Similarly, other studies (e.g. Cohen & Hanno 2000; Asare, Cohen & Trompeter 
2005) have shown that clients with low integrity managers are more likely to be 
rejected. However, some studies have found evidence that when audit firms ac-
cept high risk clients they tend to increase the level of substantive tests, bill high-
er rates or assign more experienced auditors to the engagement (Beaulieu 2001; 
Johnstone & Bedard 2003; Ethridge, Marsh & Revelt 2007). In summary, these 
results stress the importance of risk evaluations in client acceptance decisions. 
 Acta Wasaensia     69 
  
A survey conducted by Johnstone (2001) plotted the importance of various en-
gagement risk factors. She also studied whether there were differences in the per-
ceptions between less and more experienced audit partners. The results showed 
that less experienced partners had more variability in their answers than experi-
enced ones. Both respondent groups agreed that the category containing audit 
risk-related factors was the most important in decision-making. Within that cate-
gory, “management attitude towards internal controls” was ranked the most im-
portant factor by experienced partners. Less experienced partners ranked the “na-
ture of the relationship between client and their previous auditor” as the most sig-
nificant factor. Both groups valued other factors equally, such as the “subjective 
valuation of the client’s assets”,  “industry growth patterns” and “existence of an 
internal audit department”. In a client’s business risk category, “financial trends” 
were ranked as the most important factor followed by “industry comparisons”, 
“long-term planning” and “industry competition information”. In an auditor’s 
business risk category, the rank order of factors was “Expertise in the potential 
client’s industry”, “IPO information”, “spin-off work likelihood”, “timing of the 
engagement” and “pricing strategies of competing firms”. Respondents were also 
asked to name additional unlisted factors. “Information about management” and 
“discussions with management” were the most common answers. “Discussions 
with the previous auditor, client’s bankers or lawyers” were also frequently cited 
factors. 
Bell et al. (2002) developed a decision aid to make client acceptance risk assess-
ments. Based on the overall risk assessment, the appropriate level of client review 
can be determined. The authors considered a comprehensive set of risk factors 
based on numerous real cases and related interviews. Risk factors were divided 
into seven categories depending on their natures and the risk types they affected. 
The first category included client entity characteristics that affect an auditor’s 
business risk. The second category focused on engagement information mainly 
from an independence and a client relationship perspective. These factors may 
affect an auditor’s business risk. The third category included information about 
the quality of the relationship with the client, which may affect audit risk and an 
auditor’s business risk. These factors were mainly derived from the client’s third 
parties such as previous auditors, bankers and lawyers. 
The fourth and fifth categories included quantitative and qualitative factors con-
cerning the client’s business environment, industry trends, level of competition 
and legal environment. These factors affect a client’s business risk, but may also 
influence audit risk. The sixth category included risk factors that concern the 
quality of a client’s control environment and the characteristics of management 
70      Acta Wasaensia 
and board that directly impact on audit risk. Finally, the seventh category’s fac-
tors included the quality of accounting and the results of previous audits affect 
audit risk. 
In a recent survey, Ethridge, Marsh and Canfield (2007) found that audit partners 
ranked management integrity as the most important risk factor. They concluded 
that management is responsible for the entity’s control environment and thus that 
management’s attitude towards internal controls is a key factor in mitigating risks. 
Finally, some studies have compared client acceptance decision-making between 
auditors and audit firms. It is generally recognized that audit firms usually have 
written client acceptance policies that aim to standardize procedures (Huss & Ja-
cobs 1991; Gendron 2001). Huss and Jacobs’s (1991) interviews among Big-6 
audit firms showed significant differences between audit firms in specifying im-
portant information and sources. Gendron’s (2001) field study showed that even 
when decision-making procedures are established and mechanic decision aids are 
available, auditors are still most likely to make the most demanding client ac-
ceptance decisions using a considerable amount of personal judgment, i.e. not 
using decision aids extensively. 
In summary, previous client acceptance studies have demonstrated that auditors 
focus on negative information from their prospective clients. This also implies 
that their information usage does not equally consider negative and positive in-
formation. Based on these findings, it can be inferred that auditors accept a pro-
spective client if they are not able to find information that would suggest doing 
otherwise. However, it is not obvious whether these results can be generalized 
directly to this study’s settings, as all those studies were conducted in environ-
ments  where  the  risk  of  litigation  is  substantially  higher  than  that  in  this  study.  
Thus, a degree of negative/positive information usage might be conditional on the 
environment’s legal liability. 
5.2 Information cues 
Information cues for the experiment are selected based on earlier client ac-
ceptance studies (Asare & Knechel 1995; Johnstone 2000; Johnstone 2001; Bell 
et al. 2002; Ethridge, Marsh & Canfield 2007), the IFAC’s guidance of “Client 
Acceptance and Continuance” and the research problem of the present study. 
The most highlighted factor found in previous studies (Asare & Knechel 1995; 
Johnstone 2001; Ethridge, Marsh & Canfield 2007), namely the characterization 
and integrity of management, is included in RMM manipulation. The main crite-
 Acta Wasaensia     71 
  
ria for cue selection are their popularity or commonness in previous experimental 
studies of client acceptance decisions. The IFAC’s guidance is mainly used to 
form information categories. 
Based on these studies, information cues are divided into four classes (Figure 7): 
financial, industry and competition, previous audit and third parties, and organiza-
tion and internal controls information. The information set for the experiment 
aims for a systematic mix of financial and nonfinancial information on the (pro-
spective) audit client and its business environment. Financial information consists 
of publicly available cues that provide an overview of the client’s financial status. 
Nonfinancial information contains both internal and external information on the 
client. Internal information includes, for example, information on the effective-
ness of internal controls, while external information includes cues about industry 
outlook and the competitive environment. 
 
 
Figure 7. Information classes of client continuance and acceptance tasks 
 
In the experiment, each of these four information classes is operationalized by 
three separate information cues. In the information menu, cues are presented in 
alphabetical order. Information cues contain neutral or slightly positive infor-
mation about the firm or its business environment. Thus, the aim is to present 
cues that make the firm look like an average small firm without strong positive or 
negative information. The cues were pilot-tested in several rounds with experi-
enced auditors as well as other accounting professionals to ensure that there was 
sufficient appropriate information available for the given tasks. 
72      Acta Wasaensia 
Table 1 presents the 12 cues used in this study. Financial trends have been found 
to be the most important factor when evaluating a client’s business risk (John-
stone 2001). Financial information contains three cues to represent a client’s fi-
nancial position and trends. The income statement and balance sheet are presented 
in their regular forms to show the steady trends consistent with the industry’s av-
erage profitability, liquidity and solvency. Key ratios includes a summary of nine 
financial ratios covering growth, profitability, liquidity and solvency trends. All 
information is from the past three years in order to present the financial trends of 
the firm (Johnstone 2000; Johnstone 2001). 
Industry comparison and competition information have also been recognized to be 
important factors in client acceptance decisions (Johnstone 2000; Johnstone 2001; 
Bell et al. 2002). The cues in this class aim to provide information about a client’s 
business  risks.  All  presented  cues  were  ranked  next  important  after  financial  
trends (Johnstone 2001). The industry comparison cue compares the key ratios of 
the client to other firms in the same industry. The industry status and outlook cue 
describes how the whole industry has been developed compared with the previous 
year and the general expectations for future sales and estimated trends. The com-
petitive environment cue describes the qualitative analysis of the business envi-
ronment and the client’s long-term strategy to manage competitive pressure. 
Organizational and internal controls information contains cues about the client’s 
organizational structure and accounting and internal controls (Johnstone 2000; 
Bell et al. 2002; IFAC 2010). These cues aim to provide information on audit risk 
factors, i.e. factors that should be taken into account when conducting the audit 
(Johnstone 2001). The structure of organization cue presents an overview of the 
number of personnel at each organizational level, while the internal controls cue 
assesses the client’s risk factors and related controls. This cue also includes the 
previous overall assessment of the internal control system. The accounting and 
bookkeeping (financial administration) cue describes how bookkeeping and fi-
nancial planning are organized. This cue also describes the accounting system that 
is currently used in the client firm. 
Previous audit and third party information contains information on the results of 
previous audits and on the client’s relationships with external parties. Information 
on previous audit reports and the client’s relationship with a previous auditor are 
recognized to be important factors when making client acceptance decisions 
(Johnstone 2000; Johnstone 2001; Bell et al. 2002; IFAC 2010). Thus, the previ-
ous audit cue in this experiment covers information on previous audit reports and 
the reasons for changing auditor. Cues about third parties include relationship 
with the main creditor and lawyer. These cues are also perceived to be important 
 Acta Wasaensia     73 
  
by auditors (Johnstone 2001). The relationship with the main creditor cue shows 
that the client has the usual relationship with the bank and has repaid loans in 
time. This cue also indicates that clients have had no defaults in recent years. The 
relationship with the lawyer cue states the client’s lawyer confirmation that there 
are no signs of any juridical matters in the past or present. 
 
Table 1. Cues of the information menu 
Class Cue  
FINANCIAL Income Statements 
FINANCIAL Balance Sheets 
FINANCIAL Key Ratios  
INDUSTRY AND COMPETITION Industry Comparison 
INDUSTRY AND COMPETITION Industry Status and Outlook 
INDUSTRY AND COMPETITION Competitive Environment 
ORGANISATION AND INT. CTRLS Structure of Organization 
ORGANISATION AND INT. CTRLS Internal Controls 
ORGANISATION AND INT. CTRLS Accounting and Bookkeep. (financial admin.) 
PREVIOUS AUDIT AND THIRD PARTIES Predecessor Audit information/Prior Audit 
PREVIOUS AUDIT AND THIRD PARTIES Relationship with the main Creditor 
PREVIOUS AUDIT AND THIRD PARTIES Relationship with the Lawyer 
 
5.3 Web-based experiment 
The data in this study are acquired through an experiment that uses a computer-
based tracking technique. This type of experiment allows for advanced possibili-
ties to trace each subject’s information usage, which is not possible with tradi-
tional paper-and-pencil experiments (Bryant, Hunton & Stone 2004; Andersson 
2004). Alternatively, the current research problem could be studied by using ver-
bal protocol analysis (see Chapter 2.2). This method would also have generated a 
rich set of insights into the decision-making process. However, the main disad-
vantage of verbal protocol analysis is that investigating four independent varia-
bles would have required several laborious experimental sessions. These time-
consuming sessions would have been restricted to a small number of auditors, 
which might not have sampled the population representatively (Andersson 2004). 
74      Acta Wasaensia 
The advantage of web-experiments is the availability of large sample sizes. Thus, 
greater statistical power can be expected. The chances of human data entry and 
transcription errors in the analysis phase are also minimized. Furthermore, the 
risk of the experimenter’s unnecessary influence on subjects is reduced, because 
face-to-face contact is not necessary (Bryant, Hunton & Stone 2004). 
It was decided to conduct the experiment online instead of through series of su-
pervised experiments in computer rooms. In particular, a large number of treat-
ment groups would require a considerable sample of subjects for reliable statisti-
cal analyses. Thus, it would have been difficult to arrange sessions comprising 
subjects with different experience levels. These sessions would have also required 
more time from the experimenter and subjects, which would have incurred signif-
icantly more research costs in terms of travelling expense and the renting of com-
puter space (Bryant, Hunton & Stone 2004). 
5.4 Experimental design 
The experiment in this study consists of two between-subjects audit tasks. In the 
client continuance task, subjects are asked to evaluate their willingness to contin-
ue a relationship with an existing client where they would act as an (new) incum-
bent auditor. A client acceptance task is otherwise identical but now the client is 
new to the audit firm. 
There  are  two  manipulated  factors  in  the  experiment,  level  of  RMM  and  infor-
mation reliability, which are manipulated at two levels. Subjects’ task-specific 
experience is collected using the post-experimental questionnaire. Subjects are 
then classified as less experienced and experienced subjects according to their 
task-specific experience. Thus, there are eight treatment groups as presented in 
Figure 8. Subjects are randomly assigned to one of these groups. 
Both tasks are based on almost identical information. The background text and 
information cues between the tasks are only slightly modified to reflect either the 
client continuance or the client acceptance situation. The firm presented in the 
experiment is a small-sized firm operating in the boat manufacturing industry. 
Financial information on the firm is based on a real firm selected from the Voit-
to+ database, which is a large Finnish company information database. Financial 
information is modified to make the firm unidentifiable. 
 
 Acta Wasaensia     75 
  
 
Figure 8. Treatment groups of the study 
 
Client continuance task 
The semi-structured task is a client continuance task. In this task, a subject is 
asked to evaluate if her/his audit firm can continue with the existing client, where 
she/he would be a new auditor. In the background information, it is stated that the 
relationship with the current audit firm was established five years ago and that the 
client firm’s board of directors wants to continue the relationship. Furthermore, 
the Auditing standards,  such as the ISQC 1 paragraph 26 a-b (IFAC 2009b),  re-
quire audit firms to consider its competency (e.g. audit resources) as well as com-
pliance with ethical requirements (e.g. objectivity) in client continu-
ance/acceptance decisions, these concerns are relaxed by stating that the subject 
may assume that there are no conflicting issues with these requirements. 
The background information contains a short description of a (prospective) cli-
ent’s business, history, industry and management/ownership. The risk and infor-
mation reliability manipulations are also included. The risk manipulation aims to 
affect  one  part  of  the  engagement  risk,  namely  audit  risk.  Audit  risk  consists  of  
RMM and detection risk (Messier 2003). Level of RMM is manipulated at the 
beginning of the background information. The subject’s estimation of detection 
risk is indirectly examined after the judgment by asking him/her to estimate 
planned audit hours for the engagement. The information reliability manipulation 
manipulates the information reliability presented in the information menu. Ap-
pendix 1 presents all RMM and information reliability manipulations used in this 
study. 
Semi-structured
More-reliable
Reliability of
Information
Less-reliable
Task Structure
HighLow High LowRMM
Client Continuance Client Acceptance
Unstructured
Treatment 
Group 5
Treatment 
Group 7
Treatment 
Group 6
Treatment 
Group 8
Treatment 
Group 1
Treatment 
Group 3
Treatment 
Group 2
Treatment 
Group 4
76      Acta Wasaensia 
Risk 
In the low RMM manipulation, the subjects were told that the owners are happy 
with the management. It is stated that a new auditor’s preliminary discussions 
with the management indicate that it is competent, possesses high integrity and 
emphasizes high ethical values in its course of business. By contrast, in the high 
RMM manipulation, it is told that the owners are not satisfied with management 
performance and are planning to alter management compensation so that a major 
part is determined by previously reported earnings. It is also stated that “prelimi-
nary discussions with management indicate that you are not fully convinced about 
the management’s integrity and motivation to work under new terms”. 
 
Information reliability 
The information is said to have been collected by the previous auditor. In the 
more reliable information manipulation, it is stated that the previous auditor has 
one of the highest reputations in your audit firm and that s/he is known for the 
conscientiousness and preciseness of her/his work. In the less reliable reliability 
manipulation, it is stated that the previous auditor has been given several remarks 
in the audit firm’s internal quality inspections and that s/he has been blamed for 
superficial and poor documentation in her/his previous audits. 
 
Information usage 
Subjects can freely choose any information necessary for the decision-making 
from the menu and terminate information usage at any point. As observed by 
Hoffman, Joe and Moser (2003) a pre-established order of information may result 
in less effective information processing and, therefore, the unconstrained selec-
tion of information is highlighted in the instruction text. Subjects also have the 
option to go back and see the background information again. After the completion 
of the information usage stage, subjects are asked to give a probability on an 11-
point scale of recommending the continuance of the engagement. Other questions 
concern confidence with the main judgment and the estimates of audit hours and 
fees. Specifically, subjects are asked how they would adjust budgeted audit hours 
and fees compared with the previous audit. To make all these judgments, subjects 
can return to the cues and background information. 
 
  
 Acta Wasaensia     77 
  
Manipulation check questions and realism of the task 
Two manipulation questions about perceived RMM and information reliability 
were presented. The third question concerned the perceived realism of the task. 
Before the demographic information, the importance of the utilized information 
was requested. Subjects were asked to evaluate on an 11-point scale all those cues 
that they opened from the information menu. If the subject was not sure about or 
did not remember what particular information cue contained, she/he could review 
it again in a pop-up window. The research instrument is illustrated in Figure 9. 
 
Client acceptance task 
The unstructured task is a new client acceptance task that is similar to the client 
continuance task. A subject was now asked to evaluate if a totally new client can 
be accepted as a new client to the audit firm where she/he would be an incumbent 
auditor. 
 
Risk 
The auditing standards (e.g. IFAC 2009b) emphasize that the integrity of the pro-
spective client should be considered before every acceptance decision. Thus, au-
ditors should in particular assess whether the client related risks are heightened 
due to a dishonest management or principal owners. The RMM manipulation was 
similar to before (see Appendix 1). 
 
Information reliability 
Information reliability was manipulated via an audit assistant and a previous audi-
tor who was responsible for collecting information on decision-making. In the 
more reliable information manipulation, the assistant is characterized as being an 
accurate and thorough worker with a good reputation. The previous auditor is 
characterized as having a high reputation and being a conscientious auditor. In the 
less reliable information manipulation, it was stated that after information collec-
tion the assistant had pointed out to be an incompetent and negligent in her/his 
work. In addition, it was stated that the previous auditor had recently been sus-
pended by her/his audit firm because of several remarks at the audit firm’s inter-
nal quality inspections. 
  
78      Acta Wasaensia 
Post-experimental questionnaire 
After subjects had completed one of the tasks, an identical post-experimental 
questionnaire was used for follow-up questions. First, questions on the task-
specific experience of both client continuance and client acceptance were asked 
followed by auditor-specific questions, such as career length as an auditor, specif-
ic training for these tasks (number of hours), level of auditor certification, Big-
4/non-Big-4 firm, rank, time in current rank, gender and age. Subjects were also 
asked about the purpose of the task and given the option to provide feedback. 
After submitting the form, subjects had the chance to take part in a raffle by 
providing their contact information. To assure anonymity and confidentiality, 
contact information was collected in a separate form. 
 
 
Figure 9. Overview of the research instrument 
  
 Acta Wasaensia     79 
  
Technical execution 
The experiment was technically carried out using the server-side scripting, PHP 
and MySQL language. This technique allowed us to record automatically infor-
mation usage variables such as the number of used information, the length of time 
used for reading a single information cue (cumulative time), order of information 
acquisition and the total length of time spent during the experiment. A list of all 
variables that were automatically generated by the software is in Appendix 2. The 
software also incorporated some internal controls on the most critical screens to 
ensure completeness on one screen before moving to the next screen.  
5.5 Subjects and experimental procedures 
This study’s subjects consisted of Finnish CPAs, non-certified auditors and stu-
dents. Two groups of certified accountants, KHT and HTM auditors, comprised 
the majority of subjects. To ensure sufficient responses for the statistical analyses, 
the subject population included all these certified accountants. As the names of 
certified auditors and their respective audit firms are available on the (Central) 
Chamber of Commerce’s webpages, this information was used to obtain the e-
mail addresses of each auditor. The e-mail addresses of non-certified auditors 
were collected from one of the Big-4 webpages, which list people whose work is 
audit-related. The student population consisted of Master’s level students who 
had participated in at least one of the two advanced auditing and accounting 
courses in the previous semester. 
Using students as subjects is consistent with previous experimental studies of 
auditors’ information acquisition and usage (e.g. Krogstad, Ettenson & Shanteau 
1984; Kennedy 1993; Trotman & Wright 1996; Anderson & Mueller 2005). Stu-
dents were used in this study in order to have enough subjects in the less experi-
enced subjects group19. It was assumed that Master’s level students who have 
basic auditing knowledge are able to perform these tasks as well as can other less 
experienced auditors, as the studied tasks do not require complex problem repre-
sentation (i.e.  both tasks have a basic information menu and the purposes of the 
tasks are explained unambiguously) before information usage. 
                                               
 
19 Including less experienced subjects in the sample is otherwise difficult, as the contact infor-
mation of novice auditors is not publicly available. 
80      Acta Wasaensia 
Including all these subjects resulted in a total population of 1352 subjects; 1277 
certified auditors (94.5%), 36 non-certified auditors (2.7%) and 39 students 
(2.9%). To enhance the response rate, the chance to win travel gift vouchers and 
book prizes was provided in return for participation. Before the experiment was 
carried out, it was extensively pre-tested. Pre-testers included auditors and other 
accounting professionals with various experience levels. The aim of pre-tests was 
to ensure sufficient task realism as well as that the experiment’s user interface 
was easy-to-use and that the required time to complete the experiment was rea-
sonably short. 
All population subjects were randomly divided into the two groups (client contin-
uance and client acceptance) before an introductory e-mail was sent (see Appen-
dix 3). This e-mail contained general information about the purpose of the study, 
overview of the task, URL to the task and chance to participate in a raffle. When a 
subject clicked on the URL, the research instrument started automatically in a 
new window. The software randomly assigned each subject to one of the treat-
ment groups at the beginning of the experiment. 
Subjects were asked to participate in the experiment within 10 days. After 10 
days, the first follow-up email (see Appendix 4) was sent to those subjects who 
had not given their contact information for the raffle. They were again asked 
kindly to participate in the experiment, but the response time was reduced to sev-
en  days.  After  this  time  had  expired,  two  similar  follow-up  rounds  were  per-
formed at intervals of one week. 
5.6 Dependent variables 
A  rich  set  of  dependent  variables  is  used  to  measure  information  usage.  First,  
based on variables used in previous studies of information usage (e.g. Davis 1996; 
Zimbelman 1997; Turner 2001; Andersson 2004; Moroney 2007; Thayer 2011), 
these variables aim to capture the two basic dimensions of information usage: 
total time spent on the task and the number of used information. For instance, 
time-based measures have been suggested to proxy for used effort20 in previous 
information usage studies (Cloyd 1997; Zimbelman 1997). Second, alternative 
variables are used in order to overcome shortages in the traditional measures 
above and to give more insights into auditors’ information usage. 
                                               
 
20 Used time is limited to measure only one dimension of effort, namely effort duration, but not 
effort intensity, i.e. working harder per time unit (Cloyd 1997). 
 Acta Wasaensia     81 
  
The main dependent variables are total time per task (TOT_TIME), which is de-
fined as time from the beginning of the task to the submission of task-specific 
judgments (see Figure 6) and number of used cues (NUMBER_INF), which is the 
total number of cue openings in the information menu. 
The other four variables are alternative measurements for the above measures. 
These variables aim to capture the nuances in information usage caused by inde-
pendent variable manipulations. Alternative time measures divide total task time 
into the following two subcomponents. 
Total cue time (TOT_CUE_TIME) is the total time spent at cue screens (see Fig-
ure 6). This variable aims to measure whether the manipulations of independent 
variables affect a subject’s time to utilize cues from cue screens. It is expected 
that  total  cue  time  depends  on  the  number  of  used  cues.  The  variable  also  
measures the amount of effort a subject uses when reading and assimilating cues 
as well as his/her knowledge (Zimbelman 1997). 
Judgment time (JUDG_TIME) is the time spent outside of cue screens in the task. 
It is defined as TOT_TIME minus TOT_CUE_TIME. This variable measures the 
effort used to evaluate and combine (i.e. process) acquired cues and make task-
specific judgments. Thus, it is also expected that used judgment time depends on 
the number of used cues. 
Alternative count measures of used cues include very short cue visits and subjec-
tive estimates of less important cues. These measures are based on the assumption 
that plain cue visits alone do not mean a cue is fully used in a decision-making 
process. For instance, very short cue time may indicate that cue is not read to the 
end. 
Number of important cues (NUMBER_IMP_CUE) is an adjusted number of used 
cues, where the cues whose importance is self-estimated to be less than five are 
extracted from the number of used cues. This variable aims to capture cues per-
ceived to be important after reading/processing all cues and making task-specific 
judgments. 
The number of information 9 seconds (NUMBER_INF_9) has the same purpose 
as the previous variable. For this variable, cues read in less than nine seconds are 
extracted from the number of used cues. A short cue time may indicate that a cue 
is read only partly or superficially and deemed to be less important by a subject. 
82      Acta Wasaensia 
5.7 Independent variables 
Three independent variables of the study are manipulated variables: Risk, Struc-
ture and Reliability are factor (univariate) or categorical/dummy (multivariate) 
variables in the analyses. The observations from treatments 3, 4, 7 and 8 are clas-
sified as high RMM observations and those from treatments 1, 2, 5 and 6 as low 
RMM observations. Treatments 1–4 are classified as semi-structured and 5–8 as 
unstructured tasks. The observations from treatments 2, 4, 6 and 8 are classified 
as less reliable information observations and those from treatments 1, 3, 5 and 7 
as more reliable information observations. 
Subjects are classified into less experienced subject and experienced auditor 
groups based on their task-specific experience. This classification is consistent 
with previous studies in the expertise literature (e.g. Bonner & Pennington 1991; 
Bonner 1990; Earley 2002), when it is desired to examine whether there exists 
differences in judgments and information usage between “novices” and “experts”. 
Task-specific experience is chosen over general audit experience (i.e. years of 
audit experience), as neither task is conducted by all auditors on a regular basis. 
Therefore, using general experience as a proxy for knowledge of client continu-
ance and acceptance tasks would entail much noise (Tan 2001). In addition, pre-
vious studies suggest that audit partners are responsible for actual client ac-
ceptance judgments, because they have been used as subjects in these studies (e.g. 
Johnstone 2000; Wittek, van der Zee & Muhlau 2008). Thus, even a long tenure 
as an auditor may not be a suitable proxy for knowledge of these tasks if the audi-
tor has not worked at the partner-level in an audit firm. However, it is expected 
that many auditors may have assisted in these tasks, for example by gathering 
information for partners’ decision-making21 or have other second-hand experience 
(training) of tasks. 
Auditors' task-specific experiences are measured separately from client continu-
ance and acceptance tasks from all the subjects. Both measures uses the identical 
0-4  scale,  where  0  refers  to  zero  experience  and  4  refers  to  30+  times  of  task-
specific experiences (see Table 2 notes for full definition of the scale). It was de-
cided to combine task-specific experiences from both tasks to one variable as the 
purpose was to measure subjects' knowledge of these tasks broadly. Specifically 
                                               
 
21 For the sake of clarity, in the post-experimental questionnaire all “assisting experience” was 
also counted. Owing to this broad definition of task-specific experience where “pure” and “as-
sisting” experience are not separated, the cut-off point for experienced was set to relatively 
high. 
 Acta Wasaensia     83 
  
in this study, it is suggested that experience from one task might also increase 
relevant knowledge regarding the other task due to their many similarities, in oth-
er words, knowledge could be transferred across these tasks22.  
Subjects are classified as less experienced if the sum of their self-reported task-
specific experiences is less23 than 5 on a scale of 0–8. All other subjects are con-
sidered to be experienced auditors. Thus, subjects without task-specific encoun-
ters  had  a  value  of  0  and  the  most  experienced  auditor  had  a  value  of  8  on  the  
scale. Hence, the most experienced auditors have performed both tasks at least 30 
times, totaling over 60 times. On this 0-8 scale, the mean statistics (not tabulated) 
indicate that the less experienced group’s mean task-specific experience is 3.1, 
while that of experienced auditors is 5.8. 
5.8 Data and exclusion of outliers 
After the final follow-up round, 339 (25.1%) observations were gathered24. First, 
32 incomplete and duplicate observations25 were deleted from the sample. Se-
cond, data were analyzed to find significant outliers from non-diligent subjects, as 
these observations can significantly increase noise and reduce the power of an 
experiment (Oppenheimer, Meyvis & Davidenko 2009). Further, all observations 
may not represent the behavior of motivated individuals. If the motivation of the 
subject is not in line with the expectations of the experimenter, a lack of effort in 
the experiment may cause its pre-planned goals to be unmet (Kersten, Wu & Oer-
tel 2011). In particular, time-based variables are generally considered to be noisy 
measures of subjects’ efforts in task making, and a distracting environment in 
web-based experimental settings may further magnify such noise (Cloyd 1997; 
Bryant, Hunton & Stone 2004). 
                                               
 
22  For instance, it is suggested that an auditor who has only task-specific experience from a cli-
ent continuance task has more relevant knowledge for a client acceptance task than an auditor 
without any experience in these tasks. 
23 Further, as seen from Table 2, subjects in this sample are generally highly experienced audi-
tors. Thus, in order to have enough observations of both experience levels for the statistical 
analyses, the cut-off point was set higher than in previous studies (e.g. Earley 2002). In theo-
ry, one would expect even more experience effects between less experienced and experienced 
auditors if this limit had been set to a lower level (e.g. “0–10” vs. “11-“ encounter times as in 
Earley’s (2002) real-estate valuation task). 
24 E-mail addresses were manually collected from the (Central) Chamber of Commerce’s 
webpage. Because manually collecting e-mail addresses was laborious, a two-year-old collec-
tion of emails was used. This database was then updated with the auditors certified after this 
date. However, a significant number of addresses (169; 13.5%) were outdated (mainly be-
cause of retirements and job changes), which were returned to the sender. 
25 Duplicate observations were traced using IP address and time stamp match. 
84      Acta Wasaensia 
Appendix 5 reports the descriptive statistics of dependent variables (n=307). 
These statistics show a considerable range of time-based dependent variables. For 
instance, TOT_TIME ranges from 76 to 4947 seconds and TOT_CUE_TIME 
from 0 to 4340 seconds, while the mean (SD) times of these variables are 666.3 
(589.5) and 332.9 (349.1) seconds, respectively. This mean total time (about 11 
min.) is line with the time mentioned to the participants to allocate for the exper-
iment in the introductory letter. However, very short and long times indicate 
strongly that subjects were either uninterested in the actual task26 or were inter-
rupted during it. 
To control for the influence of extreme observations on the results, particular ob-
servations are removed from the sample as follows27. First, all observations where 
subjects used three minutes or less total time are removed. This exclusion is based 
on the extensive pre-testing of the tasks, which showed that it takes at least three 
minutes to read the background information of the task and to answer properly all 
four task-specific judgments. Several observations from pre-testing indicated that 
three minutes was required even for the easiest condition, i.e. when the auditor 
was experienced, RMM was low, task unstructured and information reliable. Us-
ing this exclusion removed 16 observations from the sample. 
Second, all observations taking more than 20 minutes are removed. This limit is 
defined by multiplying by two the time (10 minutes) that was mentioned to sub-
jects in the introductory letter to allocate for task making. Further, the results 
from pre-testing indicated that none of the 18 pre-testers used more than 17 
minutes for task making. However, it is acknowledged that some individuals may 
use more time as they are unfamiliar with making audit judgments on computer 
screens compared with the pre-testers. For example, older auditors may have used 
more time for this reason (see Nearon 1999). Thus, it seems reasonable to in-
crease this limit slightly as the mean age of subjects is high (48.8. years; see Ta-
ble 2). The exclusion of these observations decreased the sample by an additional 
20 observations28. 
                                               
 
26 For example, the chance to win a prize in the raffle may have encouraged participation with-
out any true intention to act professionally in the actual decision-making part of the experi-
ment (O’Neil & Penrod 2001). 
27 These two elimination steps are similar to those used by Bailey, Daily and Philips (2006), who 
measured the time spent on an audit task in a web experiment. 
28  To analyze whether experience level or background (see Table 2 for variables) of excluded 
subjects (n=36) systematically differ from included subjects, several univariate tests were car-
ried out. These results show no significant differences between these subject groups (p-values 
= > 0.400). 
 Acta Wasaensia     85 
  
Thus, the final sample size is 271 observations. All descriptive statistics and sta-
tistical analyses are carried out with the same data set to keep results comparable. 
As shown in Figure 10, observations are distributed evenly between the eight 
treatment groups, ranging from 28 to 41 observations per treatment. 
 
 
Figure 10. Distribution of observations between treatment groups (percentages 
in parentheses) 
5.8.1 Descriptive statistics of subjects 
The  total  sample  of  271  subjects  consists  of  219  certified  auditors,  32  non-
certified auditors29 and 20 students. In a post-experimental questionnaire, infor-
mation was collected about subjects’ audit experience and backgrounds. These 
descriptive statistics are summarized in Table 2. In this study, the mean age of 
subjects is considerably high (48.8 years) and 84.9% of them have completed at 
least  one  auditor  certification  (JHTT/HTM/KHT).  Slightly  less  than  one  third  
(31.0%) of subjects are female. On the task-specific experience scale of 0-4 (see 
Table 2 notes for definitions of scale), subjects indicated mean (standard devia-
tion; SD) experience levels of 2.3 (1.5) for client continuance experience and 2.7 
                                               
 
29 A closer examination of the data showed that besides “not yet certified” auditors, some retired 
auditors who previously held an auditor certification belonged to this group. 
Semi-structured Unstructured
5 7
59 72
More-reliable (21.8 %) (26.6 %)
Reliability of
Information 6 8
73 67
Less-reliable (26.9 %) (24.7 %)
132 139
(48.7 %) (51.3 %)
TREATMENT 2 TREATMENT 4
TREATMENT 5
TREATMENT 6 TREATMENT 8
TREATMENT 7
Task Structure
RMM
TREATMENT 1
(11,4 %)(10.3 %)
TREATMENT 3
Low High HighLow
28 31 31 41
(11,4 %) (15.1 %)
41 32 37 30
69 63 68 71
(15.1 %) (11.8 %) (13.7 %) (11.1 %)
(25.5 %) (23.2 %) (25.1 %) (26.2 %)
86      Acta Wasaensia 
(1.6) for client acceptance experience30. There is an almost equal number of sub-
jects working in Big-4 audit firms (30.8%) as in non-Big-4 firms (28.4%). The 
major proportion of subjects (40.8%) work outside of audit firms or are students. 
In summary, these descriptive statistics indicate that on average subjects have a 
lot of experience of both tasks. 
 
Table 2. Descriptive statistics of subjects 
 
 
In the post-experimental questionnaire, the realism of the experiment was evalu-
ated on an 11-point scale (0 = very unrealistic; 10 = very realistic). Subjects eval-
uated the realism of both the client continuance (mean 7.64, SD 1.92) and the 
client acceptance (mean 7.06, SD 2.15) tasks fairly over mean of the response 
                                               
 
30 This confounding finding that subjects had more experience with the acceptance than with the 
continuance task may be because the latter is a routine task that is passed quickly without the 
significant consideration of performing that task (especially when there has been no contro-
versy with the client). 
Descriptive Statistics of Subjects 
 
 Agea Genderb Auditor 
Certification 
Audit firmc 
Continuance 
task-specific 
experienced 
Acceptance 
task-specific 
experienced         
Treatment Mean SD Female Yes Big-4 
Non-
Big-4 None Mean SD Mean SD 
   % % % % %      
1(n=28)
 
53.4 10.5 32.1 89.3 30.8 11.5 57.7 2.3 1.6 3.0 1.6 
2(n=41) 45.1 13.6 31.7 82.9 30.8 30.8 38.5 2.3 1.6 2.7 1.5 
3 (n=31) 48.5 12.9 35.5 90.3 30.0 43.3 26.7 2.7 1.5 2.7 1.5 
4 (n=32) 51.6 11.3 40.6 81.3 14.3 32.1 53.6 1.7 1.3 2.7 1.6 
5 (n=31) 46.4 12.3 32.3 83.9 31.0 31.0 37.9 2.5 1.5 3.0 1.4 
6 (n=41) 50.5 14.2 26.8 82.9 28.9 23.7 47.4 2.2 1.6 2.4 1.6 
7 (n=37) 48.2 16.4 13.5 83.8 41.2 23.5 35.3 2.2 1.6 2.7 1.6 
8 (n=30) 47.5 14.4 40.0 86.7 38.5 30.8 30.8 2.3 1.6 2.7 1.6 
™ (n=271) 48.8 13.5 31.0 84.9 30.8 28.4 40.8 2.3 1.5 2.7 1.6 
Notes: 
 
a 10 missing observations            
b Two missing observations           
c 21 missing observations            
d Variable is defined as follows: 0 = No experience, 1 = 1–9 times, 2 = 10–19 times, 3 = 20–29 times, 4 = 30+ 
times 
 
 Acta Wasaensia     87 
  
scale (5.0). Thus, it can be assumed that the majority of subjects perceived these 
tasks to be relatively realistic for the given judgments. 
5.8.2 Descriptive statistics of the dependent variables 
Table 3 presents the descriptive statistics of the dependent variables (Appendix 6 
reports the means and SDs for each treatment and experience). It can be seen 
from Panel A that on average a subject used about 11 minutes for the task. This 
time is distributed evenly between the cue and the judgment time, indicating that 
both “phases” were time consuming on average. 
 
Table 3. Descriptive statistics of the dependent variables 
 
Panel A: Descriptive statistics of the continuous variables (All units in seconds) 
Variable 
 
Mean Std Dev Minimum Q1 Median Q3 Max. 
TOT_TIME 
 
569.6 243.9 185.0 364.0 535.0 747.0 1191.0 
TOT_CUE_TIME 298.3 164.9 27.0 163.0 275.0 409.0 926.0 
JUDG_TIME 271.3 134.5 45.0 185.0 247.0 313.0 961.0 
Panel B: Descriptive statistics of the count variables (All units in counts) 
NUMBER_INF 
 
10.9 3.5 2.0 10.0 12.0 12.0 25.0 
NUMBER_INF_9 9.7 3.3 0.0 7.0 10.0 12.0 25.0 
NUMBER_IMP_CUE 10.5 3.3 0.0 9.0 11.0 13.0 23.0 
Panel C: Pearson’s correlation coefficients of the dependent variables 
 TOT_CUE JUDG NUMBER NUMBER NUMBER   
 _TIME _TIME _INF _INF_9 _IMP_CUE   
TOT_TIME
 
0.852*** 0.768*** 0.447 *** 0.625 *** 0.382 ***   
TOT_CUE_TIME  0.320*** 0.487 *** 0.688 *** 0.423 ***   
JUDG_TIME   0.213 *** 0.289 *** 0.174 **   
NUMBER_INF    0.830 *** 0.753 ***   
NUMBER_INF_9     0.651 ***   
Notes:        
Statistical significance based on two-tailed tests at the 1%, 5% and 10% levels are denoted 
by ***, ** and *, respectively.       
The variables are defined as follows:        
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time spent reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance was self-evaluated to be more than 4, 
including multiple reads       
 
88      Acta Wasaensia 
As shown in Panel B of Table 3, there is only a slight variance in the count varia-
bles  between  subjects.  Over  one  third  (37%)  of  subjects  read  all  available  cues  
once, 22% of subjects read at least one cue more than once and four subjects used 
no cue at all (not tabulated). Additional measures, such as the number of im-
portant cues and number of information 9 seconds, show only a slightly dimin-
ished average number of used cues. Overall, Panel B suggests that cues’ names 
and contents attracted the majority of subjects to read them carefully. 
Finally,  Panel  C of  Table  3  reports  a  correlation  matrix  of  the  dependent  varia-
bles. As noted, the correlations among the continuous and count variables are 
considerably high within the same types of variables, but not between types. 
Thus, it can be inferred that different types of variables measure – at least partly – 
different dimensions of information usage. 
5.8.3 Descriptive statistics of cue usage 
Next, the descriptive statistics of cue usage are presented to assess whether some 
cues were perceived to be more important than others for decision-making. Table 
4 reports the pooled cue usage of all subjects, while in Tables 5–8 cue usage is 
examined by the level of each independent variable. 
Table 4 shows that all available cues were perceived to be relatively important for 
decision-making. All self-evaluated average importance values were above the 
mean (5.0) of the scale, the lowest average being 5.95 and the highest being 7.94. 
Similarly, the total cue visit percentage ranged from 65.7% to 97.8%, which indi-
cates that no cues were deemed to be particularly uninteresting in the menu view. 
Further, in Table 4 several variables indicate that Previous Audit was perceived to 
be the most important cue. Over 97 percent of subjects used this cue and only four 
subjects read it for less than 10 seconds31. In addition, when Previous Audit’s 
reading time is scaled by cue length, the used median time was the highest (0.43 
seconds/word) of all cues. The next important cues were Balance Sheet, Internal 
Controls and Accounting and Bookkeeping. All those cues were used by more 
than 90 percent of participants and less than five percent of them read them for 
less than 10 seconds. Consistent with that, subjects also self-evaluated these four 
cues’ average importance in the top five in the post-experimental questionnaire. 
                                               
 
31 While Previous Audit was the first cue in the Finnish information menu (alphabetic order), the 
order of the cues did not seem to fully explain their popularity, as the second most popular 
cue, Balance Sheet, was only the seventh item in the menu. 
 Acta Wasaensia     89 
  
Consequently, the results from these different measures are fairly consistent with 
each other. 
As shown in Table 4, the three least important cues were all from the industry and 
competition information category. Industry Comparison, Industry Status and Out-
look and Competitive Environment were consistently self-evaluated as being the 
least important and also read relatively superficially by subjects (i.e. over 45 per-
cent of subjects who opened these cues read them for less than 10 seconds). 
While the long length of these cues may have caused fatigue in processing them 
carefully, it must be noted that before their length was exposed to a subject, they 
also received smaller attention in the information menu compared with other cues 
(total visits varied between 65.7% and 76.0%). 
 
Table 4. Descriptive statistics of cue usage 
 
 
Tables 5–8 report the cue statistics where subjects’ cue usage is classified by ex-
perience, risk, task structure and information reliability. Table 5 shows the clear 
differences in information usage between less experienced and experienced audi-
tors. The statistics clearly show that less experienced subjects used more cues and 
more average reading time per cue than experienced auditors. By contrast, the 
average self-evaluated importance evaluations of the cues do not seem to be de-
Descriptive Statistics of Cue Usage 
Variable IS BS KR IC ISO CE SO CO AB PA RC RL AVG. 
Self-evaluated              
importance:a              
Mean 7.78 7.78 7.14 5.95 6.11 5.60 6.72 7.48 7.94 7.63 6.82 6.41 6.95 
Median 8.00 8.00 7.00 6.00 7.00 5.00 7.00 8.00 8.00 8.00 7.00 7.00 7.17 
Total visits:              
% of subjects 87.82 95.20 82.66 65.68 75.28 76.01 87.45 93.73 93.36 97.79 86.72 93.73 86.29 
Total visits (> 9 sec on cue.):             
% of subjects 74.54 93.36 68.63 54.61 59.04 60.15 72.32 89.30 88.93 96.31 72.69 77.49 75.62 
Used cue time:              
Mean time 33.61 47.03 21.39 20.52 33.57 24.32 20.40 34.06 31.45 37.21 18.48 18.69 28.39 
Median time 20.0 38.0 18.0 19.0 25.0 21.0 18.0 30.0 28.0 32.0 16.0 16.0 23.42 
Sec. per word (mean) 0.27 0.28 0.35 0.28 0.17 0.24 0.35 0.34 0.34 0.50 0.33 0.26 0.31 
Sec. per word (median) 0.16 0.22 0.30 0.26 0.13 0.21 0.31 0.30 0.30 0.43 0.29 0.23 0.26 
Notes:              
a Measured on an 11-point Likert-type scale 
Information  cues  are  defined  as  follows:  IS  =  Income  Statements,  BS  =  Balance  Sheet,  KR =  Key  Ratios,  IC  =  Industry  
Comparison, ISO = Industry Status and Outlook, CE = Competitive Environment, SO = Structure of Organization, CO = 
Internal Controls, AB = Accounting and Bookkeeping, PA = Previous audit, RC = Relationship with the main Creditor, RL = 
Relationship with the Lawyer 
 
90      Acta Wasaensia 
pendent on auditor experience, which is consistent with the findings of Johnstone 
(2001). However, less experienced subjects have slightly smaller standard devia-
tions in their importance evaluations between the cues than experienced auditors 
(0.6 vs. 0.85, not tabulated). 
 
Table 5. Descriptive statistics of cue usage – auditor experience 
 
 
Table 6 reports the cue statistics where subjects’ cue usage is classified according 
to RMM level. Unexpectedly, these results show that more information is used 
when RMM is low, as Total visits% is higher (87.8% vs. 84.6%), while cues are 
slightly less carefully read (22.7 vs. 24.2 seconds) in the decision-making process. 
This same effect is also shown by the greater drop between Total visits% and To-
tal visits (> 9 sec. on cue)% in low RMM treatments (12.9 %) than in high RMM 
treatments (9.4%). 
 
 
Descriptive Statistics of Cue Usage - AUDITOR EXPERIENCE 
Variable IS BS KR IC ISO CE SO CO AB PA RC RL AVG. 
Self-evaluated              
importance:a              
Mean (LESS) 7.95 7.84 7.23 6.16 6.40 5.91 6.62 7.35 7.89 7.52 6.61 6.36 6.99 
Mean (MORE) 7.69 7.76 7.09 5.85 5.95 5.44 6.77 7.54 7.96 7.68 6.87 6.43 6.92 
Total visits:              
% of subjects (LESS) 89.53 94.19 90.70 72.09 86.05 83.72 89.53 97.67 93.02 98.84 89.53 96.51 90.12 
% of subjects (MORE) 87.03 95.68 78.92 62.70 70.27 72.43 86.49 91.89 93.51 97.30 85.41 92.43 84.50 
Total visits(> 9 sec on cue.):             
% of subjects (LESS) 80.2 93.0 80.2 66.3 74.4 70.9 76.7 93.0 93.0 95.3 82.6 84.9 82.56 
% of subjects (MORE) 71.9 93.5 63.2 49.2 51.9 55.1 70.3 87.6 87.0 96.8 68.1 74.1 72.39 
Used cue time:              
Median time (LESS) 24.0 48.0 19.0 20.5 39.0 28.0 20.0 32.5 31.0 36.0 19.0 18.0 27.92 
Median time (MORE) 18.0 35.0 17.0 18.5 21.5 18.0 16.0 26.5 26.0 30.0 15.0 15.0 21.38 
Sec. /word median (LESS) 0.19 0.28 0.31 0.28 0.20 0.28 0.34 0.33 0.33 0.49 0.34 0.25 0.30 
Sec. /word median 
(MORE) 0.14 0.20 0.28 0.25 0.11 0.18 0.28 0.27 0.28 0.41 0.27 0.21 0.24 
Notes:              
a Measured on an 11-point Likert-type scale 
Information cues are defined as follows: IS = Income Statements, BS = Balance Sheet, KR = Key Ratios, IC = Industry 
Comparison, ISO = Industry Status and Outlook, CE = Competitive Environment, SO = Structure of Organization, CO = 
Internal Controls, AB = Accounting and Bookkeeping, PA = Previous audit, RC = Relationship with the main Creditor, RL = 
Relationship with the Lawyer 
 
 Acta Wasaensia     91 
  
Table 6. Descriptive statistics of cue usage – RMM  
 
 
Table 7 reports the cue statistics where subjects’ cue usage is classified according 
to task structure level. The only visible difference in cue usage between semi-
structured and unstructured tasks is that in the latter task more cues are initially 
used. However, the Total visits (> 9 sec. on cue) percentage is virtually the same 
for both structure levels, suggesting that cues are used more superficially in un-
structured tasks. 
 
 
 
 
 
 
Descriptive Statistics of Cue Usage - RMM 
Variable IS BS KR IC ISO CE SO CO AB PA RC RL AVG. 
Self-evaluated              
importance:a              
Mean (LOW) 7.69 7.86 7.16 6.01 6.10 5.62 6.65 7.47 7.97 7.56 6.78 6.52 6.95 
Mean (HIGH) 7.87 7.70 7.12 5.88 6.12 5.58 6.80 7.50 7.91 7.70 6.79 6.28 6.94 
Total visits:              
% of subjects (LOW) 87.94 95.04 85.82 70.92 78.01 80.14 88.65 93.62 95.04 97.16 87.94 94.33 87.88 
% of subjects (HIGH) 87.69 95.38 79.23 60.00 72.31 71.54 86.15 93.85 91.54 98.46 85.38 93.08 84.55 
Total visits (> 9 sec on cue.):            
% of subjects (LOW) 71.63 92.20 70.21 60.28 60.28 63.12 68.79 86.52 90.07 95.04 76.60 77.30 76.00 
% of subjects (HIGH) 77.69 94.62 66.92 48.46 57.69 56.92 76.15 92.31 87.69 97.69 68.46 77.69 75.19 
Used cue time:              
Median time (LOW) 21.0 37.5 17.0 18.5 24.5 20.0 17.0 27.0 27.0 31.0 15.5 16.0 22.67 
Median time (HIGH) 19.0 38.5 19.0 20.0 27.5 21.0 19.0 31.0 30.0 32.0 16.0 17.0 24.17 
Sec. /word median (LOW) 0.17 0.22 0.28 0.25 0.13 0.20 0.29 0.27 0.29 0.42 0.28 0.23 0.25 
Sec. /word median (HIGH) 0.15 0.23 0.31 0.27 0.14 0.21 0.33 0.31 0.32 0.43 0.29 0.24 0.27 
Notes:              
a Measured on an 11-point Likert-type scale 
Information cues are defined as follows: IS = Income Statements, BS = Balance Sheet, KR = Key Ratios, IC = Industry 
Comparison, ISO = Industry Status and Outlook, CE = Competitive Environment, SO = Structure of Organization, CO = 
Internal Controls, AB = Accounting and Bookkeeping, PA = Previous audit, RC = Relationship with the main Creditor, RL = 
Relationship with the Lawyer 
 
92      Acta Wasaensia 
Table 7. Descriptive statistics of cue usage – task structure 
 
 
Table 8 reports the cue statistics where subjects’ cue usage is classified according 
to information reliability level. All total visit and cue time variables are indicating 
that less reliable information is selected more and read longer time compared with 
more reliable information. This is consistent with the expectations that more less 
reliable information is needed to reach a sufficiency threshold in uncertainty re-
duction than when information is more reliable. Subjects in the more reliable in-
formation treatments evaluated cues as being only slightly more important than 
those who had less reliable information, as indicated by self-evaluated importance 
(7.06 vs. 6.85, respectively). 
 
 
 
 
Descriptive Statistics of Cue Usage - TASK STRUCTURE 
Variable IS BS KR IC ISO CE SO CO AB PA RC RL AVG. 
Self-evaluated              
importance:a              
Mean (SEMI) 7.53 7.60 7.00 5.44 5.66 5.24 6.62 7.52 7.92 7.86 6.71 6.21 6.78 
Mean (UNSTR) 8.00 7.95 7.27 6.43 6.54 5.94 6.82 7.44 7.96 7.41 6.92 6.60 7.11 
Total visits:              
% of subjects (SEMI) 84.85 94.70 79.55 63.64 72.73 75.00 86.36 94.70 92.42 96.21 84.09 93.18 84.79 
% of subjects (UNSTR) 90.65 95.68 85.61 67.63 77.70 76.98 88.49 92.81 94.24 99.28 89.21 94.24 87.71 
Total visits (> 9 sec on cue.):             
% of subjects (SEMI) 75.00 93.18 68.18 56.06 57.58 61.36 73.48 90.91 90.91 94.70 72.73 76.52 75.88 
% of subjects (UNSTR) 74.10 93.53 69.06 53.24 60.43 58.99 71.22 87.77 87.05 97.84 72.66 78.42 75.36 
Used cue time:              
Median time (SEMI) 19.0 35.0 17.0 19.0 20.0 21.0 19.0 31.0 29.0 32.0 16.0 16.0 22.83 
Median time (UNSTR) 22.5 43.0 19.0 19.0 28.5 19.0 18.0 27.0 28.0 31.0 15.0 16.0 23.83 
Sec. /word median (SEMI) 0.15 0.20 0.28 0.26 0.10 0.21 0.33 0.28 0.31 0.43 0.29 0.23 0.26 
Sec. /word median (UNSTR) 0.18 0.25 0.31 0.26 0.15 0.19 0.31 0.29 0.30 0.42 0.27 0.23 0.26 
Notes:              
a Measured on an 11-point Likert-type scale            
Information cues are defined as follows: IS = Income Statements, BS = Balance Sheet, KR = Key Ratios, IC = Industry 
Comparison, ISO = Industry Status and Outlook, CE = Competitive Environment, SO = Structure of Organization, CO = 
Internal Controls, AB = Accounting and Bookkeeping, PA = Previous audit, RC = Relationship with the main Creditor, RL = 
Relationship with the Lawyer 
 
 Acta Wasaensia     93 
  
Table 8. Descriptive statistics of cue usage – information reliability 
 
 
In parallel with the descriptive statistics of the dependent variables, these statistics 
suggest that the majority of subjects perceived all available cues to be important 
and used information extensively in their decision-making. Finally, it must be 
noted that the same cues are the generally most (e.g. financial information) and 
the least (e.g. industry information) used and ranked in Tables 5–8. This suggests 
that the “relative importance order” of information is not dependent on any par-
ticular independent variable or its level. In other words, the manipulations of the 
variables or the level of experience had no unwanted influence on the demand of 
any individual cue. 
 
Descriptive Statistics of Cue Usage - INFORMATION RELIABILITY 
Variable IS BS KR IC ISO CE SO CO AB PA RC RL AVG. 
Self-evaluated              
importance:a              
Mean (LESS) 7.75 7.81 7.15 5.81 6.10 5.47 6.59 7.26 7.85 7.21 6.83 6.38 6.85 
Mean (MORE) 7.81 7.75 7.14 6.13 6.12 5.78 6.89 7.74 8.06 8.10 6.74 6.44 7.06 
Total visits:              
% of subjects (LESS) 88.19 96.53 86.81 66.67 76.39 81.25 89.58 94.44 95.83 97.22 86.11 96.53 87.96 
% of subjects (MORE) 87.40 93.70 77.95 64.57 74.02 70.08 85.04 92.91 90.55 98.43 87.40 90.55 84.38 
Total visits (> 9 sec on cue.):             
% of subjects (LESS) 75.00 95.83 73.61 60.42 65.97 65.97 75.00 88.19 91.67 95.14 76.39 81.25 78.70 
% of subjects (MORE) 74.02 90.55 62.99 48.03 51.18 53.54 69.29 90.55 85.83 97.64 68.50 73.23 72.11 
Used cue time:              
Median time (LESS) 20.0 39.0 18.0 20.0 30.5 23.0 19.0 31.0 29.0 33.0 16.0 17.0 24.63 
Median time (MORE) 20.0 36.0 17.0 18.5 19.0 19.0 17.0 27.0 28.0 29.0 15.0 14.0 21.63 
Sec. /word median (LESS) 0.16 0.23 0.30 0.27 0.16 0.23 0.33 0.31 0.31 0.45 0.29 0.24 0.27 
Sec. /word median (MORE) 0.16 0.21 0.28 0.25 0.10 0.19 0.29 0.27 0.30 0.39 0.27 0.20 0.24 
Notes:              
a Measured on an 11-point Likert-type scale            
Information cues are defined as follows: IS = Income Statements, BS = Balance Sheet, KR = Key Ratios, IC = Industry 
Comparison, ISO = Industry Status and Outlook, CE = Competitive Environment, SO = Structure of Organization, CO = 
Internal Controls, AB = Accounting and Bookkeeping, PA = Previous audit, RC = Relationship with the main Creditor, RL = 
Relationship with the Lawyer 
 
94      Acta Wasaensia 
6 METHODOLOGY AND RESULTS 
This chapter presents the methodology used in the data analysis as well as the 
results from the empirical tests. The tests start with the univariate analyses to ex-
amine how each independent variable affects the dependent variables. In the fol-
lowing section, multivariate analyses are applied to test the hypotheses. The sup-
plementary  and  robustness  tests  are  presented  in  their  own  sections,  as  are  the  
analyses of how independent variables affect task-specific judgments. The final 
section discusses the main empirical findings of the study. 
6.1 Manipulation check  
Before the univariate and multivariate analyses are carried out, I test if subjects’ 
evaluations of RMM and information reliability are consistent with the intended 
manipulations. The first manipulation check confirms that subjects generally per-
ceived risk to be higher (mean 5.05) when the high RMM manipulation was pre-
sent than when the low RMM manipulation (mean 4.13) was present (Mann–
Whitney U = 6700.0, p =< 0.001 two-tailed)32. The second manipulation check 
confirms that subjects generally perceived information in the information menu to 
be more reliable (mean 6.65) when the more reliable information manipulation 
was present than when the less reliable information manipulation (mean 5.93) was 
present (Mann–Whitney U = 6769.50, p =< 0.001 two-tailed). These results indi-
cate that both manipulation checks are supported by the data and that the hypoth-
eses of the study can be tested. 
6.2 Univariate analyses 
The analyses start with the univariate tests, which can give insights into how the 
experience, risk, task structure, information reliability and hypothesized interac-
tion effects influence the dependent variables. Tables 9–15 present the results of 
the univariate tests for hypotheses 1–7. Because the dependent variables are not 
normally distributed, non-parametric tests are used instead of t-tests. 
                                               
 
32 Further results indicate that experienced auditors’ estimates of perceived RMM were signifi-
cantly lower than those of less experienced auditors (U = 8950.50 p =0.021 two-tailed). This 
finding is not consistent with that of Colbert (1988), who found that level of experience had 
no effect on auditors’ RMM assessments. 
 Acta Wasaensia     95 
  
The first hypothesis of this study predicts that information is used more extensive-
ly in decision-making when the subject is less experienced. A vast amount of lit-
erature has shown that experienced auditors’ information usage is different from 
less experienced auditors. For example, experienced auditors’ directed infor-
mation acquisition strategies, knowledge and better recognition of relevant cues 
are expected to decrease the number of used information and time spent on cue 
screens. 
 
Table 9. Univariate tests of hypothesis 1 
 
As shown in Table 9, five measures out of six indicate at the 1% significance lev-
el that experienced auditors use less time and cues in their decision-making than 
less experienced subjects33. Only NUMBER_IMP_CUE (p = 0.180) is statistical-
                                               
 
33 To further analyze whether students’ (n=20) information usage differs from other less experi-
enced subjects (n=66), similar univariate tests within this group were carried out. These re-
 
Hypothesis 1 – Auditor (subject) experience 
 EXPERIENCE N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME Less exper. 86 160.6      
 Experienced 185 124.6 13318.5 3.53 0.000 
TOT_CUE_TIME Less exper. 86 160.0     
 Experienced 185 125.3 13672.5 3.30 0.001 
JUDG_TIME Less exper. 86 155.7     
 Experienced 185 126.9 13386.5 2.81 0.005 
NUMBER_INF Less exper. 86 155.5     
 Experienced 185 127.0 133371.0 2.89 0.004 
NUMBER_INF_9 Less exper. 86 161.4     
 Experienced 185 124.2 13883.5 3.68 0.000 
NUMBER_IMP_CUE Less exper. 86 145.3     
 Experienced 185 131.7 12497.0 1.34 0.180 
Notes: 
The variables are defined as follows:  
EXPERIENCE = Level of experience (Less experienced/Experienced) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
96      Acta Wasaensia 
ly non-significant, which suggests that experience level has no effect on the self-
evaluation of cue importance. This is consistent with the findings of Johnstone 
(2001) who found that less and more experienced auditors' evaluations of the 
most important cues in an client acceptance task did not differ from each other. 
Overall, these results are consistent with hypothesis 1. 
The second hypothesis of the study predicts that information is used more exten-
sively in subjects’ decision-making when RMM is high than when RMM is low. 
Previous research indicates that auditors increase their audit efforts as client-
related risk increases. Specifically in a single audit task, increased risk may 
heighten conservatism, increase professional skepticism and increase exposure to 
confirmation bias, which subsequently causes available information to be used 
more extensively than otherwise. 
As shown in Table 10, all time variables show an increase in used time when 
RMM increases to a high level. Two of these variables are significant at conven-
tional levels, TOT_TIME (p = 0.068) and JUDG_TIME (p = 0.029). The other 
dependent variables that measure used information indicate that less information 
is used when RMM is high, but all these variables are statistically non-significant. 
These results imply that the type of information becomes more important when 
RMM increases. To examine this issue more carefully, total time is divided by the 
number  of  used  information  to  test  if  the  means  between  the  low  and  high  risk  
groups differ. This expectation is supported by the data (Wilcoxon = 19363.0, 
p =0.001; not tabulated). This result suggests that subjects focused on fewer cues 
in high risk treatments, but evaluated and combined those cues in a more effortful 
way than subjects in low risk treatments. In summary, the results show some sup-
port for hypothesis 2. 
 
 
 
 
 
                                                                                                                                
 
sults show that only TOT_CUE_TIME is statistically significant (p=0.035), indicating that 
students spent longer on cue screens than other subjects. 
 Acta Wasaensia     97 
  
Table 10. Univariate tests of hypothesis 2 
 
The third hypothesis of the study predicts that information is used more exten-
sively in subjects’ decision-making when the task is unstructured than when it is 
semi-structured. Arguments from the task structure literature (Bonner 1994) indi-
cate that the lack of specified procedures in unstructured tasks might lead to ex-
tensive information usage if it is used as a way to reduce uncertainty in decision-
making. 
The univariate results for hypothesis 3 are reported in Table 11. Only one mean 
rank out of the six dependent variables indicates that more cues are used in un-
structured than in semi-structured tasks. The variable NUMBER_IMP_CUE is 
significant at the 5% level (p = 0.039). Otherwise, these results are not consistent 
with hypothesis 3. 
The fourth hypothesis of the study predicts that information is used more exten-
sively in subjects’ decision-making when information is less reliable than when it 
is more reliable. Previous studies suggest that less reliable information is 
Hypothesis 2 - RMM 
 RISK N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME Low 141 127.6      
 High 130 145.1 18863.0 1.84 0.068 
TOT_CUE_TIME Low 141 131.8     
 High 130 140.5 18270.5 0.92 0.361 
JUDG_TIME Low 141 126.0     
 High 130 146.9 19097.5 2.20 0.029 
NUMBER_INF Low 141 141.8     
 High 130 129.7 16859.0 -1.32 0.189 
NUMBER_INF_9 Low 141 136.6     
 High 130 135.3 17590.5 -0.14 0.889 
NUMBER_IMP_CUE Low 141 141.4     
 High 130 130.1 16913.0 -1.20 0.232 
Notes: 
The variables are defined as follows:  
RISK = Level of RMM (Low/High) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
98      Acta Wasaensia 
weighted less than more reliable information. As auditors’ information usage 
aims to reduce uncertainty, more information is needed to reach a sufficient 
threshold to make a decision when information is less reliable. 
 
Table 11. Univariate tests of hypothesis 3 
 
As shown in Table 12, there is a significant increase in used time as indicated by 
two variables TOT_TIME (p = 0.069) and JUDG_TIME (p = 0.010) when the 
information comes from a less reliable source. The fact that the variable NUM-
BER_INF (p = 0.210) is not significant, but NUMBER_INF_9 (p = 0.020) is, 
suggests that cues are processed in a more effortful way when information comes 
from a less reliable source. Thus, information that comes from a more reliable 
source generates more “short visits” compared with less reliable information. This 
suggests that more reliable information is processed more superficially than less 
reliable information. However, used time on cue screens is not affected by infor-
Hypothesis 3 - Task structure 
 STRUCTURE N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME Semi-struc. 132 129.6      
 Unstruct. 139 142.1 17103.0 -1.32 0.189 
TOT_CUE_TIME Semi-struc. 132 133.1     
 Unstruct. 139 138.8 17565.0 -0.59 0.550 
JUDG_TIME Semi-struc. 132 129.6     
 Unstruct. 139 142.1 17111.0 -1.30 0.194 
NUMBER_INF Semi-struc. 132 130.7     
 Unstruct. 139 141.1 17247.0 -1.13 0.258 
NUMBER_INF_9 Semi-struc. 132 136.8     
 Unstruct. 139 135.2 18059.0 0.17 0.867 
NUMBER_IMP_CUE Semi-struc. 132 125.9     
 Unstruct. 139 145.6 16621.5 -2.08 0.039 
Notes: 
The variables are defined as follows:  
STRUCTURE = Level of task structure (Semi-structured/Unstructured) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
 Acta Wasaensia     99 
  
mation reliability, as the variable TOT_CUE_TIME (p = 0.290) is not significant. 
Overall, these univariate results are consistent with hypothesis 4. 
 
Table 12. Univariate tests of hypothesis 4 
 
The fifth hypothesis of the study predicts that in the context of high RMM, a less 
experienced auditor uses information more extensively than an experienced one. 
Previous research indicates that the level of conservatism and professional skepti-
cism of an auditor may be conditional on auditor experience. This is particularly 
emphasized when RMM is high and leads to different degrees of information us-
age depending on experience level. Thus, less experienced subjects may display 
greater conservatism and professional skepticism, which causes them to use addi-
tional effort to find negative information or errors from the available information. 
 
 
Hypothesis 4 - Information reliability 
 RELIABILITY N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME More reliab. 127 126.8      
 Less reliab. 144 144.1 16102.0 -1.82 0.069 
TOT_CUE_TIME More reliab. 127 130.6     
 Less reliab. 144 140.7 16589.5 -1.06 0.290 
JUDG_TIME More reliab. 127 122.9     
 Less reliab. 144 147.6 15605.5 -2.59 0.010 
NUMBER_INF More reliab. 127 129.9     
 Less reliab. 144 141.4 16492.0 -1.25 0.210 
NUMBER_INF_9 More reliab. 127 124.3     
 Less reliab. 144 146.4 15782.0 -2.34 0.020 
NUMBER_IMP_CUE More reliab. 127 131.0     
 Less reliab. 144 140.4 16639.5 -1.00 0.322 
Notes: 
The variables are defined as follows:  
RELIABILITY = Level of information reliability (More/Less reliable) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
100      Acta Wasaensia 
Table 13. Univariate tests of hypothesis 5 
 
As shown in Table 13, the results are consistent with the results of Table 9 (audi-
tor experience). Five measures out of six indicate at the 1% and 5% significance 
levels that experienced auditors use less time and cues in their decision-making 
than less experienced subjects when RMM is high. The untabulated results from 
the low RMM treatments indicate that only one variable, NUMBER_INF_9 (p = 
0.020), is significant at the 5% level, while the other four variables are significant 
at the 10% level. This finding suggests that the differences in information usage 
are stronger between less experienced subjects and experienced auditors when 
RMM is high. Thus, the results are consistent with hypothesis 5. 
The sixth hypothesis of the study predicts that the level of task structure does not 
affect the information usage of experienced auditors. Previous research indicates 
that differences between experienced and less experienced auditors in information 
usage behavior might exist only when the task is unstructured. In particular, expe-
rienced auditors’ better knowledge of task-specific information may help them 
Hypothesis 5 – High RMM and experience 
 EXPERIENCE N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME Less exper. 43 80.5      
 Experienced 87 58.1 3459.5 3.18 0.002 
TOT_CUE_TIME Less exper. 43 79.5     
 Experienced 87 58.6 3420.0 2.98 0.003 
JUDG_TIME Less exper. 43 76.2     
 Experienced 87 60.2 3278.0 2.28 0.022 
NUMBER_INF Less exper. 43 75.6     
 Experienced 87 60.5 3252.5 2.19 0.028 
NUMBER_INF_9 Less exper. 43 78.7     
 Experienced 87 59.0 3384.0 2.83 0.005 
NUMBER_IMP_CUE Less exper. 43 70.3     
 Experienced 87 63.1 3021.5 1.02 0.309 
Notes: 
The variables are defined as follows:  
EXPERIENCE = Level of experience (Less experienced/Experienced) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
 Acta Wasaensia     101 
  
perform their information usage in the same manner as in semi-structured tasks. 
By contrast, a less experienced auditor who lacks this knowledge in unstructured 
tasks may use more comprehensively available information to alleviate the in-
creased uncertainty. 
 
Table 14. Univariate tests of hypothesis 6 
 
The univariate results for hypothesis 6 are reported in Table 14. These results 
show no significant differences in information usage between semi-structured and 
unstructured tasks among experienced auditors. However, as these results are vir-
tually the same as those in Table 11 (task structure), the above finding may also 
hold among less experienced subjects. The untabulated results from the less expe-
rienced group indicate that the variable TOT_TIME (p = 0.048) is significant and 
show that they use more total time for unstructured than for semi-structured tasks. 
The variable JUDG_TIME (p = 0.117) is also almost significant at the 10% level, 
Hypothesis 6 – Experienced auditors and task structure 
 STRUCTURE N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME Semi-struc. 92 91.6      
 Unstruct. 93 94.3 8431.0 -0.34 0.732 
TOT_CUE_TIME Semi-struc. 92 93.1     
 Unstruct. 93 92.9 8568.0 0.03 0.975 
JUDG_TIME Semi-struc. 92 91.5     
 Unstruct. 93 94.5 8419.0 -0.38 0.708 
NUMBER_INF Semi-struc. 92 90.4     
 Unstruct. 93 95.6 8319.0 -0.67 0.502 
NUMBER_INF_9 Semi-struc. 92 94.0     
 Unstruct. 93 92.1 8641.5 0.24 0.814 
NUMBER_IMP_CUE Semi-struc. 92 85.3     
 Unstruct. 93 100.6 7846.5 -1.96 0.050 
Notes: 
The variables are defined as follows:  
STRUCTURE = Level of task structure (Semi-structured/Unstructured) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
102      Acta Wasaensia 
while the other variables are clearly non-significant. Thus, these results lend some 
support for hypothesis 6. 
The final hypothesis of the study predicts that in the context of less reliable in-
formation, less experienced auditors use information more extensively than expe-
rienced auditors. A general theory from psychology (Sacchi & Burigo 2008) sug-
gests that when information is reliable, experienced auditors may reject typical 
goal-directed information usage strategies, use less cognitive efforts and apply 
sequential information search strategies that are similar to those used by less ex-
perienced decision-makers. By contrast, when the available information is less 
reliable, differences in information usage strategies between less experienced and 
more experienced auditors are expected to emerge. 
As shown in Table 15, there are no significant results in information usage be-
tween less experienced and experienced auditors when information is less relia-
ble. Thus, hypothesis 7 is not supported by the univariate tests34. By contrast, the 
untabulated results of the more reliable information treatments show significant 
differences in information usage between experience levels. Five variables, 
TOT_TIME (p = 0.000), TOT_CUE_TIME (p = 0.002), JUDG_TIME (p = 
0.001), NUMBER_INF (p = 0.006) and NUMBER_INF_9 (p = 0.000), are signif-
icant at the 1% level, consistently showing that experienced auditors use less time 
and cues than less experienced subjects. The variable NUMBER_IMP_CUE (p = 
0.066) is also significant at the 10% level, further indicating differences. 
To summarize, the results of the univariate tests provide varying support for the 
study’s hypotheses. In particular, the results regarding hypotheses 1 (experience), 
4 (information reliability) and 5 (experience and RMM) support the hypothesized 
expectations. Some support is also found for hypotheses 2 (RMM) and 6 (experi-
ence and task structure), while hypotheses 3 (task structure) and 7 (experience 
and information reliability) are not supported. 
 
 
 
 
                                               
 
34  Regarding this hypothesis, further tests examining whether information reliability and auditor 
experience have an effect on the information usage order are not supported (see Chapter 6.4 
for the tests). 
 Acta Wasaensia     103 
  
Table 15. Univariate tests of hypothesis 7 
 
6.3 Multivariate analyses 
In this section, multivariate analyses are used to test the hypotheses and draw 
conclusions about whether they are supported or not. The methodologies used to 
examine the effect of the independent variables on information usage are ANO-
VA and ordered logistic regression. These two methods are chosen because there 
are two types of dependent variables. For continuous dependent variables, the 
ANOVA method is applied (TOT_TIME, TOT_CUE_TIME and JUDG_TIME) 
and for discrete dependent variables (NUMBER_INF, NUMBER_INF_9 and 
NUMBER_IMP_CUE), ordered logistic regression is applied. 
Hypothesis 7 – Less reliable information and experience 
 EXPERIENCE N Mean score 
Wilcoxon 
statistic Z 
Asymptotic 
Significance (2-
tailed) 
Variable 
TOT_TIME Less exper. 49 78.0      
 Experienced 95 69.6 3824.0 1.14 0.253 
TOT_CUE_TIME Less exper. 49 79.9     
 Experienced. 95 68.7 3915.0 1.53 0.127 
JUDG_TIME Less exper. 49 75.0     
 Experienced 95 71.3 3665.0 0.47 0.637 
NUMBER_INF Less exper. 49 79.1     
 Experienced 95 69.1 3874.0 1.39 0.164 
NUMBER_INF_9 Less exper. 49 79.7     
 Experienced 95 68.8 3903.0 1.49 0.135 
NUMBER_IMP_CUE Less exper. 49 72.9     
 Experienced 95 72.3 3570.5 0.07 0.941 
Notes: 
The variables are defined as follows:  
EXPERIENCE = Level of experience (Less experienced/Experienced) 
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time used for reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads  
 
104      Acta Wasaensia 
6.3.1 Continuous dependent variables - ANOVA 
Factorial ANOVA is used to test the differences between the group means based 
on two or more categorical independent variables with a continuous dependent 
variable. If significant main or interaction effects are found, then follow-up tests 
(planned  comparisons  or  post-hoc  tests)  can  be  used  to  find  out  which  group  
means significantly differ from each other. 
A 2×2×2×2 factorial ANOVA design is used in this study. Dependent variables 
are logarithmic transformed to improve the normality assumptions required by 
ANOVA. Then, the following models are specified: 
 
(1)                                                             Y)RELIABILITSTRUCTURERISKe(INEXP
Y)RELIABILITSTRUCTURERISK(INEXPY)RELIABILITSTRUCTURE(RISK
Y)RELIABILITSTRUCTURE(INEXPY)RELIABILITRISK(INEXP
Y)RELIABILIT(STRUCTUREY)RELIABILIT(RISKY)RELIABILIT(INEXP
YRELIABILITSTRUCTURE)RISK(INEXPSTRUCTURE)(RISK
STRUCTURE)(INEXPSTRUCTURERISK)(INEXPRISKINEXPµY
ijklm
ijkljkl
iklijl
kljlil
lijkjk
jkkijjiijklm
uuu
uuuuu
uuuu
uuu
uuu
uu 
 
Where: 
Y = Dependent variable; 
LN_TOT_TIME  =  Natural  logarithm  of  total  time  spent  
on the task (Model 1). 
LN_TOT_CUE_TIME = Natural logarithm of total time 
spent on cue screens (Model 2). 
LN_JUDG_TIME = Natural  logarithm of time spent out-
side of cue screens (Model 3). 
 
Independent variables: 
i  = The level of INEXP (0 denotes an experienced, 1 de-
notes a less experienced auditor). 
j = The level of RISK (0 denotes low risk, 1 denotes high 
risk). 
 Acta Wasaensia     105 
  
k  =  The  level  of  STRUCTURE  (0  denotes  a  semi-
structured task, 1 denotes an unstructured task). 
l = The level of RELIABILITY (0 denotes more reliable 
information, 1 denotes less reliable information). 
m = The subject within treatment (1–16). 
 
The residual plots for Models 1–3 are presented in Appendix 7. The visual inspec-
tion  of  these  plots  suggests  that  the  residuals  of  all  models  are  fairly  normally  
distributed. The results for the ANOVA models are presented in Table 16. The 
findings reported in Panel A show that INEXP and RISK are significant at the 1% 
and 5% percent levels, respectively. Further, STRUCTURE is almost significant 
at the 10% level, while RELIABILITY has no significant main effect to the used 
total time. All the signs of the mean differences are as predicted (not tabulated). 
These results seem to support H1 and H2 in that the total time for auditors’ deci-
sion-making is affected by both the level of experience and risk. 
 
 
 
 
 
 
 
 
  
106      Acta Wasaensia 
Table 16. ANOVA tables for continuous variables 
Panel A. Dependent variable LN_TOT_TIME (Model 1)  
Variable df Sum of Mean   
  Squares square F p 
INEXP 1 2.68 2.68 14.52 0.000 
RISK 1 0.72 0.72 3.87 0.050 
STRUCTURE (STR) 1 0.49 0.49 2.66 0.104 
RELIABILITY (REL) 1 0.33 0.33 1.76 0.186 
INEXP × RISK  1 0.15 0.15 0.80 0.370 
INEXP × STR  1 0.29 0.29 1.58 0.210 
INEXP × REL 1 0.52 0.52 2.81 0.095 
RISK × STR 1 0.23 0.23 1.22 0.271 
RISK × REL 1 0.08 0.08 0.41 0.524 
STR × REL 1 0.27 0.27 1.48 0.224 
INEXP × RISK × STR  1 0.56 0.56 3.05 0.082 
INEXP × STR × REL  1 0.28 0.28 1.51 0.220 
RISK × STR × REL  1 0.18 0.18 0.98 0.323 
RISK × REL × INEXP 1 0.00 0.00 0.02 0.891 
INEXP × RISK × STR × REL 1 0.10 0.10 0.54 0.463 
Model 15 6.44 0.43 2.32 0.004 
Error 255 47.14 0.19   
Total 270 53.57    
Panel B. Dependent variable LN_TOT_CUE_TIME (Model 2) 
Variable df Sum of Mean F  p 
  Squares square   
INEXP 1 3.58 3.58 9.18 0.003 
RISK 1 0.26 0.26 0.66 0.418 
STRUCTURE (STR) 1 0.06 0.06 0.14 0.704 
RELIABILITY (REL) 1 0.18 0.18 0.45 0.504 
INEXP × RISK  1 0.11 0.11 0.27 0.605 
INEXP × STR  1 0.08 0.08 0.19 0.663 
INEXP × REL 1 0.51 0.51 1.31 0.253 
RISK × STR 1 0.14 0.14 0.36 0.550 
RISK × REL 1 0.10 0.10 0.25 0.618 
STR × REL 1 1.08 1.08 2.78 0.097 
INEXP × RISK × STR  1 0.31 0.31 0.78 0.377 
INEXP × STR × REL  1 0.57 0.57 1.47 0.227 
RISK × STR × REL  1 0.00 0.00 0.01 0.930 
RISK × REL × INEXP 1 0.06 0.06 0.15 0.702 
INEXP × RISK × STR × REL 1 0.00 0.00 0.00 0.946 
Model 15 6.36 0.42 1.09 0.369 
Error 255 99.56 0.39   
Total 270 105.92    
      
 Acta Wasaensia     107 
  
Table 16. Continued 
      
Panel C. Dependent variable LN_JUDG_TIME (Model 3)  
Variable df Sum of Mean F  p 
  Squares square   
INEXP 1 1.16 1.16 6.74 0.010 
RISK 1 1.08 1.08 6.27 0.013 
STRUCTURE (STR) 1 1.14 1.14 6.62 0.011 
RELIABILITY (REL) 1 0.51 0.51 2.93 0.088 
INEXP × RISK  1 0.03 0.03 0.15 0.694 
INEXP × STR  1 0.71 0.71 4.10 0.044 
INEXP × REL 1 0.49 0.49 2.83 0.094 
RISK × STR 1 0.10 0.10 0.59 0.443 
RISK × REL 1 0.00 0.00 0.00 0.958 
STR × REL 1 0.05 0.05 0.29 0.590 
INEXP × RISK × STR  1 0.46 0.46 2.70 0.102 
INEXP × STR × REL  1 0.18 0.18 1.07 0.302 
RISK × STR × REL  1 0.94 0.94 5.43 0.021 
RISK × REL × INEXP 1 0.04 0.04 0.21 0.646 
INEXP × RISK × STR × REL 1 0.79 0.79 4.58 0.033 
Model 15 6.87 0.46 2.66 0.001 
Error 255 43.95 0.17   
Total 270 50.82    
Notes:      
The variables are defined as follows:      
LN_TOT_TIME = Natural logarithm of total time spent on the task 
LN_TOT_CUE_TIME = Natural logarithm of total time used for reading cues  
LN_JUDG_TIME = Natural logarithm of time spent outside of cue screens 
INEXP = A categorical variable with a value of 1 if a subject is less experienced, otherwise 0 
RISK = A categorical variable with a value of 1 if a treatment group contains high RMM, otherwise 0 
STRUCTURE = A categorical variable with a value of 1 if a treatment group is an unstructured task, other-
wise 0 
RELIABILITY = A categorical variable with a value of 1 if a treatment group contains the less reliable in-
formation manipulation, otherwise 0 
 
The results also show significant two- and three-way interactions with experience 
at the 10% level. For each significant interaction between the independent varia-
bles, an interaction plot is provided. In Figure 11, an interaction plot with experi-
ence and reliability is presented. The Tukey–Kramer post-hoc35 tests indicate the 
following significant observations. Compared with when an auditor is experi-
enced and information reliable, it takes less time THAN when 1) information is 
less reliable (mean difference= -0.175, p = 0.031), 2) the subject is less experi-
                                               
 
35 Alpha 0.05 is used in all post-hoc tests in this study. 
108      Acta Wasaensia 
enced (mean difference= -0.321, p = 0.002) and 3) information is less reliable and 
the subject is less experienced (mean difference= -0.300, p = 0.001). This interac-
tion suggests that when information comes from a reliable source, experienced 
auditors are able to perform tasks in significantly less time compared with all oth-
er conditions. This finding is inconsistent with hypothesis 7, which predicts that 
in the context of less reliable information, a less experienced auditor uses infor-
mation more extensively than an experienced one. Thus, consistent with the uni-
variate results, the multivariate results show the opposite to be true. 
 
 
 
 
 
 
 
Figure 11. Interaction plot – estimated marginal means for experience and in-
formation reliability 
 
In Figure 12, a three-way interaction plot with experience, risk and structure is 
presented. The Tukey–Kramer post-hoc tests indicate the following significant 
observations. Compared with when a subject is less experienced, a task is unstruc-
tured and risk is high, it takes more time  THAN  when  1)  an  auditor  is  experi-
enced, a task semi-structured and risk low (mean difference= 0.535, p = 0.000), 2) 
an auditor is experienced and risk low (mean difference= 0.475, p = 0.001), 3) an 
auditor is experienced and a task semi-structured (mean difference= 0.434, p = 
0.005) and an auditor is experienced (mean difference= 0.450, p = 0.003). This 
interaction can be interpreted that when a task is unstructured and risk is high, 
less experienced subjects use significantly more time than experienced auditors 
regardless of risk or structure level. Thus, the finding suggests that the simultane-
ous presence of high risk and an unstructured task might be necessary to see expe-
rience effects across experience levels. 
 Acta Wasaensia     109 
  
 
Figure 12. Interaction plots – estimated marginal means for experience, risk and 
structure 
 
Panel B of Table 16 presents the results from Model 236, where time spent on cue 
screens is a dependent variable. From the main effects, INEXP is significant at 
the 1% level37. This result indicates that less experienced subjects use more time 
on cue screens than experienced auditors (mean difference = 0.257, not tabulat-
ed). This result supports H1. Further, the two-way interaction between RELIA-
BILITY and STRUCTURE is barely significant at the conventional levels. The 
post-hoc tests show no significant differences in means within this interaction; 
hence, it is not analyzed further. The non-significance of the other variables and 
interactions suggests that manipulated variables have no effect on used cue time. 
Finally,  Panel C of Table 16 presents the results from Model 3,  where used cue 
time is extracted from total time. The dependent variable in this model measures 
judgment time, i.e. the time used outside of cue screens for processing cues and 
making task-specific judgments38. Panel C shows that all the main effects are sig-
nificant at the 5% level apart from RELIABILITY, which is significant at the 
                                                
 
36 While this model overall is not significant (p = 0.369), the removal of all non-significant in-
teractions changes the model to be significant at the 5% level (p = 0.035). In this model, IN-
EXP is still significant at the 1% level (p = 0.003). 
37 The multivariate analyses were also run without students. The untabulated results show that 
INEXP is now significant at the 5% level, while the other results remain virtually the same. 
Other results of tests without students are discussed in Chapter 6.4. 
38  To further point out that this variable measures the time used for processing cues, all discrete 
dependent variables are added as covariates to this model one by one. All these variables are 
significant predictors of JUDG_TIME at the 1% level with F-values ranging from 11.2 to 
43.5, indicating a significant effect size. 
110      Acta Wasaensia 
10% level. Again, all signs of mean differences are as predicted (not tabulated). 
The results also show several two-, three- and four-way interactions. In Figure 13, 
both two-way interactions are presented, where experience is on the X-axis and 
STRUCTURE or RELIABILITY is plotted one at a time (A and B). 
In Figure 13A, the Tukey–Kramer post-hoc tests indicate that less experienced 
subjects use significantly more time in unstructured tasks compared with when 1) 
an auditor is experienced and a task semi-structured (mean difference= 0.292, p = 
0.002), 2) an auditor is experienced (mean difference= 0.261, p = 0.006) and 3) a 
subject is less experienced (mean difference= 0.259, p = 0.033). This result is 
consistent with hypothesis 6 that predicts that the level of task structure does not 
affect the information usage of experienced auditors. According to this result, 
only less experienced subjects are affected by task structure when information 
usage is measured by the time used to process cues and make judgments. 
         A)      B) 
 
 
 
 
 
 
 
Figure 13. Interaction plots – estimated marginal means for experience and 
structure/information reliability 
 
Figure 13B shows a similar interaction pattern between experience and reliability 
to that presented in Figure 11 (total time as a dependent variable). Thus, the anal-
ysis is not repeated here. Taken together, these two interactions suggest that the 
observed interaction is attributable specifically to the used judgment time, but not 
to the time spent on cue screens. 
In  Figure  14,  a  three-way  interaction  plot  with  risk,  reliability  and  structure  is  
presented. The Tukey–Kramer post-hoc tests indicate that when “extreme” treat-
 Acta Wasaensia     111 
  
ments (defined as “low risk, more reliable information and semi-structured task” 
vs. “high risk, less reliable information and unstructured task”) are compared, the 
mean difference in used time is 0.514 (p = 0.002) in the expected direction. This 
complex interaction can be interpreted that used time increases significantly only 
when three “uncertainty factors”39 exist simultaneously. Thus, when one uncer-
tainty factor is present, adding a second uncertainty factor does not cause a signif-
icant increase in the dependent variable. However, the simultaneous presence of 
these three factors increases significantly the time used for processing cues. This 
finding suggests that the relationships between the uncertainty factors and the 
dependent variable are non-linear. 
 
 
 
 
 
 
 
 
Figure 14. Interaction plots – estimated marginal means for risk, structure and 
information reliability 
 
Finally, a four-way interaction is examined by comparing the significant differ-
ences in means between the cells. The Tukey–Kramer post-hoc tests indicate that 
compared with when subject is less experienced, risk high, task unstructured and 
information less reliable (hereafter “extreme”), it takes less time when: 
 
 
                                               
 
39 The term “uncertainty factor” is used to refer to any of the independent variables when they 
have a value of 1 (i.e. less experienced auditor/ high risk/unstructured task/ less reliable in-
formation). 
112      Acta Wasaensia 
1. A task is semi-structured (mean difference= -0.736, p = 0.016), 
2. an auditor is experienced and information more reliable (mean differ-
ence= -0.398, p = 0.021), 
3. an auditor is experienced, information more reliable and a task semi-
structured (mean difference= -0.636, p = 0.024), 
4. an auditor is experienced, risk low and a task semi-structured (mean 
difference= -0.585, p = 0.043), 
5. an auditor is experienced, risk low and information more reliable 
(mean difference= -0.667, p = 0.010), 
6. an auditor is experienced, risk low, a task semi-structured and infor-
mation more reliable (mean difference= -0.786, p = 0.001). 
Five out of these six observations suggest that in many cases experienced auditors 
are able to use less judgment time compared with the extreme situation. In partic-
ular, observations 3, 4 and 5 indicate that the presence of one uncertainty factor 
does not significantly increase experienced auditors’ judgment time. This result 
also holds in observation 2 where two uncertainty factors are present. Thus, the 
results indicate that experienced auditors may be able process cues in an almost 
similar way despite there being only one (or indeed no) uncertainty factors pre-
sent. However, because of the high complexity of this interaction and the few 
observations in some cells, one must be cautious when making conclusions about 
this interaction40. 
6.3.2 Discrete dependent variables – ordered logistic regression 
As the other dependent variables (NUMBER_INF, NUMBER_INF_9 and NUM-
BER_IMP_CUE) of the study are discrete variables and count data by their na-
ture, using an ordinary least squares regression would violate the normality as-
sumption of residuals. Instead, an ordered logistic regression can be used when 
modeling categorical dependent variables and categories of responses can be or-
dered. Ordered logit models are especially suitable when dependent variable val-
ues are recorded ordinally, but it is desirable to consider dependent variables as 
continuous (Flom 2010).  
Thus, this study’s dependent discrete variables are modeled using ordered logit 
models, as the number of used information can be ordered. However, as shown in 
Panel B of Table 3, these variables have very little variance, while their ranges are 
                                               
 
40 Because the smallest cell has only six observations (see Appendix 6), this interaction has a 
significant risk of Type I error and thus it is not discussed in the further parts of this study. 
 Acta Wasaensia     113 
  
considerably high (from 0 to 25). Further, the distribution of NUMBER_INF in-
dicates that 40.2% of subjects used all 12 cues at once (not tabulated). Conse-
quently, there are many levels with only few observations. Thus, in order to im-
prove model fit, the levels of each dependent variable are equalized to three simi-
lar-sized (about) categories for the number of used information: low, medium and 
high. After the combination of the levels, the following ordered logit models are 
specified: 
 
(2)                                                      )(
)()(
)()(
)()()(
)()()(
))1/((
15
1413
1211
1098
765
43210
YRELIABILITSTRUCTURERISKINEXPȕ
INEXPYRELIABILITRISKȕYRELIABILITSTRUCTURERISKȕ
YRELIABILITSTRUCTUREINEXPȕSTRUCTURERISKINEXPȕ
YRELIABILITSTRUCTUREȕYRELIABILITRISKȕSTRCTURERISKȕ
YRELIABILITINEXPȕSTRUCTUREINEXPȕRISKINEXPȕ
YRELIABILITȕSTRUCTUREȕRISKȕINEXPȕȕYYLog
uuu
uuuu
uuuu
uuu
uuu
 
 
 
Where: 
Y = Dependent variable: 
NUMBER_INF = Total number of read cues, including 
multiple reads (Model 4). 
NUMBER_INF_9 = Number of over 9 second read cues, 
including multiple reads (Model 5). 
NUMBER_IMP_CUE = Number of read cues whose im-
portance has been self-evaluated to be more than 4, in-
cluding multiple reads (Model 6). 
Independent variables: 
RISK = A dummy variable with a value of 1 if a treatment 
group contains a high RMM manipulation, otherwise 0. 
RELIABILITY = A dummy variable with a value of 1 if a 
treatment group contains a less reliable information ma-
nipulation, otherwise 0. 
STRUCTURE = A dummy variable with a value of 1 if a 
treatment group is an unstructured task, otherwise 0. 
INEXP = A dummy variable with a value of 1 if an audi-
tor is less experienced, otherwise 0. 
 
114      Acta Wasaensia 
Table 17 reports the results for Models 4–6. These results generally indicate few 
statistically significant relationships between the independent and dependent vari-
ables. RELIABILITY has a positive and significant coefficient in two models, 
suggesting that less experienced subjects select quantitatively more information 
than experienced auditors. The results also show few three- and four-factor inter-
actions at the 10% level. 
 
Table 17. Ordered logistic regression analyses for discrete variables  
 (4) (5) (6) 
 NUMBER NUMBER NUMBER_IMP 
Variable _INF _INF_9 _CUE 
Intercept1 -1.911 *** -2.623 *** -1.252 *** 
 (21.72) (34.64) (8.89) 
Intercept2 -0.009 -0.047 0.336 
 (0.00) (0.01) (0.66) 
INEXP 0.635 0.918 1.257 
 (0.67) (1.21) (2.06) 
RISK -0.571 0.008 -0.683 
 (0.89) (0.00) (1.35) 
STRUCTURE (STR) 0.451 -0.007 0.458 
 (0.69) (0.00) (0.69) 
RELIABILITY (REL) 0.872 * 1.014 * 0.198 
 (2.77) (3.35) (0.13) 
INEXP × RISK  1.135 0.715 0.418 
 (1.12) (0.40) (0.13) 
INEXP × STR  1.041 0.930 -0.901 
 (0.83) (0.58) (0.57) 
INEXP × REL -0.547 -0.598 -0.676 
 (0.30) (0.31) (0.41) 
RISK × STR 0.571 0.214 0.885 
 (0.89) (0.07) (1.24) 
RISK × REL 0.001 -0.081 0.847 
 (0.00) (0.01) (1.17) 
STR × REL -0.653 -0.254 0.655 
 (0.78) (0.11) (0.72) 
INEXP × RISK × STR  -2.113 -1.201 -0.619 
 (1.97) (0.57) (0.16) 
INEXP × STR × REL  -0.982 -0.828 -0.241 
 (0.47) (0.30) (0.03) 
RISK × STR × REL  -1.052 -0.285 -2.417 ** 
 (0.86) (0.06) (4.79) 
RISK × REL × INEXP -1.275 -0.914 -1.303 
 (0.79) (0.37) (0.80) 
INEXP × RISK × STR × REL 3.844 * 2.451 2.581 
 (3.45) (1.33) (1.59) 
-2 Log Likelihood 552.1 514.3 574.8 
Likelihood ratio, Ȥ2 23.3 *** 24.0 *** 19.8 *** 
Pseudo R2 4.1 % 4.5 % 3.3 % 
AIC 586.1 548.3 608.8 
 Acta Wasaensia     115 
  
Table 17. Continued    
    
Notes:     
Wald Chi-Squares are reported in parentheses. Statistical significance based on two-tailed tests at the 1%, 5% 
and 10% levels are denoted by ***, ** and *, respectively. 
The variables are defined as follows:    
NUMBER_INF = Total number of read cues, including multiple reads   
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads   
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads     
INEXP = A dummy variable with a value of 1 if an auditor is less experienced, otherwise 0 
RISK = A dummy variable with a value of 1 if a treat. group contains a high RMM manipulation, otherwise 0 
STRUCTURE = A dummy variable with a value of 1 if a treatment group is an unstructured task, otherwise 0 
RELIABILITY = A dummy variable with a value of 1 if a treatment group contains a less reliable infor-
mation manipulation, otherwise 0 
    
 
Owing to the complex interpretations of these interactions, all the coefficients of 
these models are re-estimated with alternative independent variables. In these 
models, TREATMENT and INEXP are independent variables. To further ease 
interpretation, treatment 1 is set as the reference category to which other treat-
ment groups are compared. This choice is made on the basis of this study’s theory 
that predicts that information usage should be at its lowest level when risk is low, 
information is reliable and a task is semi-structured (i.e. treatment 1). The follow-
ing models have identical intercepts and model statistics to the models reported in 
Table 17. 
 
(3)             
))1/((
815714613512
411310298
877665544332210
INEXPTRȕINEXPTRȕINEXPTRȕINEXPTRȕ
INEXPTRȕINEXPTRȕINEXPTRȕINEXPȕ
TRȕTRȕTRȕTRȕTRȕTRȕTRȕȕYYLog
uuuu
uuu
 
 
Where: 
Dependent variables are as previously defined. 
 
Independent variables: 
TR2 = A dummy variable with a value of 1 if a treatment group is two 
(i.e. low risk, less reliable information, semi-structured task), other-
wise 0. 
116      Acta Wasaensia 
TR3 =  A  dummy  variable  with  a  value  of  1  if  a  treatment  group  is  
three (i.e. high risk, more reliable information, semi-structured task), 
otherwise 0. 
TR4 = A dummy variable with a value of 1 if a treatment group is four 
(i.e. high risk, less reliable information, semi-structured task), other-
wise 0. 
TR5 = A dummy variable with a value of 1 if a treatment group is five 
(i.e. low risk, more reliable information, unstructured task), otherwise 
0. 
TR6 = A dummy variable with a value of 1 if a treatment group is six 
(i.e. low risk, less reliable information, unstructured task), otherwise 
0. 
TR7 =  A  dummy  variable  with  a  value  of  1  if  a  treatment  group  is  
seven (i.e. high risk, more reliable information, unstructured task), 
otherwise 0. 
TR8 =  A  dummy  variable  with  a  value  of  1  if  a  treatment  group  is  
eight (i.e. high risk, less reliable information, unstructured task), oth-
erwise 0. 
INEXP = A dummy variable with a value of 1 if an auditor is less ex-
perienced, otherwise 0. 
 
Table 18 reports the coefficients of the TREATMENT and INEXP41 variables in 
Models 4 to 6. In Model 3, TR2 is the only significant coefficient, while in Model 
4 both TR2 and TR4 are significant at the 10% level, indicating evidence that less 
reliable information is needed more for decision-making, but only in semi-
structured tasks. In Model 6, two coefficients, TR6 and TR6*INEXP, are signifi-
cant at the 5% and 10% levels, respectively. This finding suggests that less relia-
ble information attracts using important cues in this treatment group, but the ef-
fect is only limited to experienced auditors as the interaction variable’s 
(TR6*INEXP) coefficient is negative. Taken together, the results from Models 4–
6 show only weak support for hypothesis 4. 
                                               
 
41 Because the models are virtually the same, the coefficients and significance levels of INEXP 
are identical to those reported in Table 17. However, they are repeated here to ease the inter-
pretation of the interaction variables. 
 Acta Wasaensia     117 
  
Table 18. Coefficients of discrete variables with TREATMENT and IN-
EXP as the independent variables  
 
 (4) (5) (6) 
 NUMBER NUMBER NUMBER_IMP 
Variable _INF _INF_9 _CUE 
TR1 - Low risk, More rel., Semi-str. 0.000 0.000 0.000 
 (reference treatment)    
TR2 -Low risk, Less rel., Semi-str. 0.872 * 1.014 * 0.198 
 (2.77) (3.35) (0.13) 
TR3-High risk, More rel., Semi-str. -0.571 0.008 -0.684 
 (0.89) (0.00) (1.35) 
TR4-High risk, Less rel., Semi-str. 0.302 0.942 * 0.361 
 (0.27) (2.64) (0.42) 
TR5 - Low risk, More rel., Unstr. 0.451 -0.007 0.458 
 (0.69) (0.00) (0.69) 
TR6-Low risk, Less rel., Unstr. 0.670 0.753 1.311 ** 
 (1.56) (1.79) (5.18) 
TR7-High risk, More rel., Unstr. 0.451 0.215 0.659 
 (0.69) (0.14) (1.38) 
TR8-High risk, Less rel., Unstr. -0381 0.610 -0.059 
 (0.41) (1.06) (0.01) 
INEXP 0.635 0.918 1.257 
 (0.67) (1.21) (2.06) 
TR1 * INEXP 0.000 0.000 0.000 
 (reference treatment)    
TR2 * INEXP -0.547 -0.598 -0.676 
 (0.30) (0.31) (0.41) 
TR3 * INEXP 1.135 0.715 0.418 
 (1.12) (0.40) (0.13) 
TR4 * INEXP -0.686 -0.796 -1.561 
 (0.42) (0.53) (2.02) 
TR5 * INEXP 1.041 0.930 -0.901 
 (0.83) (0.58) (0.57) 
TR6 * INEXP -0.488 -0.496 -1.818 * 
 (0.25) (0.24) (2.89) 
TR7 * INEXP 0.063 0.444 -1.102 
 (0.00) (0.17) (1.03) 
TR8 * INEXP 1.104 0.555 -0.741 
 (0.90) (0.23) (0.40) 
Notes:     
Wald Chi-Squares are reported in parentheses. Statistical significance based on two-tailed tests at the 1%, 5% 
and 10% levels are denoted by ***, ** and *, respectively. 
The variables are defined as follows:   
NUMBER_INF = Total number of read cues, including multiple reads   
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads   
NUMBER_IMP_CUE = Number of read cues whose importance has been self-evaluated to be more than 4, 
including multiple reads    
TR1 = A reference treatment (i.e. low risk, more reliable information, semi-structured task), where other 
treatments are compared  
TR2 = A dummy variable with a value of 1 if a treatment group is two (i.e. low risk, less reliable information, 
semi-structured task), otherwise 0 
118      Acta Wasaensia 
TR3 = A dummy variable with a value of 1 if a treatment group is three (i.e. high risk, more reliable infor-
mation, semi-structured task), otherwise 0 
TR4 = A dummy variable with a value of 1 if a treatment group is four (i.e. high risk, less reliable infor-
mation, semi-structured task), otherwise 0 
TR5 = A dummy variable with a value of 1 if a treatment group is five (i.e. low risk, more reliable infor-
mation, unstructured task), otherwise 0 
TR6 = A dummy variable with a value of 1 if a treatment group is six (i.e. low risk, less reliable information, 
unstructured task), otherwise 0 
TR7 = A dummy variable with a value of 1 if a treatment group is seven (i.e. high risk, more reliable infor-
mation, unstructured task), otherwise 0 
TR8 = A dummy variable with a value of 1 if a treatment group is eight (i.e. high risk, less reliable infor-
mation, unstructured task), otherwise 0 
INEXP = A dummy variable with a value of 1 if an auditor is less experienced, otherwise 0 
 
To summarize, the results of these multivariate tests show consistent support for 
H1. The results also provide some support for H2 and H6, while many of the sta-
tistical tests are non-significant. There is also weak support for H3 and H4, while 
the empirical results do not support H5 or H7. The implications of these results as 
well as the other found non-hypothesized interactions are discussed in Chapter 
6.6. 
6.4 Supplementary analyses and robustness tests 
The objective of this section is to provide supplementary analyses and to test the 
robustness of the previous analyses. First, I investigate whether the independent 
variables auditor experience, risk, task structure and information reliability affect 
information usage order. As discussed in the theory part with hypothesis devel-
opment 1, the audit literature (e.g. Biggs & Mock 1983; Bonner & Pennington 
1991) suggests that experienced auditors apply directed search strategies com-
pared with less experienced auditors, who usually acquire information in the order 
in which it is presented. Furthermore, theoretical arguments behind hypothesis 
development 7 predict that particularly under less reliable information differences 
in information usage order across different experience levels are expected to 
emerge. 
To test this, an order score is calculated where a higher score indicates greater 
deviance from the usage of presentation order (i.e. from top down reading order 
in information menu). For example, a subject who read the 12 cues in presenta-
tion order would score 0, while the opposite reading order would score maximum 
points (12). Although the univariate results show that low experience is negative-
ly associated with the order score (Wilcoxon = 10715.0, p =0.096; not tabulated) 
and high RMM is positively associated with information usage score (Wilcoxon = 
 Acta Wasaensia     119 
  
18791.5, p =0.078; not tabulated), the multivariate analyses (not tabulated) fail to 
find any significant relationships between the key variables of interest and infor-
mation usage order. 
Next,  I  begin robustness tests by investigating whether the cut-off points of two 
discrete dependent variables influence the main results. As the variables NUM-
BER_IMP_CUE and NUMBER_INF_9 have somewhat arbitrary limits of im-
portance and time (i.e. more than 4 cue importance and more than 9 seconds, re-
spectively),  other  cut-off  points  are  also  tested.  Thus,  all  models  with  NUM-
BER_INF_9 are re-estimated with using alternative cut-off points of 4 and 14 
seconds and NUMBER_IMP_CUE with importance values of more than 3 and 5, 
respectively. 
The results (not tabulated) of using alternative cut-off points of 4 and 14 seconds 
(NUMBER_INF_4 and NUMBER_INF_14) are generally consistent with the 
results reported in Tables 17 and 18. The two exceptions are that with the cut-off 
point  of  14  seconds  treatment  2  is  no  longer  significant  at  the  10% level,  while  
treatment 6 becomes significant. Moreover, with different cut-off point for num-
ber of important cues, the results in both models generally hold. The only notable 
exception is that INEXP*TR6 is no longer significant when 5 is used as a cut-off 
point for the number of important cues. 
Next, I investigate whether the cut-off points for INEXP significantly affect the 
results of Models 1–3. The untabulated results show that the results are generally 
robust when task-specific experience’s cut-off point is increased or decreased by 
one unit from 5 on a scale of 0 to 8 (see Chapter 5.7). Almost all main effects are 
still significant at conventional levels when the cut-off point is decreased by one 
unit. The only exception is Model 3 where the two-way interactions related to 
auditor experience are no longer significant. However, it must be noted that using 
this cut-off point reduces the number of less experienced subjects from 86 to only 
70, which also reduces the overall fit of the models. By contrast, when the cut-off 
point is increased by one unit, the results are consistent with the main analyses 
and almost all statistical significances are even stronger than in the main analyses. 
This cut-off point classifies 101 subjects as less experienced compared to the cut-
off point 5 where number of less experienced subjects were 86. Using the alterna-
tive cut-off point of 6 improves the model fit and produces residuals that more 
closely follow the normal distribution. 
Finally, I analyze Models 1–3 without students to see whether the results are driv-
en by subjects who are not auditing professionals. The untabulated results indicate 
that RISK is no longer significant in Model 1 (i.e. risk is not affecting the amount 
of  total  time  used).  However,  consistent  with  the  main  analyses,  the  variable  
120      Acta Wasaensia 
RISK  in  the  Model  3  is  still  found  significant  at  the  5%  level.  The  other  main  
difference is that RELIABILITY is no longer significant in Models 1 and 3, indi-
cating that information reliability does not affect information usage. However, it 
should be noted again that excluding students reduces the overall fit of the models 
and leaves only 66 less experienced subjects to the analyses. To address the prob-
lem of low number of less experienced subjects, I estimate the models after re-
defining the variable INEXP by increasing variable's cut-off point from 5 to 6 
similarly  than  in  the  previous  robustness  test  (see  the  above  paragraph).  I  also  
exclude all the students from this analysis resulting in 81 subjects of low task-
specific experience. After the re-definition, the model fit improves and the esti-
mated results are generally consistent with those reported in the main analyses.  
There are however three notable exceptions. First, results reported in Appendix 8 
show that STRUCTURE is now also significant at the 10% level in the Model 1 
and has also a significant interaction with the experience. Second, interaction be-
tween INEXP and RELIABILITY is no longer significant in Models 1 and 3. 
However, both of these Models have now significant three-way interaction be-
tween INEXP, RELIABILITY and STRUCTURE at the 10% level. Thus, there is 
still an interaction between experience and reliability, but now its effect depends 
on structure as well. Also in the Model 3 there is a new significant three-way in-
teraction between INEXP, RISK and STRUCTURE at the 5% level.   Third,  IN-
EXP is barely insignificant (p = 0.119) in the Model 2 indicating that cues are not 
read longer time by less experienced subjects when the students are "replaced" by 
other less experienced auditors. Thus, robustness tests suggest that long cue times 
are related to subjects having no or little prior task-specific experience. 
To summarize, the results of robustness tests show that alternative cut-off points 
of INEXP have fairly small effect on the main results, when the model fits and 
residuals are on the similar level as in the main analyses. Results also hold when 
students are excluded. While the significances of some interactions depend on the 
used cut-off point, their number is not significantly affected by it. This consistent 
finding about the high number of interactions supports the study’s general conclu-
sion, that the relationship between uncertainty factors and the information usage 
variables are often non-linear (see Chapter 6.6 for further discussion). 
6.5 Analyses of independent variables effect on task-
 specific judgments 
The objective of this section is to investigate the role of the independent variables 
on auditors’ decision-making and judgments. While the multivariate analyses 
suggested that the independent variables have a modest effect on information us-
 Acta Wasaensia     121 
  
age, it is possible that these factors have a stronger effect on actual client continu-
ance and acceptance judgments and the pre-planning judgments of engagements. 
For instance, the client acceptance literature has showed that riskier clients are 
less likely to be accepted as new clients (Asare & Knechel 1995; Johnstone 2000) 
and are charged higher audit fees (Johnstone & Bedard 2003). 
As described in Chapter 5.4, in the client continuance task a subject was asked to 
make four task-related judgments. In the first judgment, the subject was asked to 
make a probability estimate of recommending continuance with an existing client. 
The second estimate related to the judgment confidence level. In the third esti-
mate, the subject was asked to recommend the number of audit hours for the next 
year’s audit (compared with the last audit). The fourth question asked subjects to 
estimate the audit fee. All items were presented on an 11-point scale. In the client 
acceptance task, the questions were almost identical. The notable difference was 
that in audit hour and fee estimates the reference point was now a typical same 
sized firm from the same industry. 
Owing to these slightly differently presented questions in audit hour and fee esti-
mates between the tasks, the models with the above-defined four dependent vari-
ables are estimated separately. In Panel A (B) of Table 19, treatment group 1 (5) 
is set to be the reference category in which treatments 2(6) to 4(8) are compared. 
Otherwise, these ordinal regression models have the same independent variables 
and the interaction variables of the models are the same as those in Eq. 3. The 
descriptive statistics of these variables are reported in Appendix 9. 
Panel A of Table 19 presents the results for the client continuance task.  The re-
sults of Model 1 indicate that the probability of recommending continuance de-
creases to its lowest level when RMM is high and information is less reliable, as 
the coefficient estimate is negative and significant at the 1% level. The coeffi-
cients of TR2 and TR3 are also negative and significant at the 10% level, indicat-
ing that the existence of high RMM or less reliable information alone decreases 
the probability of recommending continuance compared with treatment 1. 
The results of Model 2 show that the coefficient of TR4 is negative and significant 
at the 1% level, indicating that the existence of high RMM and less reliable in-
formation at the same time decrease subject confidence about their judgment ac-
curacy.  The  results  of  Models  3  and  4  indicate  that  the  coefficients  of  TR2 and 
TR4 are significant at the 1% level. This finding suggests that subjects regardless 
of their experience level increase their projected audit hour and fee estimates 
when information is less reliable. 
122      Acta Wasaensia 
Panel B presents the results for the client acceptance task. The results of Model 5 
indicate that only firms in TR8 are less likely to be accepted as new clients. Fur-
ther, INEXP is now significant at the 10% level, indicating that less experience 
further increases the acceptance likelihood of a prospective client. The latter find-
ing is consistent with previous studies (e.g. Libby 1995; Abdolmohammadi & 
Wright 1987) that have stated that less experienced auditors are more conserva-
tive in their judgments than experienced auditors. The results of Models 6–8 are 
similar to the corresponding models of the client continuance task. The one nota-
ble change is that high RMM (TR6) is now also significant in the HOUR_ESTIM 
and FEE_ESTIM models. 
 
Table 19. Ordinal regression models estimated to test how the independent 
variables affect task-specific judgments in semi-structured and 
unstructured tasks 
 
Panel A. Regression analyses of task-specific judgment variables for the continuance task 
 (1) (2) (3) (4) 
Variable JUDGMENT JUDG_CONF HOUR_ESTIM FEE_ESTIM 
TR1 - Low risk, More rel., Semi-str. 0.000 0.000 0.000 0.000 
 (reference treatment)     
TR2 -Low risk, Less rel., Semi-str. -1.037 * -0.375 1.363 *** 1.868 *** 
 (3.57) (0.54) (6.77) (10.71) 
TR3-High risk, More rel. Semi-str. -1.124 * -0.831 0.880 0.890 
 (3.63) (2.08) (2.63) (2.56) 
TR4-High risk, Less rel., Semi-str. -2.142 *** -2.063 *** 1.504 *** 1.789 *** 
 (13.17) (12.35) (7.35) (10.16) 
INEXP -0.443 -1.042 -0.470 -0.276 
 (0.25) (1.61) (0.32) (0.09) 
 TR1 * INEXP 0.000 0.000 0.000 0.000 
 (reference treatment)     
 TR2 * INEXP 0.336 1.138 0.792 0.456 
 (0.10) (1.61) (0.56) (0.17) 
 TR3 * INEXP -0.836 -0.183 0.701 0.223 
 (0.54) (0.03) (0.42) (0.04) 
 TR4 * INEXP 0.300 1.474 1.306 0.722 
 (0.08) (1.97) (1.46) (0.40) 
-2 Log Likelihood 395.2 397.5 388.0  
Likelihood ratio, Ȥ2 23.0 *** 17.5 *** 20.9 *** 21.5 *** 
Pseudo R2 5.5 % 4.2 % 5.1 % 5.5 % 
AIC 427.23 427.46 411.96 391.75 
  
 Acta Wasaensia     123 
  
Table 19. Continued 
Panel B. Regression analyses of task-specific judgment variables for the acceptance task  
 (1) (2) (3) (4) 
Variable JUDGMENT JUDG_CONF HOUR_ESTIM FEE_ESTIM 
TR5 - Low risk, More rel., Unstr. 0.000 0.000 0.000 0.000 
 (reference treatment)     
TR6-Low risk, Less rel., Unstr. -0.912 * -0.260 1.342 ** 1.498 *** 
 (3.26) (0.29) (6.05) (7.78) 
TR7-High risk, More rel., Unstr. 0.337 -0.066 1.678 *** 1.594 *** 
 (0.45) (0.02) (9.60) (8.75) 
TR8-High risk, Less rel., Unstr. -1.124 ** -0.944 * 1.815 *** 1.989 *** 
 (4.65) (3.01) (11.65) (12.82) 
INEXP -1.512 * -1.244 0.147 0.061 
 (3.61) (2.50) (0.04) (0.01) 
 TR5 * INEXP 0.000 0.000 0.000 0.000 
 (reference treatment)     
 TR6 * INEXP 1.405 1.305 0.173 0.634 
 (2.06) (1.82) (0.03) (0.36) 
 TR7 * INEXP 1.571 0.474 -1.181 -0.545 
 (2.40) (0.23) (1.47) (0.29) 
 TR8 * INEXP 0.472 -0.326 0.442 -0.126 
 (0.19) (0.09) (0.18) (0.01) 
-2 Log Likelihood 447.8 430.1 422.5 373.9 
Likelihood ratio, Ȥ2 12.8 ** 14.1 ** 23.0 *** 20.4 *** 
Pseudo R2 2.8 % 3.2 % 5.2 % 5.2 % 
AIC 475.85 460.01 450.48 397.94 
Notes:      
Wald Chi-Squares are reported in parentheses. Statistical significance based on two-tailed tests at the 1%, 5% 
and 10% levels are denoted by ***, ** and *, respectively. 
The variables are defined as follows (on an 11-point Likert-type scale, if not otherwise stated):  
JUDGMENT = Probability judgment of recommending client continuance (Panel A) or acceptance (Panel B) 
JUDG_CONF = Level of confidence in client continuance (Panel A) or acceptance (Panel B) judgment 
HOUR_ESTIM = Estimate of planned audit hours compared with last year (Panel A) or average number 
scaled to client size and industry (Panel B)    
FEE_ESTIM = Estimate of planned audit fee compared with last year (Panel A) or average amount 
scaled to client size and industry (Panel B)    
TR1 = A reference treatment (i.e. low risk, more reliable information, semi-structured task), where other 
treatments are compared (Panel A) 
TR2 = A dummy variable with a value of 1 if a treatment group is two (i.e. low risk, less reliable information, 
semi-structured task), otherwise 0 
TR3 = A dummy variable with a value of 1 if a treatment group is three (i.e. high risk, more reliable infor-
mation, semi-structured task), otherwise 0 
TR4 = A dummy variable with a value of 1 if a treatment group is four (i.e. high risk, less reliable infor-
mation, semi-structured task), otherwise 0 
TR5 = A reference treatment (i.e. low risk, more reliable information, unstructured task), where other treat-
ments are compared (Panel B) 
TR6 = A dummy variable with a value of 1 if a treatment group is six (i.e. low risk, less reliable information, 
unstructured task), otherwise 0 
TR7 = A dummy variable with a value of 1 if a treatment group is seven (i.e. high risk, more reliable infor-
mation, unstructured task), otherwise 0 
TR8 = A dummy variable with a value of 1 if a treatment group is eight (i.e. high risk, less reliable infor-
mation, unstructured task), otherwise 0 
INEXP = A dummy variable with a value of 1 if an auditor is less experienced, otherwise 0  
124      Acta Wasaensia 
The results of Panel B are generally consistent with previous client acceptance 
studies42 (e.g. Johnstone 2000; Johnstone & Bedard 2003; Asare, Cohen & 
Trompeter 2005) that were conducted in a different legal environment. The pre-
sent results particularly support the findings of Beaulieu (2001) and Asare, Cohen 
and Trompeter (2005) that future evidence collection strategies and the planned 
number of audit hours are adjusted depending on client riskiness. 
Overall, the pattern of results reported in Table 19 suggests that auditors respond 
to changes in RMM and information reliability in their client continuance and 
acceptance judgments as well as in the related pre-planning judgments of the en-
gagement. Specifically, the results of “extreme conditions” on both tasks (i.e. 
treatments 4 and 8) indicate strongly that high RMM and less reliable information 
together lead to significant judgment revisions. 
6.5.1 Effect of information usage on task-specific judgments 
As reasoned in the development of hypotheses 2 (RMM) and 4 (information reli-
ability), it is proposed that auditors use extensive information usage as a compen-
satory mechanism to reduce uncertainty in the decision-making process. Thus, 
subjects who perform extensive information usage may make more positive 
judgments about client continuance/acceptance and lower estimations of future 
audit hours and fees, as the information in this experiment indicates no negative 
issues about the client. By contrast, subjects who use less information compensate 
for this uncertainty by making more conservative judgments and by increasing 
their projections of future audit hours and fees. 
To test if the level of information usage and task-specific judgments are related, 
all models in Table 19 are extended with the continuous and discrete information 
usage variables. They are added as covariates one-by-one to these models to 
avoid multicollinearity problems. 
The untabulated results show that none of the continuous information usage vari-
ables is statistically significant in the models. By contrast, when three discrete 
variables are added one-by-one to these models, the results are mixed. In a few of 
the  models  where  JUDGMENT  and  JUDG_CONF  are  dependent  variables,  the  
coefficients of the models indicate that a greater quantitative usage of cues leads 
                                               
 
42 Unlike Johnstone (2000), the aim of this study was not to rank risk management strategies 
(avoiding risky clients vs. risk mitigation strategies in engagement), but rather to observe their 
usage. 
 Acta Wasaensia     125 
  
to a lower probability of client continuance and acceptance as well as lower 
judgment confidence. However, in the majority of these models, the coefficients 
are not statistically significant. 
The most robust finding in other models is that in FEE_ESTIM models, the coef-
ficients of discrete variables are positive and significant at the 5% level. Thus, a 
greater quantitative usage of cues leads to a higher estimate of planned audit fees 
than a lower usage of cues. 
Taken together, the results suggest that the extensiveness of information usage is 
not strongly related to task-specific judgments. However, some evidence of these 
analyses suggests that greater quantitative information usage implies negative 
client attractiveness, which specifically decreases the probability of positive client 
continuance and acceptance judgments and increases projections of future audit 
fees. Therefore, the results do not suggest that auditors’ more extensive infor-
mation usage compensates for uncertainty, i.e. would lead to more positive judg-
ments that are related to continuance or acceptance. 
6.6 Summary and discussion of the results 
Overall, the results show that several independent variables affect auditors’ in-
formation usage when time spent on the task is used as a measure. Specifically, I 
find that these independent variables also interact with each other, meaning that 
time spent on the task is conditional on a combination of several independent var-
iables at the same time. By contrast, when information usage is measured by the 
number of used information (i.e. count measures), this study finds only a very 
weak relationship between the dependent and independent variables. The main 
findings of the study are discussed below.  
First, the results show semi-strong support that auditor experience affects infor-
mation usage. This study finds that experienced auditors use less total time for 
tasks than less experienced subjects. Consistently, both refined measures of time 
indicate that experienced auditors use less time for reading cues on cue screens as 
well as processing those cues outside of those screens. However, robustness tests 
suggest that longer cue screen times are only related to subjects having no or very 
little prior task-specific experience. 
This shorter cue time does not seem to be explained by the number of cues select-
ed from the information menu, as experience was not significant when the count 
measures were dependent variables. This implies that experienced auditors may 
have better developed reading techniques or shorter assimilation times on cue 
126      Acta Wasaensia 
screens, which is more effortless than the approaches used by less experienced 
subjects. For example, experienced auditors may have scanned the cues to find 
critical items instead of reading each cue word-for-word43. The results also show 
that experienced auditors used significantly less time for processing cues outside 
of cue screens. Thus, these findings suggests that higher task-specific experience 
leads to the less effortful examination of each single cue when – presumably44 – 
reading these and a lower processing time of cues. 
The above results are in line with the findings of Davis (1996), who showed that 
experienced auditors used less time in control risk assessment tasks. However, he 
also found that experienced auditors acquired fewer cues, which may explain the 
decreased time. Further, the findings that time to read cues decreases as experi-
ence increases is consistent with the findings of Moroney (2007) who had similar 
findings when the audit experience was measured by industry-expertise. All this 
suggests that one of the reasons for the better efficiency of experienced auditors 
in audit tasks may be their advanced cue reading/assimilation techniques and 
evaluation/combining of cues. 
Second, this study finds some evidence that there exists a relationship between 
level of RMM and auditors’ information usage. Specifically, I find that high 
RMM increases total time spent on the task and time used outside of cue screens, 
but not time spent on cue screens. This finding implies that auditors also read and 
assimilate selected cues in a conventional way in a risky setting, but demonstrate 
greater effort in their cue processing. Thus, instead of altering their usual reading 
techniques, auditors seem to react to the environmental uncertainty by evaluating 
each used cue’s “information value” for their task-specific judgments carefully 
under high RMM. 
Third, this study finds evidence that more time is used to process information in 
unstructured tasks than in semi-structured tasks. However, a significant two-way 
interaction indicates that only less experienced subjects are affected by task struc-
ture, as they use more time outside of cue screens in unstructured tasks. The re-
sults also suggest that task structure interacts with other variables when infor-
                                               
 
43 Alternatively, they may have deemed cues more often to be irrelevant or unimportant after 
second reading and returned to the information menu. However, the descriptive statistics in 
Table 5 do not support this possibility, as the overall self-evaluated cue importance values 
were virtually the same for experienced auditors and less experienced subjects. 
44 There is an inherent limitation in cue time and judgment time measures, as some subjects may 
process cues on cue screens, while others may only read cues and process them while making 
task-specific judgments. However, I am not aware of any study that has pointed out that a 
more and less experienced auditor systematically differ in this sense. 
 Acta Wasaensia     127 
  
mation usage is measured by total time used. A significant three-way interaction 
indicates that regardless of task structure and risk, experienced auditors use less 
total time than less experienced subjects in unstructured tasks when RMM is high. 
However, as this interaction is not significant with the refined time measures (cue 
or judgment time), it is not clear which part of information usage is most affected 
by auditor experience. Taken together, these two significant interactions suggest 
that when a decision-making situation becomes difficult to less experienced sub-
jects because of unstructured tasks, experienced auditors’ expertise becomes sali-
ent. The present findings are in line with the general conclusion from the auditor 
expertise literature (e.g. Abdolmohammadi & Wright 1987; Tan & Kao 1999), 
indicating that experience effects exist only in the most difficult tasks. 
Fourth, this study finds evidence that the presence of less reliable information 
increases the time used outside of cue screens. This finding suggests that less reli-
able information is processed more effortful than more reliable information, while 
the reading and assimilation time of cues is unaffected. Surprisingly, this finding 
does not support the ad hoc expectations of increased overall uncertainty, where 
one would expect information from a less reliable source to be processed in a 
more effortful manner in all decision-making steps in order to find any contradic-
tions45. The results also show weak support that quantitatively more information 
is selected when available information is less reliable. 
The results further show some interaction between information reliability and 
auditor experience. While experience’s main effect indicates that less experienced 
subjects used more time than experienced auditors, two separate two-way interac-
tions show that when information is less reliable, experienced auditors no longer 
use significantly less time when measured by total time or time used outside of 
cue screens. These two interactions show that experienced auditors’ less effortful 
cue processing is only limited to when information is reliable. Whether experi-
enced auditors’ more efficient behavior under reliable information stems from 
strategic (e.g. intentional decisions to process information with less effortful way) 
or cognitive (e.g. more powerful processing ways) reasons (see also Joe 2003 for 
a discussion of these reasons) merits further investigation in future research. 
Overall, the results regarding more extensive information usage given less relia-
ble information are consistent with the findings of the psychological literature, 
stating that low credibility sources induce less persuasion than highly credible 
                                               
 
45 Additional comparisons between cues show some support for this expectation, as the univari-
ate tests (not tabulated) show that six out of the 12 cues are read for significantly longer when 
they are from a less reliable source. 
128      Acta Wasaensia 
sources (Pornpitakpan 2004) and that auditors are sensitive to source credibility in 
their decision-making (e.g. Goodwin 1999; Glover, Jiambalvo & Kennedy 2000). 
Fifth, a significant three-way interaction between information reliability, task 
structure and RMM indicates that used judgment time increases non-linearly as 
those factors with high values increase. In Figure 15, it is posited that the pres-
ence of one of those uncertainty factors triggers an auditor to increase attentive-
ness to information usage. For instance, this could happen because an auditor 
switches from a normal audit representation to a problem audit representation 
when there is an indication of a problem (Waller & Felix 1984; Asare & Knechel 
1995). However, adding a second uncertainty factor does not cause a linear in-
crease in judgment time. It is proposed that once an auditor has engaged in more 
careful information processing, adding a second uncertainty factor does not alter 
or alters very little his/her course of action regarding how the information will be 
processed. However, the simultaneous presence of three uncertainty factors (i.e. 
adding one more factor) significantly increases used judgment time compared 
with no factors. One explanation for this finding is that auditors do not increase 
linearly their efforts for information processing when the number of uncertainty 
factors increases, because they have few different information processing styles 
or modes for problem audits. Again, this finding merits further investigation in 
future research. 
 
 
 
 
 
 
 
 
Figure 15. Non-linear increase in judgment time depending on the number of 
uncertainty factors 
 
3 0 1 2 
Number of uncertainty factors 
Ju
dg
m
en
t t
im
e 
 
 Acta Wasaensia     129 
  
Finally, when the effect of auditor experience, RMM and information reliability 
on information usage and on the respective task-specific judgments are compared, 
the results indicate that the levels of those variables affect decision outcomes 
more compared with the antecedent decision-making process. Specifically, the 
results of this study suggest that auditors’ task-specific judgments are often ad-
justed depending on level of risk and information reliability. At the same time, 
preceding information usage is also affected, but less significantly. The several 
magnitudes of statistical significances in the analyses suggest that when RMM is 
high and information is less reliable, auditors’ task-specific judgments differ sig-
nificantly from less severe conditions, while their information usage behavior is 
also affected, but to a lesser degree compared with judgments. 
 
 
130      Acta Wasaensia 
7 CONCLUSIONS 
The final chapter begins with a summary of the study. The first section also dis-
cusses the practical implications of the results and draws conclusions about the 
study from its theoretical perspective. Possibilities for future research are also 
discussed, because the majority of the study’s findings are preliminary observa-
tions that call for additional evidence. The study concludes with the limitations of 
the study. The final section focuses on the internal and external validity issues of 
the present experiment, which should be taken into account when evaluating the 
generalizability of the results. 
7.1 Summary of the study and practical implications 
In many circumstances, an individual auditor conducts information usage alone 
without communicating his/her findings to other auditors (Hammersley 2006). 
Auditors usually work in environments where information load is very high, 
which may also reduce the likelihood of information sharing (Hammersley 2006). 
Consequently, it is important to study an individual auditor’s decision-making 
process, even when auditors work in multi-person environments. 
The present study examines auditors’ information usage in a single audit task. 
The purpose of this study was to relax the assumption that individual, environ-
mental, task-related and cue-related factors have only direct effects on infor-
mation usage. More specifically, four factors are investigated in a single study to 
examine whether these factors’ potential effects are conditional on the levels of 
other factors. 
The theoretical framework of this study consists of three parts. In the first part, 
auditors’ information acquisition and usage, its role in decision-making and 
common research approaches of the area are discussed. While there are no norma-
tive or unified theories of individuals’ information acquisition and usage process-
es, several models and theories (e.g. Waller & Felix 1984; Bonner & Pennington 
1991; Moroney 2007; Schultz, Bierstaker & O'Donnell 2010) suggest that the 
process sequentially progresses from the initial problem representation, which 
guides the extent and nature of actual information acquisition and usage. 
In the second part, previous research on the factors that affect auditors’ infor-
mation acquisition and usage are presented. Based on the auditors’ information 
acquisition and usage taxonomy proposed by El-Masry and Hansen (2008), pre-
vious studies of the determinants of information acquisition and usage are classi-
fied into four categories (individual factors, environmental factors, task-related 
 Acta Wasaensia     131 
  
factors and cue-related factors). This study contributes theoretically to the original 
taxonomy of El-Masry and Hansen (2008) by expanding it using interaction ef-
fects between these four factors. 
The expanded taxonomy suggests that a large number of factors from these cate-
gories interact with each other. Studies have produced a considerable amount of 
evidence that auditor experience and the presence of accountability mitigates or 
even reduces many judgmental biases related to auditors’ information acquisition 
and usage processes. For instance, evidence suggests that the negative effects of 
irrelevant information (Shelton 1999) and the recency effect (Kennedy 1993) on 
task-specific judgments are eliminated when a task is performed by an experi-
enced auditor or accountability is present, respectively. Overall, the expanded 
taxonomy suggests that in order to understand auditors’ information acquisition 
and usage behavior, the effects of factorial interactions should be taken into ac-
count. 
The final part of the theoretical framework begins with a presentation of the fac-
tors selected for the empirical analyses. The rationale behind this selection is 
based on capturing the factors that have gathered less attention at an audit task 
level and/or those that have important and previously identified factors in the au-
diting context that have meaningful theoretical interactions with the first-
mentioned factors. Finally, the hypotheses of the study are developed based on 
previous accounting and psychology studies. 
For the empirical part of the study, data are obtained through a computerized 
web-experiment that uses advanced process tracing methods to capture subjects’ 
information usage in multiple ways. Subjects’ information usage is measured by 
six dependent variables. It is posited that a rich set of variables captures the ex-
tensive set of effects that independent variables may have on subjects’ infor-
mation usage. These variables encompass used time for the task and the number 
of used information as well as refined measures of these variables. However, all 
empirical analyses are carried out using one dependent variable at a time in the 
models. 
The independent variables (i.e. factors) of the study, namely RMM, task structure 
and information reliability, are manipulated in a web-experiment between sub-
jects, while subjects’ task-specific experience is collected after the experiment. 
Task structure is varied between subjects by creating two similar audit tasks. Both 
tasks have 12 almost identical information cues available for decision-making. 
These cues are presented in an information menu, from where subjects choose 
information in an unconstrained way for their client continuance/acceptance 
judgments. The manipulations of RMM (low or high) and information reliability 
132      Acta Wasaensia 
(less reliable or more reliable) are carried out with the phrase manipulations with-
in task texts.  
The subject population consists of Finnish CPAs, non-certified auditors and Mas-
ter’s level auditing students. Using the data derived from 271 observations from 
the experimental tasks, the hypotheses of the study are tested using ANOVA and 
ordered logistic regression models. Table 20 presents the hypotheses tested and 
summarizes the results of these tests. 
 
Table 20. Summarized results of the hypotheses testing 
Hypotheses of the study and the results of the tests 
Hypothesis Results of the tests 
H1: Less experienced auditors use available 
information in decision-making more exten-
sively than experienced auditors 
Hypothesis is semi-strongly supported by the data. Expe-
rienced auditors use less total for tasks, time for reading 
cues and time for processing cues than less experienced 
subjects. 
H2: Information is used more extensively in 
decision-making when RMM is high, com-
pared when RMM is low  
Hypothesis is supported partly by the data. The results 
show that high RMM increases the total time used for 
tasks and the time used for processing cues. 
H3: Information is used more extensively in 
decision-making when a task is less struc-
tured  compared  with  when  a  task  is  more  
structured 
Hypothesis is supported weakly by the data. The results 
show that in unstructured tasks more time is used for 
processing cues than in semi-structured tasks. 
H4: Information is used more extensively in 
decision-making when information is less 
reliable compared with when it is more 
reliable 
Hypothesis is weakly supported by the data. The only 
statistically significant main effect indicates that when 
information is less reliable additional time is used to 
process cues outside of cue screens. 
H5:  In  the  context  of  high  RMM,  a  less  
experienced auditor uses information more 
extensively than an experienced one 
Hypothesis is not supported by the data.  
H6:  The  level  of  task  structure  does  not  
affect the information usage of experienced 
auditors 
Hypothesis is supported partly by the data. Two signifi-
cant interactions show that experienced auditors do not 
use more total time or cue processing time in unstructured 
tasks than in semi-structured tasks.  
H7: In the context of less reliable infor-
mation, a less experienced auditor uses 
information more extensively than an expe-
rienced one 
Hypothesis is not supported by the data. The results find 
the opposite to be true, namely that there exist experience 
effects when information is more reliable. Specifically, 
the results show support that less experienced subjects do 
not use more time to process cues than experienced audi-
tors when information is less reliable. 
 
The results of the hypothesis testing lead to five main findings. First, this study 
finds that experienced auditors use less overall time for tasks and spend signifi-
 Acta Wasaensia     133 
  
cantly less time on and outside of cue screens than less experienced subjects. 
These results suggest that experienced auditors read cues more quickly and pro-
cess cues more efficiently in their JDM processes than less experienced subjects. 
Second,  this study finds that when RMM is high auditors use more time to per-
form tasks, and particularly spend more time outside of cue screens, than when 
RMM is low. This finding suggests that auditors evaluate and combine cues in 
their JDM processes in a more effortful manner when RMM is high. 
Third, this study finds that task structure affects used judgment time. Specifically, 
I find that in unstructured tasks more time is used outside of cue screens. Howev-
er, the two significant interactions between structure and experience suggest that 
only less experienced subjects are affected by task structure and only they spend 
more time on unstructured tasks than on semi-structured tasks. 
Fourth, this study finds that when information is less reliable auditors use more 
time to perform tasks, and particularly spend more time outside of cue screens, 
than when information is more reliable. This finding suggests that auditors evalu-
ate and combine cues in their JDM processes in a more effortful manner when 
information is less reliable. This study also finds evidence of an interaction effect 
between auditor experience and information reliability. Contrary to expectations, 
experienced auditors’ shorter total time and time usage outside of cue screens 
diminish when information is less reliable. Thus, these interactions suggest that 
experienced auditors’ more efficient cue processing exist only when information 
is reliable. 
Fifth, this study finds a significant three-way interaction between RMM, task 
structure and information reliability. This interaction indicates that an increase in 
the time used outside of cue screens increases non-linearly when the number of 
factors with a high level of uncertainty increases. I find that the presence of one or 
two of these factors increases by about the same amount of time compared with 
when none of these factors is present. However, the presence of all three uncer-
tainty factors increases used time significantly more. This finding suggests that 
auditors do not increase linearly their efforts for information processing when the 
number of uncertainty factors increases. It is thus suggested that auditors may 
have few different information processing styles or modes for problem audits. 
The other major findings outside of the hypotheses include that risk, information 
reliability and auditor experience affect auditors’ task-specific judgments. More 
specifically, this study finds evidence that auditors adapt to high RMM and less 
reliable information by making more conservative probability estimates in their 
client continuance and acceptance judgments, especially when both these factors 
134      Acta Wasaensia 
are present at the same time. They also increase significantly auditors’ estimates 
of planned audit hours and fees for these engagements. 
The theoretical framework of this study suggests that individual, environmental, 
task-related and cue-related factors affect auditors’ information usage. From the 
perspective of this framework, the empirical results of this study suggest that an 
individual factor (auditor experience) is the most influential of the investigated 
factors, while an environmental factor (RMM), a task-related factor (task struc-
ture) and a cue-related factor (information reliability) all affect auditors’ infor-
mation usage, but their observed effect is smaller. Finding an individual factor 
(auditor experience) as the most significant factor is not surprising, as its effect on 
information usage has been widely documented in the psychological and audit 
literature (see e.g. Choo 1989). When the results are examined in light of previous 
interaction studies, the present study documents two new interactions with task 
structure and information reliability that evidence experience’s positive effects on 
auditors’ decision-making, especially on audit efficiency. 
The results of the present study also have implications for practitioners, especially 
those responsible for forming audit teams or arranging auditor training46. When 
creating time budgets that include lots of lengthy information, it should be noted 
that less experienced auditors are less efficient at reading and processing infor-
mation compared with experienced auditors. Complementary, audit firms should 
consider fine-tuning their training programs and documenting experienced audi-
tors’ reading and information processing styles in order to teach those methods to 
novice auditors. 
The time-consuming effect of high RMM and less reliable information should 
also  be  taken  into  account  when  planning  an  audit  program.  The  results  of  this  
study imply that even using highly experienced auditors in these engagements 
does not mitigate this effect. In particular, when information is suspected to come 
from a less reliable source, audit firms should allocate more time in a budget than 
is usual in order to meet audit deadlines. Audit firms should also take into account 
that in complex audit tasks less experienced auditors are likely to spend signifi-
cantly more time than experienced auditors. 
From the perspective of the interest groups of audit services, the results support 
the conclusion that auditors are relatively willing to continue with existing clients 
                                               
 
46 Audit firms should recognize that “extended” information usage may mean increased audit 
costs, which should be covered by the audit fee. This is particularly true when the need for 
greater information usage concerns experienced auditors. 
 Acta Wasaensia     135 
  
or accept new clients even when engagements are risky or contain less reliable 
information, but they are likely to change their audit conduct in those engage-
ments. Thus, the evidence suggests that in the current legal environment auditors 
are less likely to reject unattractive clients, but instead will increase their planned 
audit hours and charge higher audit fees. 
7.2 Suggestions for future research 
While the findings of this study are mostly preliminary evidence, which should be 
examined with different subjects and in other environments, the results do offer 
many insights that could be investigated in future research. First, the results re-
garding experienced auditors’ advanced reading techniques should be explored in 
more detail. For instance, an experimental study concentrating solely on reading 
cues would more effectively disentangle reading styles from actual cue pro-
cessing. A future study could employ eye-tracking techniques (see Rayner 1998 
for a review) to examine whether the reading styles of less experienced and expe-
rienced auditors differ fundamentally. 
Second, how less reliable information affects information usage should be ex-
panded to other audit tasks to investigate whether the effect is only task-specific. 
A future study could also manipulate the direction of cues to positive and nega-
tive forms to investigate whether this would amplify the processing effort of less 
reliable cues. For instance, auditors’ professional skepticism might mean that less 
reliable but positive information would be processed more carefully than negative 
information. A future study could also mix more and less reliable information 
simultaneously for selection, which would improve the overall research design as 
these groups would act their natural controls in an experiment. 
Third, some of this study’s statistically non-significant results may be because not 
all subjects used their “best” efforts for the experiment. Despite attempts to re-
move outliers, a controlled experiment might be more appropriate for examining 
the weak effects of many factors on information usage in future research. Alterna-
tively, the employment of verbal protocol analysis could extend beyond measur-
ing how plain effort duration (i.e. time) and effort intensity (see Cloyd 1997) 
change along with the current manipulations of factors. The verbalized thoughts 
of auditors may shed light on the underlying reasons for the observed behavior in 
this study. 
Finally, the theoretical framework of this study suggests several unexplored and 
meaningful interactions between factors from the four categories. The current 
136      Acta Wasaensia 
results with refined information usage measures should encourage further re-
search on this important topic. Specifically, investigating information usage in the 
most complex audit tasks (e.g. going-concern tasks) could provide insights into 
the role of cue reading/assimilating compared with actual cue processing. 
7.3 Limitations 
As in all experimental studies, there is always the concern of internal and external 
validity. Internal validity refers to the experimental design quality and specifically 
to whether observed changes in the dependent variables originated by differences 
in the independent variables (Peecher & Solomon 2001). Cook and Campbell 
(1979) identified several internal validity threats that might be present in this 
study. First, there is the concern of subject mortality, which means that systematic 
factors caused certain types of subjects to drop out during the experiment. For 
example, certain types of subjects may have experienced excessive fatigue or 
boredom in the middle of the experiment. The 25 incomplete answers partially 
support this assumption47. However, as the data were recorded in the database in 
multiple tables as the experiment progressed, this allowed us to examine the issue 
to some degree. Comparing the first and last recorded tables showed that only 
seven subjects dropped out during the experiment. 
Second, there is a concern over the self-selection of certain types of subjects as all 
participants were volunteers. As the introductory letter explained the purpose of 
the task (either client continuance or client acceptance), it is possible that auditors 
who had no interest in these tasks decided not to participate in the experiment. 
Further, busy auditors may have been omitted because of time constraints. This 
raises  the  question  of  whether  it  is  possible  to  generalize  the  results  beyond the  
sample. 
It can be argued that volunteered subjects were motivated to participate for sever-
al reasons. One reason could be their earlier first-hand experience of the task, i.e. 
task attraction. This reason is supported by the data, i.e. the average number of 
task-specific encounters of subjects is more than 20 times. However, this finding 
does not threaten the results, as task-specific experience was measured as an in-
                                               
 
47 However, this seemingly high number of incomplete answers most likely stems from incom-
patibility or technical problems between web browsers and experimental instruments, as in-
ternal controls ought to ensure the completeness of the most critical screens. This circumven-
tion  of  controls  could  also  happen if  the  subject  used  the  browser’s  back button  (they  were  
kindly requested not to at the beginning of the experiment). 
 Acta Wasaensia     137 
  
dependent variable in the study. Moreover, some subjects had no or little task-
specific experience. The second suggested reason for participation was the chance 
to win a prize in a raffle. This explanation is also supported semi-strongly by the 
data,  as the overall  response rate for the raffle was 56.6%. Although this option 
presumably increased the overall response rate, it may have worked as an extrin-
sic motivation for some subjects to participate without the true intention to use 
their “professional efforts” in the task (O’Neil & Penrod 2001; Kersten, Wu & 
Oertel 2011). However, there were attempts to control this behavior (see Chapter 
5.8). Finally, the high mean age of subjects (48.8 years) may imply that the sam-
ple is biased, suggesting that more recently certified auditors are under-
represented in the present study. 
Third, there is a threat of the imitation of treatments, i.e. that all responses are not 
independent. Owing to the nature of a web-based experiment, it is impossible to 
control the validity of observations. Although I asked respondents to answer in-
dependently, it is possible that subjects collaborated or discussed the task (during 
or after) in a way that influenced their decision-making behavior. This possibility 
cannot be ruled out. 
External validity refers to the generalizability of the results beyond the experi-
mental setting (Peecher & Solomon 2001). In experimental studies, there is usual-
ly a concern over population and ecological validity, which means that the ob-
tained results cannot be generalized to other people, situations or environmental 
settings  (Smith  2003).  A  wide  range  of  observations  from  different  audit  firms  
and auditors improves the result’s generalizability outside the sample. The subject 
population comprised Finnish CPAs and received observations divided fairly be-
tween Big-4, non-Big-4 and no firm subjects. This suggests that the results are not 
driven by certain audit firms’ policies (e.g. manuals or checklists) related to client 
continuance or acceptance. Also, using students as subjects may have distorted 
the results, as they might not be as time constrained as auditors. However, addi-
tional analyses performed without students are generally consistent with those 
reported in the main analyses. 
By contrast,  while  it  would  be  tempting  to  generalize  the  results  to  apply  to  all  
other environmental settings, it is most likely that they hold only in countries 
where the risk of litigation is low. For example, previous studies in the US (Asare 
& Knechel 1995; Johnstone 2000) have found that auditors focus on finding nega-
tive information about their prospective clients and avoid all suspicious clients. 
Thus, this study’s high RMM/less reliable information treatments may have led to 
obvious (without any additional information usage) discontinuance or client rejec-
tion decisions in a high risk litigation environment. 
138      Acta Wasaensia 
Other external validity threats concern the experiment. The exclusion of infor-
mation that would be available in a mundane environment also limits the general-
izability of the results to the real world. It is not possible for a brief client descrip-
tion and 12 cues to capture the richness available on an actual audit client contin-
uance/acceptance judgment. One must be also cautious when drawing conclusions 
from time-based dependent measures, as they may be noisy because of interrup-
tions and unfocussed behavior during the experiment by some subjects (Cloyd 
1997). While these non-laboratory settings, i.e. carrying out an experiment behind 
own work desk, might even increase external validity, as the environment is semi-
natural, a distracting environment (e.g. interruptions at home, watching television 
at same time) might reduce the focus that normally would be devoted to the task. 
To diminish this threat, the request letter for participation requested the task be 
carried out in the customary way and unbrokenly. 
However, the concerns of the ecological validity threats of this study are relaxed 
by some of its findings. These findings also give support for the construct validity 
of the study. First, the experiment was perceived to be realistic (mean 7.25 on 0–
10 scale) by subjects. Second, the results show that the majority of predicted signs 
are consistent with the theory and hypotheses, while not all results are statistically 
significant. Third, concerns that the categorization of the manipulated variables 
(two levels) may lead to the oversimplification of reality and that those levels are 
either unrealistically low or high are doubtful since the results indicate that the 
mean difference in both perceived risk and information reliability between high 
and low groups is fairly small (see Chapter 6.1). Thus, it is unlikely that the re-
sults of the study are driven by extreme conditions set in the manipulation phases. 
Finally, the estimated magnitudes of the several coefficients and R-squares are 
small and barely statistically significant at conventional levels, indicating that the 
investigated factors’ role in explaining auditors’ information usage is small. Some 
results might also be task-specific. Thus, the conclusions of this study should be 
taken cautiously before additional research is conducted on different audit tasks. 
However, according to Trotman, Tan and Ang (2011) some of these concerns are 
relaxed “They [experiments] are less beneficial if one is trying to determine the 
amount of an effect. Conducting ‘horse-races’ is usually inappropriate as the 
results will often be determined by the level at which the variables are set". 
 
 Acta Wasaensia     139 
  
REFERENCES 
Abdel-khalik, A. R. & El-Sheshai, K. M. (1980). Information choice and utiliza-
tion in an experiment on default prediction. Journal of Accounting Research 18: 
2, 325–342. 
Abdolmohammadi, M. J. (1999). A comprehensive taxonomy of audit task struc-
ture, professional rank and decision aids for behavioral research. Behavioral Re-
search in Accounting 11, 51–92. 
Abdolmohammadi,  M. J.  & Wright,  A. (1987).  An examination of the effects of 
experience and task complexity on audit judgments. Accounting Review 62: 1, 1–
13. 
Alexander,  R.  M.  (2003).  The  effect  of  source  credibility  on  tax  professionals’  
judgments in consulting engagements. Journal of the American Taxation Associa-
tion 25: Supplement, 33–50. 
Anderson,  B.  &  Maletta,  M.  M.  (1999).  Primacy  effects  and  the  role  of  risk  in  
auditor belief-revision processes. Auditing: A Journal of Practice & Theory 18: 1, 
75–90. 
Anderson,  J.  C.  &  Mueller,  J.  M.  (2005).  The  effects  of  experience  and  data  
presentation format on an auditing judgment. Journal of Applied Business Re-
search 21: 1, 53–63. 
Anderson, J. C. & Reckers, P. M. J. (1992). An empirical investigation of the ef-
fects of presentation format and personality on auditor’s judgment in applying 
analytical procedures. Advances in Accounting 10, 19–43. 
Anderson, U., Koonce, L. & Marchant, G. (1994). The effects of source compe-
tence information and its timing on auditors’ performance of analytical proce-
dures. Auditing: A Journal of Practice & Theory 13: 1, 137–148. 
Andersson, P. (2004). Does experience matter in lending? A process-tracing study 
on experienced loan officers' and novices' decision behavior. Journal of Economic 
Psychology 25: 4, 471–492. 
Arnold, V., Collier, P. A., Leech, S. A. & Sutton, S. G. (2000). The effect of ex-
perience and complexity on order and recency bias in decision making by profes-
sional accountants. Accounting & Finance 40: 2, 109–134. 
Asare, S. K. (1992). The auditor’s going-concern decision: Interaction of task 
variables and the sequential processing of evidence. Accounting Review 67: 2, 
379–393. 
140      Acta Wasaensia 
Asare,  S.  K.,  Cohen,  J.  R.  & Trompeter,  G.  M.  (2005).  The  effect  of  non-audit  
services on client risk, acceptance and staffing decisions. Journal of Accounting 
and Public Policy 24: 6, 489–520.  
Asare, S. K. & Knechel, W. R. (1995). Termination of information evaluation in 
auditing. Journal of Behavioral Decision Making 8: 1, 21–31. 
Asare, S. K., Trompeter, G. M. & Wright, A. (2000). The effect of accountability 
and time budgets on auditors’ testing strategies. Contemporary Accounting Re-
search 27: 4, 539–560. 
Ashton, A. H. & Ashton, R. H. (1988). Sequential belief revision in auditing. Ac-
counting Review 63: 4, 623–641. 
Ayers,  S.  &  Kaplan,  S.  E.  (1993).  An  examination  of  the  effect  of  hypothesis  
framing on auditors' information choices in an analytical procedure task. Abacus 
29: 2, 113–130. 
Bailey, C. D., Daily, C. M. & Philips, T. J. (2006). A study of Kruglanski's need 
for closure construct and its implications for judgment and decision making in 
accounting and auditing. Working Paper. 
Bamber, M. E. (1983). Expert judgment in the audit team: A source reliability 
approach. Journal of Accounting Research 21: 2, 396–412. 
Bamber, M. E., Ramsay, R. J. & Tubbs, R. M. (1997). An examination of the de-
scriptive validity of the belief-adjustment model and alternative attitudes to evi-
dence in auditing. Accounting, Organizations and Society 22: 3/4, 249–268. 
Beaulieu, P. R. (2001). The effects of judgments of new clients' integrity upon 
risk judgments, audit evidence, and fees. Auditing: A Journal of Practice & Theo-
ry 20: 2, 85–99. 
Bédard, J. & Mock, T. J. (1992). Expert and novice problem-solving behavior in 
audit planning. Auditing: A Journal of Practice & Theory 11: Supplement, 1–20. 
Bedard, J. C., Mock, T. J. & Wright, A. (1999). Evidential planning in auditing: A 
review of the empirical research. Journal of Accounting Literature 18, 96–142. 
Behn, B. K. & Riley, R. A. (1999). Using nonfinancial information to predict fi-
nancial performance: The case of the U.S. airline industry. Journal of Accounting, 
Auditing & Finance 14: 1, 29–56.  
Bell, T. B., Bedard, J. C., Johnstone, K. M. & Smith, E. F. (2002). KRisk: A 
computerized decision-aid for client acceptance and continuance risk assess-
ments. Auditing: A Journal of Practice & Theory 21: 2, 97–113. 
 Acta Wasaensia     141 
  
Bell, T. B., Peecher, M. E. & Solomon, I. (2005). The 21st century public compa-
ny audit: Conceptual elements of KPMG’s global audit methodology. KPMG: 
LLP. 
Bierstaker,  J.  L.  &  Brody,  R.  G.  (2001).  Presentation  format,  relevant  domain  
experience and task performance. Managerial Auditing Journal 16: 3, 124–128. 
Biggs, S. F. & Mock, T. J. (1983). An investigation of auditor decision processes 
in the evaluation of internal controls and audit scope decisions. Journal of Ac-
counting Research 21: 1, 234–255. 
Biggs, S. F., Mock, T. J. & Watkins, P. R. (1988). Auditor’s use of analytical re-
view in audit program design. Accounting Review 63: 1, 148–161. 
Blay, A. D. (2005). Independence threats, litigation risk, and the auditor's deci-
sion process. Contemporary Accounting Research 22: 4, 759–789. 
Blay, A. D., Kadous, K. & Sawers, K. (2012). The impact of risk and affect on 
information search efficiency. Organizational Behavior and Human Decision 
Processes 117: 1, 80–87. 
Blay, A. D., Sneathen, L. D. & Kizirian, T. (2007). The effects of fraud and go-
ing-concern risk on auditors’ assessments of the risk of material misstatement and 
resulting audit procedures. International Journal of Auditing 11: 3, 149–163. 
Bonner, S. E. (1990). Experience effects in auditing: the role of task-specific 
knowledge. Accounting Review 65: 1, 72–92. 
Bonner,  S.  E. (1994).  A model of the effects of audit  task complexity.  Account-
ing, Organizations and Society 19: 3, 213–234. 
Bonner, S. E. (2008). Judgment and Decision Making in Accounting. Prentice 
Hall. New Jersey. 
Bonner, S. E. & Pennington, N. (1991). Cognitive processes and knowledge as 
determinants of auditor expertise. Journal of Accounting Literature 10, 1–50. 
Boritz, J. E. (1992). DISCUSSION OF expert and novice problem-solving behav-
ior in audit planning. Auditing: A Journal of Practice & Theory 11: Supplement, 
21–32. 
Braun, R. L. (2000). The effect of time pressure on auditor attention to qualitative 
aspects of misstatements indicative of potential fraudulent financial reporting. 
Accounting, Organizations and Society 25: 3, 243–259. 
Brazel, J. F., Jones, K. & Zimbelman, M. F. (2009). Using nonfinancial measures 
to assess fraud risk. Journal of Accounting Research 47: 5, 1135–1166. 
142      Acta Wasaensia 
Brown, C. E., Peecher, M. & Solomon, I. (1999). Auditors’ hypothesis testing in 
diagnostic inference tasks. Journal of Accounting Research 14: 2, 1–26. 
Bryant, S. M., Hunton, J. E. & Stone, D. N. (2004). Internet-based experiments: 
Prospects and possibilities for behavioral accounting research. Behavioral Re-
search in Accounting 16, 107–129. 
Butt, J. L. (1988). Frequency judgments in an auditing-related task. Journal of 
Accounting Research 26: 2, 315–330. 
Butt,  J.  L.  &  Campbell,  T.  L.  (1989).  The  effects  of  information  order  and  hy-
pothesis-testing strategies or auditor's judgments. Accounting, Organizations and 
Society 14: 5/6, 471–479. 
Byström, K. (2002). Information and information sources in tasks of varying 
complexity. Journal of the American Society for Information Science and Tech-
nology 53: 7, 581–591. 
Choo, F. (1989). Expert-novice differences in judgment/decision making re-
search. Journal of Accounting Literature 8, 106–136. 
Choo, F. (1995). Auditors’ judgment performance under stress: A test of the pre-
dicted relationship by three theoretical models. Journal of Accounting, Auditing & 
Finance 10: 3, 611–641. 
Choo, F. & Trotman, K. T. (1991). The relationship between knowledge structure 
and judgments for experienced and inexperienced auditors. Accounting Review 
66: 3, 464–485. 
Christ, M. Y. (1993). Evidence on the nature of audit planning problem represen-
tations: An examination of auditor free recalls. Accounting Review 68: 2, 304–
322. 
Cianci, A. M. & Bierstaker, J. (2009). Auditors’ efficiency motivated evaluation. 
Advances in Accounting 25: 1, 20–27. 
Cloyd, B. C. (1997). Performance in tax research tasks: The joint effects of 
knowledge and accountability. Accounting Review 72: 1, 111–132. 
Cohen, J.  R. & Hanno, D. M. (2000).  Auditors’ consideration of corporate gov-
ernance and management control philosophy in preplanning and planning judg-
ments. Auditing: A Journal of Practice & Theory 19: 2, 133–146. 
Cohen, J. R., Krishnamoorthy, G. & Wright, A. (2000). Evidence on the effect of 
financial and nonfinancial trends on analytical review. Auditing: A Journal of 
Practice & Theory 19: 1, 27–48. 
Colbert, J. L. (1988). Inherent risk: an investigation of auditors’ judgments. Ac-
counting, Organizations and Society 13: 2, 111–121. 
 Acta Wasaensia     143 
  
Connolly,  T.  &  Thorn,  B.  K.  (1987).  Predecisional  information  acquisition:  Ef-
fects of task variables on suboptimal search strategies. Organizational Behavior 
and Human Decision Processes 39: 3, 397–416. 
Cook, T. D. & Campbell, D. T. (1979). Quasi-Experimentation: Design and 
Analysis Issues for Field Settings. Boston: Houghton-Mifflin. 
Cushing, B. E. & Ahlawat, S. S. (1996). Mitigation of recency bias in audit judg-
ment: The effect of documentation. Auditing: A Journal of Practice & Theory 15: 
2, 110–122. 
Davis, J. T. (1996). Experience and auditors' selection of relevant information for 
preliminary control risk assessments. Auditing: A Journal of Practice & Theory 
15: 1, 17–37. 
DeZoort, T., Harrison, P. & Taylor, M. (2006). Accountability and auditors mate-
riality judgments: The effects of differential pressure strength on conservatism, 
variability, and effort. Accounting, Organizations and Society 31: 4/5, 373–390. 
Dilla, W. N. & Stone, D. N. (1997). Representations as decision aids: The asym-
metric effects of words and numbers on auditors' inherent risk judgments. Deci-
sion Sciences 28: 3, 709–743. 
Dunn, K., Kohlbeck, M. & Mayhew, B. W. (2011). The impact of the Big 4 con-
solidation on audit market share equality. Auditing: A Journal of Practice & The-
ory 30: 1, 49–73. 
Earley, C. E. (2002). The differential use of information by experienced and nov-
ice auditors in the performance of ill-structured audit task. Contemporary Ac-
counting Research 19: 4, 595–614. 
Einhorn, H. J. & Hogarth, R. M. (1981). Behavioral decision theory: processes of 
judgment and choice. Journal of Accounting Research 19: 1, 1–31. 
El-Masry, E-H. E. & Hansen, K. A. (2008). Factors affecting auditors' utilization 
of evidential cues: Taxonomy and future research directions. Managerial Auditing 
Journal 23: 1, 26–50. 
Ethridge, J. R., Marsh, T. L. & Canfield, K. (2007). Engagement risk: A Prelimi-
nary analysis of audit firm’s client acceptance decisions. Academy of Accounting 
and Financial Studies Journal 11: 1, 1–8. 
Ethridge, J.  R.,  Marsh, T. L. & Revelt,  B. (2007).  Engagement risk: Perceptions 
and strategies from audit partners. Journal of Business & Economic Research 5: 
4, 25–32. 
Favere-Marchesi, M. (2006). “Order effects” revisited: The importance of Chro-
nology. Auditing: A Journal of Practice & Theory 25: 1, 69–83. 
144      Acta Wasaensia 
Flom, P. L. (2010). Multinomial and ordinal logistic regression using PROC LO-
GISTIC [cited 6.6.2012]. Available from World Wide Web: <URL: 
http://www.statisticalanalysisconsulting.com/wp-content/uploads/2010/10/ article 
NE SUG2010.pdf/>. 
Ford, K. J., Schmitt, N., Schechtman, S. L., Hults, B. M. & Doherty, M. L. 
(1989). Process tracing methods: contributions, problems, and neglected research 
questions. Organizational Behavior and Human Decision Processes 43: 1, 75–
117. 
Gaumnitz, B. R., Nunamaker, T. R., Surdick, J. J. & Thomas, M. F. (1982). Audi-
tor consensus in internal control evaluation and audit program planning. Journal 
of Accounting Research 20: 2, 745–755. 
Gendron, Y. (2001). The difficult client-acceptance decision in Canadian audit 
firms: A field investigation. Contemporary Accounting Research 18: 2, 283–310. 
Gibbins, M. (1984). Propositions about the psychology of professional judgment 
in public accounting. Journal of Accounting Research 22: 1, 103–125. 
Glover, S. M. (1997). The influence of time pressure and accountability on audi-
tors' processing of nondiagnostic information. Journal of Accounting Research 
35: 2, 213–227. 
Glover,  S.  M.,  Jiambalvo,  J.  &  Kennedy,  J.  (2000).  Analytical  procedures  and  
audit-planning decisions. Auditing: A Journal of Practice & Theory 19: 2, 27–45.  
Glover,  S.  M.,  Prawitt,  D. F.  & Spilker,  B. C. (1997).  The influence of decision 
aids on user behavior: Implications for knowledge acquisition and inappropriate 
reliance. Organizational Behavior and Human Decision Processes 72: 2, 232–
255. 
Glover, S. M., Prawitt, D. F., Wilks, T. J. & McDaniel, L. (2005). Why do audi-
tors over-rely on weak analytical procedures? The role of outcome and precision. 
Auditing: A Journal of Practice & Theory 24: Supplement, 197–233. 
Goodwin, J. (1999). The effects of source integrity and consistency of evidence 
on auditors' judgments. Auditing: A Journal of Practice & Theory 18: 2, 1–16. 
Goodwin, J. & Trotman, K. T. (1996). Factors affecting the audit of revalued non-
current assets - managerial initial public offerings and source reliability. Account-
ing and Finance 36: 2, 151–170. 
Guiral, A., Ruiz, E. & Rodgers, W. (2011). To what extent are auditors’ attitudes 
toward the evidence influenced by the self-fulfilling prophecy? Auditing: A Jour-
nal of Practice & Theory 30: 1, 173–190. 
Hackenbrack, K. (1992). Implications of seemingly irrelevant evidence in audit 
judgment. Journal of Accounting Research 30: 1, 126–136. 
 Acta Wasaensia     145 
  
Hackenbrack, K. & Knechel, W. R. (1997). Resource allocation decisions in audit 
engagements. Contemporary Accounting Research 14: 3, 481–499. 
Hammersley, J. S. (2006). Pattern identification and industry-specialist auditors. 
Accounting Review 81: 2, 309–336. 
Hirst, E. D. (1994). Auditors’ sensitivity to source reliability. Journal of Account-
ing Research 32: 1, 113–126. 
Hoffman, V. B., Joe, J. R. & Moser, D. V. (2003). The effect of constrained pro-
cessing on auditors’ judgments. Accounting, Organizations and Society 28: 7/8, 
699–714. 
Hoffman,  V.  B.  & Patton,  J.  M.  (1997).  Accountability,  the  dilution  effect,  and  
conservatism in auditors’ fraud judgments. Journal of Accounting Research 35: 2, 
227¬–237. 
Hogarth, R. M. & Einhorn, H. J. (1992). Order effects in belief updating: The 
belief- adjustment model. Cognitive Psychology 24: 1, 1–55. 
Houston, R. W.,  Peters,  M. F.  & Pratt,  J.  H. (1999).  The audit  risk model,  busi-
ness risk and audit-planning decisions. Accounting Review 74: 3, 281–298. 
Hurtt,  K.,  Eining, M. & Plumlee,  D. R. (2008).  An experimental  examination of 
professional skepticism. Working paper. 
Huss,  H.  F.  &  Jacobs,  F.  A.  (1991).  Risk  containment:  Exploring  auditor  deci-
sions in the engagement process. Auditing: A Journal of Practice & Theory 10: 2, 
16–32. 
IFAC (2009a). International standard on auditing 330 - the auditor’s responses to 
assessed risks [cited 12.2.2012]. Available from World Wide Web: <URL: 
http://www.ifac.org/>. 
IFAC (2009b). International Standard on Quality Control 1 [cited 02.10.2012]. 
Available from World Wide Web: <URL: http://www.ifac.org/>. 
IFAC (2010). Client acceptance and continuance [online]. International Federa-
tion of Accountants [cited 15.1.2011]. Available from World Wide Web: 
<URL:http://www.ifac.org/>.  
Jamal, K., Johnson, P. E. & Berryman, R. G. (1995). Detecting framing effects in 
financial statements. Contemporary Accounting Research 12: 1, 85–105. 
Joe, J. R. (2003). Why press coverage of a client influences the audit opinion. 
Journal of Accounting Research 41: 1, 109–133. 
146      Acta Wasaensia 
Johnstone, K. M. (2000). Client-acceptance decisions: simultaneous effects of 
client business risk, audit risk, auditor business risk, and risk adaptation. Audit-
ing: A Journal of Practice & Theory 19: 1, 1–25. 
Johnstone, K.M. (2001). Risk, experience, and client acceptance decisions. Na-
tional Public Accountant 46: 5, 27–30. 
Johnstone, K. M. & Bedard, J. C. (2001). Engagement planning, bid pricing, and 
client response in the market for initial attest engagements. Accounting Review 
76: 2, 199–220. 
Johnstone, K. M. & Bedard, J. C. (2003). Risk management in client acceptance 
decisions. Accounting Review 78: 4, 1003–1025. 
Joyce,  E. J.  & Biddle,  G. C. (1981).  Are auditors'  judgments sufficiently regres-
sive? Journal of Accounting Research 19: 1, 120–145. 
Kadous, K., Magro, A. M. & Spilker, B. C. (2008). Do effects of client preference 
on accounting professionals’ information search and subsequent judgments persist 
with high practice risk? Accounting Review 83: 1, 133–156. 
Kaplan, S.  E. (1985).  An examination of the effects of environment and explicit  
internal control evaluation on planned audit hours. Auditing: A Journal of Prac-
tice & Theory 5: 1, 12–25. 
Kaplan, S. E. (1988). An examination of the effect of presentation format on audi-
tors’ expected value judgments. Accounting Horizons 2: 3, 90–95. 
Kaplan, S. E. & Reckers, P. M. J. (1984). An empirical examination of auditors' 
initial planning processes. Auditing: A Journal of Practice & Theory 4: 1, 1–19. 
Kaplan, S. E. & Reckers, P. M. J. (1989). An examination of information search 
during initial audit planning. Accounting, Organizations and Society 14: 5/6, 539–
550. 
Kennedy, J. (1993). Debiasing audit judgment with accountability: A framework 
and experimental results. Journal of Accounting Research 31: 2, 231–245. 
Kerr, D. S. & Ward, D. D. (1994). The effects of audit task on evidence integra-
tion and belief revision. Behavioral Research in Accounting 6, 21–42. 
Kersten, G. E., Wu, S. & Oertel, C. (2011). Extrinsic or intrinsic motivation of e-
negotiation experiments’ participants. Proceedings from HICSS2011. Hawaii, 
USA. 
Kida, T. (1984). The impact of hypothesis-testing strategies on auditors' use of 
judgment data. Journal of Accounting Research 22: 1, 332–340. 
 Acta Wasaensia     147 
  
Kizirian, T., Mayhew, B. & Sneathen, L. D. (2005). The impact of management 
integrity on audit planning and evidence. Auditing: A Journal of Practice & The-
ory 24: 2, 49–67. 
Knechel, W. R. & Messier Jr., W. F. (1990). Sequential auditor decision making: 
Information search and evidence evaluation. Contemporary Accounting Research 
6: 2, 386–406. 
Koonce, L. (1993). A cognitive characterization of audit analytical review. Audit-
ing: A Journal of Practice & Theory 12: Supplement, 57–76. 
Krhisnamoorthy, G., Mock, T. J. & Washington, M. T. (1999). A comparative 
evaluation of belief revision models in auditing. Auditing: A Journal of Practice 
& Theory 28: 2, 105–127. 
Krogstad, J. L., Ettenson, R. T. & Shanteau, J. (1984). Context and experience in 
auditors' materiality judgments. Auditing: A Journal of Practice & Theory 4: 1, 
54–73. 
Krull, G. Jr., Reckers, P. M. J. & Wong-On-Wing, B. (1993). The effect of expe-
rience, fraudulent signals and information presentation order on auditors' beliefs. 
Auditing: A Journal of Practice & Theory 12: 2, 143–153. 
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin 108: 
3, 480–498. 
La Porta, R., Lopez-de-Silanes, F. & Shleifer, A. (1998). Law and finance. Jour-
nal of Political Economy 106: 6, 1113–1155. 
Lee, H., Herr, P. M., Kardes, F. R. & Kim, C. (1999). Motivated search: Effects 
of choice, accountability, issue involvement and prior knowledge on information 
acquisition and use. Journal of Business Research 45: 1, 75–88. 
Lehmann, C. M. & Norman, C. S. (2006). The effects of experience on complex 
problem representation and judgment in auditing: An experimental investigation. 
Behavioral Research in Accounting 18, 65–83. 
Libby, R. (1985). Availability and the generation of hypotheses in analytical re-
view. Journal of Accounting Research 23: 2, 648–667. 
Libby, R. (1995). The role of knowledge and memory in audit judgment. In Rob-
ert H. Ashton & Alison Hubbard Ashton (Eds). Judgment and Decision-Making 
Research in Accounting and Auditing. New York: Cambridge University Press. 
176–206. 
McDaniel, L. (1990). The effects of time pressure and audit program structure on 
audit performance. Journal of Accounting Research 28: 2, 267–285. 
148      Acta Wasaensia 
McMillan, J. J. & White, R. A. (1993). Auditors' belief revisions and evidence 
search: The effect of hypothesis frame, confirmation bias, and professional skep-
ticism. Accounting Review 68: 3, 443–465. 
Messier  Jr.,  W.  F.  (1992).  The  sequencing  of  audit  evidence:  Its  impact  on  the  
extent of audit testing and report formulation. Accounting and Business Research 
86: 1, 143–150. 
Messier Jr., W. F. (2003). Auditing & Assurance Services: A Systematic Ap-
proach. Third edition. New York: McGraw-Hill. 
Messier Jr., W. F. & Tubbs, R. M. (1994). Recency effects in belief revision: The 
impact of audit experience and the review process. Auditing: A Journal of Prac-
tice & Theory 13: 1, 57–73. 
Mock,  T.  J.  &  Wright,  A.  (1993).  An  exploratory  study  of  auditors’  evidential  
planning judgments. Auditing: A Journal of Practice & Theory 12: 2, 39–61. 
Mock, T. J.  & Wright,  A. (1999).  Are audit  program plans risk-adjusted? Audit-
ing: A Journal of Practice & Theory 18: 1, 55–74. 
Moeckel, C. (1991). Two factors affecting an auditor's ability to integrate audit 
evidence. Contemporary Accounting Research 8: 1, 270–292. 
Monroe, G. S.  & Ng, J.  (2000).  An examination of order effects in auditors’ in-
herent risk assessments. Accounting & Finance 40: 2, 153–168. 
Moroney, R. (2007). Does industry expertise improve the efficiency of audit 
judgment? Auditing: A Journal of Practice & Theory 26: 2, 69–94. 
Mueller, J. M. & Anderson, J. C. (2002). Decision aids for generating analytical 
review alternatives: The impact of goal framing and audit-risk level. Behavioral 
Research in Accounting 14, 157–176. 
Mutchler, J. F. (1986). Empirical evidence regarding the auditor’s going-concern 
opinion decision. Auditing: A Journal of Practice & Theory 6: 1, 148–163. 
Nearon, B. H. (1999). Survey reports CPA computer and internet use Available 
from World Wide Web: <URL:http://www.nysscpa.org/>. 
Nelson, M. W. (2009). A model and literature review of professional skepticism 
in auditing. Auditing: A Journal of Practice & Theory 28: 2, 41–71. 
Nelson, M. & Tan, H. (2005). Judgment and decision making research in audit-
ing: a task, person, and interpersonal interaction perspective. Auditing: A Journal 
of Practice & Theory 24: Supplement, 41–71. 
 Acta Wasaensia     149 
  
O'Keefe,  T.  B.,  Simunic,  D.  A.  & Stein,  M.  T.  (1994).  The  production  of  audit  
services: Evidence from a major public accounting firm. Journal of Accounting 
Research 32: 2, 241–261. 
O'Neil, K. M. & Penrod, S. D. (2001). Methodological variables in web-based 
research that may affect results: Sample type, monetary incentives, and personal 
information. Behavior Research Methods, Instruments & Computers 33: 2, 226–
233. 
Oppenheimer, D. M., Meyvis, T. & Davidenko, N. (2009). Instructional manipu-
lation checks: Detecting satisficing to increase statistical power. Journal of Ex-
perimental Social Psychology 45: 4, 867–872. 
Payne, J. W. (1982). Contingent decision behavior. Psychological Bulletin 92: 2, 
382–402. 
Peecher, M. E. (1996). The influence of auditors’ justification processes on their 
decisions: A cognitive model and experimental evidence. Journal of Accounting 
Research 34: 1, 125–140. 
Peecher,  M.  E.,  Schwartz,  R.  & Solomon,  I.  (2007).  It’s  all  about  audit  quality:  
Perspectives on strategic-systems auditing. Accounting, Organizations and Socie-
ty 32: 4/5, 463–485. 
Peecher, M. E. & Solomon, I. (2001). Theory and experimentation in studies of 
audit judgments and decisions: Avoiding common research traps. International 
Journal of Auditing 5: 3, 193–203. 
Popova, V. (2008). Integration of fraud risk in the risk of material misstatement 
and the effect on auditors’ planning decisions. Working paper. 
Pornpitakpan, C. (2004). The persuasiveness of source credibility: A critical re-
view of five decades' evidence. Journal of Applied Social Psychology 34: 2, 243–
281. 
Pratt, J. & Stice, J. D. (1994). The effects of client characteristics on auditor liti-
gation risk judgments, required audit evidence, and recommended audit fees. Ac-
counting Review 69: 4, 639–674. 
Quadackers, L., Groot, T. & Wright, A. (2009). Auditors’ skeptical characteristics 
and their relationship to skeptical judgments and decisions. Working paper. 
Rayner, K. (1998). Eye movements in reading and information processing: 20 
years of research. Psychological Bulletin 124: 3, 372–422. 
Rebele, J. E., Heintz, J. A. & Briden, G. E. (1988). Independent auditor sensitivity 
to evidence reliability. Auditing: A Journal of Practice & Theory 8: 1, 43–52. 
150      Acta Wasaensia 
Reimers, J. L. & Fennema, M. G. (1999). The audit review process and sensitivity 
to information source objectivity. Auditing: A Journal of Practice & Theory 18: 1, 
117–123. 
Ricchiute, D. N. (1984). An empirical assessment of the impact of alternative 
presentation modes on decision-making research in auditing. Journal of Account-
ing Research 22: 1, 341–350. 
Ricchiute, D. N. (2010). Evidence complexity and information search in the deci-
sion to restate prior-period financial statements. Journal of Accounting Research 
48: 3, 687–724. 
Robertson, J. C. (2010). The effects of ingratiation and client incentive on auditor 
judgment. Behavioral Research in Accounting 22: 2, 69–86. 
Rosman, A. J. & O’Neill, H. (1993). Comparing the information acquisition strat-
egies of venture capital and commercial lenders: A computer-based experiment. 
Journal of Business Venturing 8: 5, 443–460. 
Rosman,  A.  J.,  Seol,  I.  &  Biggs,  S.  F.  (1999).  The  effect  of  stage  development  
and financial health on auditor decision behavior in the going-concern task. Audit-
ing: A Journal of Practice & Theory 18: 1, 37–54. 
Sacchi, S. & Burigo, M. (2008). Strategies in the information search process: The 
interaction between task structure, knowledge and source. Journal of General 
Psychology 135: 3, 252–270. 
Schultz, J., Bierstaker, J. L. & O'Donnell, E. (2010). Integrating business risk into 
auditor judgment about the risk of material misstatement: The influence of a stra-
tegic-systems-audit approach. Accounting, Organizations and Society 35: 2, 238–
251. 
Schulz, A. K-D. & Booth, P. (1995). The effects of presentation format on the 
effectiveness and efficiency of auditors' analytical review judgments. Accounting 
& Finance 35: 1, 107–131. 
Sharma,  D.  S.,  Boo,  E.  & Sharma,  V.  D.  (2008).  The  impact  of  non-mandatory  
corporate governance on auditors’ client acceptance, risk and planning judgments. 
Accounting and Business Research 38: 2, 1–28. 
Shaub, M. K. & Lawrence, J. E. (1996). Ethics, experience and professional skep-
ticism: A situational analysis. Behavioral Research in Accounting 8: Supplement, 
124–157. 
Shelton, S. W. (1999). The effect of experience on the use of irrelevant evidence 
in auditor judgment. Accounting Review 74: 2, 217–224. 
Shields, M. D. (1983). Effects of information supply and demand on judgment 
accuracy: evidence from corporate managers. Accounting Review 58: 2, 284–203. 
 Acta Wasaensia     151 
  
Simnett, R. (1996). The effect of information selection, information processing 
and task complexity on predictive accuracy of auditors. Accounting, Organiza-
tions and Society 21: 7/8, 699–719. 
Simnett, R. & Trotman, K. T. (1989). Auditor versus model: Information choice 
and information processing. Accounting Review 64: 3, 514–528. 
Smith, J. F. & Kida, T. (1991). Heuristics and biases: Expertise and task realism 
in auditing. Psychological Bulletin 109: 3, 472–489. 
Smith, M. (2003). Research Methods in Accounting. London: Sage Publications 
Ltd. 
Spilker, B. C. (1995). The effects of time pressure and knowledge on key word 
selection behavior in tax research. Accounting Review 70: 1, 49–70. 
Spilker,  B. C. & Prawitt,  D. F.  (1997).  Adaptive responses to time pressure:  the 
effects of experience on tax information search behavior. Behavioral Research in 
Accounting 9: 1, 172–198. 
Tan, H. (2001). Methodological issues in measuring knowledge effects. Interna-
tional Journal of Auditing 5: 3, 215–224. 
Tan, H. & Kao, A. (1999). Accountability effects on auditors’ performance: The 
influence of knowledge, problem-solving ability, and task complexity. Journal of 
Accounting Research 37: 1, 209–223. 
Tan,  H.,  Ng,  T.  &  Mak,  B.  (2002).  The  effects  of  task  complexity  on  auditors’  
performance: The impact of accountability and knowledge. Auditing: A Journal 
of Practice & Theory 21: 2, 81–95. 
Tetlock, P. E. (1983). Accountability and complexity of thought. Journal of Per-
sonality and Social Psychology 45: 1, 74–83. 
Thayer, J. (2011). Determinants of investors' information acquisition: Credibility 
and confirmation. Accounting Review 86: 1, 1–22. 
Trotman, K. T. (2005). DISCUSSION OF judgment and decision making re-
search in auditing: a task, person, and interpersonal interaction perspective. Audit-
ing: A Journal of Practice & Theory 24: Supplement, 73–88. 
Trotman, K. T., Tan, H. C. & Ang, N. (2011). Fifty-year overview of judgment 
and decision-making research in accounting. Accounting & Finance 51: 1, 278–
360. 
Trotman, K. T. & Sng, J. (1989). The effect of hypothesis framing, prior expecta-
tions and cue diagnosticity on auditors' information choice. Accounting, Organi-
zations and Society 14: 5/6, 565–576. 
152      Acta Wasaensia 
Trotman, K. T. & Wright, A. (1996). Recency effects: Task complexity, decision 
mode and task specific experience. Behavioral Research in Accounting 8: 1, 175–
193. 
Turner, C. W. (2001). Accountability demands and the auditor's evidence search 
strategy: The influence of reviewer preferences and the nature of the response 
(Belief vs. Action). Journal of Accounting Research 39: 3, 683–706. 
Van Swol, L. M. & Sniezek, J. A. (2005). Factors affecting the acceptance of ex-
pert advice. British Journal of Social Psychology 44: 3, 443–461. 
Waller, W. S. & Felix, Jr. W. L. (1984). Cognition and the auditor's opinion for-
mulation process: A schematic model of interactions between memory and cur-
rent audit evidence. In S. Moriarity & E. J. Norman (Eds). Decision Making and 
Accounting: Current Research. Norman: University of Oklahoma Press. 
White, A. & Harding, N. (2011). Identifying auditor stopping rules in decision 
making under uncertainty. Working Paper. 
Wilks, J. T. (2002). Predecisional distortion of evidence as a consequence of real-
time audit review. Accounting Review 77: 1, 51–71. 
Wittek, R., van der Zee, T. & Muhlau, P. (2008). Client acceptance decisions of 
Dutch auditing partners. Journal of Management and Governance 12: 1, 93–115. 
Wood, R. E. (1986). Task complexity: Definition of the construct. Organizational 
Behavior and Human Decision Processes 37: 1, 60–82. 
Wright, W. F. (1995). Superior loan collectability judgments given graphical dis-
plays. Auditing: A Journal of Practice & Theory 14: 2, 144–154. 
Zimbelman,  M.  F.  (1997).  The  effects  of  SAS  No.  82  on  auditors'  attention  to  
fraud risk factors and audit planning decisions. Journal of Accounting Research 
35: Supplement, 75–97. 
 Acta Wasaensia     153 
  
APPENDICES  
Appendix 1. RMM and information reliability manipulations 
Manipulation phrases in client continuance -task: 
Risk of material misstatement manipulation 
1) Low risk: 
The main owners of Boat King Oy have retired from the executive management 
of the company and in recent years, have only participated in the Board of Direc-
tors. The main owners are satisfied with the top management of the company. On 
the basis of first impression and discussion with the management, you consider 
the Chief Executive Officer and Chief Financial Officer to be qualified and relia-
ble. In your discussions, it has become apparent that they emphasize ethical val-
ues and responsibility in the course of business. 
2) High risk 
The main owners of Boat King Oy have retired from the executive management 
of the company and in recent years, have only participated in the Board of Direc-
tors. Now they are concerned for the long-term profitability of the company. At 
the meeting of the Board of Directors earlier that year, the main owners have de-
cided to modify the compensation structure of the management so that in the fu-
ture, a significant part of the Chief Executive Officer’s and the Chief Financial 
Officer’s compensation is determined by the previously reported earnings. Your 
discussions have not entirely convinced you of the management’s integrity and 
motivation to work under new terms. 
Information reliability manipulation 
1) Low reliability: 
The incumbent auditor of the previous assignment has repeatedly been criticized 
for neglecting the duty of care in your audit firm’s internal quality inspections. 
The auditor has been accused of superficial auditing and poor documentation and 
is therefore no longer the incumbent auditor of your audit firm. 
2) High reliability 
The incumbent auditor is one of the most respected auditors in your audit firm 
and is known as a precise and conscientious auditor who takes his or her job seri-
ously.  The former incumbent auditor has withdrawn from the assignment due to 
retirement. 
 
154      Acta Wasaensia 
Manipulation phrases in client acceptance -task: 
Risk of material misstatement manipulation 
1) Low risk 
The main owners of Boat King Oy have retired from the executive management 
of the company and in recent years, have only participated in the Board of Direc-
tors. The main owners are satisfied with the top management of the company. On 
the basis of first impression, you consider the Chief Executive Officer and Chief 
Financial Officer to be qualified and reliable. In your discussions, it has become 
apparent that they emphasize ethical values and responsibility in the course of 
business. 
2) High risk 
The main owners of Boat King Oy have retired from the executive management 
of the company and in recent years, have only participated in the Board of Direc-
tors. Now they are concerned for the long-term profitability of the company. 
Therefore, at the meeting of the Board of Directors earlier that year, they have 
decided to modify the compensation structure of the management so that in the 
future, a significant part of the Chief Executive Officer’s and the Chief Financial 
Officer’s compensation is determined by the previously reported earnings. On the 
basis of first impression, you are not entirely convinced of the management’s in-
tegrity and motivation to work under new terms. 
Information reliability manipulation 
1) Low reliability 
Collecting information on Boat King Oy has been problematic because your as-
sistant, who was responsible for the collecting, has afterwards proven to be care-
less and incompetent in the field of accounting. In addition, according to your 
information, the former incumbent auditor of Boat King Oy has repeatedly been 
criticized in the internal quality inspections and is therefore no longer the incum-
bent auditor of the audit firm. 
2) High reliability 
Audit assistant at your audit firm has collected the information on Boat King Oy. 
The assistant has been working for your audit firm over six months and is regard-
ed by several auditors as an accurate and thorough worker who is able to perform 
the assigned tasks impeccably. In addition, the former auditor of Boat King Oy 
has a good reputation and is considered a conscientious auditor. The former in-
cumbent auditor has withdrawn from the assignment due to retirement. 
  
 Acta Wasaensia     155 
  
Appendix 2. Variables of the experiment 
 
 Definitions of variables      
 Definition      
1CL[CUENAME]a Self-evaluation of cue importance [11-point Likert-type scale: 
0 (not important at all) to 10 (very important)] 
1CL[CUENAME]_1a A dummy variable with a value of 1 if the cue has been read at 
least once, otherwise 0. 
1CL[CUENAME]_ORDERa Order of cues in read order [0–12] 
1CL[CUENAME]_ORDER2a Order of cues in read order, second time [0–...] 
1CL[CUENAME]_CUMULa Cumulative time spent reading cues [sec] 
1CL[CUENAME]_NUMBERa Number of cue reads [0– …] 
1CL_EXP Task-specific experience on client continuance decisions [0; 
no experience, 1; 1–9 times, 2; 10–19 times, 3; 20–29 times, 
4;30+ times] 
1CLBACK_NUMBER Number of background information reads [0– …] 
1CLBACKGROUND Self-evaluation of background information importance [11-
point Likert-type scale: 0 (not important at all) to 10 (very 
important)] 
AGE Age of subject 
AUDIT_FIRM2 Name of audit firm [0; not chosen, 1; Deloitte, 2; E&Y, 3; 
KPMG, 4; PwC, 5; other, 6; no firm/student] 
AUDIT_WORK_EXP A dummy variable with a value of 1 if the student has work 
experience in auditing, otherwise 0. 
AUDIT_WORK_TEN Tenure of a student's audit experience [months] 
AUDITOR_EXP Amount of total experience as an auditor [99; less than year, 
1–19, 20; over 19 years] 
BOOKK_WORK_EXP A dummy variable with a value of 1 if the student has work 
experience in bookkeeping, otherwise 0. 
BOOKK_WORK_TEN Tenure of a student's bookkeeping experience [months] 
CERTIFICATION Latest auditor certification [1; no cert., 2; JHTT, 3; HTM, 4; 
KHT] 
CERTIFICATION2 A dummy variable with a value of 1 if a subject has any audi-
tor certification, otherwise 0. 
CL_EXP Task-specific experience on client acceptance decisions [0; no 
experience, 1; 1–9 times, 2; 10–19 times, 3; 20–29 times, 
4;30+ times] 
CREDIT_UNITS If subject is a student, number of completed credit units 
CU_ACC_FIN If subject is a student, number of completed accounting and 
finance credit units 
CU_AUDIT If subject is a student, number of completed auditing courses 
credit units 
FEE_ESTIMATE Estimate of audit fee [11-point Likert-type scale: 0; signifi-
cantly less than last year to 10; significantly more than last 
156      Acta Wasaensia 
year, where 5 means the same amount as last year or average 
amount scaled to client size] 
GENDER Gender [1; male, 2; female] 
HOUR_ESTIMATE Estimate of planned audit hours [11-point Likert-type scale: 0; 
significantly less than last year to 10; significantly more than 
last year, where 5 means the same number as last year or aver-
age number scaled to client size] 
HTM_CERT_YEAR The year when the subject's HTM certification was approved 
[0/year] 
INFOR_COMM A subject's general feedback of the experiment [open question] 
INFOR2_COMM A subject's thoughts about the purpose of the experiment [open 
question] 
IPADDRESS IP address of the subject 
JHTT_CERT_YEAR The year when the subject's JHTT certification was approved 
[0/year] 
JUDG_CONFIDENCE Level of confidence in the continuance/acceptance judgment 
[11-point Likert-type scale: 0; fully unconfident to 10; fully 
confident] 
JUDGMENT Probability judgment of recommending client continu-
ance/acceptance [11-point Likert-type scale: 0%; would not 
recommend at all to 100%; would definitely recommend]  
KHT_CERT_YEAR The year when the subject's KHT certification was approved 
[0/year] 
MANIP_CHECK_RELIA Manipulation check question of perceived information reliabil-
ity [11-point Likert-type scale: 0; very unreliable to 10; very 
reliable] 
MANIP_CHECK_RISK Manipulation check question of perceived RMM [11-point 
Likert-type scale: 0; very low - 10; very high] 
NUMBER_INF Total number of read cues, including multiple reads [0– …] 
RANK Rank in audit firm [0; not chosen, 1; junior, 2; senior, 3; part-
ner, 4; manager, 5; other 
RANK_YEARS Years in current rank [99; less than year, 1–19, 20; over 19 
years] 
REALISM Estimate of the realism of the continuance/acceptance task [11-
point Likert-type scale: 0; very unrealistic - 10; very realistic] 
TOT_TIME Total time used for the task [sec] 
TRAINING_1CL Amount of training on client continuance decisions [hours] 
TRAINING_CL Amount of training on client acceptance decisions [hours] 
TREATMENT Number of treatment group [1; semi-struc., low risk, more rel., 
2; semi-struc., low risk, less rel., 3; semi-struc., high risk, more 
rel., 4; semi-struc., high risk, less rel., 5; unstruc., low risk, 
more rel., 6; unstruc., low risk, less rel., 7; unstruc., high risk, 
more rel., 8; unstruc., high risk, less rel.] 
Notes:  
a These variables exist for each cue 
 Acta Wasaensia     157 
  
Appendix 3. Introductory letter 
Dear auditing expert, 
Making significant evaluations concerning companies on a daily basis is part of 
your auditing work. Often evaluations are based on your own decisions and delib-
eration. However, the decision-making process of Finnish auditors has not been 
researched sufficiently. Therefore, I have compiled a substantial collection of 
source material on auditors’ decision-making process. Now I intend to extend the 
collection of data with information on auditing experts in practice, that is, Finnish 
auditors and other people whose work is audit-related. 
I am kindly asking you to participate in this study by first examining the infor-
mation provided and then answering the following questions. This will take ap-
proximately 10 minutes. You will be asked to evaluate your willingness to con-
tinue as the auditor of a described company. You can base your decision on any 
information available. Please make the decision independently and without inter-
ruptions. 
Among those who answer, 3 travel gift certificates of 120 euros in value will be 
drawn, as well as 10 book prizes signed with dedication of winners’ choice. A 
separate drawing entry form will open after you have answered the questions. 
I am conducting my research at the University of Vaasa under the direction of 
Professor Teija Laitinen. The results of my study will be published in a trade 
journal and as a dissertation. The research is a part of the Academy of Finland’s 
project on auditing (no. 126630). 
All information obtained will be kept completely confidential. Individual re-
sponses will be compiled and kept anonymous during the whole process. 
If you have any questions concerning the study, or if you are interested in the re-
sults of the study, do not hesitate to contact me. Please answer the questions by 
May 22nd, 2011. 
In order to progress through the research, please click the following link: 
http://auditresearch.org/kevat2011/________.php 
Sincerely, 
Tuukka Järvinen  
M.Sc. (Econ.), doctoral student 
tuja@uwasa.fi, tel. 06-3248542  
 
158      Acta Wasaensia 
Appendix 4. Follow-up letter 
Dear auditing expert, 
A  while  ago,  I  contacted  you  in  regard  to  my  doctoral  dissertation  on  auditors’  
decision-making process. I sent you an e-mail containing a link to a research and 
asked you to spend 10 minutes of your time to answer it. Maybe you have not had 
time to answer the questions due to for instance work-related reasons. 
Now I contact you again because every answer is important to the success of the 
research. After examining the information provided, you will be asked to evaluate 
your willingness to take a new company as a client.  You are free to use any in-
formation available as the basis of your decision. Please make the decision inde-
pendently and without interruptions. 
In order to progress through the research, please click the following link: 
http://auditresearch.org/kevat2011/________.php 
As a reward for your time, 3 travel gift certificates of 120 euros in value, and 10 
books by professor Erkki K. Laitinen signed with dedication of winner’s choice 
will be drawn among those who answer the questions. A separate drawing entry 
form will open after you have answered. Entering the drawing is entirely optional. 
Please  make  the  decision  as  soon  as  possible  but  by  Sunday,  the  29th  of  May,  
2011 at the latest. Your answers will be kept completely confidential. Individual 
responses will be compiled and kept anonymous during the whole process. 
If you have already answered the questions, I want to thank you for your answers 
and for participating in this research. 
Sincerely, 
Tuukka Järvinen  
M.Sc. (Econ.), doctoral student 
tuja@uwasa.fi, tel. 06-3248542  
 
 
 
 
 
 Acta Wasaensia     159 
  
Appendix 5. Descriptive statistics of the dependent variables (n=307) 
 
 
 
 
 
 
  
 
Panel A: Descriptive statistics of the continuous dependent variables (All units in seconds) 
 
Variable 
 
Mean Std Dev. Minimum Q1 Median Q2 Max. 
TOT_TIME 666.3 589.5 76.0 353.0 539.0 774.0 4947.0 
TOT_CUE_TIME 332.9 349.1 0.0 154.0 275.0 419.0 4340.0 
JUDG_TIME 333.4 446.6 45.0 178.0 247.0 329.0 4844.0 
Panel B: Descriptive statistics of the count dependent variables (All units in counts) 
NUMBER_INF 
 
10.7 3.5 0.0 9.0 12.0 12.0 25.0 
NUMBER_INF_9 9.4 4.0 0.0 7.0 10.0 12.0 25.0 
NUMBER_IMP_CUE 10.2 3.8 0.0 8.0 11.0 13.0 24.0 
Panel C: Pearson’s correlation coefficients of dependent variables 
 
 
TOT_CUE JUDG NUMBER NUMBER NUMBER   
 _TIME _TIME _INF _INF_9 _IMP_CUE   
TOT_TIME 
 
0.656*** 0.807*** 0.324 *** 0.364 *** 0.266 ***   
TOT_CUE_TIME  0.084 0.355 *** 0.441 *** 0.342 ***   
JUDG_TIME   0.152 *** 0.135 ** 0.083   
NUMBER_INF    0.870 *** 0.804 ***   
NUMBER_INF_9     0.729 ***   
Notes:        
Statistical significance based on two-tailed tests at the 1%, 5% and 10% levels are denoted 
by ***, ** and *, respectively.       
The variables are defined as follows:        
TOT_TIME = Total time spent on the task  
TOT_CUE_TIME = Total time spent reading cues  
JUDG_TIME = Time spent outside of cues (tot_time minus tot_cue_times) 
NUMBER_INF = Total number of read cues, including multiple reads  
NUMBER_INF_9 = Number of over 9 second read cues, including multiple reads  
NUMBER_IMP_CUE = Number of read cues whose importance was self-evaluated to be more than 4, 
including multiple reads        
160      Acta Wasaensia 
Appendix 6. Descriptive statistics of dependent variables per treatment and 
      experience 
Treatment Less- N Obs    
 experienced?  Variable (units) Mean Std Dev 
1 NO 22 TOT_TIME  (sec.) 468.77 240.82 
Semi-struc.   TOT_CUE_TIME (sec.)  266.09 186.29 
Low risk   JUDG_TIME (sec.) 202.68 67.44 
More-rel.   NUMBER_INF (counts) 10 3.07 
   NUMBER_INF_9 (counts) 8.64 3.35 
   NUMBER_IMP_CUE (counts) 9.55 3.22 
 YES 6 TOT_TIME  (sec.) 614.17 188.39 
   TOT_CUE_TIME (sec.)  369.17 121.14 
   JUDG_TIME (sec.) 245 142.79 
   NUMBER_INF (counts) 11.83 0.41 
   NUMBER_INF_9 (counts) 10.83 2.04 
   NUMBER_IMP_CUE (counts) 11.33 1.86 
2 NO 28 TOT_TIME  (sec.) 544.82 245.17 
Semi-struc.   TOT_CUE_TIME (sec.)  287.75 177.5 
Low risk   JUDG_TIME (sec.) 257.07 126.25 
Less-rel.   NUMBER_INF (counts) 10.96 3.19 
   NUMBER_INF_9 (counts) 9.82 3.58 
   NUMBER_IMP_CUE (counts) 10.07 3.28 
 YES 13 TOT_TIME  (sec.) 546.85 151.47 
   TOT_CUE_TIME (sec.)  275.38 124.61 
   JUDG_TIME (sec.) 271.46 83.74 
   NUMBER_INF (counts) 11.46 3.62 
   NUMBER_INF_9 (counts) 10.85 3.76 
   NUMBER_IMP_CUE (counts) 10.54 2.76 
3 NO 21 TOT_TIME  (sec.) 505.14 229.3 
Semi-struc.   TOT_CUE_TIME (sec.)  268.05 159.46 
High risk   JUDG_TIME (sec.) 237.1 94.16 
More-rel.   NUMBER_INF (counts) 9.43 3.36 
   NUMBER_INF_9 (counts) 8.24 3.4 
   NUMBER_IMP_CUE (counts) 8.81 3.17 
 YES 10 TOT_TIME  (sec.) 697 315.95 
   TOT_CUE_TIME (sec.)  371.6 178.38 
   JUDG_TIME (sec.) 325.4 157.37 
   NUMBER_INF (counts) 11.3 2.31 
   NUMBER_INF_9 (counts) 10.8 3.01 
   NUMBER_IMP_CUE (counts) 11.2 2.74 
4 NO 21 TOT_TIME  (sec.) 599.33 229.22 
Semi-struc.   TOT_CUE_TIME (sec.)  288.05 134.91 
High risk   JUDG_TIME (sec.) 311.29 144.62 
Less-rel.   NUMBER_INF (counts) 11.05 1.99 
   NUMBER_INF_9 (counts) 9.76 2.64 
   NUMBER_IMP_CUE (counts) 11.05 2.25 
 YES 11 TOT_TIME  (sec.) 528.09 181.87 
   TOT_CUE_TIME (sec.)  302.27 148.99 
   JUDG_TIME (sec.) 225.82 88.7 
   NUMBER_INF (counts) 10.82 2.56 
 Acta Wasaensia     161 
  
   NUMBER_INF_9 (counts) 10.09 3.45 
   NUMBER_IMP_CUE (counts) 9.27 3.44 
5 NO 24 TOT_TIME  (sec.) 497.63 221.74 
Unstruct.   TOT_CUE_TIME (sec.)  256.75 159.21 
Low risk   JUDG_TIME (sec.) 240.88 126.21 
More-rel.   NUMBER_INF (counts) 10.46 3.13 
   NUMBER_INF_9 (counts) 8.21 3.39 
   NUMBER_IMP_CUE (counts) 10.38 2.98 
 YES 7 TOT_TIME  (sec.) 610.71 265.05 
   TOT_CUE_TIME (sec.)  281.71 107.72 
   JUDG_TIME (sec.) 329 228.11 
   NUMBER_INF (counts) 11.43 3.31 
   NUMBER_INF_9 (counts) 10.14 3.48 
   NUMBER_IMP_CUE (counts) 9.86 5.43 
6 NO 24 TOT_TIME  (sec.) 565.04 234.32 
Unstruct.   TOT_CUE_TIME (sec.)  304.29 170.03 
Low risk   JUDG_TIME (sec.) 260.75 93.67 
Less-rel.   NUMBER_INF (counts) 11.58 2.47 
   NUMBER_INF_9 (counts) 10 3.86 
   NUMBER_IMP_CUE (counts) 11.83 3.25 
 YES 17 TOT_TIME  (sec.) 596.71 232.57 
   TOT_CUE_TIME (sec.)  331.59 147.27 
   JUDG_TIME (sec.) 265.12 98.9 
   NUMBER_INF (counts) 12.18 1.81 
   NUMBER_INF_9 (counts) 10.76 2.97 
   NUMBER_IMP_CUE (counts) 11.53 3.06 
7 NO 23 TOT_TIME  (sec.) 492.35 240.06 
Unstruct.   TOT_CUE_TIME (sec.)  248.74 142.12 
High risk   JUDG_TIME (sec.) 243.61 126.42 
More-rel.   NUMBER_INF (counts) 10.74 2.73 
   NUMBER_INF_9 (counts) 8.7 3.14 
   NUMBER_IMP_CUE (counts) 10.43 3.67 
 YES 14 TOT_TIME  (sec.) 746.86 291.1 
   TOT_CUE_TIME (sec.)  403.21 247.05 
   JUDG_TIME (sec.) 343.64 196.49 
   NUMBER_INF (counts) 11 3.37 
   NUMBER_INF_9 (counts) 10.21 3.98 
   NUMBER_IMP_CUE (counts) 10.64 3.43 
8 NO 22 TOT_TIME  (sec.) 593.86 216.58 
Unstruct.   TOT_CUE_TIME (sec.)  288.5 136.42 
High risk   JUDG_TIME (sec.) 305.36 155.55 
Less-rel.   NUMBER_INF (counts) 10.32 2.92 
   NUMBER_INF_9 (counts) 9.59 3.55 
   NUMBER_IMP_CUE (counts) 10.14 2.75 
 YES 8 TOT_TIME  (sec.) 906.38 169.38 
   TOT_CUE_TIME (sec.)  452.25 178.34 
   JUDG_TIME (sec.) 454.13 178.49 
   NUMBER_INF (counts) 13.63 5.34 
   NUMBER_INF_9 (counts) 12.75 6.02 
   NUMBER_IMP_CUE (counts) 12.38 6.46 
162      Acta Wasaensia 
Appendix 7. Residual plots for ANOVA models 
 
Dependent variable: LN_TOT_TIME 
 
 
 
 
 
 
 
Dependent variable: LN_TOT_CUE_TIME 
 
 
 
 
 
 
 
Dependent variable: LN_JUDG_TIME 
 
 
 
 
 Acta Wasaensia     163 
  
Appendix 8.  ANOVA tables for continuous variables without students 
 (n=251) 
 
Panel A. Dependent variable LN_TOT_TIME (Model 1)  
Variable df Sum of Mean   
  Squares square F p 
INEXP 1 1.28 1.28 6.69 0.010 
RISK 1 0.72 0.72 3.79 0.053 
STRUCTURE (STR) 1 0.67 0.67 3.53 0.062 
RELIABILITY (REL) 1 0.30 0.30 1.56 0.213 
INEXP × RISK  1 0.28 0.28 1.46 0.228 
INEXP × STR  1 0.81 0.81 4.23 0.041 
INEXP × REL 1 0.48 0.48 2.53 0.113 
RISK × STR 1 0.13 0.13 0.69 0.408 
RISK × REL 1 0.03 0.03 0.15 0.701 
STR × REL 1 0.22 0.22 1.18 0.279 
INEXP × RISK × STR  1 0.47 0.47 2.46 0.118 
INEXP × STR × REL  1 0.56 0.56 2.93 0.088 
RISK × STR × REL  1 0.23 0.23 1.22 0.270 
RISK × REL × INEXP 1 0.09 0.09 0.47 0.494 
INEXP × RISK × STR × REL 1 0.15 0.15 0.78 0.378 
Model 15 5.20 0.35 1.82 0.033 
Error 235 44.85 0.19   
Total 250     
Panel B. Dependent variable LN_TOT_CUE_TIME (Model 2) 
Variable df Sum of Mean F  p 
  Squares square   
INEXP 1 0.97 0.97 2.45 0.119 
RISK 1 0.04 0.04 0.11 0.744 
STRUCTURE (STR) 1 0.07 0.07 0.17 0.677 
RELIABILITY (REL) 1 0.08 0.08 0.20 0.655 
INEXP × RISK  1 0.03 0.03 0.07 0.796 
INEXP × STR  1 0.31 0.31 0.78 0.377 
INEXP × REL 1 0.62 0.62 1.57 0.211 
RISK × STR 1 0.00 0.00 0.00 0.975 
RISK × REL 1 0.06 0.06 0.14 0.706 
STR × REL 1 0.42 0.42 1.07 0.303 
INEXP × RISK × STR  1 0.02 0.02 0.04 0.838 
INEXP × STR × REL  1 0.42 0.42 1.06 0.303 
RISK × STR × REL  1 0.02 0.02 0.05 0.818 
RISK × REL × INEXP 1 0.58 0.58 1.48 0.225 
INEXP × RISK × STR × REL 1 0.06 0.06 0.15 0.700 
Model 15 4.11 0.27 0.70 0.787 
Error 235 92.5 0.39   
Total 250     
164      Acta Wasaensia 
Appendix 8. Continued  
  
Panel C. Dependent variable LN_JUDG_TIME (Model 3)  
Variable df Sum of Mean F  p 
  Squares square   
INEXP 1 1.11 1.11 6.72 0.010 
RISK 1 1.60 1.60 9.67 0.002 
STRUCTURE (STR) 1 1.06 1.06 6.41 0.012 
RELIABILITY (REL) 1 0.73 0.73 4.41 0.037 
INEXP × RISK  1 0.28 0.28 1.69 0.196 
INEXP × STR  1 1.03 1.03 6.23 0.013 
INEXP × REL 1 0.27 0.27 1.63 0.203 
RISK × STR 1 0.14 0.14 0.85 0.358 
RISK × REL 1 0.17 0.17 1.04 0.310 
STR × REL 1 0.12 0.12 0.75 0.388 
INEXP × RISK × STR  1 0.79 0.79 4.80 0.029 
INEXP × STR × REL  1 0.57 0.57 3.45 0.064 
RISK × STR × REL  1 0.58 0.58 3.52 0.062 
RISK × REL × INEXP 1 0.02 0.02 0.13 0.718 
INEXP × RISK × STR × REL 1 0.52 0.52 3.52 0.077 
Model 15 6.26 0.42 2.52 0.002 
Error 235 38.92 0.17   
Total 250     
Notes:      
The variables are defined as follows:      
LN_TOT_TIME = Natural logarithm of total time spent on the task 
LN_TOT_CUE_TIME = Natural logarithm of total time used for reading cues  
LN_JUDG_TIME = Natural logarithm of time spent outside of cue screens 
INEXP = A categorical variable with a value of 1 if a subject is less experienced, otherwise 0 
RISK = A categorical variable with a value of 1 if a treatment group contains high RMM, otherwise 0 
STRUCTURE = A categorical variable with a value of 1 if a treat. group is an unstructured task, otherwise 0 
RELIABILITY = A categorical variable with a value of 1 if a treatment group contains the less reliable in-
formation manipulation, otherwise 0 
 
 
 
 
 
 
 
 Acta Wasaensia     165 
  
Appendix 9. Descriptive statistics of the task-specific judgment variables per 
treatment (See Appendix 2 for definitions of variables) 
 
 Client continuance  
     
Treatment N Obs Variable Mean Std Dev 
1 28 JUDGMENT 93.21 8.63 
Semi-struc.  JUDG_CONF 9.07 1.02 
Low risk  HOUR_ESTIM 6.39 0.99 
More-rel.  FEE_ESTIM 6.11 0.96 
2 41 JUDGMENT 88.05 10.54 
Semi-struc.  JUDG_CONF 8.73 1.07 
Low risk  HOUR_ESTIM 7.39 1.20 
Less-rel.  FEE_ESTIM 7.10 1.09 
3 31 JUDGMENT 80.97 21.66 
Semi-struc.  JUDG_CONF 8.23 1.91 
High risk  HOUR_ESTIM 6.87 1.15 
More-rel.  FEE_ESTIM 6.65 1.11 
4 32 JUDGMENT 78.44 16.68 
Semi-struc.  JUDG_CONF 7.81 1.87 
High risk  HOUR_ESTIM 7.56 1.13 
Less-rel.  FEE_ESTIM 7.25 1.02 
     
 Client acceptance  
     
Treatment N Obs Variable Mean Std Dev 
5 31 JUDGMENT 87.74 11.17 
Unstruct.  JUDG_CONF 8.87 1.02 
Low risk  HOUR_ESTIM 5.55 0.85 
More-rel.  FEE_ESTIM 5.48 0.89 
6 41 JUDGMENT 82.44 13.92 
Unstruct.  JUDG_CONF 8.44 1.18 
Low risk  HOUR_ESTIM 6.41 1.38 
Less-rel.  FEE_ESTIM 6.22 1.24 
7 37 JUDGMENT 85.14 15.39 
Unstruct.  JUDG_CONF 8.70 1.31 
High risk  HOUR_ESTIM 6.32 1.25 
More-rel.  FEE_ESTIM 6.19 0.84 
8 30 JUDGMENT 78.00 17.69 
Unstruct.  JUDG_CONF 7.70 2.17 
High risk  HOUR_ESTIM 6.80 1.03 
Less-rel.  FEE_ESTIM 6.53 0.86