advances in accounting behavioral research

ADVANCES IN ACCOUNTING

BEHAVIORAL RESEARCH

i

ADVANCES IN ACCOUNTING

BEHAVIORAL RESEARCH

Series Editor: Vicky Arnold

Volumes 1–4: Series Editor: James E. Hunton

Volumes 5–7: Series Editor: Vicky Arnold

ii

ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH VOLUME 8

ADVANCES INACCOUNTING

BEHAVIORAL RESEARCHEDITED BY

VICKY ARNOLD

Department of Accounting, School of Business, University of Connecticut, USA

and Department of Accounting and Business Information Systems, University of

Melbourne, Australia

Associate Editors:

B. DOUGLAS CLINTON

Northern Illinois University, USA

PETER LUCKETT

University of New South Wales, Australia

ROBIN ROBERTS

University of Central Florida, USA

CHRIS WOLFE

Texas A&M University, USA

SALLY WRIGHT

University of Massachusetts Boston, USA

Amsterdam – Boston – Heidelberg – London – New York – Oxford

Paris – San Diego – San Francisco – Singapore – Sydney – Tokyo

2005

iii

r 2005 Elsevier Ltd. All rights reserved.

This work is protected under copyright by Elsevier Ltd, and the following terms and conditions apply to its use:

PhotocopyingSingle photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permissionof the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying,copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates areavailable for educational institutions that wish to make photocopies for non-profit educational classroom use.

Permissions may be sought directly from Elsevier’s Rights Department in Oxford, UK: phone (+44) 1865 843830, fax(+44) 1865 853333, e-mail: [email protected]. Requests may also be completed on-line via the Elsevierhomepage (http://www.elsevier.com/locate/permissions).

In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UKthrough the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, LondonW1P 0LP, UK; phone: (+44) 20 7631 5555; fax: (+44) 20 7631 5500. Other countries may have a local reprographicrights agency for payments.

Derivative WorksTables of contents may be reproduced for internal circulation, but permission of the Publisher is required for externalresale or distribution of such material. Permission of the Publisher is required for all other derivative works, includingcompilations and translations.

Electronic Storage or UsagePermission of the Publisher is required to store or use electronically any material contained in this work, including anychapter or part of a chapter.

Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in anyform or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permissionof the Publisher.Address permissions requests to: Elsevier’s Rights Department, at the fax and e-mail addresses noted above.

NoticeNo responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter ofproducts liability, negligence or otherwise, or from any use or operation of any methods, products, instructions orideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independentverification of diagnoses and drug dosages should be made.

First edition 2005

British Library Cataloguing in Publication DataA catalogue record is available from the British Library.

ISBN: 0-7623-1218-1ISSN: 1475-1488 (Series)

8 The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).Printed in The Netherlands.

ELSEVIER B.V.

Radarweg 29

P.O. Box 211

1000 AE Amsterdam

The Netherlands

ELSEVIER Ltd

84 Theobalds Road

London

WC1X 8RR

UK

ELSEVIER Ltd

The Boulevard, Langford

Lane, Kidlington

Oxford OX5 1GB

UK

ELSEVIER Inc.

525 B Street, Suite 1900

San Diego

CA 92101-4495

USA

Working together to growlibraries in developing countries

www.elsevier.com | www.bookaid.org | www.sabre.org

iv

CONTENTS

LIST OF CONTRIBUTORS vii

REVIEWER ACKNOWLEDGEMENTS ix

EDITORIAL POLICY AND SUBMISSIONGUIDELINES

xi

BELIEF REVISION IN ACCOUNTING: ALITERATURE REVIEW OF THE BELIEF-ADJUSTMENT MODEL

Jennifer Kahle, Robert Pinsker and Robin Pennington 1

AUDITOR CALIBRATION IN THE REVIEW PROCESSNoel Harding, Sally Hughes and Ken T. Trotman 41

LINGUISTIC DELIVERY STYLE, CLIENTCREDIBILITY, AND AUDITOR JUDGMENT

Christie L. Comunale, Thomas R. Sexton andTerry L. Sincich

59

CLIENT INQUIRY VIA ELECTRONICCOMMUNICATION MEDIA: DOES THE MEDIUMMATTER?

Anna Noteberg and James E. Hunton 87

ROLE MORALITY AND ACCOUNTANTS’ETHICALLY SENSITIVE DECISIONS

Robin R. Radtke 113

v

THE EFFECT OF MANAGER’S MORAL EQUITY ONTHE RELATIONSHIP BETWEEN BUDGETPARTICIPATION AND PROPENSITY TO CREATESLACK: A RESEARCH NOTE

Adam S. Maiga 139

ASYMMETRIC EFFECTS OF ACTIVITY-BASEDCOSTING SYSTEM COST REALLOCATION

M.G. Fennema, Jay S. Rich and Kip Krumwiede 167

EXAMINING THE ROLE OF CULTURE ANDACCULTURATION IN INFORMATION SHARING

Stephen B. Salter and Axel K.-D. Schulz 189

THE EFFECTS OF VALUE ATTAINMENT ANDCOGNITIVE ROLES OF BUDGETARYPARTICIPATION ON JOB PERFORMANCE

Vincent K. Chong, Ian R.C. Eggleton andMichele K.C. Leong

213

CONTENTSvi

LIST OF CONTRIBUTORS

Vincent K. Chong UWA Business School, The University ofWestern Australia, Australia

Christie L. Comunale School of Professional Accountancy, LongIsland University – C.W. Post Campus,USA

Ian R. C. Eggleton Waikato Management School, Universityof Waikato, New Zealand

M. G. Fennema Department of Accounting, Florida StateUniversity, USA

Noel Harding School of Accounting, University of NewSouth Wales, Australia

Sally Hughes School of Accounting, University of NewSouth Wales, Australia

James E. Hunton Accountancy Department, Bentley College,USA and Department of Accounting andInformation Management, UniversiteitMaastricht, The Netherlands

Jennifer Kahle School of Accountancy, University ofSouth Florida, USA

Kip Krumwiede College of Business and Economics, BoiseState University, USA

Michele K. C. Leong UWA Business School, The University ofWestern Australia, Australia

Adam S. Maiga School of Business Administration,University of Wisconsin – Milwaukee, USA

Anna Noteberg Department of Business Studies,Universiteit van Amsterdam, TheNetherlands

vii

Robin Pennington Department of Accounting andInformation Management, University ofTennessee, USA

Robert Pinsker College of Business and PublicAdministration, Old Dominion University,USA

Robin R. Radtke Department of Accounting, The Universityof Texas at San Antonio, USA

Jay S. Rich College of Business, Illinois StateUniversity, USA

Stephen B. Salter Universidad Adolfo Ibanez Escuela deNegocios, College of BusinessAdministration, University of Cincinnati,USA

Axel K-D. Schulz Department of Accounting and BusinessInformation Systems, The University ofMelbourne, Australia

Thomas R. Sexton College of Business, Stony BrookUniversity, USA

Terry L. Sincich Information Systems and Decision SciencesDepartment, University of South Florida,USA

Ken T. Trotman School of Accounting, University of NewSouth Wales, Australia

LIST OF CONTRIBUTORSviii

Mohammed AbdolmohammadiBentley College, USA

Elizabeth AlmerPortland State University, USA

Philip BeaulieuUniversity of Calgary, Canada

Jean BedardNortheastern University, USA

James BierstakerVillanova University, USA

Dennis M. BlineBryant College, USA

Wray BradleyUniversity of Tulsa, USA

Gary BraunUniversity of Texas at El Paso,USA

Rich BrodyUniversity of New Haven, USA

Shimin ChenUniversity of Louisiana atLafayette, USA

Vincent ChongThe University of WesternAustralia, Australia

Freddie ChooSan Francisco State University,USA

Janne ChungYork University, Canada

Bryan ChurchGeorgia Tech University, USA

Jeff CohenBoston College, USA

William N. DillaIowa State University, USA

Jesse DillardPortland State University, USA

Craig EmbySimon Fraser University, Canada

Dann FisherKansas State University, USA

Clark HamptonUniversity of Connecticut, USA

REVIEWER ACKNOWLEDGEMENTS

The Editor and Associate Editors at AABR would like to thank the many

excellent reviewers who have volunteered their time and expertise to make

this an outstanding publication. Publishing quality papers in a timely

manner would not be possible without their efforts.

ix

Joanne P. HealyKent State University, USA

Karen L. HooksFlorida Atlantic University, USA

Stacy KovarKansas State University, USA

Tanya LeeUniversity of North Texas, USA

Theresa LibbyWilfred Laurier University,Canada

Tim LindquistThe University of Northern Iowa,USA

Jill McKinnonMacquarie University, Australia

Mario MalettaNortheastern University, USA

Maureen MaschaMarquette University, USA

Elaine MauldinUniversity of Missouri, USA

Rob NieschwietzUniversity of Colorado at Denver,USA

Andreas NikolaouBowling Green State University,USA

Hossein NouriThe College of New Jersey, USA

Ed O’DonnellArizona State University, USA

Laurie PantSuffolk University, USA

Robert J. ParkerUniversity of New Orleans, USA

Will QuilliamUniversity of South Florida, USA

Randall RentfroFlorida Atlantic University, USA

Andrew J. RosmanUniversity of Connecticut, USA

Scott SummersBrigham Young University, USA

Steve SuttonUniversity of Connecticut, USA

Linda ThorneYork University, Canada

Kristin WentzelLa Salle University, USA

John WermertDrake University, USA

Patrick WheelerUniversity of Missouri, USA

Brett WilkinsonBaylor University, USA

Bernard Wong-On-WingWashington State University, USA

Alex YenUniversity of Connecticut, USA

REVIEWER ACKNOWLEDGEMENTSx

EDITORIAL POLICY AND

SUBMISSION GUIDELINES

Advances in Accounting Behavioral Research (AABR) publishes articles en-

compassing all areas of accounting that incorporate theory from and con-

tribute new knowledge and understanding to the fields of applied

psychology, sociology, management science, and economics. The journal

is primarily devoted to original empirical investigations; however, literature

review papers, theoretical analyses, and methodological contributions are

welcome. AABR is receptive to replication studies, provided they investigate

important issues and are concisely written. The journal especially welcomes

manuscripts that integrate accounting issues with organizational behavior,

human judgment/decision making, and cognitive psychology.

Manuscripts will be blind-reviewed by two reviewers and an associate editor.

The recommendations of the reviewers and associate editor will be used to

determine whether to accept the paper as is, accept the paper with minor re-

visions, reject the paper or to invite the authors to revise and resubmit the paper.

Manuscript Submission

Manuscripts should be forwarded to the editor, Vicky Arnold, at

[email protected] via e-mail. All text, tables, and figures

should be incorporated into a Word document prior to submission. The

manuscript should also include a title page containing the name and address

of all authors and a concise abstract. Also, include a separate Word doc-

ument with any experimental materials or survey instruments. If you are

unable to submit electronically, please forward the manuscript along with

the experimental materials to the following address:

Vicky Arnold, Editor

Advances in Accounting Behavioral Research,

Department of Accounting U41A

School of Business

University of Connecticut

Storrs, CT 06269-2041

xi

References should follow the APA (American Psychological Association)

standard. References should be indicated by giving (in parentheses) the

author’s name followed by the date of the journal or book; or with the date

in parentheses, as in ‘suggested by Earley (2000)’.

In the text, use the form Rosman et al. (1995) where there are more than

two authors, but list all authors in the references. Quotations of more than

one line of text from cited works should be indented and citation should

include the page number of the quotation, e.g. (Dunbar, 2001 p. 56).

Citations for all articles referenced in the text of the manuscript should be

shown in alphabetical order in the Reference list at the end of the man-

uscript. Only articles referenced in the text should be included in the Ref-

erence list. Format for references is as follows:

For journals:

Dunn, C.L., & Gerard, G.J. (2001). Auditor efficiency and effectiveness with

diagrammatic and linguistic conceptual model representations. International

Journal of Accounting Information Systems, 2(3), 1–40.

For books:

Ashton, R.H., & Ashton, A.H. (1995). Judgment and decision-making research

in accounting and auditing. New York, NY: Cambridge University Press.

For a thesis:

Smedley, G.A. (2001). The effects of optimization on cognitive skill acquisition

from intelligent decision aids. Unpublished doctoral dissertation, University.

For a working paper:

Thorne, L., Massey, D.W., & Magnan, M. (2000). Insights into selection-

socialization in the audit profession: An examination of the moral reasoning

of public accountants in the United States and Canada. Working paper

York University, North York, Ontario.

For papers from conference proceedings, chapters from book, etc.:

Messier, W.F. (1995). Research in and development of audit decision aids. In:

R.H. Ashton, & A.H. Ashton (Ed.), Judgment and decision making in

accounting and auditing (pp. 207–230). New York: Cambridge University Press.

EDITORIAL POLICY AND SUBMISSION GUIDELINESxii

BELIEF REVISION IN

ACCOUNTING: A LITERATURE

REVIEW OF THE

BELIEF-ADJUSTMENT MODEL

Jennifer Kahle, Robert Pinsker and Robin Pennington

ABSTRACT

The belief-adjustment model has been an integral part of accounting

research in belief revision, especially in the examination of order effects.

Hogarth and Einhorn ((1992) Cognitive Psychology, 24, 1–55) created

the belief-adjustment model to serve as a theoretical framework for

studying individuals’ decision-making processes. The model examines

several aspects of decision-making, such as encoding, response mode, and

task factors. The purpose of this chapter is to provide a comprehensive

examination of the accounting studies that have used the theoretical

framework of the belief-adjustment model in auditing, tax, and financial

accounting contexts. Roberts’ ((1998) Journal of the American Taxation

Association, 20, 78–121) model of tax accountants’ decision-making is

used as a guideline to organize the research into categories. By using

Roberts’ categorization, we can better sort out the mixed results of some

prior studies and also expand the review to include a more comprehensive

look at the model and its application to accounting. While many variables

have been examined with respect to their effect on accounting professionals’

Advances in Accounting Behavioral Research

Advances in Accounting Behavioral Research, Volume 8, 1–40

Copyright r 2005 by Elsevier Ltd.

All rights of reproduction in any form reserved

ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08001-9

1

belief revisions, most studies examine them in isolation and do not consider

the interaction effects that these variables may have. Our framework also

identifies areas of the belief-adjustment model that need further research.

INTRODUCTION

Professional judgment in accounting has been described in general terms as

a continuous and incremental process (Gibbins, 1984). Most judgment tasks

involve evidence that is evaluated in a sequential nature. Although

sequential processing can provide economy in decision-making in terms of

smaller demands on memory and information-processing load, it can also

have detrimental effects, such as order effects in belief updating. Order

effects, biases, and the use of heuristics resulting from the method by which

individuals update their beliefs have been found in various areas of

accounting, including auditing (e.g., Ashton & Ashton, 1988; Asare, 1992),

tax (e.g., Pei, Reckers, & Wyndelts, 1990), and financial reporting (Pinsker,

2004).

Much of the judgment and decision-making accounting literature has

been influenced by Hogarth and Einhorn’s (1992) theory of belief revision.

The theory accounts for order effects as they arise from the interaction of

information-processing strategies and task characteristics. In particular,

Hogarth and Einhorn’s belief-adjustment model assumes people handle

belief revision tasks by a general, sequential anchoring and adjustment

process in which current opinion, or the anchor, is adjusted by the impact of

succeeding pieces of evidence. The model predicts that under conditions of

evaluating a short series of simple information, a primacy effect will occur

(i.e., the decision-maker will place more weight on the earliest information

received) if the judgment is made after viewing all the evidence. However, if

a short series of mixed (i.e., both positive and negative relative to a current

hypothesis) information is evaluated piece-by-piece, differential weighting of

the mixed information will produce a recency effect (i.e., the decision-maker

will place more weight on the latest information received).

An abundance of accounting literature examining recency effects has been

motivated by the belief-adjustment model and findings of order effects in the

psychology literature. Findings suggest that, in the absence of some

mitigating influence, many accounting judgments are subject to recency

effects. From a normative perspective, the sequence of evidence evaluation

should not affect the conclusion drawn from the evidence (Christian &

Reneau, 1990). The implications for accounting professionals may be

JENNIFER KAHLE ET AL.2

systematic biases in judgments leading to reduced efficiency and effec-

tiveness.

The purpose of this chapter is to provide a comprehensive review of the

accounting studies that have used the theoretical framework of the belief-

adjustment model in auditing, tax, and financial accounting contexts. The

emphasis will be on the factors that have been shown to influence or mitigate

order effects. Roberts’ (1998) model of tax accountants’ decision-making will

be used as a guideline to organize the factors into five main categories,

including (1) individual psychological factors, (2) environmental factors, (3)

input task factors, (4) processing factors, and (5) output task factors. This

categorization is similar to that of Gibbins’ (1984) model of professional

judgment in public accounting, whose groups include the person, stimulus,

environment, decision, and judgment process. Categorization in this manner

recognizes the separate influences of both external (environmental and task)

and internal (individual) factors and reflects the recommendation by

Hogarth and Einhorn (1992) to study further procedural and task variables

that can affect belief revision. In an evaluation of recency effects in audit

judgments, Trotman and Wright (2000) noted that results have been mixed

across studies. By using Roberts’ (1998) categorizations, we can better

understand the mixed results of the prior auditing studies, and also expand

the review to include a more comprehensive look at the model and its

application to various areas of accounting.

The remainder of this chapter is organized into four sections. The first

section describes the belief-adjustment model as proposed by Hogarth and

Einhorn (1992). The following section discusses the general applications of

the belief-adjustment model to accounting. Next, factors tested for

association with belief revision are reviewed and categorized. The final

section provides some concluding remarks as well as suggested directions for

future research.

THE BELIEF-ADJUSTMENT MODEL

Order Effects

Bayes’ theorem was the dominant normative model of belief revision in

accounting prior to 1988. The theorem gained popularity because it is a

logical consequence of conditional probabilities. However, research in

behavioral decision-making suggests that it is incomplete as a descriptive

model of belief revision as it cannot adequately predict intuitive revision

Belief Revision in Accounting 3

(Ashton & Ashton, 1988). Some researchers suggest that the discrepancy

is due to the tendency for intuitive revisions to be driven by task charac-

teristics, such as presentation order of information, which are irrelevant to

the normative model (Pitz, Downing, & Reinhold, 1967). A large body of

literature in psychology and accounting (e.g., Hogarth & Einhorn, 1992;

Ashton & Ashton, 1988; Pei et al., 1992a, b; Kennedy, 1993) has confirmed

the existence of presentation order effects on individual belief revisions.

Hogarth and Einhorn (1992, p. 3) define order effects with the following

example:

There are two pieces of evidence, A and B. Some subjects express an opinion after seeing

the information in the order A–B; others receive the information in the order B–A. An

order effect occurs when opinions after A–B differ from those after B–A.

Primacy occurs when an individual places more weight on the earlier

evidence in the sequence, while recency occurs when an individual places

more weight on the latter (more recent) evidence in the sequence.

Hogarth and Einhorn (1992) developed a ‘‘belief-adjustment model’’ to

more fully explain how evidence is encoded and processed. They adapted the

general concept of anchoring and adjustment (i.e., forming a belief and then

adjusting it based on new information to form a new belief) to include

heuristics into the model. Research since 1988 has provided descriptive

validity for using the belief-adjustment model, rather than Bayes’ theorem,

to explain belief revision (e.g., Ashton & Ashton, 1988; Pei et al., 1990;

Bamber, Ramsay, & Tubbs, 1997).

There are four distinct aspects in which the belief-adjustment model differs

from Bayesian probability (Krishnamoorthy, Mock, & Washington, 1999).

The belief-adjustment model (1) predicts that belief revision is influenced by

the order in which evidence is evaluated; (2) entails an anchoring and

adjustment strategy where the extent of belief revision is based on the size of

the anchor (current opinion), a strategy that violates the fundamental tenets

of Bayesian probability theory; (3) explicitly models the decision-maker’s

sensitivity toward evidence; and (4) allows one to increase or decrease

support for a hypothesis (e.g., that an account is fairly stated) without

affecting support for its complement (e.g., that an account is not fairly

stated). In an examination of four theoretical models of belief revision,

including a version of Bayesian inference, Krishnamoorthy et al. found that

the belief-adjustment model is the only model that captures both the

direction and magnitude of auditors’ belief revisions. Hogarth and Einhorn’s

(1992) theory of belief revision accounts for order effects by examining the

interaction of task characteristics and information processing strategies.


Task Variables

In forming their model, Hogarth and Einhorn (1992) considered three task

variables: (1) the complexity of the individual items of evidence to be

processed, (2) the length of the series of items, and (3) the manner in which

judgments are elicited, which will be referred to as response mode.1

Complexity is a function of the amount of information for each piece of

evidence that needs to be processed as well as the lack of familiarity with the

task (i.e., less familiar equals more complex and vice versa). Complexity is

important to belief revision since it relates to human processing ability. As

complexity increases, people may resort to simplifying strategies to ease

cognitive strain (Hogarth & Einhorn, 1992).

Length of series refers to the number of pieces of evidence to be evaluated.

Based on their review of several studies, Hogarth and Einhorn (1992)

consider a series of between 2 and 12 items to be ‘‘short’’ and a series of 17

or more items ‘‘long.’’ To distinguish length of series from complexity,

remember that complexity results from the amount of information

processing required and not necessarily the length of the series. As Arnold,

Collier, Leech, and Sutton (2000) indicate, a ‘‘complex’’ task is one that is

unfamiliar to the decision-maker (as noted in the previous paragraph) or

one that requires heavy information processing (defined as either a large

number or a long series of items). Therefore, a complex task could result

from a short series of evidence (if unfamiliar or full of detail) or

automatically from a long series. A ‘‘simple’’ task results from a short

series of familiar items.

Response mode concerns the manner in which judgments are elicited. Hogarth

and Einhorn (1992) consider two types: Step-by-Step (SbS) and End-of-Sequence

(EoS). The SbS mode is a ‘‘sequential’’ procedure whereby participants express

their beliefs each time they integrate a new piece of evidence. On the other hand,

the EoS mode is a ‘‘simultaneous’’ procedure in which participants express their

opinions only after all the information has been presented.

Encoding

Hogarth and Einhorn (1992) also acknowledge the impact of the method by

which individuals encode or process information on their subsequent

judgments. Accordingly, the predictions of the belief-adjustment model are

affected by two additional encoding variables: (1) processing mode (SbS

versus EoS), and (2) task type (evaluation versus estimation).


While related, the processing mode cannot be solely determined by the

required response mode. When the task is completed in an SbS response

mode, the individual must employ an SbS process by adjusting his or her

opinion incrementally for each piece of evidence processed. However, if a

task is completed in an EoS response mode, an individual can employ either

an SbS or an EoS processing mode. The EoS processing mode requires an

individual to aggregate all items prior to integrating them with the anchor,

which can be cognitively demanding. Thus, the processing mode should

depend on the cognitive demands of the task. Generally, the SbS processing

mode will be used when the task is more complex. This allows the individual

to continually integrate information with the anchor. The EoS mode is

expected to be used for a simpler task where aggregating the latter

information is cognitively easier.

In addition to task complexity, psychology and accounting research have

examined processing mode with respect to level of experience. Specifically,

Yates (1990) indicated less experienced individuals employ a sequential

(SbS) decision-making process. As individual decision-makers gain experi-

ence, they develop more ‘‘sophisticated’’ decision-making processes,

consistent with an EoS processing mode (Anderson, 1988). Further, less-

experienced individuals may employ an SbS processing mode to reduce

cognitive load (Arnold et al., 2000) or to reduce the effort necessary to

process a task (Hunton & McEwen, 1997).

Hogarth and Einhorn (1992) also make a distinction between evaluation

and estimation tasks. In evaluation tasks, information is encoded as positive

or negative relative to the hypothesis under consideration. Here, evidence is

seen as bipolar relative to the hypothesis and can be expressed by some

value on the continuum between ‘‘false’’ and ‘‘true.’’ On the other hand,

estimation tasks involve assessing a ‘‘moving average’’ that reflects the

position of each new piece of evidence relative to current opinion (involving

a unipolar scale). Research in accounting judgments has generally required

and found support for the use of the evaluation form of the model (Ashton

& Ashton, 1988; Tubbs, Messier, & Knechel, 1990).

To our knowledge, no accounting study has used the estimation task. As

Hogarth and Einhorn (1992, p. 9) indicated, estimation tasks use data that

fit ‘‘averaging models;’’ whereas, evaluation tasks use data that fit ‘‘adding

models.’’ In the auditing domain (where the majority of belief-adjustment

model research has taken place), Messier and Tubbs (1994) contend that an

auditor would evaluate an item and its relationship to an audit assertion

before revising beliefs about the assertion. Thus, auditors in particular are

assumed to generally employ an additive model when revising beliefs


(consistent with an evaluation task). Although there have been accounting

studies that used only positive or negative information in their tasks (e.g.,

Ashton & Ashton, 1988; Tubbs et al., 1990), the results obtained were

measured and analyzed in an additive fashion.2

Anchoring and Adjustment Process

The belief-adjustment model assumes that people revise beliefs through a

sequential anchoring and adjustment process in which the current opinion (an

anchor) is adjusted by the impact of subsequent pieces of information. The

algebraic form of the belief-adjustment model can generally be written as3

Sk ¼ Sk�1 þ wk½sðxkÞ � R�

where

Sk ¼ degree of belief in some hypothesis, after evaluating k pieces of

evidence (0pSkp1)

Sk�1 ¼ anchor or prior belief. The initial strength of belief is denoted S0

sðxkÞ ¼ subjective evaluation of the kth piece of evidence (Different

people may accord the same evidence, xk, different evaluations)

R ¼ the reference point against which the impact of the kth piece of

evidence is evaluated

wk ¼ the adjustment weight for the kth piece of evidence (0pwkp1).

The adjustment weight for the kth piece of evidence, wk, can be further

defined as

wk ¼aSk�1 when sðxkÞpR

bð1� Sk�1Þ when sðxkÞ4R

(

where

a ¼ the sensitivity toward negative evidence, and

b ¼ the sensitivity toward positive evidence.

The formula implies that wk is related to the strength of the anchor

through a ‘‘contrast’’ effect such that large anchors are ‘‘hurt’’ more than

smaller ones by the same negative evidence. Thus, the magnitude of the

belief revision is proportional to the prior belief, Sk�1; for negative

evidence and proportional to the inverse of the prior belief, 1� Sk�1; forpositive evidence. The a and b variables are constants, which represent an


individual’s sensitivity toward negative and positive evidence, respec-

tively, and are posited to be functions of both individual and external

variables.

For SbS processing, the adjustment weight, wk, will depend on the sign of

the new evidence and the level of the anchor, Sk�1; as described above.

However, under an EoS processing mode, the individual makes only one

adjustment, and the model can be simplified to

Sk ¼ S0 þ wk½sðx1; :::; xkÞ � R�,

where s(x1,y, xk) is some function, possibly a weighted average of the items

that follow the anchor.

The contrast assumption makes a prediction regarding the information

being evaluated. Specifically, it predicts whether primacy, recency, or no

order effect will occur in belief revision (Table 1). Under a short series of

simple information evaluated with EoS processing, the model always

predicts primacy. Conversely, the model predicts recency for SbS processing

of mixed evidence and no effect for consistent evidence. Under a short series

of complex information (i.e., full of details), the model always predicts

recency for the evaluation of mixed evidence and no order effect for the

evaluation of consistent evidence. Finally, as more information is processed

(long series), decrements in a and b are expected, which eventually leads to

predictions of primacy.

Table 1. Order Effect Predictionsa.

Type of Evidence Response mode Mixed Consistent

EoSb SbSc EoSb SbSc

Short series

Simple Primacy Recency Primacy No effect

Complex Recency Recency No effect No effect

Long series Toward primacy Toward primacy Primacy Primacy

aPredictions assume the evaluation mode (R ¼ 0) of encoding, which is consistent with studies

in accounting (see Note 1). Predictions under the estimation mode for both mixed and

consistent evidence would be exactly the same as the above predictions for mixed evidence

under the evaluation mode.bEoS ¼ End-of-Sequence or simultaneous processing.cSbS ¼ Step-by-Step or sequential processing.


BELIEF REVISION IN ACCOUNTING

Research in accounting examining the belief-adjustment model, with the

exception of Anderson and Maletta (1999), has generally been interested in

the predicted recency effects for a short series of mixed evidence where

participants use sequential (SbS) processing. Fig. 1 shows the belief-

adjustment model’s predicted ‘‘fishtail’’ effect for this type of information.

Bayes’ theorem would dictate no order effects; however, recency has

prevailed in most short-series studies.

Information that is received in a short series and is used to adjust beliefs in

a sequential (SbS) processing mode characterizes many tax and audit tasks;

whereas, relatively longer series of information is characteristic of financial

reporting tasks. For example, prior research in tax has focused on sequential,

directed information searches of ambiguous tax law, whereby tax profes-

sionals make judgments or recommendations to clients. This type of task is

quite common and generally requires the sequential review of both positive

and negative evidence, such as court cases, legislation, and administrative

rulings. In audit, research has generally focused on the sequential

presentation of evidence related to such items as the existence of material

0 1 2 k

Time (k)

Sk: Degree of belief after k pieces of evidence, 0<Sk<1

S0: Initial belief (anchor)

+/-: Direction of evidence evaluated – (confirming or disconfirming with respect to initial belief, respectively)

S0

+

Sk

−+

−

Fig. 1. Hypothetical Recency Effects on Belief Revisions for Mixed Evidence.

Note: Recency is implied because the final position for the (+,�) order is lower than

the final position for the (�,+) order. Adapted from Ashton and Ashton (1988).


errors or fraud, going concern judgments, inventory write-downs, or internal

control evaluations. As mentioned by Trotman and Wright (2000), due to

time constraints, a short series of information is relevant since auditors

typically are not expected to obtain an extended set of evidence (e.g., over 15

items) for a particular assertion. In financial reporting, the sequential release

of a longer series of information items is characteristic of the new online

business-reporting model favored by the American Institute of Certified

Public Accountants (AICPA) and other regulatory constituents (2002).

Initial studies of the belief-adjustment model in accounting were primarily

concerned with testing the applicability of the model to a particular

accounting area (e.g., audit or tax). The initial belief revision study in

accounting by Ashton and Ashton (1988) tested the strength of the evidence

presented, the response mode, and the initial anchor. Since then, a variety of

other individual and external variables have been included as factors

important in the prediction of order effects. Recency effects consistent with

the belief-adjustment model have generally been found to be robust in the

absence of some mitigating factor.

FACTORS TESTED FOR ASSOCIATIONWITH BELIEF

REVISION EFFECTS

The factors that have been tested for association with belief revision effects

in accounting have been categorized in Table 2 using Roberts’ (1998)

economic psychology-processing (EPP) model that was developed for tax

accountants’ judgment and decision-making (JDM). The EPP model

emphasizes the cognitive decision-making process, with both individual

(internal) and economic (external) factors having varying degrees of

influence on cognitive processing for specific tasks. By using Roberts’

model as a framework, we begin with a solid foundation of the factors which

have been examined in prior literature, and we have an organization that

follows Hogarth and Einhorn’s (1992) suggestion to consider the influence

of individual and external factors on belief revision. While Roberts’ model

was created to better understand tax JDM, the categories are not tax

specific. The factors within each group (e.g., experience and task complex-

ity) also have been examined in audit JDM and, more specifically, the

overall belief revision process.

From a broad perspective, the EPP model is similar to Gibbins’ (1984)

model of professional judgment in public accounting, which is derived from


general learning models and includes the person, stimulus, environment,

decision, and judgment process. Comparable with Gibbins’ model, Roberts’

(1998) model contains five groups including individual, external, and

processing categories. We chose to use Roberts’ model because it takes a

more detailed approach than Gibbins. Specifically, Roberts identifies actual

factors that have been examined in prior JDM literature based upon his

review of 52 tax JDM papers. Roberts’ five groups include the following:

1. Individual factors: internal characteristics unique to the decision-maker.

These include cognitive factors, such as knowledge, as well as affective

factors, such as ethical attitudes or risk preferences.

2. Environmental factors: economic risks and rewards associated with the

decision-making environment. These include external influences such as

the incentives or pressures due to the professional role (e.g., tax versus

audit), client preference pressures, or the inherent risk in the environment

(e.g., fraud detection).

3. Task input factors: external characteristics of the decision-making task,

which may impact how information is processed. These include items

such as information load and task complexity.

Table 2. Categorization of Factors Tested for Association with

Belief Revision.

Individual psychological factors Input task factors

Cognitive Strength of evidence

Years of experience Type of evidence (consistent or mixed)

Style Response mode (SbS vs. EoS)

Task relevant knowledge Task complexity

Number of cues

Affective

Initial beliefs Processing factors

Sensitivity toward evidence Processing Mode (SbS vs. EoS)

Professional attitudes Accountability (cognitive effort)

Documentation (cognitive effort)

Environmental factors Review process

Professional roles Control over evidence order

Inherent risk Group discussion

Experimental markets Time pressure (cognitive effort)

Client characteristics

Economic benefit to firm Output task factors

Likelihood judgment

Tax decision (e.g., recommendation to client)

Audit decision (e.g., going concern)


4. Decision-processing factors: strategies used by decision makers to simplify

the judgment process. These strategies generally have been found to

influence cognitive effort through such pressures as accountability,

decision aids, or the review process.

5. Task output factors: type of decision required. For instance, consulting

and reviewing tasks are expected to influence the decision-making process

differently.

Similar to the broader JDM literature, the extant belief revision literature

has examined a multitude of variables and their effects on belief revision.

However, until now, there has been no clear, overall organization of these

variables. Using Roberts’ (1998) groupings will allow future research to

more directly draw conclusions about any general effects of each group and

potential interactions between the groups of factors. A solid organizational

framework will allow us to identify which areas are in need of additional

research. It also will enable researchers to better ensure that they have

considered (or controlled for) all possible influences on the belief revision

process. Finally, this categorization may allow future research to examine

the belief revision process as a whole by simultaneously examining factors

from each fundamental group mentioned earlier.

Table 3 provides chronologically a list of belief revision studies in

accounting and highlights the factors listed in Table 2. Specifically, of the 25

accounting studies listed, 19 have used the SbS processing mode. While a

few studies have used the EoS mode or consistent information, these have

generally been examined in addition to the SbS mode and mixed

information for comparison purposes. All but one of the studies have used

a short series of information (between 4 and 10 cues), with 4 cues being the

most common choice. These studies, along with the related findings, will be

discussed as they relate to the following sections.

Individual Psychological Factors

Individual psychological factors include both cognitive and affective factors.

Sixteen of the 25 studies listed examine, directly or indirectly, at least one

individual psychological factor. Studies of individual factors in accounting

have primarily focused on experience, a cognitive factor, and sensitivity

toward evidence, an affective factor. Task-relevant knowledge and

individual attitudes, including prior beliefs, have also been examined.


Table 3. Belief Revision Studies In Accounting.

Study Task Initial Belief Measurement of

Belief Revision

Order of Evidence

Manipulation

Other Factors

Manipulated

Factor

Category

Results

Ashton and

Ashton

(1988)

Likelihood judgment

regarding internal controls

over payroll and sales-

receivable

Manipulated at 0.20,

0.50, or 0.80 for

experiments 1

and 2 and 0.50

for experiment 3

Final likelihood

minus the

initial anchor

SbS and EoS

modes

4 cues

Strength /type of

evidence

Initial anchor

Response mode

3 No order effects for consistent

information. Recency effects found for

mixed evidence. Less extreme belief

revisions with EoS than with SbS

1

3

Butt and

Campbell

(1989)

Likelihood that internal

controls would detect

material error. Rate the

relevance of each piece of

information

Manipulated to be

positive or

negative

Only final

likelihood was

used

EoS mode (after

each set of 5

cues)

10 total cues

Prior beliefs

(manipulated)

Hypothesis-testing

strategy

(manipulated)

Prof. Roles

1 Subjects with high (low) prior beliefs

showed no (a marginally significant)

recency effect. Unless specifically

instructed to do so, auditors do not use

a confirming strategy

Ashton and

Ashton

(1990)

Likelihood judgment

regarding internal controls

over payroll, accounts

receivable, health, and

marbles

Manipulated at 0.50

for all four

experiments

Final likelihood

minus the

initial anchor

SbS and EoS

modes

4 cues

Response mode

Cue type

3 Auditors were more responsive to

disconfirming than confirming

evidence & revised their beliefs to a

greater extent under SbS than EoS.

Business executives did not exhibit

same effects in non-audit tasks

3

Pei et al.

(1990)

Likelihood treatment as a

dealer (investor) would be

upheld in court;

Recommendation to client

Not measured Final likelihood SbS mode

4 cuesTax professional

attitudes

Client preference

Judgment vs.

choice

1 Both the likelihood and recommendation

indicate recency effects. Client

preference does not affect likelihood or

recommendation. Individual attitudes

affect only recommendations

3

5

Christian and

Reneau

(1990)

Likelihood of excludability of

fellowships from taxable

income; Employee vs.

Independent contractor;

Deductibility of travel

expenses

Manipulated at 20

and 80%

(experiment 1).

Set at 50% for

other experiments

Final likelihood

minus initial

position

SbS mode

4 cues

Response mode-

SbS vs EoS-

with consistent

information

3 Recency for students and professionals in

SbS processing of mixed evidence.

Students (but not experienced

participants) in EoS mode had larger

revisions than those in SbS mode with

consistent information

Asare (1992) Likelihood that a firm will

continue (fail) and going

concern report decision

Measured after

review of

background

information

Final likelihood

minus initial

likelihood

SbS mode

4 cuesJudgment vs.

choice

Initial beliefs

measured

Prof. Roles

5 Recency effects found in likelihood

judgments and manifested in the

subjects’ opinion choice. Hypothesis

frame did not affect the existence of

recency effects

1

2

Belief

Revisio

nin

Acco

untin

g13

Pei et al.

(1992a)

Likelihood that dealer

(investor) treatment would

prevail if audited

Not measured Final likelihood SbS mode

4 cues

Experience

Client preference

1 Experienced professionals’ belief revisions

are affected by presentation order but

not client preference. Inexperienced

professionals display a reverse pattern

2

Kennedy

(1993)

Likelihood that a firm with

fail

Manipulated at 50% Final likelihood

judgment

EoS response

mode/SbS

processing

mode

8 cues

Accountability

Task complexity

(via subjects)

4 Recency effects found in complex tasks.

Accountability mitigated recency1, 3

Krull et al.

(1993)

Probability of need for a write

down of inventory

Not measured Final likelihood

judgment

SbS mode

4 cuesPresence of fraud

risk factors

Experience

2 Experience accentuated recency effects.

Fraud factors did not affect recency1

McMillan and

White

(1993)

Select hypothesis frame for

fluctuation in financial

statement ratio and

indicate likelihood of that

hypothesis

Measured after

selection of

hypothesis frame

1. ‘‘Absolute’’

final

likelihood

minus final

likelihood

2. ‘‘Relative’’

change in

likelihood

EoS mode

4 cues Hypothesis frame

Experience

Prof. Roles

1 Auditors react differently to evidence

depending on favored hypothesis

frame and belief extremity. Auditors

were more responsive to disconfirming

evidence than confirming evidence

when belief revision was measured on

an absolute scale, but not when relative

change in belief revision was measured

on a proportional scale

2

Messier and

Tubbs

(1994)

Probability of collection of an

account receivable; Review

evidence of subordinate

and make likelihood

judgment; Assign strengths

to evidence

Measured after

background

information

Final likelihood

minus initial

likelihood

EoS mode

4 cuesExperience

Review

1 Experience mitigates the recency effect.

Weak support that experience will

interact with the review process in

predicting recency for the review of a

subordinate’s work containing recency

4

Chan (1995) Likelihood the accounts

receivable balance is fairly

stated

Manipulated at 50 Final likelihood

minus initial

anchor

SbS mode

4 cues

Cognitive style 1 Significant interaction between cognitive

style and recency effects. Field

dependent auditors show greater

recency effects than field independent

auditors

Johnson

(1995)

Probability of need for a

write-down of inventory

Measured after

review of

background

information

Final probability

minus initial

probability

SbS mode

4 cuesGroup vs.

individuals

Age

Experience

4 Group-assisted final beliefs represented a

choice shift toward risk. Recency

found; age, but not years of experience,

significantly related to more

conservative final beliefs

1

1

Table 3. (Continued)


Belief Revision

Order of Evidence

Manipulation

Other Factors

Manipulated

Factor

Category

ResultsJE

NNIF

ER

KAHLEET

AL.

14

Cushing and

Ahlawat

(1996)

Likelihood that a firm will

continue and going

concern report decision

Measured after

background

information

Final likelihood

minus initial

likelihood

SbS mode

4 cues

Documentation

requirement

Judgment vs.

choice

4 Recency effect in the audit judgments of

no-documentation subjects, but not

among subjects who performed the

documentation task. No recency effect

on decision for type of report to issue

5

Hite and

Stock

(1996)

Probability of employee vs.

independent contractor;

Rate the strength of each

piece of evidence

Measured after

background

information

Final likelihood

minus initial

likelihood

SbS and EoS

modes

4 cues

Response mode

prior beliefs

3 Recency effects for SbS mode, even when

subjects have internalized prior beliefs.

Taxpayers were more sensitive to

negative than positive information.

Assigning (rather than measuring)

prior beliefs can lead to incorrect

inferences

1

Trotman and

Wright

(1996)

Likelihood that a firm will

continue; Assess the

strength of internal

controls

Measured after

background

information

Final likelihood

minus initial

likelihood

SbS and EoS

4 cuesResponse mode

Experience

3 Experienced managers do not display

recency effects. Students exhibit

recency for both tasks and both

response modes. Seniors exhibit

recency in all cases except EoS mode

for control evaluation task

1

Bamber et al.

(1997)

Likelihood that Inventory

fraud exists and likelihood

of collecting account

receivable; Rate each piece

of evidence

Manipulated at 50% Final likelihood

minus the

initial anchor

SbS mode

2 cuesExperience

Task (within subj)

1 Recency supported. Confirmation

proneness holds over both experience

levels and both contexts. Auditors are

also more sensitive to evidence when

the possibility of fraud exists

3

Tuttle et al.

(1997)

Stock trades in experimental

market (revise security

price beliefs)

Measured as initial

security price

Final price minus

initial price

SbS mode

4 cues

Experimental

market setting

2 Recency effect was found in experimental

asset markets

Ahlawat

(1999)

Likelihood of viability of

client (going concern)

Measured after

background

information

Final likelihood

minus initial

likelihood

SbS mode

6 cues

Group vs.

individual

accuracy,

confidence, and

severity of

recency

4 Recency found in individual auditors, but

not groups. Group memory was more

accurate and groups were more

confident than individuals

Anderson and

Maletta

(1999)

Likelihood of a material error

in sales/receivables and

number of planned audit

hours

Measured after

review of

background

information

Final likelihood

minus initial

likelihood

SbS mode

4 cuesInherent risk

Proportional vs.

Absolute

Measure

2 Auditors are susceptible to primacy effects

in settings that are relatively low in

inherent risk

1

Mayper et al.

(1999)

Likelihood that court will

uphold deduction of

worthless security; Final

recommendation to client

Measured after

background

information

Final likelihood

judgment with

initial

likelihood as a

covariate

EoS mode

4 cues

Staff conclusion 4 Order of arguments did not affect

reviewers’ likelihood assessments or

final recommendation. Staff conclusion

did affect likelihood assessment

Belief

Revisio

nin

Acco

untin

g15

Arnold et al.

(2000)

Likelihood audit would result

in going concern opinion;

insolvency

Measured after

review of

background

information

Final belief

revision minus

2nd belief

revision

compared to

2nd belief

revision minus

initial belief

Eos mode

10 cues

(experiment 1)

12 cues

(experiment 2)

Task complexity

Going concern vs,

insolvency

Experience

3 Experience did not mitigate order/recency

effects under conditions of heavy

information load

2

1

Cuccia and

McGill

(2000)

Likelihood that a court would

find a settlement

excludable from taxable

income; Reporting

recommendation

Measured after

background

information

Final likelihood

with initial

likelihood as

covariate

EoS mode

4 cues

Control over rder

Pre-evaluation

information

Task knowledge

Client advocacy

4 The ability to control the order in which

conflicting evidence is evaluated

eliminates recency, but only in a

familiar task. Advocacy is not related

to choice of order. Mere awareness of

conflicting information does not

mitigate recency

3

1

2

Monroe and

Ng (2000)

Likelihood that there were

material errors or

misstatements in the

client’s financials, in the

absence of any specific

internal controls

Measured after

review of

background

information

Final likelihood

minus initial

likelihood

Sbs mode

4 cues

Inherent risk 2 No recency. Auditors’ judgments were not

influenced by order effects. Judgments

of inherent risk may be biased toward

conservatism

Pinsker (2004) Stock price valuations Manipulated at $50 Midpoint price

minus initial

anchor

Final price minus

midpoint

Sbs and EoS

modes

20 cues

Response mode 3 Recency found for both modes, but

stronger for SbS. Practically, valuation

was more pronounced for SbS

conditions

EoS, End-of-Sequence (simultaneous); SbS, Step-by-Step (sequential).

Table 3. (Continued)


Belief Revision

Order of Evidence

Manipulation

Other Factors

Manipulated

Factor

Category

ResultsJE

NNIF

ER

KAHLEET

AL.

16

Cognitive-Experience

A large amount of accounting literature has found support for the idea that

a decision-maker’s experience has an effect on his or her judgment,

particularly for complex tasks (e.g., Davis & Solomon, 1989; Church, 1990;

Libby & Luft, 1993). Specifically, society expects more experienced

professionals to make higher quality judgments. Arguably, this should

occur because experienced decision-makers possess richer knowledge and

memory structures, as well as higher levels of confidence, which leads to

lower levels of sensitivity toward evidence received (Trotman & Wright,

2000). Based on this prior research, studies in belief revision posit that

greater order effects (lower performance) should occur among less

experienced individuals. However, findings in the belief revision literature

generally have not supported this view.

Experience in belief revision has been studied by examining the number of

years of experience (Pei et al., 1992a; Krull et al., 1993), age (Johnson, 1995),

experience titles (McMillan & White, 1993; Messier & Tubbs, 1994;

Trotman & Wright, 1996; Bamber et al., 1997), and task-related

experience/knowledge (Kennedy, 1993; Arnold et al., 2000; Cuccia &

McGill, 2000). Task-related experience, however, may be confounded with

input task factors. For example, an individual who has less experience with a

task is likely to consider the task more complex. Further, Trotman and

Wright (2000) argue that besides the task, experience is likely to have an

interactive effect with response mode and amount of information.

Measuring any single factor in isolation may be a cause of the lack of

consistent results supporting recency effects in experience studies. The

studies that examined task-related experience with task complexity will be

discussed later in conjunction with the discussion of input task factors.

Pei et al. (1992a) examined the impact of tax professionals’ experience on

belief revisions about ambiguous tax treatments. They found a significant

recency effect for experienced (5–13 years) tax managers, but not for

inexperienced (2–4 years) tax managers. One explanation for this result is

that a second manipulated factor, client preference, may have dominated the

results for the inexperienced participants. (Client preference is an environ-

mental factor and will be discussed in greater detail in the section

‘‘Environmental Factors.’’)

Krull et al. (1993) found similar results, whereby more experienced audit

managers exhibited greater recency effects than less experienced managers.

They, however, expected these somewhat counter-intuitive results. They

provided their participants with little background information, leading them

to predict that experienced participants would be more likely to recognize


the inadequacy of this information and place greater weight on the

subsequent information received.

Asare (1992) investigated order effects with different hypothesis frames

(viable or failing audit client) using audit partners and managers. He used

four pieces of evidence (two contrary and two mitigating) in both frames.

Significant recency order effects were found for both frames for these

experienced auditors.

Messier and Tubbs (1994) contend that the magnitude of the contrast

effect could be reduced if the participants had additional experience with the

stimuli. Consistent with this reasoning, they found that experience mitigated

recency effects. Only weak support was found for the proposition that

experience and the review process interacted to reduce recency effects. The

inconsistencies were arguably due to their use of different experience levels

(seniors and managers), task differences (more versus less background

information), and response mode (SbS versus EoS).

While not specifically hypothesized, McMillan and White (1993) found

that experience level (staff, seniors, and manager/partners) had no signi-

ficant effect on belief revisions for auditors after they evaluated confirming

or disconfirming evidence relative to a chosen hypothesis frame. Similarly,

Bamber et al. (1997) examined experience as it related to evidence sensi-

tivity. They concluded that inexperienced auditors (staff) were not more

sensitive to confirming evidence than experienced auditors (seniors).

Johnson (1995) examined two measures of experience: years on the job

and age of the participant. Age, but not years of experience, was signifi-

cantly associated with more conservative final beliefs. Despite the conser-

vatism related to age effect, recency did exist in an inventory write-down

task.

In sum, the majority of studies find either no recency effect for experi-

enced participants or more of a recency effect with experience. Therefore, we

conclude that the evidence suggests that experience does not necessarily

mitigate recency. However, task and response mode differences, the use of

different experience manipulations, and interactions between individual and

external factors make interpretation of the results across studies difficult.

Future research should continue to consider experience as an important

cognitive factor and seek to determine if experience only fails to mitigate

recency bias or actually increases recency bias.

Cognitive–Style

One accounting study manipulated cognitive style to find out if there was an

interaction with order effects. Chan (1995) had professional auditors assess


the likelihood that the accounts receivable balance of a fictional company

was fairly stated. Using the Group Embedded Figure Test (GEFT), he

separated auditors into field-dependent (those that have ‘‘spectator’’

learning approaches and are not able to separate items of information

from their contexts) and field-independent (those who have ‘‘active’’

learning approaches and tend to differentiate information into well-

distinguished components). Results indicated a significant interaction

between cognitive style and order effects, as the field-dependent auditors

showed greater recency effects than their field-independent counterparts.

These results add credence to understanding the learning and processing

behaviors of decision-makers and perhaps even training them to be more

active in learning about task procedures.

Affective–Sensitivity to Evidence

Hogarth and Einhorn (1992) discuss the importance of an individual’s

sensitivity to negative and positive evidence. These parameters are often

described as sensitivity toward disconfirming and confirming evidence,

respectively. Confirming or disconfirming evidence is determined with

respect to a professional’s initial hypothesis. Similarly, an individual who is

sensitive to negative or positive evidence may be paralleled to someone who

is an advocate or a skeptic, respectively. That is, the individual views

information as positive or negative with respect to the client-favored

position.

Sensitivity is a critical element of the belief-adjustment model. It is posited

to be a constant for each individual decision-maker. The contrast effect,

which determines whether there will be an order effect, is diluted as one’s

sensitivity to evidence decreases. Sensitivity, as posited by Hogarth and

Einhorn (1992), is a function of both individual factors (e.g., experience) and

external factors (e.g., client preference).

The model suggests that in the absence of prior information or biases,

people will be highly sensitive to evidence, but as they become more

committed to their beliefs, as with a long series of information, sensitivity

will decline. For example, as discussed by Messier and Tubbs (1994),

individuals with more experience (an individual factor) should be more

confident in an initial impression of a problem. This higher confidence

should lead the individual to be less sensitive toward confirming and

disconfirming evidence, resulting in a smaller recency effect. Messier and

Tubbs’ research supports this theory. However, Krull et al. (1993) found

opposite results for their experienced versus inexperienced participants.

They suggest that differences in the ambiguity of the initial information


provided to the participants may help explain the seemingly conflicting

results. Experienced auditors may have been better able to appreciate the

inadequacy of the initial information and were more sensitive, in general, to

subsequent information.

An individual may inherently be more prone to a particular sensitivity

toward evidence, or the individual’s professional role may affect his or her

sensitivity. There is evidence from psychology as well as accounting that

individuals have a general tendency to search for or give more weight to

evidence that confirms an initial hypothesis (e.g., Church, 1990; Klayman &

Ha, 1987) and is often referred to as a confirmation bias. Conversely, some

evidence from the auditing literature supports disconfirmation bias,

arguably due to the auditor’s role requiring professional skepticism (e.g.,

Ashton & Ashton, 1990; McMillan & White, 1993). These biases affect an

individual’s weighting of evidence, and consequently, can affect the

predictions of Hogarth and Einhorn’s (1992) model. In fact, recency effects

may become negligible if the decision-maker maintains an extremely

asymmetrical sensitivity toward evidence (Pei et al., 1990).

Eight accounting studies in belief revision have looked at an individual’s

sensitivity toward evidence. These studies primarily looked at sensitivity as

it is influenced by the accountant’s professional role. For example, several

studies examined whether auditors were more sensitive to disconfirming

evidence or whether tax professionals exhibited biases related to their

professional roles as client advocates. The studies will be detailed in the

discussion for ‘‘Professional Roles’’ in the section on ‘‘Environmental

Factors.’’

Bamber et al. (1997) attempted to measure an individual’s sensitivity to

evidence. By controlling for the subjective strength of evidence and the

degree of the contrast, they examined whether auditors were more sensitive

to a particular type of evidence. Bamber et al. discussed five possible

sensitivities, including evidence neutrality, confirmation bias, disconfirma-

tion bias, positive evidence bias, and negative evidence bias. They found that

auditors were more sensitive to confirming evidence, which is consistent with

the confirmation bias findings from the psychology literature (e.g., Lord,

Ross, & Lepper, 1979; Nisbett & Ross, 1980).

Affective–Professional Attitudes

Pei et al. (1990) suggested that tax preparers’ attitudes might affect their

recommendations to clients and possibly their decision processes as well.

They explored tax professionals’ attitudes relative to (1) the relative

frequency of evasion, (2) the morality of evasive or aggressive tax reporting,


(3) the individual’s self-insight as to their aggressiveness, and (4) the

perception of the tax professional’s responsibilities to client and govern-

ment. In the study, tax professionals were asked to express a belief about the

likelihood of an ambiguous tax treatment (dealer versus investor classifica-

tion) prevailing upon an audit and to provide a recommendation to the

client. A significant recency effect was found for both the likelihood and

recommendation. Only the morality and self-insight attitudes were

associated with the client recommendation, and none were significantly

related to the likelihood judgment. Therefore, this study provides some

evidence that individual attitudes can impact a tax professional’s final

judgment.

Affective–Initial Beliefs

Accounting studies have recognized the importance of correctly measuring

the initial anchor (belief) and the participants’ belief revisions. The belief-

adjustment model relies heavily on prior beliefs as an input into the model.

Prior beliefs provide an anchor from which decision-makers adjust when

presented with additional evidence (Butt & Campbell, 1989).

Originally, many studies used background facts to manipulate a

participant’s initial belief. Ashton and Ashton (1988) manipulated initial

belief by describing the likelihood that controls would prevent or detect

material error to be 20, 50, or 80%. They found support for Hogarth and

Einhorn’s (1992) contention that smaller (larger) anchors are helped (hurt)

more by positive (negative) evidence than larger (smaller) anchors. Many

later studies (Ashton & Ashton, 1990; Kennedy, 1993; Bamber et al., 1997;

Pinsker, 2004) assigned an initial belief to the participants in order to

provide a consistent starting point for measuring recency.

Rather than assigning an initial belief, Butt and Campbell (1989)

recognized that having participants come to their own conclusions about

the initial belief gave more assurance that the researcher had actually

captured their priors. They measured participants’ initial beliefs after

presenting them with background information designed to induce either low

or high priors, and found recency effects only for participants with induced

low prior beliefs. Asare (1992) recognized that there also could be possible

source credibility issues in manipulating initial beliefs and similarly,

measured participants’ initial beliefs on likelihood scales. Unlike Butt and

Campbell, Asare found recency effects in both groups examined.

Hite and Stock (1996) found significant differences between belief

revisions for participants receiving assigned prior beliefs and those whose

prior beliefs were measured. They theorized that participants do not


internalize an assigned prior belief, therefore, the mean belief revision for

these participants may include measurement error. Hite and Stock

concluded that assigned prior beliefs are appropriate only when: (1) actual

priors are close to those assigned, (2) participants have no prior beliefs, or

(3) participants have weak internal priors.

McMillan and White (1993) recognized the possible scale effects that

could result from the strength of the initial belief in measuring a

participant’s belief revision. They pointed out, for example, that if a

participant’s initial likelihood score was 70% and the participant evaluates

confirming evidence, a 30-point range is available for upward revision.

Conversely, if the participant evaluates disconfirming evidence, a 70-point

range is available for downward revision. To control for scale effects, they

used a proportional measure for the participants’ belief revisions in addition

to the usual absolute measure. They found that the two measures of belief

revision do not provide consistent results for the effect of evidence direction

on the magnitude of belief revision. However, Anderson and Maletta (1999)

found that the scaled (proportional) measure of belief revision is consistent

with the unscaled (absolute) measure in their study of primacy effects.

With the exception of the last two studies mentioned above, the vast

majority of the accounting studies have used the final judgment minus the

beginning judgment as the measurement of belief revision. However, it is

important for future research to recognize the potential limitation of an

absolute measure due to the scale effects discussed above. Furthermore, a

comparison of the proportional measure to the belief-adjustment model’s

measurement of belief revision (as used by Bamber et al., 1997) reveals that

both have a similar method for examining belief revision processes when the

strength of evidence is constant for all evidence.

In general, the research to date suggests that individual psychological

factors, such as experience, fail to mitigate recency bias. However, limited

research concerning cognitive learning style shows active learning approaches

may mitigate some recency bias. Affective factors such as sensitivity to

evidence, professional attitude, and initial belief have been shown to affect

final judgments either directly or indirectly. Both confirmation and

disconfirmation biases have been documented with respect to sensitivity to

evidence and professional role. Initial beliefs have been shown to effect belief

revision differentially depending on whether the initial belief has been

assigned or measured. With respect to cognitive and affective factors, limited

research demonstrates that experience does not necessarily reduce sensitivity

to evidence. Studies have provided conflicting results, which indicates the need

for future research in the area of individual affective psychological factors.


Environmental Factors

As previously discussed, Hogarth and Einhorn’s (1992) model outlines a

framework explaining how external variables and processing factors interact

to produce order effects in belief updating. There are two types of external

variables – environmental and input task factors. Environmental factors

concern risks and rewards that are present in the decision-maker’s

environment. This may be due to the role of the decision-maker, the

particular characteristics of the client she/he is serving, or the inherent risk

present in the decision environment. Decision-makers do not operate in a

vacuum, and cognitive models should reflect the potential influence of items

such as the client, regulatory agencies, and professionals’ employers

(Roberts, 1998). Studies which have examined these environmental factors

are discussed below. Input task factors are described and discussed in the

following section.

Professional Roles

Hogarth and Einhorn (1992) suggest that an individual’s professional role

might produce differential ‘‘sensitivities’’ to evidence, and they discuss how

these differential sensitivities are considered to be advocate or skeptic

attitudes. Within each decision context, accounting professionals face

pressures that may cause them to be differentially sensitive to positive and

negative evidence. Accountants are charged with professional skepticism in

audit and assurance services (AICPA, 1995, AU, 316.16). On the other

hand, tax professionals have an obligation to act as a client advocate and to

interpret the law in the client’s best interests (AICPA, 1995). As discussed by

Cuccia and McGill (2000), the tax professional is a goal-oriented decision-

maker with an incentive to propose and defend client-favored propositions.

All accounting professionals must maintain objectivity in their professional

judgments (AICPA, 1995) and face litigation risks in both audit and tax for

non-objectivity. (Tax professionals face penalties for overly aggressive

positions under Code Section 6694.) Thus, the role of the accountant should

influence the process of belief revision. Seven studies in addition to the

Bamber et al. (1997) study mentioned previously, examine professional roles

and their influence on sensitivity to evidence in the belief-adjustment

process. The studies that have examined auditor belief revision will be

discussed first followed by the studies that have examined tax professionals’

belief revision.

Ashton and Ashton (1990) used both auditors and business executives for

auditing and non-auditing tasks. Their results indicated auditors were more


sensitive to disconfirming than confirming evidence, but business executives

were statistically indifferent.

Butt and Campbell (1989) had participants use a confirming, disconfirm-

ing, or neutral strategy to evaluate the strength of an internal control

system. They found recency effects with participants who had low initial

beliefs. However, they found no support for the use of a confirming strategy,

and concluded that auditors were more naturally disconfirming.

Asare (1992) manipulated the participants’ hypothesis frame to investi-

gate the effects of confirmation bias and order effects. He had participants

make judgments about the failure (viability) of a firm and choose the type of

opinion they would issue. In support of the belief-adjustment model, he

found that recency exists in both belief revisions and audit report choices.

Hypothesis frame did not affect the existence of recency, and thus,

experienced auditors did not exhibit a confirmatory strategy.

McMillan and White (1993) found that auditors’ belief revisions were

significantly larger for disconfirming evidence than for confirming evidence

when the absolute measure is used (similar to Ashton & Ashton, 1988,

1990), but not significantly different when the proportional measure is used.

This raises issues about whether the absolute measure, which has been used

by most studies in belief revision, is the most appropriate way to measure

belief revisions.

Overall, studies examining audit professionals tend to find that auditors

are more sensitive to disconfirming evidence. This is not surprising in light

of their role and their requirement to exhibit professional skepticism.

Conversely, Christian and Reneau (1990) found that professional tax

participants were much more influenced by evidence supporting than by

evidence opposing the client’s preferred treatment. This was contrary to the

equal sensitivity displayed by student participants in the same ambiguous

tax judgment case. Christian and Reneau suggested that the professionals’

asymmetric sensitivity may be due in part to the client advocacy relationship

between tax professionals and their clients. Consistent with the idea that the

advocacy role of tax professionals may influence belief revision, Pei et al.

(1992a) found that the client preference induced an attention directing effect

such that the tax professional was more sensitive to evidence favoring the

client’s preferred treatment.

Cuccia and McGill (2000) included a measure of client advocacy in their

experiment examining tax professionals’ judgments of an ambiguous tax

issue, and the potential mitigating influence of control over order of

evidence evaluation. They found that 57% of participants chose to examine

positive evidence first, but that client advocacy was not related to that


choice. Thus, confirmation proneness was ruled out as an alternative

explanation to the mitigating influence of control.

Tax professionals appear to be driven by the client preference in their

judgments, which seems reasonable given their charge to be client advocates.

Importantly, these studies appear to find differences in belief revision

between auditors and tax professionals, indicating the important influence

of professional roles on sensitivity to evidence. The evidence regarding tax

professionals reflects a tendency toward a client advocacy in most cases,

rather than the tendency to be biased toward disconfirming evidence as

shown by auditors. Future research should also examine whether profes-

sional role effects are dependent upon the particular decision environment

(audit or tax task).

Inherent Risk

Four studies have examined audit fraud or inherent risk as an environ-

mental factor that can influence the robustness of the belief-adjustment

model. Krull et al. (1993) suggested that perceptions of very high (low) risk

of fraud factors should foster heightened (lessened) sensitivity to evidence

suggestive that fraud exists. Results indicated that the presence or absence of

fraud signals did not mitigate or accentuate order effects. Thus, auditors’

sensitivity to information may not be differentially affected by the perceived

risk of fraud. However, this is based on the effect of fraud factors on the

belief revision. They note the limitation that no attempt was made to

separately measure the effect of the fraud factors on sensitivity to evidence.

Bamber et al. (1997) investigated whether auditors exhibited different

sensitivity toward evidence when the initial hypothesis involved a possible

irregularity. They measured sensitivity to evidence for a non-fraud case and

a fraud case and found that auditors’ attitudes changed in response to the

audit context. Auditors exhibited a heightened sensitivity (conservatism)

when confronted with the possibility of fraud.

In a somewhat atypical belief revision study in accounting, Anderson and

Maletta (1999) examined auditors’ susceptibility to primacy. They predicted

that primacy would result when there was a low-risk audit environment and

an information sequence with late positive information. Their argument was

based on prior psychology research, which states that primacy is a function

of diminishing cognitive effort in the evaluation of a sequence of

information, and that cognitive effort depends upon the risk associated

with the decision task (Anderson, 1981; Fiske, Kinder, & Larter, 1983). In

addition, Anderson and Maletta argued that auditors, by nature, have a

heightened sensitivity to negative information, and thus, may not exhibit


diminished cognitive effort with late negative information. The results were

consistent with their predictions for primacy in a low-risk, late positive

information situation.

Finally, Monroe and Ng (2000) had auditors investigate the likelihood

there were material errors or misstatements in a fictional client’s financial

statements, in the absence of any specific, related internal controls. No

recency effects were found, as auditor judgments were not influenced by

order. Similar to Bamber et al. (1997), judgments of inherent risk may have

been biased toward conservatism, and thus outweighed recency effects. It

appears that auditors tended to be more sensitive toward disconfirming

evidence, potentially due to their role requirement to maintain professional

skepticism. In other words, they placed more weight on disconfirming

evidence regardless of presentation order. Thus, it appears that increased

inherent risk may exacerbate the tendency toward conservatism.

Experimental Markets

In a unique application of the belief-adjustment model to a more complex

setting, Tuttle, Coller, & Burton (1997) examined the basic order effect

predictions using an experimental assets market. Contrary to the efficient

market hypothesis, which states that the market impounds new information

quickly and without bias into security prices, they found the existence of

recency effects in a market setting. Their study provides support for the

robust nature of the belief-adjustment model and demonstrates that

systematic, individual biases can survive in a market setting.

Client Characteristics

Client characteristics, such as the size or importance of a client, revenue

potential, and risk preference/aggressiveness are expected to influence

accountants’ decisions. For instance, studies have found a link between

client risk preference and aggressiveness of tax recommendations (e.g.,

Schisler, 1994). While these are important environmental characteristics, we

could find no studies that specifically examined the impact of these variables

on belief revision. This is an area that future research may want to address.

Input Task Factors

Task factors are the second type of external variables discussed. Beginning

with Ashton and Ashton (1988), some of the basic predictions for task

variables were tested in accounting. As the literature has developed, belief


revision predictions have been examined under more complex environments,

including those where the individual has control over the order of evidence,

where documentation is required, where accountability is salient, where the

information load is higher, and even in an experimental market setting.

Nineteen of the 25 accounting belief revision studies cited have examined at

least one of the following task factors.

Type of Evidence and Response Mode

It may be recalled that response mode, in combination with the complexity

of the information, should dictate the processing mode an individual will

use. According to the belief-adjustment model, SbS processing must be used

with SbS response mode. However, SbS or EoS processing can be used with

EoS response mode depending on cognitive demands. Predictions of the

belief-adjustment model also depend on the type of the evidence (e.g.,

whether it is mixed or consistent).

Ashton and Ashton (1988) were the first to test Hogarth and Einhorn’s

(1992) model in an accounting environment. They used a series of five

experiments to test the basic model predictions for consistent versus mixed

information, EoS versus SbS response modes, and the strength of initial

evidence with consistent information (two task variables). No order effects

were found for consistent information, but recency was found for mixed

evidence. Results generally supported the belief-adjustment model predic-

tions of no order effects for consistent information, recency effects for mixed

evidence, and less extreme revisions with the EoS than the SbS response

mode. A later study by Ashton and Ashton (1990) using a series of two

‘‘accounting’’ studies and two ‘‘non-accounting’’ studies confirmed the

results.

Most studies in accounting have investigated the SbS mode with mixed

evidence, and have generally found the existence of recency effects. A few

studies, in addition to Ashton and Ashtons’ (1988, 1990), have specifically

looked at comparing EoS versus SbS response mode and mixed versus

consistent evidence. Consistent with the model predictions, Christian and

Reneau (1990) found that recency effects existed for both students and

experienced tax professionals when mixed evidence regarding an ambiguous

tax case is presented in a SbS mode. However, they found conflicting results

for consistent evidence. Students exhibited larger belief revisions in the EoS

mode than in the SbS mode, but experienced participants were unaffected by

response mode. Christian and Reneau pointed to the need to measure an

individual’s sensitivity to evidence before conclusions regarding response

mode can be appropriately drawn.


Hite and Stock (1996) also supported model predictions with their study

involving taxpayer judgments of an ambiguous tax issue (independent

contractor versus employee). In the participants’ evaluation of mixed

evidence, results were consistent with a recency effect in the SbS response

mode and a tendency toward primacy in the EoS mode. Similarly, Trotman

and Wright (1996) found that, for less experienced participants, recency

existed for a going concern task in either response mode, but not in the EoS

mode for a control evaluation task. However, they did not find recency

effects for more experienced participants in either response mode.

Pinsker (2004) examined stock price valuations from unsophisticated

investors given a different frequency of consistent information releases.

Continuous (SbS) conditions had significantly different valuations in the

direction of the information provided than periodic (EoS) conditions at the

midpoint evaluation of information releases, but not after all of the

information had been released. However, recency was detected after all

information items were released.

Making accurate conclusions regarding the EoS response mode is difficult

since the particular response mode does not automatically guarantee that

the EoS processing mode will be used. Overall, it appears that the SbS

processing model predictions are supported for recency in the evaluation of

mixed evidence and mixed for consistent evidence. The results of these

studies have important implications for practice in both audit and tax tasks.

In particular, when a task requires SbS processing and responses, a recency

bias may be inherent. Structuring the review process as EoS may be helpful

in alleviating some of the bias introduced during the evidence collection

stage. Future research should consider task processing characteristics and

examine task structure as a means of potentially mitigating the recency bias.

Task Complexity

As previously discussed, predictions regarding task complexity are often

commingled with experience predictions. Hogarth and Einhorn (1992)

define complexity as a function of (1) the amount of information that needs

to be processed (task factor), as well as (2) the lack of familiarity with the

task (individual factor). Familiarity with a task can overcome task

characteristics and render an inherently complex task to be one that is

relatively ‘‘simple’’ for an experienced person (Trotman & Wright, 1996).

Although the belief-adjustment model predicts recency effects for the SbS

mode regardless of task complexity, several authors (e.g., Kennedy, 1993;

Trotman & Wright, 1996; Arnold et al., 2000) have suggested that

complexity or experience may mitigate the effect.


Kennedy (1993) explored task complexity by using two groups of

participants; those who should be unfamiliar (M.B.A. students) and those

who should be familiar (auditors) with a going concern judgment task. She

suggested that task complexity can affect the processing mode used by an

individual. When the task is complex, individuals are likely to use a SbS

processing strategy to ease the cognitive demands of the task. Conversely,

easier tasks should elicit an EoS processing strategy and be less susceptible

to recency effects. She found evidence consistent with EoS processing and

no recency effects for auditors. Consistent with expectations, she did find

recency effects for the MBA students.

Similarly, Trotman and Wright (1996) examined task complexity and

experience, suggesting that the interaction of these variables was likely to be

instrumental in determining whether a task was ‘‘simple’’ or ‘‘complex’’ for

an individual to perform. They considered the judgments of students,

seniors, and managers in both a going concern case and an internal control

evaluation case. Trotman and Wright expected that the familiarity of each

task would depend on the participants’ experience level, and that

participants performing more familiar (simpler) tasks would exhibit less

recency. Their results support the predicted experience effect. A recency

effect was found for students and seniors for both tasks. The use of EoS

response mode mitigated this effect for seniors in the more familiar task. In

contrast, managers did not exhibit recency effects for either task.

One of the more recent studies to include the effect of task experience on

belief revision is Cuccia and McGill (2000). They examined tax profes-

sionals’ likelihood judgments in a familiar versus an unfamiliar task. Similar

to Trotman and Wright (1996), Cuccia and McGill found that if an

individual has control over the order in which evidence is examined, recency

is mitigated in a familiar task, but not in an unfamiliar task.

Arnold et al. (2000) took a different approach to task complexity. They

operationalized complexity as high information load (number of cues). They

argued that even given a high level of task familiarity, a task could still be

considered complex within the definition set forth by the belief-adjustment

model if the information load received by the decision-maker was high (a

condition consistent with many accounting tasks). Under both going

concern and insolvency experiments, Arnold et al. found experience did not

mitigate recency under conditions of high information load. This result

raises some questions as to the methodology for measuring task complexity

in future research. Specifically, it would appear that an interaction of

response mode, task familiarity/experience, and information load is present

and needs to be considered simultaneously in future studies.


Environmental factors and input task factors are the external factors that

influence the belief revision process of accounting professionals. As

discussed above, research has found that these factors influence how

individuals interpret information and ultimately make decisions. The

specific environment created by items such as client factors, the decision-

maker’s role, and inherent risk in the decision-maker’s environment can

differ for each decision-maker. To completely understand the belief revision

process, researchers must account for the effects of these factors. By

grouping the factors according to Roberts’ classifications, we separate these

external factors from those that are internal (unique characteristics of the

specific decision-maker). Thus, we can more appropriately identify what

factors may be able to be externally controlled (e.g., choosing more

conservative clients) versus those that may be need to be internally

controlled (e.g., gaining more experience or knowledge with an issue).

Processing Factors

Seven studies have attempted to identify factors that may mitigate or

eliminate the recency bias. Aside from the mixed results for experience as a

mitigating factor, other potential mitigating factors that are particularly

meaningful to accounting have been examined. These factors include

accountability, required documentation, the review process, use of groups,

and control over the order of evidence evaluation.

Kennedy (1993) examined whether accountability, defined as the

requirement to justify one’s judgments to others, mitigates recency.

Generally, accountability is thought to have the ability to influence the

information attended to, the complexity and type of information processing,

and ultimately the decision or judgment made. Kennedy suggested that

accountability may induce individuals who resort to effort-saving strategies,

such as SbS processing, to supply the requisite effort for EoS processing,

thereby overcoming recency. She asked MBA students and audit managers

to make going concern judgments using EoS response mode. MBA students

who were less familiar with the task exhibited recency effects. However,

when accountability was imposed, no recency effects were found, suggesting

that accountability mitigates recency. Audit managers who were familiar

with the task were not expected to, and did not, exhibit recency under either

condition.

Using similar logic (e.g., that greater effort will mitigate recency), Cushing

and Ahlawat (1996) examined documentation as a potential mitigating


factor. Documentation can be viewed as a processing factor eliciting greater

cognitive involvement, similar to accountability. Cushing and Ahlawat

asked auditors to draft a memorandum to the audit partner providing

reasons supporting the recommended audit opinion. The documentation

task was expected to induce greater effort resulting in better information

recall and comprehension, greater levels of attention to the task, and

improved problem-solving representation. All of these factors should

contribute to a process that did not place undue weight on information

received most recently. As predicted, a recency effect was present in the

audit judgments of no-documentation participants, but disappeared among

those who performed the documentation task.

Two studies have examined how the review process affects the predictions

of the belief-adjustment model. Messier and Tubbs (1994) proposed that a

review responsibility produces an averaging process when integrating the

beliefs of one’s own with a subordinate. They predicted that experience and

the review process would interact in predicting recency effects when an

auditor reviewed a subordinate’s judgment that contained a recency effect.

However, their results did not support their prediction.

Mayper, Anderson, and Kilpatrick (1999) examined the impact of the

review process in a tax setting using a staff’s tax memorandum. The

memorandum included the staff’s arguments for and against the deduction

of a worthless security and their final conclusion on the judgments of a tax

manager or partner reviewer. Mayper et al. found that the order of

argument presentations in the memorandum did not produce a recency (or

primacy) effect in the reviewers’ judgments. This provides support for the

effectiveness of the review process in reducing recency effects. However,

Mayper et al. were careful to point out that experience (partner and

managers were used here) may be an alternative explanation for the

mitigated recency effect.

Two auditing studies examined whether or not use of groups in the

decision-making process would mitigate recency. Johnson (1995) investi-

gated the probability of a need for the write-down of inventory in both

group and individual decision-making contexts. Although recency was

discovered overall, information order did not result in recency for the

group-assisted condition. In fact, for the bad news–good news sequence, the

group-assisted condition actually revised their final valuation downward,

indicating a primacy effect. Additionally, the manner of the revision for the

group-assisted conditions was in the direction of greater risk. The groups

actually revised their probabilities of inventory obsolescence downward for

all information orders.


Ahlawat (1999) had auditors state their likelihood of client viability in a

going concern case using either group or individual decision-making

processes. Results indicated recency effects for individual auditors, but

not for groups. Additionally, group memory was more accurate and groups

were more confident as compared to individuals.

A final processing factor, control over the order in which evidence is

evaluated, was examined by Cuccia and McGill (2000). Control over order

is important in tasks such as the search of tax databases, which allow

organized access to selected pieces of evidence. For example, a search of tax

court cases may allow the professional to review all the positive cases first or

all of the negative cases first. As discussed by Cuccia and McGill, an

individual’s ability to process information depends on the congruence of the

information’s organization with its intended use. Allowing control over

evidence evaluation order would facilitate the use of a preferred decision

strategy, thereby reducing the complexity of the task and the susceptibility

to recency. Results of the study found ability to control the order of

evaluation of conflicting evidence eliminated recency in a familiar task.

Recency was observed only when participants had no context-relevant

knowledge or were precluded from structuring the task. Additionally, mere

knowledge (pre-evaluation information) that conflicting evidence would

exist was not sufficient to mitigate recency.

The factors discussed above examine order effects in a more complex and

realistic environment. Results indicate that accountability, documentation,

reviewing, using groups, and controlling evidence order can be effective in at

least reducing, if not eliminating, recency effects. Given that many of these

characteristics (e.g., accountability, review) are present in both audit and tax

contexts, the previously observed recency biases may be easily mitigated in a

real-world environment.

Justification requirements and accountability pressures influence the

amount of cognitive effort individuals will exert on a decision task. Thus,

research on processing factors provides further evidence that researchers

cannot look at the effect of one variable in isolation of other factors.

Inclusion of these processing factors provides researchers a path for

examining ways to overcome many effort-related decision biases. It is

important for future research to continue to consider the effects of recency

and other biases within the context of these processing factors to enable

accurate conclusions to be drawn about the true consequences to accounting

decisions.

While recent research has examined more complex environments, there

are still additional variables as recommended by Hogarth and Einhorn


(1992) that have not been previously examined. These variables include:

time pressure, temporal delays, the effects of interrupting updating tasks,

and how roles and incentives might affect differential sensitivity to negative

and positive information. These factors are relevant in an accounting

environment and deserve attention in belief revision research.

Output Task Factors

An output task factor refers to the actual decision of the participant in an

experiment (as opposed to the judgment process used to arrive at the

decision). In the accounting literature, the most common decisions

requested from participants have been likelihood judgments. Audit

participants have judged items such as the likelihood of material error in

an account, the likelihood of collection of an account, and the likelihood

that a firm will continue. Tax participants have been asked to judge the

likelihood of events such as whether a court will rule in favor of an

ambiguous expense deduction (income exclusion), whether an individual

should be considered an independent contractor, or whether a client meets

the requirements of a real estate investor or dealer. Financial participants

have judged the price for one share of stock. Although all studies have an

output task factor, three studies have explicitly discussed this factor as a

variable in their results.

Asare (1992) pointed out that there is a distinction between judgment and

choice. The observed differences in likelihood assessments do not allow an

inference that different actions always will result. Therefore, he examined

the separate effects of order presentation on likelihood judgments and

choices. In addition to asking participants to assess the likelihood that a firm

will continue, Asare also asked them to make a going concern report

decision. He found support for recency in both the judgment and the choice.

Inconsistent with Asare’s (1992) results, Cushing and Ahlawat (1996)

were not able to replicate the recency effects for the decision in a similar

going concern task. However, consistent with Asare (1992), in a tax

experiment where subjects provided the client recommendation as well as a

likelihood assessment, Pei et al. (1990) found that both dimensions exhibited

recency effects.

To date, the limited research in this area suggests that perhaps recency

effects carry over from the judgment task and influence output task factors,

such as choice. Although most researchers have been concerned with

whether presentation order affects the judgment processes of individuals, it


is often a dichotomous decision (e.g., recommendation whether or not to

deduct and expense, decision to issue a going concern opinion, etc.) that

affects the client. Thus, there is value in understanding how likelihood

judgments are manifested into actions or decisions. Future research should

examine if recency effects automatically carry over from the judgment

process to the action or if this occurs in only certain situations.

CONCLUDING REMARKS

In the absence of some mitigating factor, recency predictions of the belief-

adjustment model have generally been supported in auditing, tax, and

financial reporting tasks. Of the 25 studies we have included in this paper, 21

found recency effects for at least some combination of the factors

researchers included in their studies. This is consistent with the basic

predictions of Hogarth and Einhorn’s (1992) belief-adjustment model.

Many of these 21 studies, along with the other 4 studies, also found specific

factors that either outweighed or mitigated recency effects or created

entirely different biases (e.g., confirmation bias).

Trotman and Wright (2000) suggested that the Hogarth and Einhorn

(1992) model needs to be adapted and revised to include the uniqueness of

auditing (accounting) tasks. While the belief-adjustment model appears to

be a reasonable model for examining accountant’s belief revisions, we agree

that the model can be expanded to better fit accounting tasks. While

Hogarth and Einhorn (1992) suggest that individual and external variables

influence the belief-adjustment process, they only distinguish their overall

predictions for three task variables (complexity, length of the series of items,

and response mode) and two encoding variable (processing mode and task

type). The model does not make specific predictions about experience, client

pressures, accountability, professional roles, time pressure, the review

process, etc. These are factors that are specifically important to belief

revision in accounting and should be incorporated into the model.

The 25 studies provide clear evidence that (1) individual psychological

factors such as experience, initial beliefs, and sensitivity toward evidence;

(2) environmental factors such as individual role and inherent risk; (3) input

task factors such as type of evidence and response mode; (4) processing

factors such as processing mode and accountability; and (5) output task

factors such as likelihood judgment and recommendation (or choice), all can

impact the process by which accountants update their beliefs. The current


chapter’s application of Roberts’ (1998) framework to the belief-adjustment

model literature seems appropriate as a framework and represents a much-

needed summation of the research stream.

Prior studies have not provided a general framework from which to view

the belief revision process. By using Roberts’ (1998) categorization, we have

provided such a framework. As can be seen from the studies reviewed in the

current chapter, each of these categories of factors affects the way

accountants make decisions. More importantly, as several of the studies

indicate, the factors can interact with one another, thus, providing detailed

information about the belief revision process.

The factor categories may be thought of as an overall framework where

each category influences other categories within the model. A true

understanding of the process by which individuals revise beliefs should be

examined with all of these categories in mind. For instance, an individual’s

experience level may influence his or her initial belief (as well as subsequent

belief revisions) in a sequential decision task. The subsequent belief revisions

also may be impacted by that individual’s sensitivity to evidence, which may

first be influenced by their professional role (e.g., as an advocate or a

skeptic). Additionally, task factors, such as response mode, and processing

factors that may mitigate biases (e.g., accountability) or otherwise influence

how individuals revise their beliefs should be considered in understanding

the complete process.

As presented in Table 3, our review indicates that 17 of the studies

included one or more variables from at least two of the factor categories.

Sixteen studies examined individual psychological factors, with six of those

studies examining multiple individual factors. Ten studies examined at least

one environmental factor, while 11 studies examined at least one task factor.

Seven studies specifically investigated processing factors, and three studies

directly compared output task factors.

Most studies examined variables from either the individual or the external

factor categories. These processing factors were examined primarily to

understand whether they had any mitigating influence on the biases

associated with the belief-adjustment process (e.g., recency effects). There

is room for additional examination of processing and mitigating factors,

especially in conjunction with the other factor categories. These mitigating

factors are important in forming more realistic expectations about how

presentation order might affect accounting professionals in their true

environment. Accountability pressures, the review process, control over

order evaluation, and group discussion are all relevant factors that influence

an accounting professional’s decisions. These items should be considered in


order to more accurately generalize the findings from behavioral experi-

ments to accounting practice.

Some of the conflicting evidence found in the accounting studies,

especially in the area of experience, may be attributable to differences in

the researchers’ experimental designs. The differences include the actual task

used, response mode variations, number and length of cues used, control of

the initial anchor, and belief revision measurement. By controlling for these

differences in future studies, researchers can begin to understand better

under what conditions experience, and other factors, influence accountants’

belief revisions. Conflicting results may also be due to the moderating

influences of some factors on other factors. For example, accountability

may mitigate order effects for inexperienced accountants, but not

experienced accountants. Knowledge of the particular individual, environ-

mental, input task, processing, and output factors present in the

environment under investigation will enable more accurate conclusions to

be drawn about these variables’ effects on accountants’ decisions.

Future Directions

Based upon the research in this area, all individuals, regardless of experience

level, appear to be susceptible to biases in judgment and decision making.

Thus, we suggest rather than focusing on whether a particular bias exists,

the aim of future research should be not only to better understand the

judgment process, but more importantly to improve judgment and decision-

making.

Using our framework, we can identify areas of the belief-adjustment

model that need further research. One aspect of the belief-adjustment model

that has not been fully examined is the effect of individual and

environmental factors on an individual’s sensitivity to evidence. Prior

research has generally failed to separate this effect from the impact that

these factors have on belief revision. An individual’s sensitivity to evidence is

a large factor in whether the predicted contrast effect will occur. While

Bamber et al. (1997) measured sensitivity, they did not examine whether

greater sensitivity translates into a larger recency effect. Additionally,

examining sensitivity more closely may provide more insight into how

information is actually processed.

Related to this issue is the use of process-tracing measures to more closely

examine cue usage by individuals. If individuals use the anchoring and

adjustment heuristic, the contrast effect may result in recency effects.


However, individuals may not necessarily spend more time on these later

cues. Thus, training accountants to allocate their time spent evaluating

evidence in a particular manner may be able to mitigate recency effects.

The effects of temporal delay and interruptions are additional factors that

are particularly interesting in the accounting environment. Information in

audit, reporting, and tax tasks is often received piecemeal. Interruptions

during tasks may cause delays or pauses between the review of evidence, and

possibly reduce the effect of the contrast. As Kennedy (1993) indicated,

recency can result from an effort-related judgment bias. If recency can be

mitigated by effort-inducing strategies, it is possible that time pressure and

payment incentives may induce the requisite effort needed to mitigate

recency as well.

Much more research is needed in examining plausible methods for

mitigating recency effects. As Arnold et al. (2000), Krull et al. (1993), and

Asare (1992) found, experience alone does not mitigate recency. Strategies

are necessary to avoid potentially disastrous results (e.g., not identifying a

going concern problem when one exists (Asare, 1992)).

Previous strategies examined for mitigating recency effects have included

making group decisions, explanation, (which is related to accountability) as

well as counterexplanation. As Ahlawat (1999) found, group decision-

making helped ease cognitive load, so that effort-reducing strategies (e.g.,

recency) were not needed. Additionally, making auditors (for example)

accountable for their decisions (by writing explanations/reasons for their

decisions) did result in reduced recency effects as compared to those not held

accountable (Kennedy, 1993). However, Koonce (1992) argued that

auditors could still arrive at lower quality decisions (including order

effects), because they would not investigate the reasons why a result

occurred (i.e. counterexplanation). Psychology research has shown that the

use of counterexplanation will negate any observed explanation effect

(Anderson & Sechler, 1986). Thus, one potentially rich area for future

research in all accounting domains is to follow the psychological literature

by examining whether or not counterexplanation ‘‘recreates’’ recency effects

previously mitigated by use of accountability/explanation.

NOTES

1. ‘‘Response mode’’ and ‘‘presentation mode’’ have been used interchangeably inthe literature. Both, essentially, are referring to the same concept. Similar to Arnoldet al. (2000), we use the term ‘‘response mode’’ in this chapter to be consistent withHogarth and Einhorn’s (1992) terminology.


2. Psychology studies have extensively used estimation tasks (see Lopes, 1985,1987; Anderson, 1981). Results have generally supported Hogarth and Einhorn’s(1992) model predictions.3. The evaluation form of the model is presented (see Hogarth & Einhorn, 1992,

pp. 9–12.) The reasons, similar to those cited by Bamber et al. (1997), are that (1) inevaluation tasks, evidence is often seen as bipolar relative to the hypothesis, which issimilar to many accounting tasks; (2) previous research in accounting has tended toemploy the evaluation form; and (3) previous research in accounting supports thepredictions of the evaluation form, but has not examined the estimation form of themodel (Ashton & Ashton, 1988; Tubbs et al., 1990).

ACKNOWLEDGMENTS

This chapter has benefited from the comments and suggestions from Vicky

Arnold, the editor. We are also grateful to Rich White for his suggestions on

an earlier version of this chapter.

REFERENCES

Ahlawat, S. S. (1999). Order effects and memory for evidence in individual versus group

decision making in auditing. Journal of Behavioral Decision Making, 12, 71–88.

American Institute of Certified Public Accountants (1995). AICPA Professional Standards and

Code of Professional Conduct.

American Institute of Certified Public Accountants. (2002). The business reporting model of the

future. The CPA Letter: Education Supplement, 82, G1–G2.

Anderson, B. H., & Maletta, M. J. (1999). Primacy effects on the role of risk in auditor belief-

revision processes. Auditing: A Journal of Practice and Theory, 18, 75–89.

Anderson, C., & Sechler, E. (1986). Effects of explanation and counterexplanation on the

development and use of social theories. Journal of Personality and Social Psychology, 50,

24–34.

Anderson, M. (1988). A comparative analysis of information search and evaluation behavior of

professional and non-professional financial analysts. Accounting, Organizations and

Society, 13, 431–446.

Anderson, N. H. (1981). Foundations of information integration theory. New York, NY:

Academic Press.

Arnold, V., Collier, P., Leech, S. A., & Sutton, S. (2000). The effect of experience on order and

recency bias in decision making by professional accountants. Accounting and Finance, 40,

107–132.

Asare, S. K. (1992). The auditor’s going concern decision: Interaction of task variables and the

sequential processing of evidence. The Accounting Review, 67, 379–393.

Ashton, A. H., & Ashton, R. H. (1988). Sequential belief revision in auditing. The Accounting

Review, 63, 623–641.

Ashton, A. H., & Ashton, R. H. (1990). Evidence-responsiveness in professional judgment:

Effects of positive versus negative evidence and presentation mode. Organizational

Behavior and Human Decision Processes, 46, 1–19.


Bamber, E. M., Ramsay, R. J., & Tubbs, R. M. (1997). An examination of the descriptive

validity of the belief-adjustment model and alternative attitudes to evidence in auditing.

Accounting, Organizations and Society, 22, 249–268.

Butt, J., & Campbell, T. L. (1989). The effects of information order and hypothesis-testing

strategies on auditors’ judgments. Accounting, Organizations and Society, 14, 471–479.

Chan, M. M. K. (1995). The moderating effects of cognitive style and recency effects on the

auditors’ belief revision process. Managerial Auditing Journal, 10, 22–28.

Christian, C., & Reneau, H. (1990). Updating beliefs in legal tax research: An experimental test

of the belief revision model. Working paper, Arizona State University.

Church, B. K. (1990). Auditors’ use of confirmatory processes. Journal of Accounting Literature,

9, 81–112.

Cuccia, A. D., & McGill, G. A. (2000). The role of decision strategies in understanding

professionals’ susceptibility to judgment biases. Journal of Accounting Research, 38,

419–435.

Cushing, B. E., & Ahlawat, S. S. (1996). Mitigation of recency bias in audit judgment: The effect

of documentation. Auditing: A Journal of Practice and Theory, 15, 110–122.

Davis, J., & Solomon, I. (1989). Experience, expertise, and expert-performance research in

public accounting. Journal of Accounting Literature, 8, 150–164.

Fiske, S. T., Kinder, D. R., & Larter, W. M. (1983). The novice and the expert: Knowledge-

based strategies in political cognition. Journal of Experimental Social Psychology, 19,

381–400.

Gibbins, M. (1984). Propositions about the psychology of professional judgment in public

accounting. Journal of Accounting Research, 22, 103–125.

Hite, P. A., & Stock, T. (1996). Taxpayer belief revision and the importance of controlling for

prior beliefs. Advances in Taxation, 8, 81–99.

Hogarth, R. M., & Einhorn, H. J. (1992). Order effects in belief updating: The belief-adjustment

model. Cognitive Psychology, 24, 1–55.

Hunton, J. E., & McEwen, R. A. (1997). An assessment of the relation between analysts’

earnings forecast accuracy, motivational incentives and cognitive information search

strategy. The Accounting Review, 72, 497–515.

Johnson, E. N. (1995). Effects of information order, group assistance, and experience on

auditors’ sequential belief revision. Journal of Economic Psychology, 16, 137–160.

Kennedy, J. (1993). Debiasing audit judgment with accountability: A framework and

experimental results. Journal of Accounting Research, 31, 231–245.

Klayman, J., & Ha, Y. (1987). Confirmation, disconfirmation, and information in hypothesis

testing. Psychological Review, 94, 211–228.

Koonce, L. (1992). Explanation and counterexplanation during audit analytical review. The

Accounting Review, 67, 59–76.

Krishnamoorthy, G., Mock, T. J., & Washington, M. T. (1999). A comparative evaluation of

belief revision models in auditing. Auditing: A Journal of Practice and Theory, 18,

105–127.

Krull, G., Jr., Reckers, P. M. J., & Wong-on-Wing, B. (1993). The effect of experience,

fraudulent signals and information presentation order on auditors’ beliefs. Auditing: A

Journal of Practice and Theory, 12, 143–153.

Libby, R., & Luft, J. (1993). Determinants of judgment performance in accounting settings:

Ability, knowledge, motivation, and environment. Accounting Organizations, and

Society, 18, 425–450.


Lopes, L. L. (1985). Averaging rules and adjustment processes in Bayesian influence. Bulletin of

the Psychonomic Society, 23, 509–512.

Lopes, L. L. (1987). Procedural debiasing. Acta Psychologica, 64, 167–185.

Lord, C., Ross, L., & Lepper, M. (1979). Biased assimilation and attitude polarization: The

effects of prior theories on subsequently considered evidence. Journal of Personality and

Social Psychology, 37, 2098–2109.

Mayper, A. G., Anderson, U., & Kilpatrick, B. G. (1999). Tax workpaper review: The impact of

staff conclusions and argument presentation. Advances in Taxation, 11, 145–167.

McMillan, J. J., & White, R. A. (1993). Auditors’ belief revisions and evidence search: The

effect of hypothesis frame, confirmation bias, and professional skepticism. The

Accounting Review, 68, 443–465.

Messier, W. F., Jr., & Tubbs, R. M. (1994). Recency effects in belief revision: The impact of

audit experience and the review process. Auditing: A Journal of Practice and Theory, 13,

57–72.

Monroe, G. S., & Ng, J. (2000). An examination of order effects in auditors’ inherent risk

assessments. Accounting and Finance, 40, 151–166.

Nisbett, R., & Ross, L. (1980).Human inference: Strategies and shortcomings of social judgment.

Englewoods Cliffs, NJ: Prentice-Hall.

Pei, B. K. W., Reckers, P. M. J., & Wyndelts, R. W. (1990). The influence of information

presentation order on professional tax judgment. Journal of Economic Psychology, 11,

119–146.

Pei, B. K. W., Reckers, P. M. J., & Wyndelts, R. W. (1992a). Tax professionals’ belief revision:

The effects of information presentation sequence, client preference, and domain

experience. Decision Sciences, 23, 175–199.

Pei, B. K. W., Reed, S., & Koch, B. (1992b). Auditor belief revision in a performance auditing

setting: An application of the belief adjustment model. Accounting, Organizations, and

Society(17), 169–183.

Pinsker, R. (2004). Order effects in a more frequently reported information environment:

Laboratory evidence opposing the belief-adjustment model’s primacy prediction. Working

paper, Old Dominion University.

Pitz, G. F., Downing, L., & Reinhold, H. (1967). Sequential effects in the revision of subjective

probabilities. Canadian Journal of Psychology, 21, 381–393.

Roberts, M. (1998). Tax accountants’ judgment-decision-making research: A review and

synthesis. Journal of the American Taxation Association, 20, 78–121.

Schisler, D. L. (1994). An experimental examination of factors affecting tax preparers’

aggressivness – A prospect theory approach. The Journal of the American Taxation

Association, 16(2), 124–142.

Trotman, K., & Wright, A. (1996). Recency effects: Task complexity, decision mode, and task-

specific experience. Behavioral Research in Accounting, 8, 175–193.

Trotman, K., & Wright, A. (2000). Order effects and recency: Where do we go from here?

Accounting and Finance, 40, 169–182.

Tubbs, R. M., Messier, W. F., Jr., & Knechel, W. R. (1990). Recency effects in the auditor’s

belief revision process. The Accounting Review, 65, 452–460.

Tuttle, B., Coller, M., & Burton, F. G. (1997). An examination of market efficiency:

Information order effects in a laboratory market. Accounting, Organization and Society,

22, 89–103.

Yates, F. J. (1990). Judgment and decision making. Englewood Cliffs, NJ: Prentice-Hall.


AUDITOR CALIBRATION IN

THE REVIEW PROCESS

Noel Harding, Sally Hughes and Ken T. Trotman

ABSTRACT

A recent change to audit workpaper review has been the movement toward

delegating more review tasks to senior auditors and including more staff

auditors in the review process. This study investigates the efficiency and

effectiveness implications of this change. It considers the calibration of

reviewers of different levels of experience on both conceptual and me-

chanical errors. The results reveal that reviewers are miscalibrated (over-

confident) in their workpaper error judgments. No differences are found

in the calibration of staff and senior auditors across hierarchical level or

type of error. The implications for audit effectiveness are discussed in the

paper.

INTRODUCTION

The review process plays a critical role in maintaining the quality of an audit

(Bamber & Ramsay, 1997; Rich, Solomon, & Trotman, 1997a, b). It pro-

vides additional assurance that sufficient and appropriate audit work has

been performed and that all audit objectives have been achieved. Tradi-

tionally, a team of auditors performed an audit via a sequential and iterative





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08002-0

41

process with multiple, hierarchical layers of review (Solomon, 1987). Al-

though seniors were involved in the review process, managers and partners

performed the majority of the review activities. Due to competitive pressures

for cost-effective audits and the expense associated with hierarchical reviews

performed by experienced auditors, the review process has become a target

for possible cost savings. In this regard, some audit firms have redesigned

their review practices and procedures. These changes include a review proc-

ess now characterized by real-time coaching, the review of audit work being

conducted by interview, the delegation of more review tasks to seniors and

the inclusion of staff auditors in the review process (Rich et al., 1997b;

Winograd, Gerson & Berlin, 2000). Several studies have examined the ef-

fectiveness and efficiency implications of these changes (Ramsay, 1994;

Bamber & Ramsay, 1997; Harding & Trotman, 1999; Bamber & Ramsay,

2000). This study extends these earlier studies by examining reviewer cal-

ibration, a measure of reviewer performance that has not previously been

examined in the literature.

When conducting workpaper review, the reviewer is required to examine

large amounts of workpaper documentation that must be recalled when con-

sidering the probity of subsequent workpaper documentation and the con-

clusions contained therein. While reviewers have the opportunity to refer back

to the previously reviewed workpapers, they will often rely on their memory

in order to improve review efficiency (Libby & Trotman, 1993). This reliance

on memory, however, can jeopardize review effectiveness and efficiency if the

reviewer’s memory is incomplete and/or inaccurate. A reviewer’s confidence

in their memory for workpaper documentation is, therefore, important as it

will determine how the reviewer will use the judgment, and the subsequent

action they will take (Sniezek & Henry, 1989). Calibration is a measure of the

appropriateness of the expressed level of confidence. It assesses the proficiency

of a judge (Oskamp, 1962) since it measures the relationship between judg-

ment accuracy and judgment confidence.

Ideally, confidence should increase with accuracy to give a well calibrated

reviewer. An auditor who is miscalibrated can be either overconfident or

underconfident in their judgment. Overconfidence occurs when, on average, a

reviewer’s confidence is greater than their accuracy. Underconfidence occurs

when, on average, a reviewer’s accuracy is greater than their confidence. A

miscalibrated reviewer is likely to make poor judgments, which in turn, will

impact upon the effectiveness and efficiency of the review process (Beck,

Solomon, & Tomassini, 1985; Sprinkle & Tubbs, 1998).

There can be several costs incurred if a decision-maker is miscalibrated.

Overconfidence might mean that a decision-maker disregards contradictory

NOEL HARDING ET AL.42

evidence or, alternatively, fails to search for additional information that

would lead them to question their original decision (Pincus, 1991). In con-

trast, underconfidence may result in excessive collection of evidence to sup-

port a judgment, leading to a reduction in efficiency. Therefore, delegation

of review tasks to miscalibrated reviewers can lead to poor performance

and/or increased audit hours, both of which are undesirable in an environ-

ment increasingly characterized by competition, fee, and legal pressures

(e.g., Winograd et al., 2000).

A consistent finding in the literature is that individuals, including audi-

tors, are overconfident (e.g., Plous, 1993). If workpaper reviewers; are sim-

ilarly overconfident in their review judgments, the reliance that audit firms

place on the review process may not be justified. An overconfident reviewer

might, for example, place excessive reliance on inaccurate memory for audit

evidence contained in the workpapers previously reviewed.

A number of recent studies have examined the accuracy of reviewers in

detecting different types of workpaper errors (Ramsay, 1994; Bamber &

Ramsay, 1997; Harding & Trotman, 1999; Tan & Trotman, 2003). Ram-

say (1994) suggested that seniors and managers use different review tem-

plates to guide their review of the workpapers, arguing that managers

focus more on conceptual errors while seniors focus more on mechanical

errors.1 He found that managers performed better than seniors on con-

ceptual errors and seniors performed better than managers on mechanical

errors. Harding and Trotman (1999) extended these results to audit staff.

They found that staff auditors identified more mechanical errors than

seniors, while the opposite was true for conceptual errors. Bamber and

Ramsay (1997) investigated the effect of specialized reviews on reviewers’

performance. That is, the effects of directing an auditor to perform a

review that focuses on identifying particular types of errors, either me-

chanical (seniors) or conceptual (managers). The results revealed that re-

viewers were more accurate when directed to perform a comprehensive as

opposed to a focused review. Tan and Trotman (2003) have recently

demonstrated that these hierarchical differences in review performance

may be influenced by preparer stylization and a reviewer’s sensitivity to

that stylization.

Despite its importance in understanding the effectiveness and efficiency of

the review process, only one study has, to date, examined reviewer confi-

dence in workpaper error judgments with no study examining the relation-

ship between confidence and accuracy (i.e., calibration). Bamber and

Ramsay (2000) examined the confidence that seniors and managers had in

performing all encompassing versus specialized reviews. They found that

Auditor Calibration in the Review Process 43

seniors were more confident when performing specialized as compared to all

encompassing reviews. There was no difference in confidence for managers.2

The present study extends earlier research examining hierarchical differ-

ences in reviewer accuracy (e.g., Ramsay, 1994; Harding & Trotman, 1999)

by investigating hierarchical differences in reviewer calibration. This study

allows us to assess more completely the appropriateness of moving the re-

view process to lower levels in the audit firm hierarchy.

The results of our study reveal that reviewers are miscalibrated (over-

confident) in their workpaper error judgments. Contrary to expectations,

there are no differences in the level of overconfidence across hierarchical

level or type of error. The results suggest that while the inclusion of staff

auditors in the review process might improve review efficiency, their inclu-

sion in this study neither adds to nor detracts from review effectiveness.

HYPOTHESIS DEVELOPMENT

Calibration has been extensively studied in the psychology literature (see

Lichtenstein, Fischhoff & Phillips (1982) and Plous (1993) for reviews). With

a few exceptions, when making judgments for at least moderately difficult

tasks, miscalibration in the form of overconfidence is the norm. Undercon-

fidence, however, is often identified for simple tasks where the percentage of

subjects making the correct judgment is above �70%. This is referred to as

the hard–easy effect.

In auditing, initial research suggested that auditors were generally well

calibrated and might even tend towards underconfidence (Tomassini,

Solomon, Romney, & Krogstad, 1982). Mladenovic and Simnett (1994),

consistent with the hard–easy effect identified in the psychology literature,

found that overconfidence increases as the task becomes less predictable

(more difficult). Given the difficult nature of audit judgments, it is perhaps

not surprising that recent studies show auditors to be overconfident. For

example, Simnett (1996) found auditor overconfidence on a prediction of

failure task. Kennedy and Peecher (1997) found that audit staff, seniors, and

managers were all overconfident in their technical knowledge. Moeckel and

Plumlee (1989) investigated auditor’s confidence in the recognition of audit

evidence and found that auditors can be just as confident in their inaccurate

memories as their accurate memories.

Workpaper review is of moderate to high difficulty. It requires a consid-

erable amount of time and is characterized by multiple levels of review. It


involves the examination and integration of large amounts of information

that, of itself, might not be important, but could be critical when combined

with other information. Given the difficulty of the review process and the

consistent findings of overconfidence, we anticipate that the auditors in the

present study will be miscalibrated (overconfident) in their recognition of

workpaper errors. We test the following hypothesis.

H1. Auditors are overconfident in their recognition of workpaper errors.

While we argue that auditors will be overconfident in their workpaper

judgments, the long standing hierarchical nature of the review process, and

research findings that report hierarchical differences in reviewer accuracy

across error types (e.g., Ramsay, 1994; Bamber & Ramsay, 1997; Harding &

Trotman, 1999), suggest that there might be differences across hierarchical

levels.

A reviewer’s confidence in their memory for workpaper documentation

is affected by their ability to retrieve information from their memory,

which in turn, is a function of the type of elaboration that occurred during

the encoding of that information (Craik & Tulving, 1975). Elaboration at

encoding is a function of the depth of processing. Craik and Tulving iden-

tified two types of processing; deep and shallow processing. Deep process-

ing at encoding occurs when evidence receives full attention, is entirely

analyzed and ‘‘enriched by association’’ (p. 270). Shallow processing of

information occurs when evidence is not attended to fully and is only

analyzed at a surface level. If a reviewer has performed an in depth analysis

of certain aspects of the workpapers, their ability to retrieve information

from their memory is greater for those areas and likely to lead to increased

confidence.

Depth of processing is also argued to impact accuracy. Indeed, this has

been the theoretical foundation of studies reporting hierarchical differences

in reviewer accuracy (e.g., Ramsay, 1994; Bamber & Ramsay, 1997; Harding

& Trotman, 1999). These studies argue that accuracy for conceptual errors

increases with experience due to increased attention directed toward this

aspect of the workpapers. Accuracy for mechanical errors, they argue, de-

creases with experience, as these errors are no longer the primary focus of

the reviewer’s attention. It is, therefore, anticipated that confidence and

accuracy will increase with the amount of elaboration at encoding. If, for

example, mechanical errors are the focus of the reviewer’s attention, both

confidence and accuracy will increase for these types of errors.

In terms of the relationship between accuracy and confidence, Glenberg

and Epstein (1985), Brothwell, Deffenbacher, and Brigham (1987) and


Glenberg and Epstein (1987) all provide evidence of a positive association. It

might, therefore, be expected that there are no hierarchical differences in

overconfidence as changes in accuracy are associated with corresponding

changes in confidence. However, highlighting that the relationship may be

more complex in an auditing context, Moeckel and Plumlee (1989) report

results showing both a positive and negative association.

In addition, experience may have an effect in addition to influencing

accuracy and the amount of elaboration. If this is the case, hierarchical

differences in the level of overconfidence might be expected. There is only a

limited literature examining the effect of experience on confidence. In ac-

counting domains, the results have been mixed. Raiborn and Estes (1986)

(cited in Pincus, 1991) observed a positive relationship, while Snowball

(1980) observed a negative relationship. Weber (1978) and Pincus (1991)

both observed no significant relationship.

While Pincus (1991) found no relationship between experience and con-

fidence across her entire sample, she did observe a positive relationship when

the decision made was consistent with an unqualified opinion and a negative

relationship when the decision made was consistent with a modified opinion.

This, she speculated, might be the result of an availability heuristic in that

auditors are more commonly faced with a situation where an unqualified

opinion is issued.

An alternative explanation for the results reported in Pincus (1991) relates

to the implications of the decision (see Solomon, Ariyo, & Tomassini, 1985;

Mladenovic & Simnett, 1994). A decision leading to a modified opinion

might be argued to have greater implications than one leading to an un-

qualified opinion. Auditors in Pincus’ study may have been less confident in

a decision leading to a modified opinion and this lack of confidence was

amplified with experience as auditors become more aware of the implica-

tions of their decisions. In this regard, Sprinkle and Tubbs (1998) found that

auditors were less willing to rely on their memory (less confident) for in-

formation of greater importance.

Seniors being more experienced than staff auditors at workpaper review

are argued to be more confident than staff auditors in their review judg-

ments. This is particularly the case for mechanical errors, as there are also

likely to be differences in confidence derived from the perceived importance

and mechanical errors across hierarchical levels. While there may be some

variations depending on the next level of review (Rich et al., 1997a; Tan &

Trotman, 2003) the importance for seniors of mechanical errors, by com-

parison to conceptual errors, is low. Staff auditors, however, view mechan-

ical errors as critical to their career progression. The prevention and/or


identification of mechanical errors is an important factor in staff auditors’

performance evaluation. We argue that these factors will lead to a decrease

in confidence for staff auditors and an increase in confidence for seniors. If

reviewers are overconfident (as proposed in H1), these changes in confidence

will reduce overconfidence for staff auditors and increase overconfidence for

seniors. We test the following hypothesis.

H2. Staff auditors will be better calibrated (less overconfident) in the

recognition of mechanical errors than senior auditors.

On the other hand, seniors are acutely aware of the importance of con-

ceptual errors thereby reducing their confidence. Staff auditors, on two

counts, are likely to be more confident. First, these errors are less the focus

of their work compared to mechanical errors. In addition, staff auditors are

unlikely to have the same appreciation of the importance of conceptual

errors as their senior colleagues. Given that experience is argued to temper

confidence on review matters relating to conceptual errors (and inexperience

increases confidence) we argue that staff auditors will be more confident

(and more overconfident) than seniors on conceptual errors. We test the

following hypothesis.

H3. Senior auditors will be better calibrated (less overconfident) in the

recognition of conceptual errors than staff auditors.

METHOD

The same participants and research instrument in Harding and Trotman’s

(1999) study were used in this study. However, Harding and Trotman only

analyzed the accuracy of participants’ responses to workpaper error ques-

tions. This paper analyzes the calibration of reviewers’ workpaper error

judgments. A 2� (2) experiment was designed to examine the impact of

experience and the type of error on reviewer calibration. The between-subject

variable was experience (staff, senior) and the within-subject variable was the

error type (mechanical, conceptual).

Subjects

Our analysis is based on a sample of 20 senior auditors and 20 staff auditors

from a then ‘‘big 6’’ accounting firm.3 The mean experience of the staff and


senior auditors in our sample was 2.13 and 4.77 years respectively. The

experience of senior and staff auditors was significantly different

(t ¼ 10.653, two-tailed p ¼ 0.001). From the original sample of 42 in

Harding and Trotman (1999), one auditor was dropped because of incom-

plete confidence levels and one was dropped because their review notes were

not legible.

Research Instrument and Administration

The research instrument used in Harding and Trotman (1999) and this study

is the same as that developed by Ramsay (1994).4 Subjects were asked to

perform a review of a set of hypothetical workpapers relating to the pro-

vision for doubtful debts account for a manufacturing firm. Subjects were

given three envelopes, each individually labeled (A, B, C) and a sheet out-

lining the general instructions. After the subjects read the general instruc-

tions they were instructed to open envelope A, which contained the

workpapers and blank pages upon which to write review notes. Subjects

were given 40min in order to complete the review. Subjects were instructed

to review the workpapers as they would in a real audit, but were not told

that they would be using their review notes or that they would be tested on

the content of the workpapers. Following the review, the workpapers were

collected and a line was ruled under the last review note on each page.

Thereafter, subjects were asked to open envelope B, which contained a true/

false questionnaire containing 16 questions. These included errors that ex-

isted in the workpapers and filler items, that is, errors that did not exist in

the workpapers. The recognition test included seven conceptual errors, four

mechanical errors, four mechanical filler items, and one conceptual filler

item. This was the same true/false test as used by Ramsay (1994). Following

Ramsay (1994) and Bamber and Ramsay (2000) subjects were also asked to

indicate the confidence they had in their response to each of the 16 questions

on a 3-point scale: 1 – I am certain of my answer; 2 – I believe my answer is

correct; and 3 – I guess my answer is correct. Materials in envelope C elicited

demographic information.

Subjects completed the research materials in one of three sessions con-

ducted at the offices of the then ‘‘big 6’’ firm involved. At least one of the

authors was present on each occasion in order to administer the materials.

Each session began with an introduction of the author(s) by a senior mem-

ber of the firm who also emphasized the firm’s support for the research being


conducted. In each session, subjects worked diligently and appeared to take

the task seriously.

Method of Analysis

Calibration refers to the accuracy of the decision-maker’s confidence. That

is, is the confidence level assigned appropriate given the accuracy of the

responses? The greater the confidence, the greater should be the accuracy.

While there are a number of methods of calculating calibration (see Yates,

1990) we chose the method that distinguishes between over and under con-

fidence given the differences in consequences for auditing. This method

which allows an assessment of direction and magnitude of any miscalibrated

judgments has been used in other accounting/auditing studies (e.g., Dilla,

File, Solomon, & Tomassini 1991; Pincus, 1991; Simnett, 1996; Kennedy &

Peecher, 1997). It provides an over/underconfidence score calculated using

the following equation:

Over=underconfidence ¼1

N

X

T

i¼1

ni Pi � Cið Þ

where N is the total number of probability assessments, ni the number of

times a probability response was used, Pi the probability response category

(i.e., 1.0, 0.75, 0.50), Ci the percentage of correct responses for each prob-

ability category and T the total number of probability response categories

(three in the present study).

A positive score indicates overconfidence while a negative score indicates

underconfidence. Interpretation of this score, however, must proceed with

caution. A score of zero may represent one of two possibilities. It may

indicate that the decision maker is perfectly calibrated or that overconfi-

dence/underconfidence at one probability response category is perfectly

offset by underconfidence/overconfidence at another probability response

category. Similarly, a positive (negative) score should be interpreted as in-

dicating that, over the three probability response categories, the decision-

maker is, on average, overconfident (underconfident). It is therefore

necessary to interpret this score in conjunction with calibration curves,

which plot perfect calibration (called the ‘‘identity line’’) against the average

percentage correct answers at each probability level. These calibration

curves are shown in Table 2.

As noted above, subjects indicated their confidence on a 3-point scale: 1 –

I am certain of my answer; 2 – I believe my answer is correct; and 3 – I guess


my answer is correct. It was therefore necessary to assign percentages to

each of these responses. Responses 1, 2 and 3 were coded 100%, 75% and

50%, respectively.5 Subjects had access to their review notes (but not the

workpapers) when completing the recognition test. In order to calculate

calibration, it was necessary to exclude the responses to questions that were

noted in the subject’s review notes. In doing so, the questions that remained

represented a true test of memory and calibration. Two of the authors

independently examined each of the review notes with a view to identifying

those that related to a question on the recognition test. The small number of

disagreements between coders were resolved. For each subject, all recog-

nition test questions identified in the workpapers were removed from the

data set. This meant that from zero to three questions were removed for

each staff auditor (average 1.45) and zero to five questions removed for each

senior (average 2.15).

RESULTS

Consistent with Bamber and Ramsay (2000), our analysis is based on the

actual errors contained in the true–false test.6 Table 1 reports the over/

underconfidence score and the number of subjects who were overconfident,

perfectly calibrated, and underconfident.

Table 1 (panel A) reveals that subjects were generally overconfident in

their workpaper review judgments. Of the 40 subjects, 29 were overconfi-

dent.7 The mean over/underconfidence score of 0.151 was significantly dif-

ferent from zero (t ¼ 5.124, two-tailed po0.001). There was no difference in

the over/underconfidence score between senior (0.139) and staff auditors

(0.163) (t ¼ 0.396, p ¼ 0.694). While subjects were well calibrated at the

50% probability response category (percent correct 51.06), the percentage

correct was significantly different from that expected for perfect calibration

at the 75% probability response category (percent correct 59.33; two-tailed

p ¼ 0.005) and 100% probability response category (percent correct 80.73;

two-tailed po0.001). These results provide support for H1. Auditors in our

study were overconfident in their recognition of workpaper errors. We dis-

cuss the implications of this finding in the following section.

Table 1 (panel B) further disaggregates this overconfidence across type of

error and hierarchical level. Table 2 reports the percentage of correct re-

sponses (accuracy) at each probability response category for each of the four

research conditions. The table also plots accuracy against the identity line


representing perfect calibration. Points above and below the identity line

represent overconfidence and underconfidence, respectively.

Tables 1 and 2 reveal that seniors were overconfident for both conceptual

errors (t ¼ 2.013, p ¼ 0.06) and mechanical errors (t ¼ 2.70, p ¼ 0.01).

While staff auditors were overconfident for conceptual errors (t ¼ 3.047,

p ¼ 0.01) the over/underconfidence score for mechanical errors, although in

Table 1. Over/Underconfidence Score Number of Subjects

Underconfident Overconfident Perfectly Calibrated.

Panel A: Aggregated Results (n ¼ 40)

Over/underconfidencea 0.151

t statisticb 5.124

Significanceb po0.001

Number of subjects

Overconfident 29

Perfectly calibrated 3

Underconfident 8

Panel B: Disaggregated Results

Staff Auditors (n ¼ 20) Senior Auditors (n ¼ 20)

Conceptual errors

Over/underconfidencea 0.18 0.09

T statisticb 3.047 2.013

Significanceb p ¼ 0.01 p ¼ 0.06

Number of subjects

Overconfident 15 14

Perfectly calibrated 0 2

Underconfident 5 4

Mechanical errors


T statisticb 1.644 2.70

Significanceb p ¼ 0.12 p ¼ 0.01

Number of subjects

Overconfident 10 13

Perfectly calibrated 2 4

Underconfident 8 3

aOver/underconfidence is calculated with reference to the formula provided in the text. A

positive (negative) score indicates overconfidence (underconfidence).bThis t statistic tests whether the calibration score is significantly different from zero.


Table 2. Calibration Curves.

Staff Auditors Senior Auditors

Conceptual Errors


Probability response cat. 50 75 100 50 75 100

Percentage correct 48.7 55.8� 82.3� 62.5 67.4 84.4�

0102030405060708090

100

50 75 100

Probability response category

Pe

rce

nta

ge

co

rre

ct

Identity line % Correct

0102030405060708090

100

50 75 100


Pe

rce

nta

ge

co

rre

ct


Mechanical Errors


Probability response cat. 50 75 100 50 75 100

Percentage correct 45.8 50.0 88.2� 57.3 30.0� 83.3�

0102030405060708090

100

50 75 100


Pe

rce

nta

ge

co

rre

ct


0102030405060708090

100

50 75 100


Pe

rce

nta

ge

co

rre

ct


aOver/underconfidence is calculated with reference to the formula provided in the text. A positive (negative) score indicates overconfidence

(underconfidence).�Significant at p ¼ 0.01 (significantly different from the identity line).

NOELHARDIN

GET

AL.

52

the direction of overconfidence, was not statistically significant (t ¼ 1.644,

p ¼ 0.12). In each of the four research conditions, at least 50% of the sub-

jects were overconfident. As was the case for the aggregate analysis, our

subjects were well calibrated at the 50% probability response category, and

overconfident at the 75% and 100% probability response categories in each

of the four research conditions.

H2 and H3, taken together, predict a significant interaction between re-

viewers’ calibration for the nature of the item in the recognition test (me-

chanical or conceptual) and the experience of the reviewer (staff or senior).

Contrary to expectations, the results from a 2� (2) repeated measures

ANOVA revealed no significant interaction between error type and expe-

rience (F ¼ 2.125, p ¼ 0.153). H2 predicted that staff auditors would be

better calibrated than senior auditors in the recognition of mechanical errors

while H3 predicted that the opposite would be true for conceptual errors.

Although in the expected direction, there was no significant difference in the

calibration scores between staff and senior auditors for mechanical errors

(t ¼ 1.056, one-tailed p ¼ 0.149) or conceptual errors (t ¼ 1.102, one-tailed

p ¼ 0.139). Therefore, our data do not support H2 or H3.8

DISCUSSION AND CONCLUSIONS

The literature on the ability of reviewers to identify various workpaper

errors in the review process (Ramsay, 1994; Bamber & Ramsay, 1997, 2000;

Harding & Trotman, 1999) has provided useful insights into differences in

reviewer accuracy and efficiency across different experience levels of re-

viewers. These studies, however, have not examined the calibration of

reviewer judgments. Calibration measures the relationship between judg-

ment accuracy and judgment confidence. A well-calibrated judge is more

likely to act appropriately on their judgment, irrespective of whether that

judgment is correct or not. An overconfident reviewer, on the other hand,

believes they are correct when, in fact, they are incorrect. Overconfidence

could mean, for example, that the reviewer chooses not to reinspect the

workpapers or ignores contradictory evidence. Given the importance of

the review process in ensuring that errors are identified and corrected before

the final opinion is issued, calibration is an important element of a review-

er’s performance. This study provides insights into the effectiveness of

the review process and the appropriateness of delegating specific aspects

of the review process to less experienced auditors.


The results revealed that senior and staff auditors were miscalibrated

(overconfident) in their workpaper review judgments. When reviewers

confidently rely on their inaccurate memory, they are less likely to re-inspect

the workpapers. An incomplete or inaccurate recall of workpaper

contents might mean that deficient or incomplete audit procedures are not

identified. Given that the identification and correction of deficient or in-

complete work is the primary purpose of the review process, reviewer over-

confidence is an issue that should be addressed by audit firms. Audit firms

should exercise caution when delegating review tasks to seniors and staff

auditors.

The results show that there was no difference in the levels of overcon-

fidence between seniors and staff auditors on either conceptual or mechan-

ical errors. That is, staff and senior auditors are equally aware (or more

precisely based on the results of this study, equally unaware) of the accuracy

and completeness of their memory for workpaper contents. Combining

these results with Bamber and Ramsay (1997, 2000) and Harding and Trot-

man (1999), there are some important practical implications for audit firms

with respect to changes in the review process. The effectiveness of staff

auditors in identifying mechanical errors, the fact that they have similar

calibration levels to seniors on these errors, and their lower cost means that

it is appropriate for staff auditors to play a role in the review process. As

noted above, the results do not suggest that a greater number of mechanical

errors will be found with the inclusion of staff auditors. Rather, they suggest

that a similar outcome will be achieved, at lower cost, if staff auditors

instead of seniors perform the review work; however, the same conclusion

could not be drawn for conceptual errors. While staff and senior auditors

might be equally aware (or unaware) of when they are incorrect, only senior

auditors are likely to be able to correct their inaccurate understanding by re-

inspecting the workpapers. Staff auditors will not have this ability, even if

they do re-inspect the workpapers.

From these results, it is tempting to suggest a specialized review with staff

considering mechanical errors and seniors considering conceptual errors.

However, Bamber and Ramsay (1997, 2000) indicate that these specialized

reviews are less effective and efficient than comprehensive reviews. There-

fore, our overall conclusion based on the combined research is that there are

benefits of having both staff auditors and seniors carrying out comprehen-

sive reviews for a client. The addition of the staff auditors will add to the

quality of documentation and provide training in review. These benefits

need to be compared to the additional cost of the time they spend on re-

views.


NOTES

1. Examples of mechanical errors include missing tick marks and numbers onone workpaper not agreeing with the original calculation of the numbers locatedon another workpaper. Examples of conceptual errors include use of an impropermateriality threshold for the type of client and the inadequate explanation or jus-tification of audit conclusions (Bamber & Ramsay, 2000).2. As part of additional analysis, Bamber and Ramsay (2000) calculated a sur-

rogate for calibration. This measure involved examining accuracy when subjectsindicated they were certain of their answer. No analysis was performed when subjectsindicated they were less than certain.3. Subjects were selected by the firm involved on the basis of experience and

ability.4. Minor ‘‘semantic changes’’ were made to the workpapers and the True/False

questionnaire in order to make them appropriate to Australian auditors.5. The descriptions attached to responses 1 and 3 meant that coding did not

represent a major problem. Coding for response 2, however, was problematic. Giventhe absence of any compelling reason to code otherwise, a response of 2 was coded75%. It is recognized, however, that subjects may have inferred that I believe myanswer is correct to mean something other than being 75% confident. With this inmind, we re-calculated all scores using 80% and 70%. Our statistical inferences wereunchanged with the use of these alternative figures.6. We re-analyzed the data including both actual errors and filler items. All in-

ferences were unchanged.7. Of the eight subjects who were underconfident, four were seniors and four were

staff auditors. Two staff auditors and one senior were perfectly calibrated.8. Similar results were obtained when the analysis was repeated for all questions

(actual errors and filler items).

ACKNOWLEDGMENTS

We thank Michael Bamber, Amna Khalifa, and Roger Simnett for their

useful comments. We also acknowledge the financial support of an

Australian Research Council Discovery Grant to Ken Trotman.

REFERENCES

Bamber, E. M., & Ramsay, R. J. (1997). An investigation of the effects of specialization in audit

workpaper review. Contemporary Accounting Research, 14(3), 501–513.

Bamber, E. M., & Ramsay, R. J. (2000). The effects of specialization in audit workpaper review

on review efficiency and reviewer’ confidence. Auditing: A Journal of Practice and The-

ory, 19(2), 147–157.


Beck, P. J., Solomon, I., & Tomassini, L. A. (1985). Subjective prior probability distributions

and audit risk. Journal of Accounting Research, 23(1), 37–56.

Brothwell, R., Deffenbacher, K., & Brigham, J. (1987). Correlation of eyewitness accuracy and

confidence: Optimality hypothesis revisited. Journal of Applied Psychology, 72(Novem-

ber), 691–695.

Craik, F., & Tulving, E. (1975). Depth of processing and the retention of words in episodic

memory. Journal of Experimental Psychology: General, 104(3), 268–294.

Dilla, W., File, R. D., Solomon, I., & Tomassini, L. A. (1991). Predictive bankruptcy judgments

by auditors. In: L. Poneman & D. Gabhat (Eds), Probabilistic approach auditing: ad-

vances in behavioral research. Berlin: Springer.

Glenberg, A., & Epstein, W. (1985). Calibration of comprehension. Journal of Experimental

Psychology: Learning Memory and Cognition, 11(October), 702–718.

Glenberg, A., & Epstein, W. (1987). Inexpert calibration of comprehension. Memory and Cog-

nition, 15(January), 84–93.

Harding, N., & Trotman, K. (1999). Hierarchical differences in audit workpaper review per-

formance. Contemporary Accounting Research, 16(4), 671–684.

Kennedy, J., & Peecher, M. (1997). Judging auditors’ technical knowledge. Journal of Ac-

counting Research, 35(2), 279–293.

Libby, R., & Trotman, K. T. (1993). The review process as a control for differential recall of

evidence in auditor judgments. Accounting, Organizations and Society, 18(6), 559–574.

Lichtenstein, S., Fischhoff, B., & Phillips, L. D. (1982). Calibration of probabilities. The state of

the art to 1980. In: D. Kahneman, P. Slovic & A. Tversky (Eds), Judgment under

uncertainty: Heuristics and biases. Cambridge: Cambridge University Press.

Mladenovic, R., & Simnett, R. (1994). Examination of contextual effects and changes in task

predictability on auditor calibration. Behavioral Research in Accounting, 6, 178–203.

Moeckel, C. L., & Plumlee, R. D. (1989). Auditors’ confidence in accurate and inaccurate

recognition of audit evidence. The Accounting Review, 64(4), 653–666.

Oskamp, S. (1962). The relationship of clinical experience and training methods to several

criteria of clinical prediction. Psychological Monographs, 76(28), 547.

Pincus, K. V. (1991). Audit judgment confidence. Behavioral Research in Accounting, 3, 39–64.

Plous, S. (1993). The psychology of judgment and decision making. New York: McGraw Hill.

Raiborn, D. D., & Estes, R. (1986). The impact of personal characteristics on materiality

decisions. In: M. Massoud (Ed.), Proceedings-1986 western accounting association annual

meeting: the economics of accounting.

Ramsay, R. (1994). Senior/manager differences in audit workpaper review performance. Jour-

nal of Accounting Research, 32(1), 127–135.

Rich, J. S., Solomon, I., & Trotman, K. T. (1997a). The audit review process: A character-

ization from the persuasion perspective. Accounting, Organizations and Society, 22(5),

481–505.

Rich, J. S., Solomon, I., & Trotman, K. T. (1997b). Multi-auditor judgment/decision making

research: A decade later. Journal of Accounting Literature, 16, 86–126.

Simnett, R. (1996). The effect of information selection, information processing and task com-

plexity on predictive accuracy of auditors. Accounting, Organizations and Society, 21(7/

8), 699–719.

Sniezek, J. A., & Henry, R. A. (1989). Accuracy and confidence in group judgments. Organ-

izational Behavior and Human Decision Processes, 43(1), 1–28.


Snowball, D. (1980). Some effects of accounting expertise and information load: An empirical

study. Accounting Organizations and Society, 5, 323–338.

Solomon, I. (1987). Multi-auditor judgment/decision making research. Journal of Accounting

Literature, 6, 1–25.

Solomon, I., Ariyo, A., & Tomassini, L. A. (1985). Contextual effects on the calibration of

probabilistic judgments. Journal of Applied Psychology, 70(3), 528–532.

Sprinkle, G. B., & Tubbs, R. M. (1998). The effect of audit risk and information importance on

auditor memory during working paper review. The Accounting Review, 73(4), 475–502.

Tan, H. T., & Trotman, K. T. (2003). Reviewers’ responses to anticipated stylization attempts

by preparers of audit workpapers. The Accounting Review, 78(2), 581–604.

Tomassini, L. A., Solomon, I., Romney, M. B., & Krogstad, J. L. (1982). Calibration of

auditor’s probabilistic judgments: Some empirical evidence. Organizational Behavior and

Human Performance, 30(3), 391–406.

Yates, J. F. (1990). Judgment and decision making. Englewood Cliffs, NJ: Prentice-Hall.

Weber, R. (1978). Auditor decision making on overall system reliability: Accuracy, consensus

and the usefulness of a simulation decision aid. Journal of Accounting Research, 16(2),

368–388.

Winograd, B. N., Gerson, J. S., & Berlin, B. L. (2000). Audit Practices of Pricewaterhouse-

Coopers. Auditing: A Journal of Practice and Theory, 19(2), 175–182.


This page intentionally left blank

58

LINGUISTIC DELIVERY STYLE,

CLIENT CREDIBILITY, AND

AUDITOR JUDGMENT

Christie L. Comunale, Thomas R. Sexton and

Terry L. Sincich

ABSTRACT

This chapter introduces linguistic delivery style to auditing research,

demonstrates how linguistic delivery style relates to client credibility, and

shows how linguistic delivery style and client credibility influences audi-

tors’ judgment. Two hundred auditors participated in an analytical pro-

cedures task. The results indicate that high client credibility and powerful

linguistic delivery style increase the auditor’s assessed likelihood that the

explanation accounts for the fluctuation and decrease their intent to per-

form additional testing. Moreover, powerless linguistic delivery style from

an otherwise high credibility client leads to auditor judgments and inten-

tions that are indistinguishable from those that arise from a low credibility

client. Finally, evidence indicates that linguistic delivery style is a fourth

component of credibility.





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08003-2

59

INTRODUCTION

SAS No. 56 (America Institute of certified Public Accountants, AICPA,

1988) requires that auditors use analytical procedures in the planning and

review stages of the audit. In addition, SAS No. 99 (AICPA, 2002) requires

that auditors gather information necessary to identify risks of material

misstatement due to fraud by considering the results of the analytical pro-

cedures performed in planning the audit. Frequently, while employing an-

alytical procedures, the auditor must engage in management inquiry to

investigate unexpected findings. In addition, SAS No. 99 directs auditors to

inquire of management and others within the entity about the risks of fraud.

Such inquiry often takes the form of oral exchanges of questions and re-

sponses.

The effectiveness and efficiency of the management inquiry process de-

pend heavily on the perceived credibility of the management source (Mautz

1958). Nonaudit research indicates that perceived credibility is composed of

three factors: competence, trustworthiness, and objectivity (Berlo, Lemert,

& Mertz, 1969; McCroskey, 1966). Audit research demonstrates that au-

ditors are sensitive to all three factors and illustrates a positive relationship

between each factor and perceptions of messenger credibility and accept-

ability of audit evidence (Joyce & Biddle, 1981; Bamber, 1983; Rebele,

Heinz, & Briden, 1988; Anderson, Koonce, & Marchant, 1994; Hirst, 1994;

Peecher, 1996; Ayers & Kaplan, 1998; Goodwin, 1999; Beaulieu, 2001).

Research in communications suggests that linguistic delivery style – the

presence of hedges, hesitations, and faulty grammar – also influences per-

ceived credibility. Interestingly, audit research literature has not explored

linguistic delivery style. This study introduces linguistic delivery style to the

audit literature by showing the effects of powerful or powerless language on

the auditor’s evaluation of the client’s credibility and hence on the auditor’s

likelihood assessments and planning decisions in the management inquiry

process.

Three possible outcomes may result from the influence of linguistic delivery

style. First, linguistic delivery style may not have a discernible effect. Second,

the auditor may be inefficient if he or she appropriately questions a valid

explanation when the client uses a particularly powerless form of speech.

Finally, the auditor may be ineffective if he or she inappropriately accepts an

insufficient explanation because the client uses a particularly powerful form of

speech. While the two latter outcomes are important, this study examines only

the first kind, that is, the effects of client credibility and linguistic delivery style

on auditors’ likelihood assessments and planning decisions in the case where

CHRISTIE L. COMUNALE ET AL.60

the client’s explanation is valid. This outcome was chosen as the focus of this

study primarily because both auditors and clients seek to streamline the audit

process, thereby reducing audit effort and costs.

Consistent with the existing audit literature, this study provides

evidence that client credibility influences auditors’ likelihood assessments

and planning decisions. In addition, the evidence suggests that the

client’s linguistic delivery style influences auditors’ likelihood assessments

and planning decisions. Finally, while researchers have recognized compe-

tence, trustworthiness, and objectivity as components of credibility,

linguistic delivery style appears to be a fourth component. Moreover, the

supplemental results suggest that powerless linguistic delivery style from an

otherwise high-credibility client leads to auditor judgments and intentions

that are indistinguishable from those that arise from a low-credibility client,

thereby requiring the auditor to perform additional work.

MODELING FRAMEWORK AND

HYPOTHESIS DEVELOPMENT

This study relies on research in psychology, communications, and

auditing to generate the hypotheses, identify the components of credibility

and linguistic delivery style, and build the research model (shown in Fig. 1).

The model posits that client credibility and linguistic delivery style have

direct effects on auditors’ likelihood assessments and planning decisions

during analytical procedures. In addition, the model suggests that linguistic

delivery style influences the auditor’s perceptions of client credibility.

Message Content

During analytical procedures, the auditor may request an explanation from

the client for a significant account balance fluctuation. The client’s oral

explanation constitutes the message in this study. Message content is held

constant and thus is not shown in the research model. Message content is

theoretically independent of the linguistic delivery style of the client and

refers to the literal meaning of the words used in the explanation, separate

and distinct from any linguistic cues regarding the strength of the client’s

belief and confidence in the explanation. Specifically, the power of the cli-

ent’s linguistic delivery style does not alter the message content although it

Linguistic Delivery Style, Client Credibility, and Auditor Judgment 61

may be an imperfect indicator of the strength of the client’s belief and

confidence in the explanation.

Client Credibility

Following O’Keefe (1990), client credibility refers to the auditor’s percep-

tion of the client’s believability. Early factor-analytic investigations of cred-

ibility reveal three dimensions: competence, trustworthiness, and objectivity

(Berlo et al., 1969; McCroskey, 1966). In the following three subsections,

these dimensions are discussed as they relate to auditor judgment and

behavior.

Client Credibility

Competence

Trustworthiness

Objectivity

Linguistic

Delivery Style

Hedges

Hesitations

Grammar

Likelihood Assessments

and Planning Decisions Likelihood Assessment Confidence in

Assessment Likelihood of

Substantive Tests Further Client Inquiry

Fig. 1. Research Model.


Competence

Audit research suggests that auditors are sensitive to the expertise or com-

petence of the source (Bamber, 1983). Rebele et al. (1988) find that auditors

place more reliance on evidence obtained from a high-expertise source than

on evidence obtained from a low-expertise source in an accounts receivable

collections task. In the same light, Anderson et al. (1994) find that auditors

judge evidence gained from a more competent source as more reliable than

evidence gained from source of lesser competence.

Trustworthiness

Audit research finds mixed results concerning auditor sensitivity to source

trustworthiness. Bernardi (1994) finds no difference in fraud detection rates

among auditors considering information from a low- versus a high-integrity

client. Likewise, Kaplan and Reckers (1984) find that auditors are insen-

sitive to client integrity in the initial audit planing process. On the other

hand, Goodwin (1999) finds that auditors evaluating management-provided

evidence involving obsolete inventory are sensitive to management’s integ-

rity. Beaulieu (2001) finds that judgments of client integrity relate negatively

to risk judgments, audit evidence extent recommendations, and fee recom-

mendations. Ayers and Kaplan (1998) discover that audit-firm partners

utilize their assessments of client integrity in client acceptance decisions.

Finally, Peecher (1996) reports that management’s integrity influences the

auditor’s acceptance of a client-provided explanation in analytical proce-

dures.

Objectivity

Audit research demonstrates that auditors are more sensitive to the objec-

tivity of the source when the source is an individual within, as opposed to

outside, the firm under investigation. Hirst (1994) finds that auditors assess

explanations received from fellow auditors as more diagnostic than those

provided by the client’s chief financial officer. Joyce and Biddle (1981) dis-

cover that auditors assess information from an independent credit agency as

more diagnostic than information received from the client’s credit manager.

Finally, Brown (1983) finds that auditors evaluating internal auditor reli-

ability consider objectivity more important than competence and

performance.

Linguistic Delivery Style

O’Keefe (1990) notes that the factors of competence, trustworthiness, and

objectivity represent only the most general dimensions of perceived


credibility. He argues that the delivery characteristics of the message also

may influence perceived credibility and thus warrant further investigation.

Erickson, Lind, Johnson, and O’Barr (1978), Lind and O’Barr (1979), and

O’Barr (1982) examine courtroom transcripts and develop linguistic clusters

that, they argue, exhibit high and low communicative power. They find that,

on an average, uneducated witnesses use ‘‘low-social power’’ forms of

speech – hedges, hesitations, intensifiers, tag questions, deictic words or

phrases,1 and polite form – whereas lawyers, judges, and expert witnesses

typically use ‘‘high-social power’’ forms of speech. Their findings suggest

that speakers using high-social power speech are perceived as possessing a

greater depth of knowledge about the topic and therefore more credible than

speakers using low-social power speech. Haleta (1996) finds that students

rate teachers using language devoid of hedges, intensifiers, deictic phrases,

and hesitation forms as significantly more credible than teachers who use

these linguistic characteristics.

A number of studies find that hedges and hesitations are often

perceived as lowest in power and more strongly associated with lower per-

ceptions of messenger credibility and message believability. Wright and

Hosman (1983) discover that high levels of hedging significantly reduce

credibility ratings. Vinson and Johnson (1989) reveal that both hedges

and hesitations have negative effects on perceptions of messenger credibility.

Hosman and Wright (1987) find that lower levels of hedges and

hesitations produce more positive speaker and message evaluations. More

recently, Adkins and Brashers (1995) show that the user of a powerful

language style in a computer-mediated group is generally perceived as more

credible, attractive, and persuasive than the user of a powerless language

style.

In addition to hedges and hesitations, linguistic research indicates that

other delivery characteristics such as mispronunciation, poor organization,

slow speech rate, low levels of diversity, and faulty grammar are associated

with decreased ratings of speaker effectiveness and credibility (Harms, 1961;

Miller & Hewgill, 1964; McGuire, 1973). Koonce and Phillips (1996) find in

an analytical procedures task that when information pertaining to the cli-

ent’s suggested non-error cause is easy to comprehend, auditors judge that

cause more plausible than when the same information is difficult to com-

prehend. Thus, research in both communication and audit judgment suggest

that linguistic delivery style may influence the auditor’s likelihood assess-

ments and planning decisions, a phenomenon that has not been studied in

the audit literature.


Research Hypotheses

Based on the research cited above, this study investigates whether the ma-

nipulation of linguistic delivery style influences the auditor’s perception of

the client’s credibility. In other words, while the manipulation of the client’s

credibility is expected to influence the auditor’s perception of the client’s

credibility, the manipulation of the linguistic delivery style of the client is

also expected to affect the auditor’s perception of the client’s credibility.

H1. The auditor’s perception of client credibility will be higher when

manipulated linguistic delivery style is more powerful.

Also based on the research cited above, this study examines the effects of

client credibility and linguistic delivery style on auditor likelihood assess-

ments and planning decisions. The research hypotheses are as follows:

H2. High client credibility will:

� Increase the auditor’s likelihood that the client-provided explanation ac-

counts for the fluctuation.� Increase the auditor’s confidence in his/her likelihood assessment.� Decrease the auditor’s intention to perform additional substantive testing.� Decrease the auditor’s intention to gather additional client-provided ex-

planations.

H3. Powerful linguistic delivery style will:

� Increase the auditor’s likelihood that the client-provided explanation ac-

counts for the fluctuation.� Increase the auditor’s confidence in his/her likelihood assessment.� Decrease the auditor’s intention to perform additional substantive testing.� Decrease the auditor’s intention to gather additional client-provided ex-

planations.

RESEARCH METHODOLOGY

Experimental Procedure

The context of this study is preliminary analytical procedures. Two hundred

auditors employed by two of the Big 4 accounting firms participated in the


experiment at the firms’ annual professional training workshops. One

hundred eighty-one study participants were audit seniors; the remaining 19

participants were audit managers.

Each auditor was asked to assume that he or she was an audit team

supervisor of a new manufacturing client and was performing preliminary

analytical procedures. Each auditor was provided with the firm’s financial

information including comparative financial statements of the current year’s

unaudited account balances and the prior year’s audited (by a predecessor

firm) account balances. The account balances included a sizable inventory

fluctuation (19.3%), although no reason was given for the cause of the

fluctuation or whether it reflected a change or error in the account balance.

The auditors were asked to assume that this fluctuation warranted inves-

tigation through a client-provided explanation.2

The auditors were then provided with background information about the

assistant controller, the member of the firm who would provide the expla-

nation for the fluctuation. This information referenced the assistant con-

troller’s competence, trustworthiness, and objectivity – the three factors

comprising credibility. These factors were manipulated as either consistently

high or consistently low. The high client credibility condition portrayed the

assistant controller as qualified, candid, and impartial, while the low client

credibility condition portrayed the assistant controller as unqualified, not

candid, and potentially biased. The wordings of the credibility manipula-

tions are shown in the appendix.

Next, the auditor groups were provided with the assistant controller’s

explanation for the fluctuation, delivered once using a compact disc player.

There was no opportunity for clarification of the explanation and no as-

sumption was made regarding its validity. The linguistic delivery style of the

explanation was manipulated as either powerful or powerless. Powerful lin-

guistic delivery style is free of hedges, hesitations, and faulty grammar. In

contrast, powerless linguistic delivery style includes hedges, hesitations, and

faulty grammar. The wordings of the linguistic delivery style manipulations

are shown in the appendix.

Auditors then indicated their assessment of the likelihood that the client’s

explanation accounted for the fluctuation and their confidence in that as-

sessment. They also indicated the likelihoods that they would perform sub-

stantive tests and that they would conduct further client inquiry. The first

three responses were on 0–100 scale and the client inquiry response was on a

7-point Likert-type scale, with endpoints of ‘‘unlikely’’ and ‘‘likely.’’ See

Table 1. Next, they replied to the items designed to check the manipulations

of credibility and linguistic delivery style (shown in Table 2). Finally, the


auditors provided demographic information. The entire experiment took

�1 h to complete.

Data were collected on three dates at three US locations, with two data-

collection sessions conducted at the same time on the same day (in different

rooms) in the Northeast, a third data-collection session in the Southeast,

and a fourth in the West. In the Northeast, the manipulated linguistic de-

livery style (powerful or powerless) was randomly assigned to the rooms in

which the auditors were already present.3 At the data-collection sessions in

the Southeast and in the West, all the auditors were in the same room and

thus were constrained to hear the same manipulated linguistic delivery style.

However, participants were randomly assigned to either the high or low

manipulated credibility group.4 Table 3 presents sample demographics

across locations and across treatments.

For quantitative demographic measures (e.g., age and experience), the

means for the four experimental conditions were compared using a non-

parametric analysis of variance (Kruskal–Wallis test). Table 3, panel B,

shows statistically significant differences (po0.001) for age and months of

overall audit experience, but not months of manufacturing experience. For

qualitative demographic measures (e.g., gender and current position), the

proportions of auditors in each of the four experimental conditions were

compared using a w2: Table 3, panel B, shows statistically significant dif-

ferences (po0.001) for current position (supervising senior or manager), but

not for gender or audit experience in manufacturing.

Table 1. Questions Designed to Elicit the Dependent Variables.

Likelihood Assessment How likely do you think it is that Joe Smith’s explanation accounts

for the fluctuation in inventory? Please indicate with a percentage

of 0–100%, where 0 is least likely and 100 is most likely:

_________%

Confidence in

Likelihood

Assessment

How confident are you in your likelihood percentage? Please

indicate with a percentage of 0–100%, where 0 is least confident

and 100 is most confident: _________%.

Change in Substantive

Tests

Given Joe’s explanation, how likely are you to change your

substantive test of details relating to the inventory account?

Please indicate with a percentage of 0–100%, where 0 is least

likely and 100 is most likely: _________%

Further Client Inquiry How likely are you to inquire of another member of WEBER’s

management regarding the significant fluctuation in the

inventory account?


There are statistically significant differences in some of the demographic

variables across different levels of both linguistic delivery style and client

credibility. Specifically, the findings indicate that participants who experi-

enced powerful linguistic delivery were 1.3 years younger (p ¼ 0:001) andhad 14.4 months less experience (po0.001) on an average and were less

Table 2. Items Relating to Client Credibility and Linguistic Delivery

Style.

Panel A: Items Relating to Client Credibility

Competence� I think that Joe Smith possesses an adequate amount of experience to provide an accurate

explanation for the financial statement fluctuation.� It is my perception that Joe Smith is qualified to provide an accurate explanation for the

financial statement fluctuation.� I think that Joe Smith possesses the requisite knowledge to provide an accurate explanation

for the fluctuation.

Trustworthiness� I feel that Joe Smith was candid in response to my inquiry.� Joe Smith may not have been totally honest when responding to my inquiry.� I feel that Joe Smith was truthful in response to my inquiry.

Objectivity� I feel that Joe Smith was impartial when providing his explanation.� Joe Smith may have been biased when responding to my inquiry.� Joe Smith was probably objective when providing his explanation.

Panel B: Items Relating to Linguistic Delivery Style

Hedges� Joe Smith’s explanation seemed to be unclear.� I feel Joe Smith was confident in the explanation he provided.� Joe Smith seemed to be direct when providing his explanation.

Hesitation� Joe Smith’s explanation flowed smoothly.� Joe seemed to hesitate when providing his explanation.� I think that Joe Smith was fluent when providing his explanation.

Grammar� I think that Joe Smith’s explanation was grammatically correct.� I feel that Joe Smith’s explanation was articulate.� Joe Smith’s explanation seemed to be well spoken.

Note: All competence items were adapted from Leather’s Personal Credibility Scale (1992). All

other credibility items were adapted from McCroskey’s Scales for the Measurement of Ethos or

developed by the authors. All linguistic delivery style items were adapted from scales used in

Bradac, Konsky, and Davies (1976) or developed by the authors.


Table 3. Participant Demographics by Location and Treatment.

Panel A: Participant Demographics by Location

Firm 1 Firm 2

Demographic Item 1st Collection 2nd Collection 1st Collection 2nd Collection All

Frequencies

Number of Participants 69 70 29 32 200

Male 31 29 17 14 91

Female 38 41 12 18 109

Current Position: Supervising

Seniors169 69 11 32 181

Current Position: Managers 0 1 18 0 19

Number with Audit Experience

in Manufacturing

49 48 23 22

Means (Standard deviations)

Months of Manufacturing

Experiencea,227.0 (19.9) 24.1 (18.7) 37.3 (29.1) 12.0 (8.1) 25.4 (21.1)

Age (years)3 27.0 (2.63) 27.4 (3.44) 27.8 (1.98) 25.1 (1.54) 26.9 (2.86)

Months Overall Audit

Experience341.7 (10.6) 43.1 (10.4) 67.2 (21.6) 26.7 (5.43) 43.6 (16.6)

Linguistic

Delivery

Style,

Clien

tCred

ibility

,andAudito

rJudgment

69

Table 3. (Continued )

Panel B: Participant Demographics by Treatment

Client Credibility High Low High Low All

Linguistic Delivery Style Powerful Powerful Powerless Powerless All

Frequencies

Number of Participants 52 53 49 46 200

Male 19 22 28 22 91

Female 33 31 21 24 109

Current Position

Supervising Seniors4 52 52 35 42 181

Managers 0 1 14 4 19

Number with Audit Experience in Manufacturing 38 35 36 33 142

Means (Standard deviations)

Months of Manufacturing Experience 23.3 (20.2) 19.9 (15.6) 28.8 (24.4) 29.8 (22.6) 25.4 (21.1)

Age (years)3 26.3 (1.91) 26.3 (3.09) 28.2 (3.45) 27.0 (2.36) 26.9 (2.86)

Months of Overall Audit Experience3 37.6 (9.72) 35.8 (12.1) 53.9 (21.6) 48.0 (13.7) 43.6 (16.6)

aRestricted to those auditors with audit experience in manufacturing.1po0.001 in a 2� 4 w2 test for current position by location.2p ¼ 0.009 in a Kruskal–Wallis nonparametric ANOVA.3po0.001 in a Kruskal–Wallis nonparametric ANOVA.4po0.001 in a 2� 4 w2 test for current position by treatment.

CHRISTIE

L.COMUNALE

ET

AL.

70

likely to be managers (po0.001) relative to those participants who expe-

rienced powerless linguistic delivery style. In addition, the findings show that

participants who experienced high credibility were more likely to be man-

agers (p ¼ 0:034).

Independent Variables

As described above, the two key independent variables in the study are client

credibility (high or low) and linguistic delivery style (powerful or powerless).

To determine whether the client credibility manipulation was successful, the

auditors’ perceptions of client credibility in the high and low groups were

compared. The auditors’ perceptions of linguistic delivery style in the pow-

erful and powerless groups were compared to determine whether the lin-

guistic delivery style manipulation was successful.

Three items were used to represent each of the three dimensions of client

credibility, resulting in nine items, as shown in Table 2. Descriptive statistics

for these items are presented in Table 4, panel A. Similarly, three items were

used to represent each of the three dimensions of linguistic delivery style,

again resulting in nine items, as shown in Table 2. Descriptive statistics for

these items are shown in Table 4, panel B.

The nine items related to client credibility were summed to obtain an

overall construct score for perceived client credibility. (This summation is

justified by an inter-item reliability of a ¼ 0:84:) Similarly, the nine items

related to linguistic delivery style were summed to obtain an overall con-

struct score for perceived linguistic delivery style. (This summation is jus-

tified by an inter-item reliability of a ¼ 0:96:)

Statistical Analysis

To perform a manipulation test on linguistic delivery style, a one-way

ANOVA using perceived linguistic delivery style as the dependent variable

and using manipulated linguistic delivery style (powerful and powerless) as

the factor was conducted. Moreover, separate one-way ANOVAs for each

of the three components of linguistic delivery style were conducted.

To perform a manipulation test on client credibility and to test H1, a two-

way ANOVA with perceived client credibility as the dependent measure and

with manipulated client credibility (high and low) and manipulated linguis-

tic delivery style (powerful and powerless) as the independent factors was


conducted. Moreover, separate two-way ANOVAs for each of the three

components of client credibility were performed.

To test H2 and H3 concerning the impacts of client credibility and lin-

guistic delivery style on auditors’ likelihood assessments and planning de-

cisions, a two-way ANOVA was employed. The two independent factors,

each at two levels, were manipulated – client credibility (high and low) and

manipulated linguistic delivery style (powerful and powerless). A test for the

presence of interaction between the independent factors was also conducted.

Four dependent measures were investigated – likelihood assessment, con-

fidence in likelihood assessment, change in substantive tests, and further

client inquiry.

Table 4. Descriptive Statistics for the Components of Client Credibility

and Linguistic Delivery Style.

Manipulated Client Credibility High Low High Low

Manipulated Linguistic Delivery Style Powerful Powerful Powerless Powerless

Panel A: Client Credibility

Sample Size 52 53 49 46

Perceived Credibility 40.9 31.8 36.5 25.8

7.4 7.9 8.8 6.6

Perceived Competence 17.6 10.7 14.6 7.2

2.5 3.6 4.6 3.3

Perceived Trustworthiness 12.9 11.0 11.3 9.3

3.8 3.6 3.9 3.3

Perceived Objectivity 10.4 10.1 10.6 9.3

3.2 3.0 4.1 2.8

Panel B: Linguistic Delivery Style


Perceived Linguistic Delivery Style 48.5 44.8 20.1 16.7

7.4 7.8 7.4 5.7

Perceived Hedges 15.7 14.6 6.9 5.7

3.2 3.3 3.0 2.3

Perceived Hesitations 16.7 15.4 6.4 5.1

2.7 3.1 2.8 2.0

Perceived Grammar 16.1 14.8 6.9 6.0

2.8 2.9 2.9 2.0

Note: Top entry in each cell is the mean; bottom entry is the standard deviation.


To remove unwanted sources of variation attributable to the participant

demographic variables, gender, current position, existence of audit experi-

ence in manufacturing, months of manufacturing experience, age, and

months of overall audit experience were included in the ANOVA models as

covariates. Results were substantively equivalent in models run with and

without covariates. Therefore, all results are shown without covariates.

RESULTS

Manipulation Testing of Linguistic Delivery Style

The F-value for manipulated linguistic delivery style is 733.0 (po0.001). In

addition, the findings indicate manipulated linguistic delivery style is sta-

tistically significant in all three models (po0.001) for the separate compo-

nents of linguistic delivery style. This confirms that the manipulation of

linguistic delivery style was successful.

Manipulation Testing of Client Credibility and Test of H1

The F-value for manipulated client credibility is 81.71 (po0.001). In addi-

tion, manipulated client credibility is statistically significant in the models

for competence and trustworthiness (po0.001), but not in the model for

objectivity (p ¼ 0:154). With statistically significant differences in the overall

means and in the competence and trustworthiness components, the manip-

ulation of credibility was partially successful. While the manipulation of

objectivity was not successful, the manipulation of the other two compo-

nents of credibility did have the intended effect.

H1 specifies that auditors’ perceptions of client credibility are higher when

manipulated linguistic delivery style is powerful. The F-value for manipu-

lated linguistic delivery style is 22.56 (po0.001). The means for perceived

client credibility for the four experimental conditions and the overall means

for each main effect are shown in Table 4, panel A. As hypothesized by H1,

auditors assigned to the powerful linguistic delivery style condition perceive

a higher mean client credibility rating (36.3) than auditors in the powerless

linguistic delivery style condition (31.3). The results support hypothesis H1

that when linguistic delivery style is powerful, the auditors perceive client

credibility at a higher level, on average.


Tests of H2 and H3

H2 and H3 postulate that high client credibility and powerful linguistic de-

livery style, respectively, influence the four dependent measures. The means

of each dependent variable for each experimental condition are shown in

Table 5. The ANOVA results5 are shown in Table 6. Table 5 shows that the

differences in mean values across the four dependent measures are all in the

direction hypothesized by H2 and H3. Table 6 indicates that the main effects

for client credibility and linguistic delivery style are statistically significant

for three of the four dependent measures, but the interaction term is not

statistically significant in any of the models. Specifically, client credibility is

statistically significant in the models for likelihood of client explanation

(F ¼ 8:26; p ¼ 0:005), likelihood of substantive tests (F ¼ 8:18; p ¼ 0:005),and further client inquiry (F ¼ 6:48; p ¼ 0:012) but not in the model for

confidence in assessment (F ¼ 0:44; p ¼ 0:508).In addition, linguistic delivery style is statistically significant in the models

for likelihood of client explanation (F ¼ 42:00; po0.001), likelihood of

substantive tests (F ¼ 12:46; po0.001), and further client inquiry

(F ¼ 11:92; po0.001) but not in the model for confidence in assessment

(F ¼ 1:11; p ¼ 0:294). Therefore, there is support for the components of H2

and H3 related to likelihood of client explanation, likelihood of substantive

tests, and further client inquiry.

Table 5. Comparison of Means across Experimental Conditions.

Manipulated Client Credibility High Low High Low

Manipulated Linguistic Delivery Style Powerful Powerful Powerless Powerless


Likelihood of Client Explanation 54.5 42.6 30.1 21.9

29.9 26.6 19.1 20.1

Confidence in Assessment 75.7 74.2 73.0 70.3

19.3 18.9 22.4 28.1

Likelihood of Substantive Tests 51.3 65.0 68.0 79.6

32.6 33.8 29.5 28.3

Further Client Inquiry 6.15 6.74 6.86 6.96

1.55 0.90 0.41 0.21

Note: Top entry in each cell is the mean; bottom entry is the standard deviation.


Supplemental Analysis: Investigation of Perceptions and the Dependent

Variables

In addition to testing the effects of the experimental manipulations on the

dependent variables, the effects of perceived client credibility and linguistic

delivery style on the dependent variables are also examined. Multiple re-

gression models are built for each dependent variable using the following

form:

EðY Þ ¼ b0 þ b1CP þ b2LP þ b3ðCPnLPÞ (1)

where

Y ¼ dependent measure;

CP ¼ measure of perceived client credibility;

LP ¼ measure of perceived linguistic delivery style.

Table 7 presents the regression results. Note that, for each dependent

measure, the interaction term in the model is statistically significant. The

Table 6. ANOVA Results.

Source DF Type III SS Mean Square F Value p4F

Likelihood Assessment

Manipulated credibility 1 4978.3 4978.3 8.26 0.005

Manipulated Linguistic delivery style 1 25309.2 25309.2 42.00 o0.001

Interaction 1 166.3 166.3 0.28 0.600

Confidence in Assessment


Manipulated Linguistic delivery style 1 548.0 548.0 1.11 0.294

Interaction 1 17.0 17.0 0.03 0.853

Likelihood of Substantive Tests



Interaction 1 50.1 50.1 0.05 0.821

Further Client Inquiry



Interaction 1 2.9 2.9 3.25 0.073


Table 7. Multiple Regression Results for Each Dependent Variable.

Least Squares Prediction Equation Model Statistics

Likelihood of client explanation Y ¼ 15:87þ 0:037C�p � 0:6781L�

p þ 0:036ðCp � LpÞ F ¼ 55:35 (po0:001)

Adjusted R2 ¼ 0:450

po0:005 for interaction

Confidence in assessment Y ¼ 98:94� 0:0915C�p � 0:744L�

p þ 0:25ðCp � LpÞ F ¼ 2:73 (po0:001)

Adjusted R2 ¼ 0:025

p ¼ 0:010 for interaction

Likelihood of substantive tests Y ¼ 76:88� 0:007C�p þ 0:617L�

p � 0:027ðCp � LpÞ F ¼ 10:10 (po0:001)Adjusted R2 ¼ 0:121

p ¼ 0:054 for interaction

Further client inquiry Y ¼ 6:06þ 0:031C�p þ 0:056L�

p � 0:002ðCp � LpÞ F ¼ 24:42 (po0:001)

Adjusted R2 ¼ 0:261po0:005 for interaction

Note: The p value for the global F-test of H0: b1 ¼ b2 ¼ b3 ¼ 0 are given in parentheses next to the F value.

CHRISTIE

L.COMUNALE

ET

AL.

76

estimates of b3 given in the least squares prediction equation have the ex-

pected signs (positive for likelihood of client explanation and confidence in

assessment; negative for likelihood of substantive tests and further client

inquiry).

If b3 ¼ 0; then the rate of change in the dependent variable with respect to

perceived linguistic delivery style equals b2 in Eq. (1) – a rate of change that

is independent of the level of perceived client credibility. On the other hand,

when b3 differs from zero, the rate of change in the dependent variable with

respect to perceived linguistic delivery style is represented by b2 þ b3CP in

Eq. (1) – a rate of change that depends on the level of perceived client

credibility.

To understand better the nature of the interactions detected, the regres-

sion models are examined at both a low and high level of perceived client

credibility (Neter, Kutner, Wasserman, & Nachtsheim, 1996). The 10th

percentile of the distribution is selected for the low level – the value 22. The

90th percentile of the distribution is selected for the high level – the value 46.

These percentile values are referred to as ‘‘low’’ and ‘‘high’’ perceived client

credibility. At each of these levels, the estimated model parameters are used

to estimate the slope of the regression line relating the dependent measure to

perceived linguistic delivery style. From Eq. (1), this slope is b2 þ b3CP;which represents the change in the dependent variable for a unit increase in

perceived linguistic delivery style when perceived client credibility is fixed at

the value CP. Thus, the slope is b2 þ 22b3 at the low level of credibility and

b2 þ 46b3 at the high level of perceived credibility.

Table 8 shows the estimated slopes and corresponding significance tests

for each regression model; and Fig. 2 presents graphs of the estimated lines

for low and high perceived client credibility. For three of the four dependent

measures, the estimated slopes of the regression lines are not statistically

significantly different from zero when the auditors perceived client credi-

bility is low. In the case of the dependent measure further client inquiry, the

estimated slope is positive and statistically significant. However, this result is

misleading due to the bounded nature of the dependent variable. In fact, the

sample data reveal that all auditors who perceived client credibility at 27 or

below answered the ‘‘inquiry’’ question the same (at a level of ‘‘7’’).6 There-

fore, linguistic delivery style has little influence on any of the four dependent

variables when perceived client credibility is low.

When the auditors perceive client credibility as high, the estimated slopes

of the regression lines for all four dependent measures are statistically sig-

nificantly different from zero. Note also that the slopes are positive for

likelihood of client explanation and confidence in assessment, and negative


for likelihood of substantive tests and further client inquiry. Thus, linguistic

delivery style is associated in the predicted direction with all four dependent

variables when perceived client credibility is high.

The findings indicate that the influence of perceived linguistic delivery

style on the auditor’s likelihood assessments and planning decisions is

greater when perceived client credibility is high. An interaction of this nature

is plausible because, when perceived client credibility is low, there is a con-

servative tendency on the part of the auditor to obtain additional evidence

beyond a client-provided explanation regarding an unusual account fluctu-

ation. Hence, when perceived client credibility is low, a powerful linguistic

delivery style may be insufficient to overcome the auditor’s skepticism and

the auditor will continue to gather additional evidence. On the other hand,

when perceived client credibility is high, powerful linguistic delivery style

may reinforce the faith in the client-provided explanation and limit further

investigation by the auditor.

DISCUSSION

This study introduces linguistic delivery style to auditing research, demon-

strates how linguistic delivery style relates to client credibility, and shows

how the two act independently (and together) to influence auditors’ like-

lihood assessments and planning decisions. The evidence in this study

indicates that client credibility and linguistic delivery style influence

Table 8. Estimated Regression Slopes for Linguistic Delivery Style.

Likelihood of

Client

Explanation

Confidence in

Assessment

Likelihood of

Substantive

Tests

Further Client

Inquiry

Perceived low

credibility

(10th

percentile)

Slope ¼ 0:123 Slope ¼ �0:183 Slope ¼ 0:032 Slope ¼ 0:013

P ¼ 0:414 p ¼ 0:262 p ¼ 0:889 p ¼ 0:037

Perceived high

credibility

(90th

percentile)

Slope ¼ 0:990 Slope ¼ 0:421 Slope ¼ 20:061 Slope ¼ �0:033

P ¼ 0:001 p ¼ 0:007 p ¼ 0:006 p ¼ 0:001

Note: Values are estimated slopes of the dependent variables with respect to linguistic delivery

style, holding client credibility fixed at a low (10th percentile) and high (90th percentile) level.

The observed significance levels (p values) for a test of zero slope are given in parentheses.


auditors’ likelihood assessments and planning decisions in the expected di-

rections.

Moreover, the supplemental results suggest that powerless linguistic de-

livery style from an otherwise high credibility client leads to auditor judg-

ments and intentions that are indistinguishable from those that arise from a

low-credibility client. The interaction, which is statistically significant using

the perceived variables but not statistically significant using the manipulated

0102030405060708090

100

9 13 17 21 25 29 33 37 41 45 49 53 57 61


Lik

elih

oo

d

0

10

20

30

40

50

60

70

80

90

100

9 13 17 21 25 29 33 37 41 45 49 53 57 61


Con

fid

ence

High Credibility Low Credibility


Panel A: Likelihood of Client Explanation

Panel B: Confidence in Assessment

Fig. 2. Graphs Depicting the Interaction Term in Each Dependent Variable.


variables, may be because the perceived variables are measured on a con-

tinuous, rather than a binary, scale, thereby providing more precise meas-

ures of client credibility and linguistic delivery style. This greater precision

allows us to detect an effect that would otherwise remain unseen. Finally,

while researchers have recognized competence, trustworthiness, and

0102030405060708090

100

9 13 17 21 25 29 33 37 41 45 49 53 57 61


Lik

elih

ood

1

2

3

4

5

6

7

8

9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63

Inte

nti

on

To

Req

ues

tFu

rth

er

Cli

ent

Inq

uir

y



Panel D: Request Client Inquiry

Panel C: Likelihood of Substantive Tests


Fig. 2. (Continued).


objectivity as components of credibility, the evidence from this study in-

dicates that linguistic delivery style may be a fourth component.

The good news is that the results indicate that a low-credibility client

cannot alter the auditor’s likelihood assessments and planning decisions

using powerful linguistic delivery style. The bad news, for the client, is that a

high-credibility client with powerless linguistic delivery style may induce

auditor responses equivalent to those of a low-credibility client. This may

occur due to obscurities introduced by the elements of powerless linguistic

delivery style. This may result in greater audit work and greater audit cost to

the client.

These results imply that clients should communicate in powerful linguistic

delivery style. Failure to do so may result in increased audit effort and costs,

the allocation of which will depend on the extent to which the auditor can

charge the client for additional audit work arising from this inefficiency.

Whatever the ultimate allocation, powerless linguistic delivery style may

lead to higher overall audit costs when the client, ironically, has otherwise

high credibility. While formal communication skills have long been empha-

sized as important for auditors (Brune, 2003; Canadian Institute of Char-

tered Accountants, 2000), this study also demonstrates the need for good

interviewing skills, from the interviewee’s perspective.

LIMITATIONS AND FUTURE RESEARCH

This study has two primary limitations that provide opportunities for future

research. First, the scenario limits the examination of linguistic delivery style

to the issue of efficiency (Type I error), which is studied only when sufficient

explanations exist. In other words, it is not possible to address whether

linguistic delivery style influences effectiveness (Type II error) and whether

linguistic delivery style matters when insufficient explanations exist.

Second, the auditor’s likelihood assessment of the client’s explanation and

the confidence in that explanation occur simultaneously with the auditor’s

planning decisions, which are intermediate judgments regarding the need for

additional information. Therefore, we cannot be sure that all of the

dependent measures are not simply capturing the same underlying uncer-

tainty at this intermediate point. Thus, the iterative nature of the process

leading to the final judgment (can the auditor believe the client’s represen-

tation after additional audit work) cannot be studied given the current

design.


The influence of the individual components of linguistic delivery style and

client credibility on the dependent variables cannot be assessed due to the

experimental design utilized in this study. However, the primary concern

was that the manipulation of linguistic delivery style and client credibility

was sufficiently strong to produce the hypothesized effects if indeed they

existed. In fact, the powerless linguistic delivery style and low client cred-

ibility manipulations used in this study were quite strong; hence, the re-

search findings may be amplified.

Every effort was made to ensure that the message content was independ-

ent of the linguistic delivery style. For example, the sentences ‘‘Sales were

lower in the third quarter relative to the second quarter’’ and ‘‘Sales were

um y lower in the third quarter relative to the second quarter’’ carry the

same content despite the introduction of the hesitation ‘‘um.’’

All the participants in the Southeast heard the same linguistic delivery

style manipulation. Similarly, all the participants in the West heard the

other linguistic delivery style manipulation. Since participants may vary in

their sensitivity to linguistic delivery style across geographic locations, this

unintended crossing of linguistic delivery style and region could explain in

part the absence of an interactive effect using the manipulated variables

while an interactive effect was observed using the perceived variables.

In addition, a 7-point scale was used to measure the dependent variable

‘‘Further Client Inquiry,’’ which creates endpoint problems that violate

statistical assumptions used in the analysis. Consequently, this may present

difficulties in analyzing and interpreting results.

Only one scenario was used in this study so the results may not generalize

to other scenarios. Moreover, participants were exposed to audio explana-

tions only, while a video explanation might have been more realistic. In

addition, the auditors in this experiment were not subject to review as they

are in a typical audit engagement. Finally, the focus is on one kind of

undesirable outcome (performing additional audit work when no errors

exist) and a second kind of undesirable outcome (accepting a client’s faulty

explanation when errors do exist) may be of even greater importance to the

auditor.

Despite these limitations, the results of this study represent an important

contribution to the understanding of the client–auditor relationship

by demonstrating the role of linguistic delivery style in the client

inquiry process. In an era of increasing scrutiny of the audit function,

client–auditor communication has become even more crucial to audit

success, and these results can assist auditors in improving this vital audit

process.


NOTES

1. These are words or phrases that show or point out directly, such as the wordsthis, that, and those. For example, ‘‘I never work that hard’’ is weaker than ‘‘I neverwork hard.’’2. Three separate pilot studies were conducted using student participants to test

the efficacy of the linguistic delivery style and client credibility manipulations, as wellas to assess overall understandability of the instrument. Based on pilot study feed-back, the wording of several variable items was refined to improve the clarity ofexperimental scripts.3. Auditors were assigned to rooms according to training session, which were

unrelated to the purposes of this study. There is no reason to suspect that theauditors in one room differed from those in the other room with respect to assess-ment of credibility or sensitivity to linguistic delivery style.4. Initial credibility assessments were not captured at this stage to avoid potential

biases that might occur if participants anchored on these assessments when their finalassessments were made later in the study.5. Because the dependent variables violate the normal distribution and constant

variance assumption, an ANOVA based on ranks is also conducted. The results areessentially identical to the parametric ANOVAs. Therefore, the ANOVA resultsbased on the actual data are reported.6. There are very few observations at the ‘‘low’’ level of credibility. Consequently,

slope estimates at the ‘‘low’’ level yield potentially unreliable results.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the guidance and advice of Drs. Gary L.

Holstrum, James Hunton, Rosann W Collins, and Jacqueline Reck of the

University of South Florida, who were members of Dr. Comunale’s disser-

tation committee. We are also indebted to the referees whose comments and

suggestions greatly improved the quality and the presentation of this chapter.

REFERENCES

Adkins, M., & Brashers, D. E. (1995). The power of language in computer-mediated groups.

Management Communication Quarterly, 8(3), 289–322.

American Institute of Certified Public Accountants (AICPA). (1988). Statement on Auditing

Statements No. 56: Analytical Procedures. New York: AICPA.

American Institute of Certified Public Accountants (AICPA). (2002). Statement on Auditing

Statements No. 99: Consideration of Fraud in a Financial Statement Audit. New York:

AICPA.


Anderson, U. L., Koonce, L., & Marchant, G. (1994). The effects of source-competence in-

formation and its timing on auditor’s performance of analytical procedures. Auditing:

A Journal of Practice and Theory, 13(Spring), 137–148.

Ayers, S., & Kaplan, S. (1998). Potential differences between engagement and risk review

partners and their effect on client acceptance judgments. Accounting Horizons (June),

12(2), 139–153.

Bamber, E. M. (1983). Expert judgment in the audit team: A source reliability approach.

Journal of Accounting Research, 21(Autumn), 396–412.

Beaulieu, P. R. (2001). The effects of judgments of new client’s integrity upon risks judgments,

audit evidence, and fees. Auditing: A Journal of Practice and Theory, 20(September),

85–99.

Berlo, D. K., Lemert, J. B., & Mertz, R. J. (1969). Dimensions for evaluating the acceptability

of message sources. Public Opinion Quarterly, 33(4), 563–576.

Bernardi, R. A. (1994). Fraud detection: The effect of client integrity and competence and

auditor cognitive style. Auditing: A Journal of Practice and Theory, 13(Supplement),

68–84.

Bradac, J. J., Konsky, C. W., & Davies, R. A. (1976). Two studies of the effects of linguistic

diversity upon judgments of communicator attributes and message effectiveness. Com-

munication Monographs, 4, 70–79.

Brown, P. R. (1983). Independent auditor judgment in the evaluation of internal audit func-

tions. Journal of Accounting Research, 21(Autumn), 444–455.

Brune, C. (2003). The artful interviewer. The Internal Auditor, 60(2), 25–27.

Canadian Institute of Chartered Accountants (CICA). (2000). Audit enquiry: Seeking more

reliable evidence from audit enquiry. Toronto, Canada. (CICA).

Erickson, B., Lind, A. E., Johnson, B. C., & O’Barr, W. M. (1978). Speech style and impression

formation in a court setting: The effects of ‘‘powerful’’ and ‘‘powerless’’ speech. Journal

of Experimental Social Psychology, 14(May), 266–279.

Goodwin, J. (1999). The effects of source integrity and consistency of evidence on auditors’

judgments. Auditing: A Journal of Practice and Theory, 18(Fall), 1–16.

Haleta, L. L. (1996). Student perceptions of teachers’ use of language: The effects of powerful

and powerless language on impression formation and uncertainty. Communication in

Education, 45(January), 16–28.

Harms, L. S. (1961). Listener judgments of status cues in speech. Quarterly Journal of Speech,

47, 164–168.

Hirst, D. E. (1994). Auditor’s sensitivity to source reliability. Journal of Accounting Research,

32(Spring), 113–126.

Hosman, L. A., Wright, J. W. (1987). The effects of hedges and hesitations on impression

formation in a simulated courtroom context. Western Journal of Speech Communication,

51(Spring), 173–188.

Joyce, E. J., Biddle, G. C. (1981). Are auditor’s judgments sufficiently regressive? Journal of

Accounting Research, 19(Autumn), 323–349.

Kaplan, S. E., & Reckers, P. M. J. (1984). An empirical examination of auditors’ initial plan-

ning process. Auditing: A Journal of Practice and Theory (Fall), 4(1), 1–19.

Koonce, L., & Phillips, F. (1996). Auditor’s comprehension and evaluation of client-suggested

causes in analytical procedures. Behavioral Research in Accounting, 8, 32.


Lind, E. A., & O’Barr, W. M. (1979). The social significance of speech in the courtroom. In:

H. Giles & R. St Clair (Eds), Language and social psychology (pp. 66–87). England: Basil

Blackwell (U.S. distribution, Baltimore MD: University Park Press).

Mautz, R. K. (1958). The nature and reliability of audit evidence. The Journal of Accountancy

(May), 105(5), 40–47.

McCroskey, J. C. (1966). Scales for the measurement of ethos. Speech Monographs, 33, 65–72.

McGuire, W. J. (1973). Persuasion, resistance, and attitude change. In: I. Pool (Ed.), Handbook

of Communication (pp. 16–252). Chicago, IL: McNally College Publishing Company.

Miller, G. R., & Hewgill, M. A. (1964). The effect of variations in nonfluency on audience

ratings of source credibility. Quarterly Journal of Speech, 50, 36–44.

Neter, J., Kutner, M. H., Wasserman, W., & Nachtsheim, C. J. (1996). Applied linear statistical

models (4th ed.). New York, NY: McGraw-Hill/Irwin.

O’Barr, W. M. (1982). Linguistic evidence: Language, power, and strategy in the courtroom.

New York: Academic Press.

O’Keefe, D. J. (1990). Persuasion: Theory and research. Newbury Park: Sage Publications.

Peecher, M. E. (1996). The influence of auditors’ justification processes on their decisions:

A cognitive model and experimental evidence. Journal of Accounting Research (Spring),

34(1), 125–140.

Rebele, J. E., Heintz, J. A., & Briden, G. E. (1988). Independent auditor sensitivity to evidence

reliability. Auditing: A Journal of Practice and Theory, 8 (Fall), 43–52.

Vinson, L., & Johnson, C. (1989). The use of written transcripts in powerful and powerless

language research. Communication Reports, 2, 16–21.

Wright, J. W., & Hosman, L. A. (1983). Language style and sex bias in the courtroom: The

effects of male and female use of hedges and intensifiers on impression formation.

Southern Speech Communication Journal, 48, 137–152.

APPENDIX

Client Credibility Manipulations

High Client Credibility

During your controls testing, you had one other interaction with Joe Smith.

At that time, it was revealed to you that Joe possessed both a CPA and an

MBA and has over 20 years of accounting experience (competence). More-

over, during this previous interaction, he seemed, in your opinion, to pro-

vide candid responses to your questions (trustworthiness). As far as you

know, Joe had no vested interest that would have caused him to provide

biased responses (objectivity).

Low Client Credibility

During your controls testing, you had one other interaction with Joe Smith.

At that time, it was revealed to you that this is Joe’s first year as assistant

controller. Moreover, it is your understanding that Joe does not possess any


professional certifications and has not completed any post-graduate work

(competence). During this previous interaction, you were not quite sure

whether Joe was entirely candid when responding to your questions (trust-

worthiness). As far as you know, Joe could have a vested interest that would

have caused him to provide biased responses (objectivity).

Linguistic Delivery Style Manipulations

Powerful Linguistic Delivery Style

The year over year increase in inventory is part of our overall strategic plan.

One of our goals at Weber is to be the leading provider of appliances in the

Southeast. To accomplish this goal we must deliver quality products to our

customers, on time, every time. Unfortunately, we received numerous com-

plaints last year regarding product shortages. Therefore, we increased the

production of our more popular models. The elimination of product short-

ages at our retail outlets will greatly improve customer satisfaction.

Powerless Linguistic Delivery Style

Hmmm (hesitation)yI’m thinkin’ (grammar and hedge) that, that has

somethin’ (grammar) to do with our um (hesitation)y.overall plan. A goal

here, you know, (hedge) is to be ah (hesitation)yleading provider of ap-

pliances. So, I believe (hedge) we’ve been tryin’ (grammar) to deliver prod-

ucts, quality products, on time. I’m pretty sure (hedge) that we got

(grammar) some complaints last year about runnin’ outta (grammar) our

stuff. Maybe (hedge) we started increasin’ (grammar) the production, prob-

ably (hedge) of our better sellers. By steppin’ (grammar) up production, we

ah (hesitation) ycan maybe (hedge) make our customers happy.


CLIENT INQUIRY VIA

ELECTRONIC COMMUNICATION

MEDIA: DOES THE

MEDIUM MATTER?

Anna Noteberg and James E. Hunton

ABSTRACT

Face-to-face meetings between auditors and clients are becoming increas-

ingly more difficult and expensive to arrange, due in large part to the

ceaseless expansion of commerce across the globe. Relying on electronic

communication media such as e-mail messaging or video-conferencing for

auditor–client inquiry purposes is one way to enhance the timeliness of

such communications; however, questions arise with respect to potentially

biasing influences of certain technical aspects of electronic media on au-

ditors’ judgment and decision-making processes. Drawing on information

processing theories, the current study posits that media and message at-

tributes can interact, thereby differentially affecting auditors’ belief re-

visions – holding information content constant. The media attributes

examined in the current study are cue multiplicity (i.e., the range of

central and peripheral cues a medium is capable of transmitting) and

message reprocessability (i.e., the extent of archival and retrieval features

a medium is capable of handling); and the message attribute studied is

evidence strength (e.g., the credibility of client-provided evidence).





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08004-4

87

Research findings from a laboratory experiment with 189 graduate ac-

counting students indicate the following: (1) when client-provided evi-

dence is strong, neither message reprocessability nor cue multiplicity

significantly affect the auditors’ belief revisions; (2) when evidence is

weak and reprocessability is present, higher cue multiplicity leads to sig-

nificantly greater belief revision in favor of the client; (3) when evidence

is weak and reprocessability is absent, lower cue multiplicity results in

significantly greater belief revision in favor of the client. Study results

suggest theoretical and practical implications for globally distributed

auditor–client communications.

INTRODUCTION

The use of electronic communication media within and among organiza-

tions is increasing at an exponential rate (Strauss & McGrath, 1994; Baltes,

Dickson, Sherman, Bauer, & LaGanke, 2002). The integration of digital

communication technologies into business entities provides opportunities

for people in geographically dispersed locations to rapidly contact one an-

other for a variety of purposes, such as sharing information, discussing

ideas, forming relationships, and making decisions. The primary driver of

this technological trend is the growing dispersion of commercial enterprises

across the globe.

Audit firms are expanding their organizational boundaries across the

world as well, primarily because their clients are globally extending their

commercial reach. Thus, auditors are making increasing use of electronic

communication media for intra-firm and auditor–client communications, as

the relatively slow speed and high cost of face-to-face exchanges are quite

constraining. While one objective of incorporating electronic media into

auditor–client communications is to enhance the timeliness of such ex-

changes, questions arise as to whether certain media and message attributes

can bias the interpretation of the message content.

The purpose of the current study is to examine the potentially biasing

effects of two media attributes and one message attribute. The first media

attribute, cue multiplicity, reflects the range of central and peripheral mes-

sage cues that can be transmitted by a particular medium. The second media

attribute, message reprocessability, refers to the extent to which a medium is

capable of archiving and retrieving messages. The message attribute under

examination is evidence strength, or the credibility of evidential matter

ANNA NOTEBERG AND JAMES E. HUNTON88

underlying a client’s argument. Information processing theories suggest that

the media and message attributes will interact such that auditors’ belief

revisions will be differentially affected, depending on the attribute states.

In this study, we develop hypotheses based on the belief adjustment model

(Einhorn & Hogarth, 1985; Ashton & Ashton, 1988; Hogarth & Einhorn,

1992) and dual processing models (Chaiken, 1980; Petty & Cacioppo, 1981)

of information processing. We conduct a 2� 2� 2 between-subjects exper-

iment to test the hypotheses. The independent variables are cue multiplicity

(low, high), message reprocessability (absent, present), and evidence

strength (weak, strong). The scenario involves an auditor who identifies a

potentially serious problem – a large part of the client’s inventory is seem-

ingly obsolete and the auditor believes that the inventory may be over-

valued by a material amount. Using electronic media for communication,

the chief financial officer (CFO) expresses disagreement with the auditor’s

conclusions and explains why an inventory write-down is not necessary. The

dependent variable reflects the extent of belief revision, in favor of the cli-

ent’s position, exhibited by the participants.

Study findings indicate that belief revision is unaffected by cue multiplic-

ity, message reprocessability or the interaction of both when the client-

provided evidence is strong. However, when evidence is weak, the effect of

cue multiplicity inverts depending on the state of message reprocessability.

Specifically, in the presence of reprocessability, higher cue multiplicity leads

to greater belief revision toward the client’s preferred position, as predicted;

unexpectedly however, when reprocessability is absent, lower cue multiplic-

ity results in greater belief revision in favor of the client. This is the first

study to examine how auditors’ belief adjustments can be differentially af-

fected by interactions among certain electronic communication media and

message attributes. The value of this research lies in its contribution to

extant information processing theory and audit practice.

The next section discusses background literature and develops study hy-

potheses. The following sections explain the research method, present the

study findings, and discuss the implications for information processing the-

ory and audit practice.

BACKGROUND AND HYPOTHESES

Auditor–Client Inquiry

The planning phase of an audit includes performing analytical review pro-

cedures aimed at directing attention to potential problem areas. Analytical

Client Inquiry Via Electronic Communication Media 89

review procedures involve four phases: (1) developing an understanding of

the client’s business activities and searching for unusual fluctuations in a

company’s financial statements (mental representation); (2) generating po-

tential causes for any observed inconsistencies (hypothesis generation);

(3) gathering diagnostic information (information search); and (4) analyzing

findings to identify the most likely hypothesis and arrive at a diagnosis

(hypothesis evaluation) (Koonce, Walker, & Wright, 1993). The current

study focuses on the information search phase of analytical review proce-

dures. More specifically, this study examines a particular information source

– client inquiry (Hirst & Koonce, 1996).

During analytical review, auditors might notice one or more unusual

relationships in the clients’ account representations. When this happens,

auditors often ask managers to explain how and why such situations arose.

During inquires, auditors are aware that managers’ explanations can be

biased toward supporting the underlying transactions giving rise to the un-

usual relationships. Accordingly, auditors are professionally skeptical in

such situations. Possible reasons why managers can bias their arguments

toward supporting the current state of accounts are found in agency theory

(e.g., Jensen & Meckling, 1976).

One assumption in agency theory is that managers (agents) will attempt to

take self-serving advantage of private information they possess of which

owners (principals) are unaware. Intervening in the principal–agent rela-

tionship are independent auditors who represent the principals’ interests.

Agency theory suggests that managers with a higher (lower) motivation to

maintain their information-asymmetric advantage over the principals will

likely be less (more) honest with and forthcoming toward auditors (Hirst &

Koonce, 1996). Under such circumstances, when auditors attempt to gather

information via client inquiry about unusual account fluctuations, managers

with higher levels of self-serving interests might try harder to persuade the

auditors to accept current account representations.

From both academic and practical standpoints, it is important to under-

stand how persuasive intentions of this nature can influence auditors’ belief

revision processes and judgments, as auditors might unknowingly make

suboptimal choices and decisions under certain conditions (e.g., Jensen &

Meckling, 1976). The condition of interest in the current study involves the

electronic communication media through which auditor–client communi-

cations are exchanged.

Although auditors and clients may be scattered across the world, they

nevertheless need to communicate with each other about important audit

issues in a timely manner. One way for auditors and clients to establish


prompt communication exchanges is through the use of electronic commu-

nication media, such as e-mailing or video-conferencing. While this solution

might be efficient from time and cost perspectives, questions arise regarding

the effectiveness of such media. The media-task fit model developed by

Noteberg, Benford, and Hunton (2003) proposes that the effectiveness of

auditor–client inquires are affected by the fit between task and technology

attributes. In the current study, we test a proportion of the media-task fit

model, and predict that certain media attributes can unintentionally bias

auditors’ belief revisions and judgments, as next discussed.

Belief Adjustment

According to the anchoring-and-adjustment heuristic, individuals anchor on

initial beliefs regarding a subject matter, obtain and evaluate additional

evidence, and then adjust their initial anchor accordingly. The belief ad-

justment model (BAM) (Einhorn & Hogarth, 1985; Ashton & Ashton, 1988;

Hogarth & Einhorn, 1992) predicts that individuals revise their beliefs on

the basis of evidence direction and strength. The current study holds ev-

idence direction constant, meaning that the client presents evidence that

attempts to disconfirm the auditor’s initial belief state. This factor is held

steady because it reflects the audit condition of interest: a client is attempt-

ing to persuade an auditor to accept a current account representation.

However, the client may provide either strong or weak evidential cues to the

auditor in this circumstance. While the positive main effect of evidence

strength on judgments is not a new issue to investigate (e.g., Hogarth &

Einhorn, 1992), the current study is unique, as it examines the potential

interaction of evidence strength with two electronic communication media

attributes – reprocessability and cue multiplicity.

Reprocessability

Reprocessability refers to an electronic communication media feature that

allows communicators to re-examine messages (Dennis & Valacich, 1999).

Reprocessability is particularly prominent in electronic media that allow for

vast information storage, such as e-mail or voice mail. It is a media attribute

that acts as an ‘‘externally recorded memory’’ (Sproull, 1991) and thereby

aids in understanding the situation, particularly as the volume, complexity


and equivocality of messages increase (Noteberg et al., 2003). The current

study examines the effect of reprocessability on auditors’ belief revisions.

Two processing strategies implicit in the BAM (sequential and simulta-

neous) are reviewed to explain potential effects of reprocessability on belief

revision. According to the BAM, individuals use one of two information-

processing strategies when faced with multiple pieces of evidential matter.

Sequential processing takes place when individuals adjust their beliefs in-

crementally after evaluating each piece of evidence. Simultaneous processing

means that individuals adjust their initial anchor only after evaluating the

full aggregate set of evidential matter.

According to the ‘‘dilution effect’’ (Ashton & Ashton, 1988), disconfir-

mation-prone individuals make stronger belief revisions when information is

elicited in a sequential format, as compared to a simultaneous format.

Ashton and Ashton (1988) note that auditors are generally disconfirmation-

prone; therefore, sequential evaluation of consistently negative evidence will

result in more extreme belief change than simultaneous processing of the

same evidence. Consequently, belief revision should be stronger when in-

dividuals process evidence sequentially as opposed to simultaneously. The

primary reason for the dilution effect is that sequential evidence processing

offers more opportunities to anchor and adjust than does simultaneous

processing, thereby resulting in more extreme revisions in the direction of

the evidence (Francis & Schipper, 1999).

Returning to the media attribute of interest, let us first contemplate the

use of a medium that lacks reprocessability, i.e., messages are not stored for

later re-examination. For instance, if an auditor holds a telephone conver-

sation with a client during which multiple issues are discussed, the auditor is

likely to process the evidential matter simultaneously upon reflection of the

whole conversation. One could argue that the auditor sequentially processes

information as received during the conversation, but research evidence sug-

gests that cognitive conversation processes (e.g., listening, speaking, adjust-

ing, and coordinating) effectively block sequential anchoring and adjusting

(Schober & Brennan, 2003). Even if some degree of sequential processing

takes place during conversation, the lack of a message archive effectively

prevents post-conversation sequential adjusting (Schober & Brennan, 2003).

Hence, simultaneous post-conversation processing of all evidential matter is

likely to occur.1

Next, assume that the medium used for message conveyance allows the

auditor to re-examine evidential cues after an initial exposure to evidence

collection. To the extent that auditors utilize a medium’s reprocessing ca-

pability, the subsequent assessment of each piece of information can evoke a


sequential processing strategy.2 Hence, we expect that the extent of belief

revision will increase with the presence, as compared to the absence, of

reprocessability, as stated in the first Hypothesis

H1. There is a positive relationship between reprocessability and belief

revision.

While we assume that the second media attribute examined in this study (cue

multiplicity) will also positively affect belief revision, we posit that its impact

will vary depending on the interaction of evidence strength and the medi-

um’s reprocessing capability, as next discussed.

Cue Multiplicity

While the BAM only considers content-related evidence cues in its predic-

tion of individuals’ belief revision processes, we extend the model by incor-

porating peripheral (i.e., non-content related) cues. If given the opportunity,

it is likely that message senders with persuasive intentions (e.g., clients) will

employ peripheral cues (e.g., body language and voice intonation) to per-

suade the recipients (e.g., auditors) of their messages. Given that some

electronic media allow for the conveyance of such peripheral cues while

others do not, it is important to consider both central and peripheral cues, in

order to understand decision-makers’ potential reactions to persuasive

evidence.3

According to dual processing models, such as the elaboration likelihood

model (Petty & Cacioppo, 1981) and the heuristic-systematic model

(Chaiken, 1980), message recipients can sequentially or simultaneously fol-

low two routes when processing persuasive information – a central route

(systematic) and a peripheral (heuristic) route.4 When following the central

route, message recipients systematically scrutinize the validity of arguments

(i.e., central cues) contained in a persuasive message, and revise their initial

beliefs based on the content-related cues included in the message. The cur-

rent study incorporates two levels of evidence strength (weak and strong)

into the experimental design. The purpose of manipulating the validity of

arguments or central cues in this manner is to discriminate message effects,

which are processed via the central route, from media effects, which trigger

peripheral route processing.

When following a peripheral information-processing route, decision-

makers factor secondary, personal, and social cues into their belief revisions,

which may or may not be related to the central message (Chen & Chaiken,


1999). For instance, peripheral cues such as source attractiveness and voice

inflections can hold persuasive powers without demanding considerable

mental effort. In a laboratory experiment, Comunale, Sexton, and Sincich

(2005) demonstrated that auditors were affected by the linguistic manner

(e.g., hedges and hesitations in speech) in which a client delivered expla-

nations. They found that auditors incorporate linguistic delivery style into

their evaluation of client credibility, thereby supporting the assumption that

peripheral cues play an important role in auditors’ belief revisions.

In the current study, cue multiplicity represents the extent to which the

medium makes available various peripheral cues. Communication media

may hold a high or low level of cue multiplicity. Media that restrict the

availability of peripheral factors offer a low level of cue multiplicity (e.g.,

e-mail messages), whereas those that provide access to such cues provide a

high level of cue multiplicity (e.g., audio–video presentations).

Previous research has examined the impact of cue multiplicity on various

communication outcomes, such as effectiveness and efficiency of managers’

decisions (e.g., Daft & Lengel, 1986; Daft, Lengel, & Klebe Trevino, 1987;

Kraut, Galegher, Fish, & Chalfonte, 1992; Rice, 1992; Hollingshead,

McGrath, & O’Connor, 1993; Dennis & Kinney, 1998; Suh, 1999), attitude

change (Matheson & Zanna, 1989), and communicator likeability

(Weisband & Atwater, 1999). The current study examines the effect of cue

multiplicity on auditors’ belief revisions, assuming that the peripheral cues

support the central message.

As suggested by Chaiken and Eagly (1983), we posit that high cue mul-

tiplicity enables peripheral processing by drawing the message recipient’s

attention to the sender’s peripheral cues. Thus, assuming that peripheral

signals are directionally aligned with the central cues, a medium with high

cue multiplicity is expected to enhance the persuasive power of the message,

as compared to a medium with low cue multiplicity. However, we also

hypothesize that the belief revision effect of cue multiplicity is moderated by

the strength of the evidence and the presence or absence of reprocessability,

as reviewed next.

Interactive Effects

When evidence is weak and reprocessability is absent, greater cue multi-

plicity is likely to have a positive effect on belief revision because the pe-

ripheral cues are supporting the central message. When evidence is weak and

reprocessability is present, the positive effect of cue multiplicity will be


amplified significantly because individuals are repeatedly exposed to the

(weak) central message and (supportive) peripheral cues. Thus, we offer

Hypothesis 2 (see Fig. 1 for a graphic depiction):

H2a. When evidence is weak and reprocessability is absent, belief revision

is greater when cue multiplicity is high as compared to low.

H2b. When evidence is weak and reprocessability is present, belief revi-

sion is greater when cue multiplicity is high as compared to low.

H2c. When evidence is weak, the difference in belief revision between high

and low cue multiplicity is greater in the presence, as compared to absence,

of reprocessability.

Given strong evidence and the absence of reprocessability, increased cue

multiplicity is still expected to result in significantly greater belief adjust-

ment. However, we predict no effect of cue multiplicity when strong evi-

dence can be reprocessed. This prediction considers the occurrence of a

ceiling effect when evidence is strong and reprocessability is present. We

argue that the combined effect of strong evidence and reprocessability will

Reprocessability

absent

Reprocessability

present

Pre

dic

ted

po

st-

test

belief

Low C

Weak Evidence

ue Multiplicity

High Cue Multiplicity

H2c = H2a < H2b

H2a

H2b

Fig. 1. Hypothesis 2: Predicted Downward Belief Revision in Favor of the Client’s

Desired Position. (Scenario: The Auditor Initially Sets a Relatively High Belief An-

chor with Regard to a Proposed Adjusting Journal Entry. Afterward, the Client

Explains Why the Adjusting Journal Entry is Unnecessary. The Auditor then Revises

His/Her Initial Belief. A Downward Revision Suggests that the Auditor is Lowering

His/Her Initial Belief in Favor of the Client’s Position.)


result in decision-makers reaching a maximum belief revision. Adding pe-

ripheral cues to the message will not make a significant difference because

the decision-makers have reached a belief revision ceiling. Accordingly,

Hypothesis 3 is offered (see Fig. 2 for an illustration).

H3a. When evidence is strong and reprocessability is absent, belief revi-

sion is greater when cue multiplicity is high as compared to low.

H3b. When evidence is strong and reprocessability is present, belief re-

vision is not significantly different when cue multiplicity is high as com-

pared to low.

H3c. When evidence is strong, the difference in belief revision between

high and low cue multiplicity is greater in the absence, as compared to

presence, of reprocessability.

Reprocessability

absent

Reprocessability

present

Pre

dic

ted

po

st-

test

belief

Low Cue Multiplicity


H3c = H3a > H3b

H3a

H3b

Strong Evidence

Fig. 2. Hypothesis 3: Predicted Downward Belief Revision in Favor of the Client’s

Desired Position. (Scenario: The Auditor Initially Sets a Relatively High Belief An-

chor with Regard to a Proposed Adjusting Journal Entry. Afterward, the Client

Explains Why the Adjusting Journal Entry is Unnecessary. The Auditor then Revises

His/Her Initial Belief. A Downward Revision Suggests that the Auditor is Lowering

His/Her Initial Belief in Favor of the Client’s Position.)


METHOD

We designed and administered a computerized laboratory experiment to test

the research hypotheses. The experiment reflects a 2 (cue multiplicity: low,

high) �2 (message reprocessability: absent, present) �2 (evidence strength:

weak, strong) between-subjects design. Participants were randomly assigned

to treatment conditions.

Experimental Procedure

Participants were instructed to assume the role of an audit partner of a large

accounting firm. The client firm was a computer manufacturer called ‘‘Mi-

croClone.’’5 The senior manager on the audit had uncovered a potential

problem with the client’s finished goods inventory, which was currently

valued at h2 million. Specifically, due to the recent introduction of 5th

generation computers, part of the inventory (4th generation computers)

might be over-valued by about h400,000 (see the appendix for complete

wording of the background information presented to all participants). Next,

the company’s CFO reacted to the auditor’s concern with five arguments

defending why the 4th generation computers should not be written down.6

All arguments, whether strong or weak, attempted to persuade the auditor

not to book the recommended inventory write-down.

Experimental Manipulations

Evidence strength is a function of source objectivity (e.g., Abdel-Khalik,

Snowball, & Wragge, 1983), source independence (e.g., Brown, 1983), and

evidence verifiability (e.g., Spires, 1991; Goodwin, 1999). All three aspects

were used to manipulate the relative strength of CFO-provided arguments.

Following each of the five arguments, the CFO referred to a secondary

source who confirmed the argument. In the weak evidence condition, the

CFO referred to an internal source (i.e., a mid-level employee). This treat-

ment was designed to weaken the perception of source objectivity and

independence. Strong arguments contained the same message content as the

weak arguments, but the CFO referred to ‘well-known experts’ in the field

who confirmed the arguments in published articles. The use of an external

source was intended to strengthen the perception of source objectivity and

independence. To further reinforce the strong argument manipulation, the

senior manager on the audit verified the genuineness of the published article;


on the other hand, participants in the weak message condition were not ex-

plicitly told that the evidence provided by the secondary source was verified.7

Following the first round of arguments, participants in the reprocessabil-

ity absent (present) condition were not (were) allowed to re-examine the

CFO-provided arguments. To ensure that participants in the ‘reprocess-

ability present’ condition actually reprocessed the messages, the computer

software presented for a second time each of the five arguments (in random

order) before allowing the participants to record their post-argument belief

state.8

We manipulated cue multiplicity by presenting the CFO’s arguments in

either e-mail or audio–video format. E-mail is low in cue multiplicity, as it

does not allow for any peripheral cues, such as body language and voice

intonation. In the audio–video clips, the CFO conveyed the arguments using

the same wording as the e-mails. The actor was male, around 50 years of

age, dressed in a dark suit and wore glasses, i.e., a stereotypical business

professional. The video clips showed his head and upper torso in an office

environment, i.e., seated at a desk in daylight next to a computer. The actor

asserted the same serious tone in his voice across the five clips. His pres-

entation style was free of hedges, hesitations, and faulty grammar in an

attempt to induce a perception of high source credibility (Comunale et al.,

2005). Thus, the peripheral cues present in the audio–video clips, but not

available in the e-mail treatment, were (1) professional looking businessman,

(2) sincere look on his face, (3) credible grammar, (4) serious voice tone, and

(5) orderly business office environment.

Dependent Measure

Recall that the CFO in the case provided evidence aimed at persuading the

auditor to revise his/her initial belief from a relatively high (strong) position

to a lower (weaker) state. Hence, the dependent variable used to test the

hypotheses was the extent of downward belief revision regarding the pro-

posed inventory valuation write-down, after being exposed to the client’s

arguments. The participants responded to the following statement: ‘‘Given

the available information, I strongly believe that MicroClone should write

down its 4th generation inventory by h400,000’’ (1 ¼ strongly disagree, 7 ¼

strongly agree). The participants’ initial belief anchor was first measured

after reading the background information. Then, once the participants were

exposed to the five counter-arguments offered by the chief executive officer

(CEO) (i.e., the treatments), they responded to the same statement once


again. Thus, the dependent variable metric is calculated as the post-test

belief, adjusted for the effect of the covariate (pre-test belief). This metric is

referenced as ‘‘belief revision.’’

RESULTS

Sample Demographics

The sample is comprised of 189 part-time graduate students in accounting

with prior work experience. Individual cell sizes for all eight conditions

range from 21 to 26, with a median cell size of 23 and a mode of 26. The

mean (s.d.) age of participants is 26.67 (5.21) and the age range is between

20 and 52 years. There are 123 male participants (65.1%) and 66 (34.9%)

female participants. The number of participants in their first, second, and

third year of graduate education, respectively, are 131 (69.3%), 38 (20.1%),

and 20 (10.6%). Most participants have some work experience in the field of

accounting and/or auditing; specifically, 13 (6.9%) indicate no work expe-

rience, 21 (11.1%) record less than a year of experience, 44 (23.3%) specify

between 1 and 2 years of experience, 96 (50.8%) note between 3 and 5 years

of experience, 13 (6.9%) hold between 5 and 10 years of experience, and 1

(0.5%) possesses more than 10 years of work experience. At the time of the

experiment, each participant was enrolled at one of two Dutch universities.9

Manipulation Checks

Participants were asked the extent to which they agreed with a statement

indicating that the messages were strong (1 ¼ strongly disagree, 7 ¼

strongly agree). The response means (s.d.) are as follow: weak evidence

treatment 4.44 (1.81) and strong evidence treatment 5.16 (1.46). An

ANOVA indicates statistical significance between the two treatments

(F ¼ 8:88; po0:01). As a further check on evidence strength, participants

were asked to state the extent to which they agreed (1 ¼ strongly disagree,

7 ¼ strongly agree) with the following statements: (1) ‘‘The arguments in

favor of a lower write-down provided by Tom van Breukelen were con-

vincing,’’ and (2) ‘‘The arguments in favor of a lower write-down provided

by Tom van Breukelen were credible.’’ The mean (s.d.) for the two questions

are: (1) convincing – 3.90 (4.58) for weak evidence, 4.76 (4.58) for strong evidence; (2)

credible – 4.57 (1.63) for weak evidence, 5.28 (1.35) for strong evidence. Based on

ANOVA tests, the means for both questions are significantly different (po0:001)


between the weak and strong evidence conditions, and they are directionally consistent

with expectations. Hence, the manipulation of evidence strength was successful.

Participants also stated the extent to which they agreed with a statement

indicating that they were allowed to review the five arguments after receiving

the first exposure but before recording their second belief (1 ¼ strongly

disagree, 7 ¼ strongly agree). The response mean (s.d.) is 3.72 (2.46) for the

‘reprocessability absent’ group and 6.31 (1.46) for the ‘reprocessability

present’ group. An ANOVA indicates a significant difference between the

two means (F ¼ 78:16; po0:01), suggesting a successful manipulation of

reprocessability.

To test for cue multiplicity, participants were asked whether they were

exposed to the CFO’s arguments via e-mail or audio–video media. All par-

ticipants in the e-mail condition correctly responded to this question. How-

ever, out of 98 subjects in the high cue multiplicity condition, 16 stated that

they had been exposed to e-mail messages. One possible reason for some

participants responding incorrectly could be that they misinterpreted the

question; that is, all participants were exposed to written attachments fol-

lowing each e-mail or audio–video message (i.e., either an internal or ex-

ternal source verified each argument in writing); hence, participants in the

high cue multiplicity condition might have thought that the verifying at-

tachments were e-mail messages when responding to this manipulation

check question. A Pearson w2 test (w2 ¼ 136:99; po0:01) indicates a signif-

icant difference between treatment groups, and additional sensitivity testing

of participants in the high cue multiplicity condition who did and did not

respond correctly to this question reveals no significant difference in their

pre–post belief revisions (p40:90). Thus, the manipulation of cue multi-

plicity was successful.

Parametric Assumptions

To assess whether the post-test belief scores meet the normality assumption,

Kolmogorov–Smirnov and Shapiro–Wilk tests were conducted. Both tests

are significant at po0:001; indicating that the data are not normally dis-

tributed. However, between-subjects ANCOVA models are robust to non-

normality if the skewness of each treatment condition is in the same direc-

tion and the largest variance is o4 times the smallest variance (Torrance,

2002). The study meets both conditions, as response distributions in all eight

treatment conditions are positively skewed and the largest variance

(s ¼ 4:06) is 3.87 times the smallest variance (s ¼ 1:05). An additional


analysis for homogeneity of variance (Levene’s test) is non-significant

(F ¼ 1:139; p ¼ 0:341), indicating that cell variances are statistically equiv-

alent across experimental conditions. Based on the results of normality and

variance testing, we conclude that the use of ANCOVA for hypothesis test-

ing is appropriate.10

The design employed in this study is a pre–post measurement design, for

which ANCOVA is a commonly employed method. The initial belief anchor

measure is incorporated as a covariate in the final model to adjust for pre-

test differences. ANCOVA was used to test for main and interactive effects

of message reprocessability (present or absent), evidence strength (high or

low), and cue multiplicity (high or low) on belief revision.

ANCOVA Model

Treatment means, standard deviations, and sample sizes are shown in

Table 1, and the ANCOVA results are presented in Table 2. The covariate,

pretest belief, is significant (po0:001), the main effect of evidence strength is

marginally significant (p ¼ 0:092), a two-way interaction between cue mul-

tiplicity and reprocessability is significant (p ¼ 0:009), and the three-way

interaction among message reprocessability, evidence strength and cue mul-

tiplicity is significant (p ¼ 0:025). The results of hypothesis testing are

shown next.

Hypothesis Testing

H1 predicts a positive effect of reprocessability on belief revision. This hy-

pothesis is not supported (see Table 2), as the main effect for

reprocessability is non-significant (p ¼ 0:834). However, as indicated by

the significant three-way interaction, the effect of reprocessability is con-

tingent on the state of the other two variables – evidence strength and cue

multiplicity. To test H2 and H3, planned comparison tests were conducted

and are shown in Table 3.

H2a predicts that, in light of weak evidence and no reprocessability, higher

cue multiplicity will result in greater belief revision. As illustrated in Fig. 3

and reported in Table 3, the means are significantly different (difference

score ¼ �0:90; p ¼ 0:022), but in the opposite direction as expected; hence,

H2a is not supported. Testing for H2b indicates significantly different means

(difference score ¼ 1:02; p ¼ 0:006) in the anticipated direction; thus, H2b is

supported. To test for H2c, the high cue multiplicity mean was subtracted


from the low cue multiplicity mean in each reprocessability treatment, and

the difference-in-difference between the reprocessability present and absent

conditions was compared. As predicted, the ‘reprocessability present’ mean

difference (1.02) is greater than the ‘reprocessability absent’ mean difference

ð�0:90Þ; as the difference-in-difference of 1.92 ½1:02� ð�0:90Þ� is significant(t ¼ 10:69; po0:001). Thus, H2c is conditionally supported; that is, when

interpreting H2c, one must consider that the relationship in H2a is opposite

from expectations.

H3 holds evidence strength constant at strong (see Fig. 4 for an illustration

of the results). We first examine the difference in belief revision means

between high and low cue multiplicity when reprocessability is absent

(Table 3). The difference in means ð�0:31Þ is in the opposite direction from

expectations and non-significant (p ¼ 0:434); hence, H3a is not supported.

Difference in means between high and low cue multiplicity when reproc-

essability is present is tested next. Once again, the mean difference ð�0:16Þ isin the opposite direction as anticipated and non-significant (p ¼ 0:691);therefore, H3b is not supported. While the difference-in-difference between

reprocessability conditions of �0:15 ½ð�0:31Þ2ð�0:16Þ� is significant

Table 1. Means, Standard Deviations (s.d.) and Sample Sizes by

Treatment Condition.

Evidence

Strength

Message

Reprocessability

Cue

Multiplicity

Belief Pre-

Test Mean

(s.d.)

Belief Post-

Test Mean

(s.d.)

Sample

Size

Weak Absent High 4.64 3.41 22

(1.53) (2.02)

Weak Absent Low 4.54 2.46 26

(1.94) (1.58)

Weak Present High 4.27 2.08 26

(1.89) (1.20)

Weak Present Low 4.12 3.04 26

(2.10) (1.78)

Strong Absent High 4.30 2.39 23

(1.43) (1.53)

Strong Absent Low 4.14 2.00 22

(1.55) (1.02)

Strong Present High 4.14 2.52 21

(1.56) (1.66)

Strong Present Low 4.30 2.43 23

(1.82) (1.50)


(t ¼ �3:68; po0:001), the finding does not support H3c, as the differential

effects between reprocessability conditions is in the opposite direction as

predicted.11

DISCUSSION

The current study provides theoretical contributions to both information

systems (IS) and auditing literatures, as it combines IS research examining

media attributes with auditing theory involving belief adjustment. Whereas

previous media studies in IS have primarily examined the impact of com-

munication media attributes on users’ preferences, this study is unique in

that it investigates belief revision effects on auditors’ judgments. Some re-

search findings support extant theory, while other findings disconfirm

theoretical propositions.

When strong evidence was presented to the participants, belief revision

was unaffected by cue multiplicity. While statistical testing revealed

Table 2. ANCOVA Results.

Source Type III Sum

of Squares

d.f. Mean Square F p-value

Corrected model 150.075 8 18.759 10.330 0.000

Intercept 9.336 1 9.336 5.141 0.025

Pretest belief 114.079 1 114.079 62.819 0.000

Evidence strength 5.201 1 5.201 2.864 0.092

Reprocessability 0.080 1 0.08029 0.044 0.834

Cue multiplicity 0.359 1 0.359 0.198 0.657

Evidence strength �

Reprocessability

2.712 1 2.712 1.493 0.223

Evidence strength � Cue

multiplicity

1.073 1 1.073 0.591 0.443

Reprocessability � Cue

multiplicity

12.784 1 12.784 7.040 0.009

Evidence strength �

Reprocessability �

Cue multiplicity

9.296 1 9.296 5.119 0.025

Error 326.877 180 1.816

Total 1,696.000 189

Corrected total 476.952 188

Note: R2 ¼ 0:315 (adjusted R2 ¼ 0:284).


Table 3. Planned Comparisons.

Hypothesis Evidence

Strength

Message

Reprocessability

Cue

Multiplicity

Adjusted

Post-Test

Meana,b

Cue

Multiplicity

Differencec

p-value

Weak Absent High 3.26

H2a Weak Absent Low 2.36 �0.90 0.022

Weak Present High 2.10

H2b Weak Present Low 3.12 1.02 0.006

Strong Absent High 2.39

H3a Strong Absent Low 2.08 �0.31 0.434

Strong Present High 2.60

H3b Strong Present Low 2.44 �0.16 0.691

aAdjusted for the covariate (pre-test belief).bNote that lower mean scores indicate stronger belief revision in favor of the client, as the experimental task concerned the auditors’ downward

belief revision from their initial (relatively high) anchor toward the client’s desired position.cLow cue multiplicity minus high cue multiplicity.

ANNA

NOTEBERG

AND

JAMESE.HUNTON

104

significant cue multiplicity differences between reprocessability conditions

(H3c), we question the practicality of this finding, as it is inconsistent with

expectations and there are no significant cue multiplicity effects within each

reprocessability condition. Thus, from a pragmatic point of view, we suggest

that belief revision was unaffected by cue multiplicity, message reprocess-

ability or their interaction in the presence of strong evidence. The most

likely explanation for this finding is a ceiling effect; that is, the central

message of the argument was so convincing that neither the presence of

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Reprocessabilityabsent

Reprocessabilitypresent

Adj

uste

d po

st-t

est

belie

f



Weak Evidence

Fig. 3. Hypothesis 2: Observed Downward Belief Revision.

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Reprocessability

absent

Reprocessability

present

Ad

jus

ted

po

st-

tes

t b

eli

ef



Strong Evidence

Fig. 4. Hypothesis 3: Observed Downward Belief Revision.


congruent persuasive peripheral cues in the high multiplicity condition nor

the ability to reprocess the messages multiple times afforded significant

incremental weight in the participants’ cognitive processes. Results obtained

in the strong evidence condition may be an experimental artifact or it might

reflect a theoretically sound inference that can be generalized to different

populations, times and settings. Future research in this area is needed to

tease apart these two possible explanations.

When evidence was weak, however, some of the media attributes exam-

ined herein played a significant role in the auditors’ belief revision processes;

specifically, research findings reveal that the effect of cue multiplicity in-

verted, depending on whether participants reprocessed evidence cues. When

the reprocessability feature was available, the extent of belief revision to-

ward the client’s position was greater in the audio–video (high cue multi-

plicity) condition, as compared to the e-mail (low cue multiplicity)

condition. This finding is consistent with information processing theory.

However, when reprocessing was not available, the opposite results were

unexpectedly obtained.

We offer the following possible explanation for the latter finding. When

individuals are exposed to weak evidence yet fairly persuasive peripheral cues

(e.g., a professional looking and sounding businessperson situated in an or-

derly office environment), the seeming inconsistency between the strength of

the central message (weak) and the strength of the peripheral cues (strong)

might trigger professional skepticism in the auditors’ minds with respect to

the believability of the central message. Such skepticism might become man-

ifest in an unexpected way; that is, the central message is perceived as weaker

than it would have been had the inconsistently strong peripheral cues not

been available. This might explain why belief revision was greater in the e-

mail as opposed to the audio–video condition when evidence was weak and

reprocessing was not available. However, one must reconcile this observation

with the opposite effect in the presence of reprocessability.

The cognitive processing weight given to a one-time exposure to incon-

sistently strong peripheral cues might dissipate or disappear altogether with

repeated exposure. We term this phenomenon ‘peripheral cue attenuation’ –

an effect that has not been recognized or tested in prior research. To the

extent that the peripheral cue weights are diminished, the cognitive process-

ing weight assigned to the central cue increase relatively. Consistent with

anchoring and adjustment theory, multiple exposures to central messages

result in more opportunities to anchor and adjust (i.e., sequential updating),

hence theoretically, belief revision would be greater in the high cue mul-

tiplicity condition when reprocessing was available. Perhaps this could


partially account for the higher belief revision in the audio–video condition,

as compared to the e-mail condition, when reprocessability was present and

evidence strength is weak. Future research should focus more attention on

the peripheral cue attenuation effect.

We admit, however, that the unexpected negative effect of cue multiplicity

when reprocessability was absent could be due to an experimental artifact;

that is, in the absence of reprocessability, the participants’ cognitive capacity

was possibly overtaxed when they processed high cue multiplicity (audio–

video) messages. Participants had no control over the speed of audio–video

message presentation, other than pushing the play or stop button. Thus,

audio–video clips were played at a pre-determined speed. These factors

could have increased complexity and uncertainty among participants due to:

(1) the presence of peripheral cues that taxed the cognitive processing load

and (2) the participants’ inability to control the speed of information re-

ception. As a result, participants may have preferred to be conservative in

their beliefs, thereby revising their beliefs only to a very limited extent. On

the other hand, e-mail recipients in the absence of reprocessability were able

to read messages at their own pace and they revised their beliefs more

strongly than the video recipients. This may have occurred because (1) they

had no peripheral cues to process thereby focusing more cognitive process-

ing attention on the central, albeit weak, message and (2) the technology did

not force them to process the information at a certain speed. As a result,

they were less uncertain about the conveyed information and less conserv-

ative in their belief revisions. Preliminary evidence for the suggested expla-

nation of our unexpected finding can be found in post-hoc testing. Namely,

given weak evidence and the absence of reprocessability, higher response

variance was found when cue multiplicity was high as compared to low. This

finding suggests that participants who were exposed to weak messages in audio–

video format were significantly less consistent in their judgments than e-mail

recipients, suggesting some level of confusion and possibly information overload

caused by the combination of high cue multiplicity and no reprocessability.

From a practical standpoint, results from the current study suggest that the

auditing profession should consider various advantages and disadvantages

offered by computer-mediated communication with clients. While the use of

electronic communication media to collect evidential matter from clients is

timely and cost efficient, auditors should be aware that certain media at-

tributes might unintentionally bias their judgments, particularly when the

clients hold persuasive intentions. Naturally, definitive conclusions in this

regard cannot be made from a single study, as more research is needed in this

area.


For instance, future studies might focus more on peripheral cues. Only

peripheral cues that generally support the direction of the central message

cues were considered in this study. It would be interesting to investigate the

interactive effect of cue format (central vs. peripheral), cue strength (strong

or weak), and cue direction (confirming or disconfirming an initial belief) on

belief revision. We also recommend that future research study the impact

and use of electronic media for two-way communication, i.e., where com-

munication parties interact with each other. In such scenarios, attributes

other than those examined in this study could be investigated. Most of the

other attributes would revolve around the concept of synchronicity, i.e.,

whether interaction occurs in real-time (synchronous) or with a time delay

(asynchronous) (e.g., Burgoon et al., 2000). Related to the concept of sync-

hronicity, one could also study the impact of another media attribute, re-

hearsability (e.g., Dennis & Valacich, 1999) or the extent to which the

medium allows for message rehearsal before transmission, on audit judg-

ment and decision making. Finally, future research should explore how

audit teams can effectively use computer-mediated communication technol-

ogy during all phases of the audit. Indeed, this is an exciting line of research

– one that is relevant to the burgeoning demand for and growing use of

electronic communication media in global commerce.

NOTES

1. One could argue that some auditors would write down what they discuss withclients via a telephone conversation and such notes would serve as a reprocessablearchive. To the extent that auditors re-examine and reflect on such written records,post-conversation sequential reprocessing of evidence might occur. However, even ifthis were the case, personally recorded evidence of this nature is often incomplete,inaccurate and biased to some degree due to simultaneous demands of mental fil-tering, linguistic writing, and cognitive conversation processing (Schober & Brennan,2003). One could further argue that the auditor could mentally recall and re-examineeach piece of evidence gathered during a conversation. However, memory decay andrecall filtering would likely dilute and bias this form of sequential processing.2. This would be the case even if auditors engaged in some degree of sequential

processing during conversation, and/or simultaneous anchoring and adjusting post-conversation before subsequent reprocessing of the messages.3. Note that this study is limited in that it considers only peripheral cues that

support the underlying, disconfirming message provided by central cues. Thus, ourpredictions apply only to scenarios where the message source is perceived as per-suasive both in terms of central as well as peripheral message cues.4. Although the Elaboration Likelihood Model and the Heuristic–Systematic

Model differ in some important aspects (see Eagly & Chaiken (1993) for a detailed


review), these differences are not central to our study. For this reason, the termscentral and peripheral are used interchangeably with the terms systematic and heu-ristic here.5. The scenario is based on the classroom case ‘‘MicroClone, Inc.’’ (Kistler &

Strickland, 1997).6. The five arguments provided by MicroClone’s CFO were presented in random

order.7. Pilot tests revealed that explicitly mentioning verification in the ‘strong evi-

dence’ treatment and not mentioning verification in the ‘weak evidence’ treatmentwas an effective manipulation. However, we recognize that participants in the ‘weakevidence’ condition may have believed that the internal source was either easilyverifiable or presumed verified. Either belief would drive the results toward the null,not the alternative, hypothesis.8. Participants could re-examine the arguments as many times as they desired;

however, they were all purposefully exposed to the messages for at least a secondtime because the cognitive aspect of the reprocessability hypothesis assumes that theelectronic media is capable of reprocessability and that the message recipient takesadvantage of such functionality.9. There are no significant differences ðp40:10Þ across treatment conditions for

the following demographic variables: university, experimental session, age, gender,year of graduate study, educational background, and work experience.10. Additional testing using non-parametric tests agree with the results of up-

coming parametric tests.11. The difference-in-difference analysis (H3c) results in a considerably smaller

standard deviation than the single difference scores for H3a ð�0:31Þ and H3b ð�0:16Þ:Hence, we find statistical significance regarding the former but not the latter meancomparisons.

ACKNOWLEDGMENTS

We thank the reviewers and participants of the European Conference of

Information Systems, where an earlier version of this chapter was presented

in 2004 and many useful comments were provided.

REFERENCES

Abdel-Khalik, A. R., Snowball, D. A., & Wragge, J. H. (1983). The effects of certain internal

audit variables on the planning of external audit programs. Accounting Review, 58(2),

215–227.

Ashton, A. H., & Ashton, R. H. (1988). Sequential belief revision in auditing. Accounting

Review, 63(4), 623–641.

Baltes, B. B., Dickson, M. W., Sherman, M. P., Bauer, C. C., & LaGanke, J. S. (2002).

Computer-mediated communication and group decision making: A meta-analysis. Or-

ganizational Behavior and Human Decision Processes, 87(1), 156–179.


Brown, P. R. (1983). Independent auditor judgment in the evaluation of internal audit func-

tions. Journal of Accounting Research, 21(2), 444–455.

Burgoon, J. K., Bonito, J. A., Bengtsson, B., Cederberg, C., Lundeberg, M., & Allspach,

L. (2000). Interactivity in human–computer interaction: A study of credibility, under-

standing and influence. Computers in Human Behavior, 16, 553–574.

Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus

message cues in persuasion. Journal of Personality and Social Psychology, 39, 752–766.

Chaiken, S., & Eagly, A. H. (1983). Communication modality as a determinant of persuasion:

The role of communicator salience. Journal of Personality and Social Psychology, 45(2),

241–256.

Chen, S., & Chaiken, S. (1999). The heuristic-systematic model in its broader context. In:

S. Chaiken & Y. Trope (Eds), Dual-process theories of social psychology. New York:

Guildford Press.

Comunale, C. L., Sexton, T. R., & Sincich, T. (2005). Linguistic delivery style, client credibility,

and auditor judgment. Advances in Accounting Behavioral Research, 8, 61–90.

Daft, R. L., & Lengel, R. H. (1986). Organizational information requirements, media richness

and structural design. Management Science, 32, 554–571.

Daft, R. L., Lengel, R. H., & Klebe Trevino, L. (1987). Message equivocality media selection

and manager performance: Implications for information systems. MIS Quarterly,

September 1987, 11(3), 335–366.

Dennis, A. R., & Kinney, S. T. (1998). Testing media richness theory in the new media: The

effects of cues, feedback, and task equivocality. Information Systems Research, 9(3),

256–274.

Dennis, A. R., & Valacich, J. S. (1999). Rethinking media richness: Toward a theory of media

synchronicity. Paper presented at the 32nd Hawaii international conference on system

sciences, Maui, Hawaii.

Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX: Harcourt

Brace Jovanovich.

Einhorn, H. J., & Hogarth, R. M. (1985). A contrast/surprise model for updating beliefs.

Working Paper, University of Chicago.

Francis, J., & Schipper, K. (1999). Have financial statements lost their relevance? Journal of

Accounting Research Autumn, 37(2), 319–352.

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.

Hirst, E., & Koonce, L. (1996). Audit analytical procedures: A field investigation. Contem-

porary Accounting Research, 13(2), 457–486.

Hogarth, R. M., & Einhorn, H. J. (1992). Order effects in belief updating: The belief-adjustment

model. Cognitive Psychology, 24, 1–55.

Hollingshead, A. B., McGrath, J. E., & O’Connor, K. M. (1993). Group task performance and

communication technology: A longitudinal study of computer-mediated versus face-to-

face work groups. Small Group Research, 24(3), 307–333.

Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Behavior, agency costs and

ownership structure. Journal of Financial Economics, 3, 305–360.

Kistler, L. H., & Strickland, S. (1997). Instructional case: Microclone Inc. Issues in Accounting

Education, 12(2), 427–434.

Koonce, L., Walker, N. R., & Wright, W. F. (1993). A cognitive characterization of audit

analytical review. Auditing: A Journal of Practice & Theory, 12(Supplement), 57–76.


Kraut, R. E., Galegher, J., Fish, R., & Chalfonte, B. (1992). Task requirements and media

choice in collaborative writing. Human–Computer Interaction, 7(4), 375–407.

Matheson, K., & Zanna, M. P. (1989). Persuasion as a function of self-awareness in computer-

mediated communication. Social Behavior, 4, 99–111.

Noteberg, A., Benford, T. L., & Hunton, J. E. (2003). Matching electronic communication

media and audit tasks. International Journal of Accounting Information Systems, 4,

27–55.

Petty, R. E., & Cacioppo, J. T. (1981). Attitudes and persuasion: Classic and contemporary

approaches. Dubuque, Iowa: W.C. Brown.

Rice, R. E. (1992). Task analyzability, use of new media, and effectiveness: A multi-site ex-

ploration of media richness. Organization Science, 3, 457–500.

Schober, M. F., & Brennan, S. E. (2003). Processes of spoken discourse: The role of the partner.

In: A. C. Graesser, M. A. Gernsbacher & S. R. Goldman (Eds), Handbook of discourse

processes (pp. 123–164). Hillsdale, NJ: Lawrence Erlbaum.

Spires, E. (1991). Auditors’ evaluation of test-of-control strength. The Accounting Review, April,

66(2), 259–276.

Sproull, R. F. (1991). A lesson in electronic mail. In: L. Sproull & S. Kiesler (Eds), Connections:

New ways of working in the networked organization (pp. 177–184). Cambridge, MA: MIT

Press.

Strauss, S. G., & McGrath, J. E. (1994). Does the medium matter? The interaction of task type

and technology on group performance and member reactions. Journal of Applied Psy-

chology, 79, 87–97.

Suh, K. S. (1999). Impact of communication medium on task performance and satisfaction: An

examination of media-richness theory. Information & Management, 35, 295–312.

Torrance, M. (2002). Introduction to analysis of variance, from http://www.staffs.ac.uk/person-

al/sciences/smt15/mt15/rm3/lecture_web/index.htm. Accessed on December 1, 2004.

Weisband, S., & Atwater, L. (1999). Evaluating self and others in electronic and face-to-face

groups. Journal of Applied Psychology, 84, 632–639.

APPENDIX

Experimental Case

Please assume the role of a partner at a large audit firm. You are the partner

in charge of the audit of MicroClone Inc., a manufacturer of IBM-com-

patible personal computers.

The senior manager on the audit, Steven de Wit, who reports directly to

you, has extensively analyzed the company’s finished goods inventory. 25%

(2,000 units) of the finished goods inventory is comprised of 4th generation

computers and the remaining 75% (6,000 units) is represented by 5th gen-

eration computers (the latest microprocessor chip available on the market).

MicroClone’s inventory of 2,000 units of 4th generation computers is pres-

ently valued at cost (h1,000 per unit), for a total of h2,000,000.


Steven de Wit has identified what might be a problem concerning the 4th

generation inventory. He noted that, during the 4th quarter, MicroClone

began shipping personal computers using the 5th generation chips; thus, he

is concerned that the value of the 4th generation computers remaining in the

inventory might be overstated.

After listening to Steven, you are also concerned about a potential over-

valuation problem because 4th generation computers will eventually become

obsolete by industry standards. Your and Steve’s analyses suggest that each

4th generation computer in the inventory should be valued at h800, not

h1,000. Thus, you believe that MicroClone’s estimate of 4th generation

computers could be overstated by as much as h200 per unit, yielding a

potential over-valuation of h400,000 – at the most.

You have decided to take the inventory issue directly to Tom van

Breukelen, the chief financial officer (CFO) of MicroClone. Your aim is to

negotiate with van Breukelen the appropriate over-valuation estimate,

which can range from h0 to h400,000. You ask van Breukelen to explain his

point of view regarding the 4th generation issue by sending him an e-mail

message with the following content:

To: Tom van Breukelen

From: You

Dear Mr van Breukelen,

I have some questions about the finished goods inventory.

Specifically, I want to ask you about the h2 million valuation of the 4th

generation microcomputers. By industry standards, the 4th generation

computers will likely become obsolete as the new, faster 5th generation

models become more widely adopted.

According to your records, about 25% (2,000 units) of your

company’s finished goods inventory is in 4th generation models, which

are valued at h1,000 each, for a total inventory valuation of h2,000,000.

Based on my preliminary analyses, your 4th generation computers

might be over-valued by around h200 per unit. Therefore, the finished

goods inventory appears to be overvalued by h400,000 at the most,

which means that your pre-tax profits also could be overstated by that

amount.

I am concerned about this situation because h400,000 is material

to your financial statements taken as a whole.

Kind regards,


ROLE MORALITY AND

ACCOUNTANTS’ ETHICALLY

SENSITIVE DECISIONS

Robin R. Radtke

ABSTRACT

If individuals exhibit less ethical behavior in the workplace than in their

personal decisions, this may constitute evidence of role morality behavior.

Role morality can be defined as ‘‘claim(ing) a moral permission to harm

others in ways that, if not for the role, would be wrong’’ (Applbaum,

1999. Ethics for adversaries: The morality of roles in public and profes-

sional life (p. 3). Princeton, NJ: Princeton University Press.) To inves-

tigate this issue, 55 practicing accountants completed and returned the

experimental survey. Results show that in many situations, business de-

cisions were less ethical than personal decisions, consistent with the theory

of role morality. The implications and limitations of this study as they

relate to practicing accountants are discussed.

INTRODUCTION

The importance of ethical decision-making in the accounting profession has

recently received renewed attention. Concerns about unethical activities





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08005-6

113

related to the failure of Enron is one of the latest examples. Accountability

to the client in terms of pressure to present a favorable financial picture of a

corporation’s health may often lead to ethical breaches (Chartier, 2002).

Because of the Enron example concerning dubious accounting practices and

other documented cases of fraud over the last few years, public confidence in

the accounting profession may be diminishing. While calls for effective

codes of conduct and increased monitoring and accountability of the ac-

counting profession are not new (National Commission on Fraudulent

Financial Reporting, 1987), the credibility crisis remains. New calls for reg-

ulation both within and outside the accounting profession are specifically

concerned with violations of ethical standards (Pitt, 2002). Within this cli-

mate, investigation of accountants’ ethical decision-making processes is

particularly important.

Since recent examples suggest that at least some accountants are making

unethical business decisions, research is needed to investigate whether these

accountants make unethical decisions in both personal and business situ-

ations or whether they exhibit more ethical decisions in personal situations,

consistent with the theory of role morality. Implications for ethical

training and/or screening programs may vary significantly based on these

findings.

The remainder of the chapter is organized as follows. In the first section of

the chapter, relevant literature is reviewed and research questions are de-

veloped. Next, the experimental design is described. The data analysis and

results are presented in the third section. The chapter concludes with a

discussion of the limitations and implications for future research.

RELEVANT THEORY AND RESEARCH QUESTIONS

Role Morality and Ethically Sensitive Decisions

Mostly everyone deals with situations that contain ethical issues on a regular

basis in both personal and business decisions. Undoubtedly, many of these

situations are quite complex and difficult to reconcile. When pondering an

ethically sensitive situation, many factors may come into play. One com-

monly accepted ethical principle is universalism, which suggests that once a

certain action, such as theft, is determined to be wrong, it should always be

considered wrong. Thus, if one ascribes to the principle of universalism and

agrees that theft is wrong, one would not steal regardless of whether the

situation is of a personal or business nature.

ROBIN R. RADTKE114

If individuals exhibit less ethical behavior in the workplace than they do

in their personal decisions, this is consistent with the theory of role morality.

Role morality can be defined as ‘‘claim(ing) a moral permission to harm

others in ways that, if not for the role, would be wrong’’ (Applbaum, 1999,

p. 3). Role morality at its extremes can be seen in such examples as Sanson,

the executioner of Paris (Applbaum, 1999, p. 15) and Adolf Eichmann, a

middle-level Schutzstaffel (SS) officer who was responsible for the depor-

tation of millions of European Jews to Nazi death camps (Applbaum, 1995,

p. 470). Less extreme examples are present everyday in many professional

realms and include a doctor denying a patient involved in a clinical trial, a

treatment that the doctor has reason to believe is beneficial (Applbaum,

1999, p. 48), a CEO engaging in a strategy of secrecy wherein certain in-

formation is concealed from the public to advance the purposes of the CEO

(Applbaum, 1999, p. 183), and a campaign strategist who willfully distorts

an opponents’ record to smear his/her reputation in the eyes of the public

(Applbaum, 1999, p. 3).

An alternative, less extreme definition of role morality is ‘‘the extent to

which one succeeds in meeting the demands and obligations of one’s role’’

(Werhane & Freeman, 1999, p. 3). This definition is consistent with the view

that role morality and common-sense morality often prescribe the same ac-

tion. Specifying in more detail how role morality relates to common-sense

morality is the starting point in developing an understanding of how indi-

viduals can undertake actions consistent with the former and not the latter.

Fig. 1 shows a graphical depiction of the overlaps among role morality,

common-sense morality, and self-interest, as suggested by Kerssens-van

Drongelen and Fisscher (2003). Much of role morality overlaps with com-

mon-sense morality, such that only infrequently will individuals be faced

with situations where they must choose between the two. In these situations,

however, in order for an individual to choose role morality over common-

sense morality, the individual must ‘‘excuse oneself from a common moral

obligation by appealing to a role or the institution that creates the role’’

(Luban, 1988, p. 116). One way to achieve this end is through moral dis-

engagement.

Bandura, Barbaranelli, Caprara, and Pastorelli (1996) suggest that moral

disengagement allows individuals to displace their moral control in order to

engage in detrimental conduct. Several methods may be used to achieve this

disengagement including displacement of responsibility, moral justification,

disregarding or distorting the consequences of actions, and dehumanization.

Displacement of responsibility allows individuals to pass on responsibility

for actions to a legitimate authority who has dictated the actions to be

Role Morality and Accountants’ Ethically Sensitive Decisions 115

undertaken. In this situation, the individual eliminates self-censuring reac-

tions normally attributable to reprehensible conduct. When detrimental

conduct is portrayed as being in the service of a valued social purpose, moral

justification makes the culpable action righteous. By focusing only on the

positive benefits of personal gain while ignoring the harm caused to others,

one disregards or distorts consequences. Dehumanization treats people as

being devoid of human qualities. McPhail (2001, p. 280) suggests that be-

cause accounting is technical and mathematical in nature, it ‘‘dehumanises

individuals and consequently makes it easier for some people to treat other

people cruelly.’’ Taken together, these and several other disengagement

practices may sufficiently disinhibit an individual such that detrimental ac-

tion is undertaken.

Multiple studies from the 1970s provide evidence of the detrimental ef-

fects of moral disengagement. Tilker (1970) found that an individual who is

forced to feel responsible for the condition of another who is providing

feedback on their well-being is most likely to act socially responsible.

Diener, Dineen, Endresen, Beaman, and Fraser (1975) found that the pres-

ence of several disinhibiting forces may lead to increases in antisocial

Organizational factors:

1) Company type 2) Length of

employment

Accountability

Personal factors:

1) Gender 2) Age

Common-sense

morality

Role morality

Self-interest

Fig. 1. Theoretical Model. (Based on Kerssens-van Drongelen and Fisscher’s

(2003) Graphical Representation of the Relationships among Role Morality,

Common-Sense Morality, and Self-Interest. This Figure Incorporates the Factors

of the Current Study and Depicts their Posited Impacts on both Role Morality and

Common-Sense Morality.)

ROBIN R. RADTKE116

behavior, while Bandura, Underwood, and Fromson (1975) found that

lower personal responsibility and dehumanization led to increased aggres-

siveness. More recent efforts in the area have found that moral disengage-

ment reduced anticipatory self-censure and prosocialness and promoted

affective and cognitive reactions conducive to detrimental conduct and ag-

gression (Bandura et al., 1996). Additionally, Bandura, Caprara, Barba-

ranelli, Pastorelli, and Regalia (2001) found that both perceived academic

and self-regulatory efficacy reduced transgressiveness both directly and by

increasing adherence to moral self-sanctions for harmful conduct and pro-

socialness. Bandura (2002, p. 114) also comments that industrywide collec-

tive moral disengagement practices such as those associated with the

tobacco industry ‘‘require a large network of otherwise considerate people

performing jobs drawing on their expertise and social influence in the service

of a detrimental enterprise.’’

Thus, moral disengagement provides a basis for understanding how role

morality behavior could depart from common-sense morality. Within the

accounting profession examples of role morality may include such behavior

as earnings management or allowing a questionable audit adjustment or tax

treatment.

Other Factors Potentially Impacting Ethically Sensitive Decisions

Accountability

Accountability is one factor that may impact individuals’ ethically sensitive

decision-making and role morality behavior. Defined as having to justify a

decision to an external party, accountability may alter one’s thought process

(Weigold & Schlenker, 1991) and compel an individual to conform to the

expectations of the evaluator, when those expectations are made clear to the

individual (Tetlock, 1985; Tetlock, Skitka, & Boettger, 1989). This induces

the individual to make that decision, which the evaluator is expecting when

the individual knows he or she must explain the decision to the evaluator.

Absence of accountability leaves the individual without certain knowledge

of the expectations of the evaluator and without the perceived benefits of

conformity.

Ashton, Kleinmuntz, Sullivan, and Tomassini (1988, p. 130) attest to the

importance of accountability within an auditing setting by stating that ‘‘be-

cause of the environment, the professional auditor must be prepared to

justify, document, and take responsibility for his/her judgments and deci-

sions.’’ Related research has shown that auditors have a good understanding


of the inherent accountability in their profession (Emby & Gibbins, 1988).

Accountability has also been shown to affect various auditing decisions in

experimental settings. Johnson and Kaplan (1991) found that accountable

auditors showed higher consensus and self-insight (the extent to which an

auditor is aware of his/her own judgment process) than those who were not

accountable. Accountable audit managers were found to be more likely to

issue qualified opinions in a study by Lord (1992), while Buchman, Tetlock,

and Reed (1996) found that the audit report chosen by experienced auditors

matched the expectation of the evaluator (the client or the partner in

charge). In a study involving both executive MBA students and auditors,

Kennedy (1993) concluded that effort-related biases such as recency can be

mitigated by accountability. While several earlier studies found that ac-

countability exacerbates the dilution effect (Hackenbrack, 1992; Messier &

Quilliam, 1992), more recent results have not supported this finding (Glover,

1997; Hoffman & Patton, 1997). More recently, Turner (2001) found support

for the results of Peecher (1996) in that accountability demands may represent

bias-instigators instead of bias-mitigators, while Asare, Trompeter, and

Wright (2000) found accountable auditors focused more on testing potential

error hypotheses and increased the breadth of hypotheses tested.

In summary, accountability can have varying impacts in different decision

contexts. Several accounting studies have shown, however, that accountable

subordinates often make judgments consistent with the views of their su-

pervisors or clients in situations such as inventory valuation (Ponemon,

1995), financial reporting (Hackenbrack & Nelson, 1996), interpretation of

tax standards (Cuccia, Hackenbrack, & Nelson, 1995), and litigation dis-

closure (Buchman et al., 1996). These results suggest that accountability

may compel accountants to make decisions they otherwise might not make.

If these decisions are unethical, then accountability may be a factor that

impacts role morality behavior as is posited in Fig. 1.

Additional Organizational and Personal Variables

While accountability may be the primary variable affecting role morality

behavior, various organizational and personal variables may also be related

to accountants’ reactions to ethically sensitive situations of a personal and

business nature. Specifically, two organizational and two personal factors

are included in the regression analyses of this study – company type (public

accounting versus private industry), and length of employment with the

firm, and gender, and age.1 A discussion of each of these variables follows

and a graphical depiction is included in Fig. 1.

ROBIN R. RADTKE118

Undoubtedly, many differences exist between accountants working in

public accounting and private industry. For example, the hierarchical nature

and low retention rates of the public accounting profession suggest that

conformity pressures may be particularly acute. Although accountants

working in private industry may also face various pressures within the

company, these pressures may not be as severe as in public accounting.

Thus, accountants may self-select into public or private accounting as a

result of the firm socialization process. Ponemon (1992a) found that the

ethical development of staff and seniors was higher than that of managers

and partners and posited that accountants with higher levels of ethical de-

velopment (than the firm norm) may leave public accounting for private

industry before entering the management ranks. Jeffrey and Weatherholt

(1996) suggest that this migration may cause accountants in private industry

to exhibit higher levels of ethical development, despite similar educational

and experience bases. This may cause fewer differences in responses to eth-

ically sensitive situations of a personal and business nature in accountants

working in private industry.

Length of employment with the firm may also impact whether accountants

are susceptible to role morality behavior. Results of firm socialization studies

suggest that promotion constitutes an employee screening process, signaling

to upper management, which employees should be retained (Nystrom &

McArthur, 1989). More specifically, individuals who are promoted are per-

ceived by upper management to have personal characteristics similar to

those of the corporate culture (Weick, 1979; Lockheed, 1980; Smircich,

1983). Individuals’ cognitive characteristics, including moral reasoning,

have been found to become more homogeneous as they advance in man-

agement levels within an organization (Fisher, Merron, & Torbert, 1987;

Avolio & Gibbons, 1988). The result of this firm socialization process may

promote different responses to ethically sensitive situations of a personal

and business nature in those accountants with a longer tenure with the firm.

The personal variables of gender and age may also affect an accountant’s

susceptibility to role morality behavior. In most cases, age may be correlated

with length of employment with the firm (r ¼ 0:8930; p ¼ 0:0001 in the

current study), causing the effects of these variables to be similar. Gender is

a variable that has traditionally been shown to affect many decision con-

texts. Specifically, males and females in the general population have been

found to be virtually identical with respect to moral development (Rest,

1986), while several studies of accounting professionals and students, alter-

natively, show that female accountants demonstrate a higher level of moral

reasoning (Shaub, 1994; Ameen, Guffey, & McMillan, 1996). Further,


recent evidence of a growing number of female whistle-blowers (Moak,

2004; www.cnn.com, 2002) lends support to the contention that females may

be less susceptible to firm socialization and may exhibit fewer differences in

responses to ethically sensitive situations of a personal and business nature.

Development of Research Questions

In order to measure whether accountants exhibit behavior consistent with

the theory of role morality, two sets of responses to ethically sensitive sit-

uations are necessary for each accountant. Specifically, comparison of ac-

countants’ responses to ethically sensitive situations framed in both personal

and business settings would allow identification of those accountants whose

decisions differ. Thus,

RQ1. Do accountants react to ethically sensitive situations of a personal

and business nature in a similar manner (which is inconsistent with the

theory of role morality)?

RQ2. Is potential role morality behavior affected by accountability,

company type, length of employment with the firm, gender, or age?

METHODOLOGY

Participants

Accountants from both public accounting and private industry were solic-

ited to act as participants. Companies of both types with local offices in a

metropolitan city in the southwest were randomly selected. Appropriate

contact personnel were requested to garner participation from their staffs.

These personnel were asked to provide the number of staff who would

participate in a study of ‘‘reactions to ethically sensitive situations.’’ The

surveys were then sent to contact personnel for distribution. Of the 116

surveys sent for distribution, 55 (47%) were returned completed.

Instrument

The survey instrument was comprised of two main parts. In one section,

participants responded to two matched sets of ethically sensitive situations:

ROBIN R. RADTKE120

of a personal and business nature.2 The second section elicited demographic

and work environment data, including measures of perceived accountability

pressure.3 The ethically sensitive situations were used in a previous study by

Radtke (2000) and were generally consistent with the definition of an ethical

issue as proposed by both Velasquez and Rostankowski (1985) and Jones

(1991): an ethical issue occurs when a person’s actions, freely performed,

may harm or benefit others. The 16 ethically sensitive situations corre-

sponded to two matched sets of eight situations each, when personal and

business situations dealing with the same issue were paired together. Of the

eight issues, four were patterned after those of the Defining Issues Test

(Rest, 1979) – situations dealing with theft, withholding information from

authorities, freedom of speech, and racial discrimination. The other four

issues were similar, but more situation specific, dealing with copying soft-

ware, cheating on taxes, deception/honesty, and trading on inside informa-

tion. In order to obtain two matched sets of eight situations each, every

situation was chosen specifically for its ability to be framed in both a per-

sonal and business setting (although ensuring that participants perceive the

matched situations to be comparable is impossible). Also, each business

situation was general enough such that a practicing accountant working in

any business area should be able to formulate an answer. For each situation,

participants were asked to indicate their probable action on the issue on the

five-point scale of yes, probably yes, unsure, probably no, and no. The

resulting 16 ethically sensitive situations are presented in the appendix.4

Consistent with Dreike and Moeckel (1995), the survey instrument in-

cluded a combination of situations representing some actions that are ob-

viously illegal, some that would fall under issues addressed by the AICPA

Code of Professional Conduct (AICPA, 2003), and some that may be con-

sidered as inappropriate or gray areas. Including situations of varying de-

grees of ethical content makes the survey instrument richer and less

repetitive. The issue of whether the participants perceived the situations to

have significant ethical content is difficult to ascertain and control. Dreike

and Moeckel (1995) point out that even though they used a survey con-

taining ethical situations consistent with the Velasquez and Rostankowski

(1985) definition, the 66 auditors in their study who rated eight ethical

situations each only viewed 56% of the situations as actually containing an

ethical issue.

Pretesting the ethically sensitive situations consisted of several steps. First,

several colleagues familiar with ethics research reviewed the situations for

general readability and design. Second, the entire instrument was completed

by 44 cost accounting students (32 undergraduate and 12 graduate students


who would be reasonably familiar with the types of situations presented).

Third, a number of colleagues and doctoral students reviewed the compa-

rability of the situations that were matched in terms of personal and busi-

ness settings. In nearly all cases, the vast majority of the matched situations

were easily judged as comparable.

RESULTS

Preliminary Analyses

For the 55 participants, the average age was 34, while 20 of the participants

who reported their gender were female and 31 were male. Nearly 40% of the

participants were from public accounting, while most of the private industry

participants worked at oil and gas firms. The sample was almost equally

split between CPA and non-CPA respondents. Table 1 shows the remaining

sample characteristics.

A summary of the responses to the 16 ethically sensitive situations is

shown in Table 2. Interestingly, responses to some of the situations were

quite homogeneous (situations 1, 4, and 8 had one response chosen by over

half the participants), while responses to the majority of the situations were

quite diverse. At the extremes, 40 (72.7%) respondents answered ‘‘No’’ to

situation 8, while responses to other situations were extremely diverse (e.g.,

situations 7, 10, and 14 had at least 10% of the respondents choosing each

of the possible responses, while nearly half of the respondents chose yes or

probably yes and over half of the respondents chose no or probably no for

situation 2).

For further statistical analyses, the responses ranging from yes to no were

coded from 1 to 5, with 1 representing the least ethical response and 5 rep-

resenting the most ethical response. Due to the phrasing of some questions,

the order of the responses was reversed (such that a response of yes would be

representative of the most ethical response and would be coded as 5), so that

participants would not always expect the least ethical response to appear first.

Specifically, questions 5, 6, 7, 8, 11, 13, 14, 15, and 16 were reverse ordered

and coded accordingly. Based on this coding, a mean response was calculated

for each question and for each participant. The mean responses (shown in

Table 2) for the questions range from 1.42 on situation 8 to 4.29 on situation

1. The mean response of 1.42 on situation 8 represents a somewhat unethical

response about reporting/withholding information in a business setting. The

mean response of 4.29 on situation 1 represents a somewhat ethical response

ROBIN R. RADTKE122

about cheating on taxes in a personal setting. For the 55 sample participants,

the overall mean response for all 16 situations was 2.93, which represents a

middle response between the two endpoints of ethical and unethical be-

haviors. The range of mean responses across participants was from 1.94 to

3.75, ranging from somewhat unethical to ethical behavior.

Tests of Research Questions

Univariate tests were first used to address the first research question of

the study, that accountants will react to ethically sensitive situations of a

personal and business nature in a similar manner. For these tests, responses

Table 1. Sample Demographics.

Male Female

Company type

Public accounting 21 10 10

Oil and gas 24 16 7

Insurance 5 3 1

Manufacturing 2 1 1

Other 2 1 1

Ethnic background

Caucasian 35

Hispanic 5

African American 1

Asian 2

Undergraduate degree

Accounting 27

Non-accounting 20

Graduate degree

Yes 12

No 40

CPA

Yes 28

No 26

Mean Median Standard Deviation

Years with the company 8.36 5.0 8.55

Years in current position 3.32 1.0 4.59

Years of accounting experience 9.93 6.5 8.59

Note: All respondents did not answer all questions.


Table 2. Sample Responses to Ethically Sensitive Situations.

Situation/Response Mean Yes Probably Yes Unsure Probably No No

1. Personal case of cheating on taxes 4.29 2 (3.6%) 6 (10.9%) 1 (1.8%) 11 (20.0%) 35 (63.7%)

2. Business case of cheating on taxes 3.22 7 (12.7%) 17 (30.9%) 3 (5.5%) 13 (23.6%) 15 (27.3%)

3. Personal case of copying software 2.36 13 (23.6%) 27 (49.1%) 3 (5.5%) 6 (10.9%) 6 (10.9%)

4. Business case of copying software 2.00 17 (30.9%) 31 (56.4%) 1 (1.8%) 2 (3.6%) 4 (7.3%)

5. Personal case of theft 2.42 8 (14.6%) 11 (20.0%) 0 (0%) 13 (23.6%) 23 (41.8%)

6. Business case of theft 1.76 0 (0%) 3 (5.5%) 5 (9.1%) 23 (41.8%) 24 (43.6%)

7. Personal case of reporting/withholding information 3.18 7 (12.7%) 22 (40.0%) 9 (16.4%) 8 (14.5%) 9 (16.4%)

8. Business case of reporting/withholding information 1.42 0 (0%) 3 (5.5%) 2 (3.6%) 10 (18.2%) 40 (72.7%)

9. Personal case of trading on inside information 4.02 0 (0%) 8 (14.5%) 8 (14.5%) 14 (25.5%) 25 (45.5%)

10. Business case of trading on inside information 2.93 8 (14.5%) 19 (34.6%) 8 (14.5%) 9 (16.4%) 11 (20.0%)

11. Personal case of deception/honesty 3.84 17 (30.9%) 26 (47.3%) 3 (5.4%) 4 (7.3%) 5 (9.1%)

12. Business case of deception/honesty 3.89 4 (7.3%) 9 (16.4%) 3 (5.4%) 12 (21.8%) 27 (49.1%)

13. Personal case of racial discrimination 3.53 13 (23.6%) 19 (34.5%) 10 (18.2%) 10 (18.2%) 3 (5.5%)

14. Business case of racial discrimination 2.82 6 (10.9%) 15 (27.3%) 7 (12.7%) 17 (30.9%) 10 (18.2%)

15. Personal case of freedom of speech 2.16 1 (1.8%) 8 (14.6%) 7 (12.7%) 22 (40.0%) 17 (30.9%)

16. Business case of freedom of speech 3.05 9 (16.3%) 16 (29.1%) 4 (7.3%) 21 (38.2%) 5 (9.1%)

ROBIN

R.RADTKE

124

to personal and business situations for each participant were matched by

situation type. This resulted in a total of eight matched situations. The

difference between the personal and business response (which could range

between �4 and 4) was then calculated for each type of situation. The

average difference was then calculated for each participant. For the 55 par-

ticipants, the mean average difference was 0.59, indicating that business

responses were on average less ethical than personal responses. This differ-

ence was statistically significantly different from 0 at p ¼ 0:0001 using a

univariate test.

Further analysis of the differences between personal and business re-

sponses by type of situation shows interesting results. Table 3 shows average

differences by situation type, as well as the p values of the univariate tests.

Only differences 2 and 6, dealing with copying software and deception/

honesty, respectively, are insignificant at a ¼ 0:05: This suggests that par-

ticipants’ responses on these two issues were similar for personal and busi-

ness situations. All other differences were positive and statistically

significantly different except difference 8, which was negative. The positive

differences imply that business responses were less ethical than personal

responses in situations dealing with cheating on taxes, theft, reporting/

withholding information, trading on inside information, and racial

Table 3. Differences in Responses to Personal and Business Ethically

Sensitive Situations by Situation Type.

Differences by Situation Type Average

Difference

t-Test on Sample

Means (p value)

Wilcoxon Test on Sample

Medians (p value)

D1 ¼ S1�S2 1.05 0.0001 0.0001

Cheating on taxes

D2 ¼ S3�S4 0.36 0.0534 0.0516

Copying software

D3 ¼ S5�S6 0.64 0.0029 0.0022

Theft

D4 ¼ S7�S8 1.73 0.0001 0.0001

Reporting/withholding information

D5 ¼ S9�S10 1.07 0.0001 0.0001

Trading on inside information

D6 ¼ S11�S12 �0.05 0.8286 0.9778

Deception/honesty

D7 ¼ S13�S14 0.70 0.0012 0.0007

Racial discrimination

D8 ¼ S15�S16 �0.88 0.0001 0.0001

Freedom of speech


discrimination. The one negative difference implies that business respons-

es were more ethical than personal responses in situations dealing with

freedom of speech. These results suggest that accountants react differently

to ethically sensitive situations of a personal and business nature. Specif-

ically, in the majority of the situations examined, business responses were

less ethical than personal responses, consistent with the theory of role mo-

rality.

A more complete analysis of these differences requires the use of regres-

sion analysis with control variables.5 The dependent variables for regression

purposes are the matched responses by situation type. This allows for com-

parison of responses for each participant for each situation type based on

varying the situation setting between the two options of personal and

business. The control variables previously identified are accountability,

company type (public accounting versus private industry), length of em-

ployment with the firm, gender, and age. A total of nine regressions were

run, one for each type of ethically sensitive situation previously identified,

and one across all participants’ responses to all situations. The regression

equation is as follows:

Matched Responses ¼ aþ b1P=Bþ b2ACCTþ b3COTþ b4LENGTH

þ b5GENDERþ b6AGEþ x

Variables included in the regression have been previously discussed in the

development of research questions section of the chapter. Variable coding is

as follows: P/B (0 ¼ business setting and 1 ¼ personal setting), COT

(0 ¼ private industry and 1 ¼ public accounting), and GENDER

(0 ¼ male and 1 ¼ female). The accountability variable, ACCT, represents

the accountability pressure felt by each participant in either the business or

personal setting. The business accountability equals the sum of each par-

ticipants’ responses to the four accountability scales for the business setting

(accountability to superiors/company, accounting profession, clients/other

companies, and general public), while the personal accountability represents

the accountability felt to oneself.6 Each of these scaled measures can po-

tentially range from 1 (low) to 7 (high); the business measure can thus range

from 4 to 28. For the sample participants the average for the personal

measure is 6.64, with a range of 4–7, while the average for the business

measure is 22.95, with a range of 14–28, indicating that most participants

felt a relatively sizable amount of accountability pressure both to themselves

and to the business-related entities included in the study. Since the personal

and business measures are based on different scales (i.e., a 4 on the personal

ROBIN R. RADTKE126

measure represents the midpoint of the scale, while a 4 on the business

measure represents the minimum value), the scales were standardized by

dividing the personal measure by 7 and the business measure by 28.

Results of the regression are shown in Table 4 and support the propo-

sition that business responses were less ethical than personal responses,

consistent with the theory of role morality. The personal or business setting

variable is significant in six of the nine regressions. In five of these cases,

including across all situations, the parameter estimate was positive, indi-

cating that the business response was less ethical than the personal response.

The significance of the regression across all situations including all partic-

ipants’ responses addresses the potential concern over the matching of sit-

uations by situation type. The only control variables that were significant

were accountability in the situation dealing with copying software, and age

in the situation of trading on inside information. Taken as a whole, the

results of this study suggest that business responses to ethically sensitive

situations are less ethical than personal responses, consistent with the theory

of role morality.

To further investigate the effect of accountability in the business situa-

tions, a regression was run including only the business responses, excluding

the setting (P/B) variable and including the other five variables mentioned

above. As such, the accountability measure in this regression represents the

accountability felt by each participant in the business setting. Results show

that the accountability measure is positive and significant (b ¼ 0:050;t ¼ 2:12), indicating that those participants who felt greater accountability

pressure at work were more likely to choose more ethical responses, con-

trary to the expectations of this study. None of the other control variables

show significant results.

DISCUSSION

The results of this study support the premise that accountants react differ-

ently to ethically sensitive situations of a personal and business nature,

consistent with the theory of role morality and inconsistent with the concept

of universalism. Specifically, in the majority of the situations examined,

business decisions were less ethical than personal decisions, consistent with

the theory of role morality. Whether accountants work in public accounting

or private industry, length of employment with the firm, age, and gender did

not significantly affect accountants’ reactions to ethically sensitive situa-

tions. Additional results support the contention that greater perceived


Table 4. Coefficient Estimates of Regressions by Situation Type (b Coefficient and t-Statistic Reported).

Situation Type/Variable Personal or

Business

Accountability Company Type Length of

Employment

Gender Age Adjusted R2

Cheating on taxes 0.925 (2.84)� 0.127 (0.77) 0.051 (0.19) 0.007 (0.22) �0.449 (�1.45) �0.028 (�0.79) 0.1263

Copying software �0.041 (�0.14) 0.297 (2.02)� �0.336 (�1.40) 0.006 (0.22) �0.039 (�0.14) 0.009 (0.30) 0.0700

Theft 0.738 (2.42)� �0.048 (�0.31) 0.118 (0.46) �0.046 (�1.47) 0.338 (1.16) 0.027 (0.81) 0.0845

Reporting/withholding

information

2.031 (7.36)� �0.190 (�1.35) �0.093 (�0.41) 0.022 (0.78) 0.474 (1.80) �0.011 (�0.38) 0.4084

Trading on inside

information

0.812 (2.89)� 0.187 (1.31) �0.434 (�1.85) �0.025 (�0.87) 0.416 (1.55) 0.085 (2.75)� 0.3315

Deception/honesty �0.264 (�0.85) 0.251 (1.59) �0.442 (�1.71) 0.032 (1.02) �0.169 (�0.57) �0.015 (�0.43) 0.0531

Racial discrimination 0.595 (1.87) �0.012 (�0.07) �0.388 (�1.46) 0.005 (0.16) 0.455 (1.50) 0.007 (0.21) 0.0345

Freedom of speech �0.637 (�2.15)� �0.200 (�1.32) 0.045 (0.18) �0.015 (�0.50) 0.102 (0.36) 0.014 (0.42) 0.0972

Across all situations

ðn ¼ 800Þ

0.520 (4.04)� 0.052 (0.79) �0.185 (�1.72) �0.002 (�0.12) 0.141 (1.15) 0.011 (0.78) 0.0421

Note: The sample size for the regression analysis was 50 participants, since some participants did not supply all needed data. Thus, for the first

eight regressions, n ¼ 100; which includes both the personal and business response for each participant for each situation type.

Matched Responses ¼ aþ b1P=Bþ b2ACCTþ b3COTþ b4LENGTH

þ b5GENDERþ b6AGEþ x�t-value is significant at a ¼ 0:05:

ROBIN

R.RADTKE

128

accountability pressure is associated with more ethical responses in the

business setting.

This evidence implies that accountants may tend to view ethically sen-

sitive situations of a business nature differently from those of a personal

nature. Specifically, they may not perceive acting less ethically in business

situations negatively; they may believe they are expected to act in a certain

manner that is consistent with their role. Firm socialization may be one

factor related to this behavior. As previously discussed, firm socialization

theory suggests that individuals who are promoted are perceived by upper

management to have personal characteristics similar to those of the corpo-

rate culture (upper management) (Weick, 1979; Lockheed, 1980; Smircich,

1983). The results of the current study suggest that accountants may not

actually change their moral beliefs to be consistent with those of upper

management, but may merely change their business ethics so as to appear to

have consistent beliefs and increase their promotion potential. These results

are somewhat conflicting with those of Ponemon (1990, 1992a) and Shaub

(1994) who found that moral reasoning level and position level within the

firm are inversely related. These studies suggest that those who are retained

by public accounting firms have lower moral reasoning levels.

These results are at least partially tempered by the accountability findings

that greater perceived accountability pressure is associated with more ethical

responses in the business setting. This suggests that one way to try to avoid

role morality behavior is to make individuals accountable for their actions.

This is consistent with an increase in perceptions of personal responsibility

for actions as a means to avoid moral disengagement. In the current climate

of diminishing confidence in the accounting profession, the policy implica-

tions of this result are promising, given calls for increased monitoring of

auditors by the Sarbanes-Oxley Act, for example.

Focusing on additional factors shown to be consistent with moral disen-

gagement could also aid in avoiding role morality behavior. Specifically,

avoiding distortion of consequences and dehumanization by being aware of the

harm caused to others and focusing on the end users of the accounting process

would minimize chances for moral disengagement, whereby an individual

would be more prone toward role morality behavior. Firm training programs

that recognize the importance of these sorts of factors may prove beneficial.

The results of the current study represent a new avenue of investigation

into understanding accountants’ decision-making in ethically sensitive sit-

uations. Previous studies have focused on associating certain behaviors such

as making questionable independence judgments (Ponemon & Gabhart,

1990), underreporting audit time (Ponemon, 1992b), and avoidance of


whistle-blowing (Brabeck, 1984; Arnold & Ponemon, 1991) with lower levels

of moral reasoning. The present study shows that the nature of the decision

(whether the decision is personal or business) may be another critical factor

in influencing accountants’ ethically sensitive decisions.

Another interesting finding of the study is the lack of differences between

public accountants and non-public accountants, as well as among individuals

based on the length of tenure with the firm. Since these factors have been

shown to be significant in other studies, additional analysis is warranted.

Results of this study should be interpreted with caution due to several

limitations. First, the survey elicited responses to eight matched sets of eth-

ically sensitive scenarios. These responses may not be representative of ac-

countants’ responses to other types of ethically sensitive situations. Second,

the majority of the sample participants from private industry were from oil

and gas firms. Inclusion of more participants from other types of private

industry firms would broaden the generalizability of the results. Third, per-

ceptions of the ethical nature of the ethically sensitive situations may have

varied across participants (i.e., not all participants may have considered all

situations to contain significant ethical issues). A related concern is that the

magnitude of the ethically sensitive situations may have been interpreted to be

different across the personal and business settings. Also, participants may not

have perceived situations that were paired together as a personal and business

example of a similar situation to really be representative of similar situations

(although obviously, a personal and business setting cannot be identical).

Additionally, limitations typically associated with exploratory research and

survey instruments apply to this study as well.

Notwithstanding these general limitations, the results of this study support the

premise that accountants react differently to ethically sensitive situations of a

personal and business nature. Specifically, in the majority of the situations ex-

amined, business decisions were less ethical than personal decisions, consistent

with the theory of role morality. This evidence implies that accountants may be

susceptible to role morality behavior in the workplace by means of moral dis-

engagement, as well as the effects of firm socialization. Whether this affects the

quality of accountants’ ethically sensitive decisions has yet to be determined.

NOTES

1. Education level was also considered as a control variable, since previousresearch has found that an individual’s level of moral reasoning increases witheducation (Thoma, 1986; McNeel, 1994). In the current sample, 40 participantsreported having an undergraduate degree, while only 12 reported having a graduate

ROBIN R. RADTKE130

degree. Despite this small variation in education level across participants, theregressions were run including a variable representing education level. Resultsshowed no significant variation from those not including an education level variable.2. Ethically sensitive scenarios (or vignettes) were used in which the participants

are asked to place themselves in the situations and indicate their probable reaction.Vignettes were deemed to be the preferable research technique by Cavanaugh andFritzsche (1985, p. 291) in their study of techniques available to investigate indi-viduals’ ethical principles and behavior and have been used in many previous studieson ethical behavior (Reidenbach & Robin, 1988, 1990; Flory, Phillips, Reidenbach,& Robin, 1992; Hunt & Vasquez-Parraga, 1993; Dreike & Moeckel, 1995).3. Additional measures pertaining to daily activities on the job and whether the

questionnaire was completed at work or at home were also included in this section.Statistical analysis of these measures showed that they were not significantly relatedto the variables investigated in the study.4. Note that the ethically sensitive situations as they appear in the appendix are

ordered such that the personal and business scenarios dealing with each type ofsituation are grouped together. This ordering is solely for presentation purposes; the16 situations were randomly ordered in the actual questionnaire.5. Given the continuous nature of two of the regression variables, natural log

transformations were performed on age and length of employment. Regression re-sults using the transformations were virtually identical to those without the trans-formations. Consequently, regression results with the transformations are notreported in the chapter. As mentioned earlier in the chapter, age and length ofemployment were highly positively correlated at 0.8930. To ensure that multicol-linearity was not a problem in the regressions, variance inflation factors (VIFs) werecomputed. Since none of the VIFs were greater than the threshold value of 10 (Neter,Wasserman, & Kutner, 1990, p. 409), both variables were included in the regressions.6. Generally, accountability is thought of in terms of external parties. Given the

wide range of settings in the personal ethically sensitive situations included in thestudy, however, it was considered to be quite difficult to determine to whom anindividual may feel responsible across the various settings. Thus, the measure ofpersonal responsibility or accountability included in the study represents a constantmeasure across all personal ethically sensitive situations.

ACKNOWLEDGMENTS

The author expresses appreciation for the helpful comments of Scott Jackson,

Timothy Louwers, Austin Reitenga, Richard Scamell, editor Vicky Arnold,

associate editor Robin Roberts, and two anonymous reviewers.

REFERENCES

Ameen, E., Guffey, D., & McMillan, J. (1996). Gender differences in determining the ethical

sensitivity of future accounting professionals. Journal of Business Ethics, 15(5), 591–597.


American Institute of Certified Public Accountants (AICPA). (2003). Code of professional con-

duct. New York, NY: AICPA.

Applbaum, A. (1995). Professional detachment: The executioner of Paris. Harvard Law Review,

109(1), 458–486.

Applbaum, A. (1999). Ethics for adversaries: The morality of roles in public and professional life.

Princeton, NJ: Princeton University Press.

Arnold, D., & Ponemon, L. (1991). Internal auditors’ perceptions of whistle-blowing and the

influence of moral reasoning: An experiment. Auditing: A Journal of Theory and Practice,

10(2), 1–15.

Asare, S., Trompeter, G., & Wright, A. (2000). The effect of accountability and time budgets on

auditors’ testing strategies. Contemporary Accounting Research, 17(4), 539–560.

Ashton, R., Kleinmuntz, D., Sullivan, J., & Tomassini, L. (1988). Audit decision making. In:

A. R. Abdel-khalik & I. Solomon (Eds), Research opportunities in auditing: The second

decade (pp. 95–132). Sarasota, FL: The American Accounting Association.

Avolio, B., & Gibbons, T. (1988). Developing transformational leaders: A life span approach.

In: J. Conger & R. Kanungo (Eds), Charismatic leadership: The elusive factor in organ-

izational effectiveness (pp. 276–308). San Francisco, CA: Jossey-Bass.

Bandura, A. (2002). Selective moral disengagement in the exercise of moral agency. Journal of

Moral Education, 31(2), 101–119.

Bandura, A., Barbaranelli, C., Caprara, G., & Pastorelli, C. (1996). Mechanisms of moral

disengagement in the exercise of moral agency. Journal of Personality and Social

Psychology, 71(2), 364–374.

Bandura, A., Caprara, G., Barbaranelli, C., Pastorelli, C., & Regalia, C. (2001). Sociocognitive

self-regulatory mechanisms governing transgressive behavior. Journal of Personality and

Social Psychology, 80(1), 125–135.

Bandura, A., Underwood, B., & Fromson, M. (1975). Disinhibition of aggression through

diffusion of responsibility and dehumanization of victims. Journal of Research in

Personality, 9, 253–269.

Brabeck, M. (1984). Ethical characteristics of whistle-blowers. Journal of Research in Person-

ality, 18(1), 41–53.

Buchman, T., Tetlock, P., & Reed, R. (1996). Accountability and auditors’ judgments about

contingent events. Journal of Business Finance and Accounting, 23(3), 379–398.

Cavanaugh, G., & Fritzsche, D. (1985). Using vignettes in business ethics research. In:

L. Preston (Ed.), Research in corporate social performance and policy (pp. 279–293).

Greenwich, CT: JAI Press.

Chartier, J. (2002). Accounting fraud on the rise (January 11). CNNMoney. http://money.cnn.com/

2002/01/11/companies/acct_scandals

Cuccia, A., Hackenbrack, K., & Nelson, M. (1995). The ability of professional standards to

mitigate aggressive reporting. The Accounting Review, 70(2), 227–248.

Diener, E., Dineen, J., Endresen, K., Beaman, A., & Fraser, S. (1975). Effects of altered

responsibility, cognitive set, and modeling on physical aggression and deindividuation.

Journal of Personality and Social Psychology, 31(2), 328–337.

Dreike, E., & Moeckel, C. (1995). Perceptions of senior auditors: Ethical issues and factors

affecting actions. Research on Accounting Ethics, 1, 331–348.

Emby, C., & Gibbins, M. (1988). Good judgment in public accounting: Quality and justifi-

cation. Contemporary Accounting Research, 4(2), 287–313.

ROBIN R. RADTKE132

Fisher, D., Merron, K., & Torbert, W. (1987). Human development and managerial effective-

ness. Group and Organizational Studies, 12(3), 257–273.

Flory, S., Phillips, T., Reidenbach, R., & Robin, D. (1992). A multidimensional analysis of

selected ethical issues in accounting. The Accounting Review, 67(2), 284–302.

Glover, S. (1997). The influence of time pressure and accountability on auditors’ processing on

nondiagnostic information. Journal of Accounting Research, 35(2), 213–226.

Hackenbrack, K. (1992). Implications of seemingly irrelevant audit evidence in audit judgment.

Journal of Accounting Research, 30(1), 126–136.

Hackenbrack, K., & Nelson, M. (1996). Auditors’ incentives and their application of financial

accounting standards. The Accounting Review, 71(1), 43–59.

Hoffman, V., & Patton, J. (1997). Accountability, the dilution effect, and conservatism in

auditors’ fraud judgments. Journal of Accounting Research, 35(2), 227–237.

Hunt, S., & Vasquez-Parraga, A. (1993). Organizational consequences, marketing ethics and

salesforce supervision. Journal of Marketing Research, 30(1), 78–90.

Jeffrey, C., & Weatherholt, N. (1996). Ethical development, professional commitment, and rule

observance attitudes: A study of CPAs and corporate accountants. Behavioral Research

in Accounting, 8, 8–31.

Johnson, V., & Kaplan, S. (1991). Experimental evidence on the effects of accountability on

auditor judgments. Auditing: A Journal of Practice and Theory, 10(Supplement), 96–113.

Jones, T. (1991). Ethical decision making by individuals in organizations: An issue-contingent

model. Academy of Management Review, 16(2), 366–395.

Kennedy, J. (1993). Debiasing audit judgment with accountability: A framework and exper-

imental results. Journal of Accounting Research, 31(2), 231–245.

Kerssens-van Drongelen, I., & Fisscher, O. (2003). Ethical dilemmas in performance measure-

ment. Journal of Business Ethics, 45(1), 51–63.

Lockheed, G. (1980). Know then decide. In: T. Wallsten (Ed.), Cognitive processes in choice and

decision behavior (pp. 143–154). Hillsdale, NJ: Lawrence Erlbaum Associates Inc.

Lord, A. (1992). Pressure: A methodological consideration for behavioral research in auditing.

Auditing: A Journal of Practice and Theory, 11(2), 89–108.

Luban, D. (1988). Lawyers and justice. Princeton, NJ: Princeton University Press.

McNeel, S. (1994). College teaching and student moral development. In: J. Rest & D. Narvaez

(Eds), Moral development in the professions (pp. 27–43). Hillsdale, NJ: Lawrence

Erlbaum Associates, Inc.

McPhail, K. (2001). The other objective of ethics education: Re-humanising the accounting

profession – a study of ethics education in law, engineering, medicine and accountancy.

Journal of Business Ethics, 34(3/4), 279–298.

Messier, W., Jr., & Quilliam, W. (1992). The effect of accountability on judgment: Development

of hypotheses for auditing. Auditing: A Journal of Practice and Theory, 11(Supplement),

123–151.

Moak, E. (2004). Female whistleblowers: Contributions, motivations, and society’s response.

Georgetown University Law Center www.law.georgetown.edu.

National Commission on Fraudulent Financial Reporting (the Treadway Commission). (1987).

Report of the national commission on fraudulent financial reporting. Washington, DC:

NCFFR.

Neter, J., Wasserman, W., & Kutner, M. (1990). Applied linear statistical models (3rd ed.). Burr

Ridge, IL: Richard D. Irwin, Inc.


Nystrom, P., & McArthur, A. (1989). Organizations’ roles in career theory. In: M. Arthur,

D. Hall & B. Lawrence (Eds), Handbook of career theory (pp. 490–505). Cambridge,

England: Cambridge University Press.

Peecher, M. (1996). The influence of auditors’ justification processes on their decisions: A cog-

nitive model and experimental evidence. Journal of Accounting Research, 34(1), 125–140.

Pitt, H. (2002). Public statement by SEC chairman: Regulation of the accounting profession

(January 17). Washington, DC: U.S. Securities & Exchange Commission.

Ponemon, L. (1990). Ethical judgments in accounting: A cognitive-developmental perspective.

Critical Perspectives in Accounting, 1(2), 191–215.

Ponemon, L. (1992a). Ethical reasoning and selection-socialization in accounting. Accounting,

Organizations and Society, 17(3–4), 239–258.

Ponemon, L. (1992b). Auditor underreporting of time and moral reasoning: An experimental

lab study. Contemporary Accounting Research, 9(1), 171–189.

Ponemon, L. (1995). The objectivity of accountants’ litigation support judgments. The Ac-

counting Review, 70(3), 467–488.

Ponemon, L., & Gabhart, D. (1990). Auditor independence judgments: A cognitive-developmental

model and experimental evidence. Contemporary Accounting Research, 7(1), 227–251.

Radtke, R. (2000). The effects of gender and setting on accountants’ ethically sensitive deci-

sions. Journal of Business Ethics, 24(4), 299–312.

Reidenbach, R., & Robin, D. (1988). Some initial steps toward improving the measurement of

ethical evaluations of marketing activities. Journal of Business Ethics, 7(11), 871–879.

Reidenbach, R., & Robin, D. (1990). Toward the development of a multidimensional scale for

improving evaluations of business ethics. Journal of Business Ethics, 9(8), 639–653.

Rest, J. (1979).Development in judging moral issues. Minneapolis, MN: University ofMinnesota Press.

Rest, J. (1986). Moral development: Advances in research and theory. New York, NY: Praeger.

Shaub, M. (1994). An analysis of factors affecting the cognitive moral development of auditors

and auditing students. Journal of Accounting Education, 12(1), 1–24.

Smircich, L. (1983). Concepts of culture and organizational analysis. Administrative Science

Quarterly, 28(3), 339–358.

Tetlock, P. (1985). Accountability: The neglected social context of judgement and choice. In:

B. Staw & L. Cummings (Eds), Research in organizational behavior, 9 (pp. 279–332).

Greenwich, CT: JAI Press.

Tetlock, P., Skitka, L., & Boettger, R. (1989). Social and cognitive strategies for coping with

accountability: Conformity, complexity, and bolstering. Journal of Personality and Social

Psychology, 57(4), 632–640.

Thoma, S. (1986). Estimating gender differences in the comprehension and preference of moral

issues. Developmental Review, 6(2), 165–180.

Tilker, H. (1970). Socially responsible behavior as a function of observer responsibility and

victim feedback. Journal of Personality and Social Psychology, 14(2), 95–100.

Turner, C. (2001). Accountability demands and the auditors’ 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.

Velasquez, M., & Rostankowski, C. (1985). Ethics: Theory and practice. Englewood Cliffs, NJ:

Prentice-Hall.

Weick, K. (1979). Cognitive processes in organizations. In: B. Staw (Ed.), Research in organ-

ization behavior 1 (pp. 41–74). Greenwich, CT: JAI Press.

ROBIN R. RADTKE134

Weigold, M., & Schlenker, B. (1991). Accountability and risk taking. Personality and Social

Psychology Bulletin, 17(1), 25–29.

Werhane, P., & Freeman, R. (1999). Business ethics: The state of the art. International Journal

of Management Reviews, 1, 1–16.

www.cnn.com. (2002). Time names whistleblowers as persons of the year (December 23). http://

www.cnn.com/2002/US/12/23/time.persons.of.year

APPENDIX

Ethically Sensitive Situations

(1) You are preparing your personal tax return. During the year you made

several substantial charitable contributions which totaled $1,700. Ad-

ditionally, you made several smaller contributions for which you did

not receive receipts. You estimate the value of these items at no more

than $150. When preparing this year’s tax return, you notice that by

claiming just over $2,000 in charitable expenses (instead of $1,850) you

would drop to a lower tax bracket which would save you about $100 in

taxes. Do you claim over $2,000 in charitable contributions?

(2) Assume you are a tax accountant within your company. During the

prior year, the company purchased an asset for $55,000. Normal ship-

ping, handling, insurance, and setup costs totaled $7,000. This makes the

cost basis of the asset $62,000. Owing to a problem installing the asset,

however, an additional cost of $5,000 was incurred. You are aware that

unusual expenditures on asset installment should be capitalized (making

the cost basis of the asset $67,000), but expensing these costs for tax

purposes lowers the company’s taxes for the current year. The controller

of your company has made it clear that she favors aggressive tax po-

sitions and that the probability of these expenditures being audited is

minimal. Do you expense the additional cost of $5,000 for tax purposes?

(3) You and your brother regularly enjoy competing at computer games. He

recently purchased a relatively costly new game and installed it on his

PC. You have since been playing the game at his house, but he has

gotten the upper hand, since he can practice anytime he wants. He offers

to install the game on your PC from the same disk. You’re not sure if

you should do this since the software is licensed to be installed on only

one computer, but he says everyone does it all the time. Since you hate

losing to him, you consider it. Do you copy the game onto your PC?

(4) At work you use many different software packages. Several weeks ago

your supervisor ordered a new package for you that several of your


colleagues are currently using. The software is now late in arriving.

The package would aid you tremendously in completing your current

project, but is not absolutely necessary. Earlier today your supervisor

brought her copy of the software over to you and suggested that you

copy it onto your computer for use until your copy arrives. You know

that the software is licensed to be installed onto only one computer.

Do you copy the software?

(5) Yesterday you drove to the store with your neighbor and her young

son. When you got back to the car, your neighbor noticed that her

son picked up a small item from the store worth about $5 that wasn’t

paid for. Your neighbor reprimanded the child and then turned to

you and said she was ready to go. You asked her if she was going to

go back into the store to pay for the item. She said it’s not worth the

hassle. Do you refuse to drive her home unless she goes back to the

store and pays for the item?

(6) While at a colleague’s home you notice several piles of supply items

from the office including notepads, diskettes, and boxes of pens.

When you ask your colleague about this, she explains that she often

brings work home from the office and even if she uses some of the

work supplies for personal projects, she sometimes uses personal

items for work-related efforts, so it all washes out in the end anyway.

Although the supplies don’t total a substantive amount, you are

concerned that they have been removed from the office and wonder

what else your colleague may ‘‘borrow’’ for personal use. Do you tell

your superior about your colleague’s actions?

(7) You and your friend are watching the evening news when a picture of

a man wanted for questioning is shown. You recognize the man as an

old acquaintance of your friend. You ask your friend when was the

last time he saw the man. He tells you he actually ran into him last

week, but had no idea of any illegal activities he may have been

engaged in. You ask your friend whether he is going to call the au-

thorities to report his encounter with the man and he says that he

doesn’t want to get involved. Do you call the authorities to report

your friend’s association with the man?

(8) You and a colleague of yours started work at the same time after

graduating college. Both of you are quick learners, but he seems to

always get things done faster and is continually receiving praise from

your superior who supervises you both. Since your company operates

on flexible hours, you and your colleague rarely leave the office at the

same time. Yesterday you left at the same time, however, and you

ROBIN R. RADTKE136

noticed that he was taking files home with him to work on. Since your

company considers you salaried employees, you don’t receive over-

time. On the one hand you think that if your colleague wants to spend

his personal time on work that’s his problem, but on the other hand

he’s making you look bad. Do you tell your superior about your

colleague’s actions?

(9) Last week you discussed your personal investments over lunch with a

friend of yours from college. Your friend told you that since he has

been working at a brokerage firm, he’s privy to a wealth of inside

information about a variety of stocks. Over the course of the meal,

several pieces of information about stocks that you own are disclosed.

Your friend also tells you how to use this information discreetly and

suggests that you may want to use it to your advantage. He divulges

that he has done this several times without being caught. Do you

trade your stock based on this inside information?

(10) While at lunch with several of your colleagues last week you over-

heard a discussion about a client company’s financial situation. An

accountant working closely with the company noticed significant de-

creases in sales and receivables. He wasn’t sure exactly how bad it was

until he heard a rumor at the company about the possibility of filing

for bankruptcy. You’re now worried because you own a significant

block of shares in the company. Do you sell the shares based on this

inside information?

(11) Two of your good friends are engaged in negotiations for the sale/

purchase of a used car. Several years ago the seller of the car was

involved in an accident which you witnessed, but the buyer is unaware

of. The car sustained substantial damage, but was repaired. Now the

seller is representing the car as never having been in an accident. The

difference in the appraisal value of the car is difficult to measure. Do

you tell the buyer of the car about the accident?

(12) While on a trip out of town on business you had dinner with your sister.

Your company has a policy of reimbursing dinner expenses up to $50

per meal. The total cost for this meal for both you and your sister was

$35.70. The cost of your meal alone was $16.30. You know that others in

your company routinely submit claims for dinner expenses for nonbusi-

ness parties. Do you claim the entire amount for reimbursement?

(13) Yesterday your second grader came home and started watching car-

toons. During one of the programs he said ‘‘see that little boy, he’s

not as smart as I am.’’ The little boy was of a different race and when

you asked your son why he thought that, he said his teacher had told


him so during school. Do you take action to reprimand the teacher on

the basis of racial discrimination?

(14) A couple of weeks ago your division posted a job opening. You told a

good friend of yours who was looking to switch jobs about the po-

sition and she submitted a resume. She met the job qualifications and

was one of three candidates interviewed. She told you that she

thought the interview went very well and was excited about the pos-

sibility of being hired. Your direct superior also told you that he

enjoyed interviewing your friend and that in his opinion, she should

get the job. Yesterday when you came into the office you found out

that one of the other candidates got the job. When inquiring as to the

reasons that your friend didn’t get an offer, you were told that al-

though her credentials were the best of the three, her personality just

didn’t fit with the rest of the employees in the division. You are

somewhat suspicious of this explanation, since your friend is a mi-

nority and you have heard some of the senior personnel in the division

make racial slurs. Do you investigate the matter further on the

grounds of racial discrimination?

(15) While talking with your neighbor he tells you that at the next neigh-

borhood association meeting he intends to use the open forum time to

plug his company’s new product line. The neighborhood association

has rules about this, but the president is often somewhat negligent in

enforcing them. When you tell your friend that he probably shouldn’t

use the meeting time in this way, he responds by saying that it’s a free

country and he can speak about anything he wants to. Do you at-

tempt to contact the president and others in the association to try to

block your friend from speaking?

(16) Based on your recent work-team experiences you were asked by your

division head to draft a memo highlighting the good and bad points.

You recalled many of each and drafted the memo accordingly. After

your supervisor reviewed the memo she told you to tone it down a bit.

You ask her what she means by that and she basically tells you that in

her opinion it’s not good to give either too much praise or too many

admonishments. You say that is counterproductive to the purpose of

the memo to not disclose the actual occurrences. She tells you to print

the memo as is at your own risk. You feel that you’ve done the job

that was asked of you and you have a right to tell it like it is. Do you

print the memo as is?

ROBIN R. RADTKE138

THE EFFECT OF MANAGER’S

MORAL EQUITY ON THE

RELATIONSHIP BETWEEN BUDGET

PARTICIPATION AND PROPENSITY

TO CREATE SLACK: A RESEARCH

NOTE

Adam S. Maiga

ABSTRACT

This chapter uses agency theory and ethics literature to assess the mod-

erating effect of manager’s moral equity on the relation between budget

participation and propensity to create slack. Moral equity is the major

evaluative criterion for ethical judgment, is based on the overall concept of

fairness, justice and right and is often very influential in contemporary

moral thought (Robin & Reidenbach (1996) Journal of Business, 5(1)

17–28). The results indicate that a manager’s moral equity moderates the

effect of budget participation. For managers with high moral equity, the

relationship between participation and manager’s propensity to create

slack is significantly negative while, for managers with low moral equity,

the relationship is significantly positive. Further analyses indicate that





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08006-8

139

high budget participation and high moral equity result in less propensity

to create slack than high budget participation and low moral equity.

INTRODUCTION

A major concern expressed in the literature is that the use of participative

budgetary control processes may result in the generation of slack budgets

(Antle & Eppen 1985). Budgetary participation is the means by which sub-

ordinate managers influence plans and their methods of implementation,

thereby sharing in the decision-making process with their superiors on mat-

ters that affect their areas of responsibility (Milani, 1975; Brownell, 1982).

Budget slack, defined as overstated expenses, understated revenues, or un-

derestimated performance capabilities, allows managers to obtain excess

resources and to shirk more effectively (Lukka, 1988).1

Several accounting studies invoke the principal-agent framework to argue

that subordinates have more accurate information than their superiors re-

garding local conditions, i.e., private information about local conditions

(Merchant, 1981; Christensen, 1982; Chow, Cooper, & Waller, 1988; Waller,

1988). Subordinates’ participation in the budget-setting process may give

superiors the opportunity to gain access to local information (Baiman, 1982;

Baiman & Evans, 1983; Magee, 1980). Participation allows subordinates to

communicate or reveal some of their private information that may be in-

corporated into the standards or budgets against which their performance

would be assessed (Baiman & Evans, 1983). But subordinates may misrep-

resent or withhold some of their private information, resulting in under-

stated revenues and overstated expenses, which could lead to budgetary

Slack (Christensen, 1982; Merchant, 1985; Pope, 1984; Young, 1985).

The results of prior studies examining the relationship between subordi-

nates’ participation and budgetary slack have not been consistent. For ex-

ample, Dunk (1993), Merchant (1985), Cammann (1976), and Onsi (1973)

report evidence suggesting that participation reduces the amount of budget

slack. On the other hand, Lukka (1988) and Young (1985) report that

budget participation and the creation of slack may be positively related; and

Collins (1978) reports that the relationship between budgetary participation

and budgetary slack is not significant.

The above studies view budgetary slack as an organizational or behavioral

issue (Douglas & Wier, 2000) based on moral hazard2 and are not designed to

distinguish wealth maximization from alternative preferences such as ethics,

ADAM S. MAIGA140

fairness, trust, accountability or integrity, and conscience (Luft, 1997). In

agency theory, moral hazard is the basis for a manager’s economic decision

behavior, while moral development theory suggests that individuals’ decision

behavior is influenced by their level of ethical principles (Rutledge & Karim,

1999; Douglas & Wier, 2000; Luft, 1997; Noreen, 1988). Forsyth (1992) pro-

poses that individuals judge or approach business practices, such as the cre-

ation of budgetary slack, as ethical or unethical, and then decide whether or

not to engage in those practices. Noreen (1988) suggests that while there may

be some people who are unreservedly opportunistic, others do constrain their

own behavior out of an ethical sensibility or conscience.

As depicted in Fig. 1, this paper suggests that the relation between man-

ager’s budget participation and his/her propensity to create budgetary slack

may be moderated by his/her moral equity. The moral equity factor used in

past empirical research accounts for the major explanatory impact in the

ethical evaluation process. Moral equity represents a universal ethics con-

struct, and it is the major evaluative criterion for ethical judgment (Robin &

Reidenbach, 1996; Flory, Phillips, Reidenbach, & Robin, 1992).

The philosophy of moral equity is based on the overall concept of fair-

ness, justice and right, and has been very influential in contemporary moral

thought. Moral equity can be used as the basis for arguing against unethical

business practices. Organizational policies are likely to be most effective if

they can be justified according to the ethical concepts used by the people

involved (Arrow, 1985; Baiman, 1990; Douglas & Wier, 2000). Jones (1991)

suggests that concerns for ethics are jointly determined by characteristics of

the situation and the individual. Hence, moral equity could be a useful tool

for looking at how managers make certain ethical judgments in a parti-

cipative budget setting. Essentially, as both the subject and consequence of

unethical business behavior grow in importance, so does the need to know

Budget SlackBudget Participation

Moral Equity

Fig. 1. Model Showing Moral Equity as Moderating Variable.

Effect of Manager’s Moral Equity 141

the grounds on which managers tend to make ethical judgments and their

impact of those judgments. Such insights will help to expand upon previous

studies. Also, the empirical investigation of this study will help to under-

stand how the manager’s level of moral equity impacts the relationship

between budget participation and propensity to create slack.

Many scholars have highlighted, from both the theoretical and empirical

perspectives, the impact of a manager’s budget participation on budget slack.

Most contributions, however, did not analyze the link between a manager’s

budget participation and his/her propensity to create slack within the context

of the manager’s moral equity. The purpose of this study is to assess the

moderating effect of manufacturing business unit3 managers’ moral equity on

the relationship between the extent of their budget participation and propen-

sity to create slack. This study proposes that moral equity influences managers’

behavior in participative budgeting. Specifically, managers with high moral

equity may use their influence primarily to reduce slack while managers with

low moral equity may use their influence primarily to create slack.

The remainder of this chapter is organized as follows. The next section

draws upon previous literature to develop a theoretical framework linking

the variables of interest. The methodology and statistical results are dis-

cussed in the third and fourth sections, respectively. The paper concludes

with a discussion of the findings and suggestions for future research.

LITERATURE REVIEW AND HYPOTHESIS

DEVELOPMENT

Agency theory posits that information asymmetry may systematically in-

fluence the extent to which participation leads to slack budgets. Information

asymmetry arises when subordinates (agents) are in possession of informa-

tion that affects the decision process between subordinates and superiors

(principals) (e.g., Baiman & Evans, 1983; Penno, 1984; Coughlan & Sch-

midt, 1985). When the agent has private information that is not available to

the firm (i.e., information asymmetry), the principal can no longer verify if

the agent’s decisions are in accordance with the firm’s interests. This pro-

vides the agent with the opportunity to shirk by making decisions that

conflict with the interests of the firm. When the agent is under conditions of

an incentive to shirk and an opportunity to shirk (e.g., private information),

the problem of moral hazard can occur.

Magee (1980) proposes that budgets could be improved if principals were

aware of local information held by subordinates prior to the budgets being set,

ADAM S. MAIGA142

i.e., information asymmetry is eliminated. Baiman (1982), Chow (1988),

Blanchard and Chow (1983), and Waller (1988) argue that subordinates in

many organizational settings have more accurate information than their su-

periors on the factors influencing performance. Baiman and Evans (1983)

propose that firms in which subordinates have such information, participa-

tion-based management control systems allow subordinates to reveal or com-

municate some of their private information that may then be incorporated into

the standards or budgets against which their performance is evaluated.

Unfortunately, agents may misrepresent or withhold from their principals

some or all of their locally- based information, which could lead to budgets

containing slack (Christensen, 1982; Baiman & Sivaramakrishnan, 1991). The

argument for agents either falsifying or withholding their private information

is that managers plan to have slack in their budgets to enable budgeted

objectives to be met and improve the likelihood that they will be compensated

for their efforts. For example, Waller (1988) argues that if subordinates be-

lieve their communicated private information will be employed in standard

setting for the purpose of performance evaluation, they may have an incentive

to bias their communications to facilitate the setting of relatively easy stand-

ards. This problem, Waller (1988) stresses, is particularly salient when sub-

ordinates’ pay schemes are budget-based and the budget-setting process is

participative. As managerial compensation is often based on budget achieve-

ment and information provided by agents is likely to be used in their per-

formance evaluation (Christensen, 1982; Baiman & Evans, 1983), the

prospect of dysfunctional consequences arising from the presence of infor-

mation asymmetry may be non-trivial. Young (1985) warns that the existence

of private information in conjunction with participation may result in sub-

ordinates intentionally building excess resource requirements into budgets or

consciously understating production capabilities.

These agency-based studies capture behavior shifts in response to incen-

tives in particular settings, but do not provide decisive tests of the simple

self-interest model vs. plausible alternative utility functions. Luft (1997)

questions the commonly cited belief that self-interest is a good approxima-

tion for behavior and argues that prior tests have lacked the power to

distinguish between self-interested and ethical models of behavior. Thus,

while the agency-based studies offer evidence for the existence of a pref-

erence for wealth, they do not offer evidence against the existence of other

significant preferences (Luft, 1997).

Hence, this study proposes that whether managers are predisposed to

pursuing self-interests or organizational interests depend upon their level of

moral equity. Ethical concerns typically arise in situations where self-interest


conflicts with a moral duty to others (Bowie & Duska, 1990). DeGeorge

(1992) asserts that ethically motivated agents exercise effective self-control

that no amount of external control can match, and that researchers should

utilize, promote, and incorporate such motivation.

A person who fails to recognize a moral issue will fail to employ moral

decision-making schemata (Jones, 1991). The consideration of budget slack

from the perspective of ethical decision-making assumes that a person must

be able to recognize budget slack creation as a moral issue. Slack creation

may be inconsistent with role-related norms and desired virtues of profes-

sional managers, and the resource misallocation that results is detrimental to

other organizational units and to investors (Merchant, 1995). Thus, creation

of budget slack is an ethical dilemma – a predicament with a moral com-

ponent (Douglas & Wier, 2000). Opportunistic behavior on the part of the

agent may be controlled in part by the agents’ concerns for reputation or

ethics (see Arrow, 1985; Baiman, 1990), and the idea that individuals can be

strongly motivated to pursue organizational interests (i.e., without the pros-

pect of self-interest) has been shown in the literature. For example, in an

experimental study, Stevens (2002) investigated the effects of two potential

controls for opportunistic self-interest – reputation and ethics. The results

provide strong evidence that reputation and ethics reduce budgetary slack.

While slack levels under a slack-inducing pay scheme were higher than in

prior experimental studies, subjects still restricted the amount of slack in

their budgets below the maximum, and thereby failed to maximize their pay.

This result is consistent with findings of Evans, Hanna, Krishnan, Moser

(2001) that subjects often sacrifice wealth to make honest reports of pro-

ductive capability. In Evans et al. (2001), budgetary slack is negatively as-

sociated with a measure of ethical responsibility from the pre-experiment

personality questionnaire as well as reputation and ethical concerns ex-

pressed in the exit questionnaire. As information asymmetry regarding pro-

ductive capability increased, subordinates expressed lower reputation

concerns, thereby reducing the superior’s ability to monitor the slack in

their budget. Ethical concerns, however, were not diminished with increases

in information asymmetry. These results suggest that reputation is a socially

mediated control whereas ethics is an internally mediated control for op-

portunistic self-interest.

In an experimental study, Douglas, Davidson, & Schwartz (2001) inves-

tigated auditors’ ethical judgments in situations typical of those they face in

practice. Results indicate that ethical orientation is related to ethical judg-

ments in high (but not low) moral intensity situations. These results support

Jones’ (1991) issue-contingent argument that suggests that differences in

ADAM S. MAIGA144

characteristics of a moral issue itself, its moral intensity, affect individuals’

responses to the issue.

The chapter paper proposes that in participatory budgeting, the individ-

ual’s level of moral equity influences his or her attempts to create budgetary

slack. For subordinates with high moral equity, budget participation and

budgetary slack may be inversely related. In a high budget participation

setting, managers are likely to make use of all possible sources of informa-

tion in order to increase the accuracy of the budget decision (Gul, 1991).

Since information availability is improved through participation, parti-

cipative budgeting will lead to more analytical and accurate decision-mak-

ing. Therefore, in a high participatory budget, subordinates with high moral

equity will use their private information to produce an accurate budget, i.e.,

to reduce slack and benefit the organization. In contrast, when budget par-

ticipation is low, subordinates with high moral equity will have little op-

portunity to share their private information, and superiors will have only

limited success in attempting to reduce slack. Therefore, there is less op-

portunity to reduce budgetary slack.

For individuals with low moral equity, budget participation provides op-

portunities for the subordinate to create slack. For these individuals, the

relation between budget participation and budgetary slack is positive, i.e., as

budget participation increases, budgetary slack increases. Since the subor-

dinate has low moral equity, he/she may use participation to introduce slack

and gain favorable future evaluations (i.e., the subordinate seeks to max-

imize self-interest).

The proposed effects of moral equity on the relationship between budget

participation and budgetary slack results in the following hypothesis, stated

in alternative form:

H1. There is an interaction between moral equity and participative

budgeting that affects budgetary slack. For managers with high moral

equity, increasing budget participation will decrease budgetary slack. For

managers with low moral equity, increasing budgetary participation will

increase budgetary slack.

RESEARCH METHOD

Sample

A questionnaire was administered to a sample of managers (plant managers,

manufacturing managers, operations managers, marketing managers,


research managers, distribution managers) from manufacturing companies

in the USA Manufacturing organizations were selected for study, as the use

of budgets in such organizations is common. The primary source of sample

selection was the Industry Week series. For this study, an initial mailing list

of 1,103 business units was obtained and a random sample of 650 names was

selected.4 A cover letter explained the purpose of the study with an exhor-

tation for participation and cooperation. Based on the survey responses, two

criteria were used to select the participants: (1) each participant had budget

responsibility in the subunit and (2) each unit was a profit-center (see ap-

pendix, part II). An abbreviated copy of the questionnaire used in the study

appears in the appendix.

In the first 3 weeks, 167 questionnaires were returned; that was followed

by a second mailing which resulted in 56 new responses. Of the 223 returned

questionnaires, only 193 were usable.5 In an attempt to increase the number

of respondents, 150 non-respondents were chosen randomly and contacted

by telephone; that resulted in a return of 78 questionnaires of which 58 were

usable.6 Overall, this data collection led to 251 usable responses7 with a

38.61% response rate.

Measurement and Validation of Variables

The variables used to answer the research question are budget slack, budget

participation, and moral equity. The measurement of the variables is ob-

tained from average responses from the questionnaire results. The factor

loadings, explained variances, and reliability measures are reported in

Table 1. The appendix contains an abbreviated copy of the research ques-

tionnaire used to measure the self-reported variables in this study.

Propensity to Create Slack

Propensity to create slack is operationalized using the three-item scale used in

Kren (1993) and adapted from Merchant (1985). Merchant’s original four-

item scale was examined by Hughes and Kwon (1990) who suggested deleting

one item to improve the scale’s reliability. Thus this study uses the three items

suggested by Hughes and Kwon (1990). The response scale is a 7-point

Likert-type scale ranging from one (strongly disagree) to seven (strongly

agree). Principal component analysis with varimax rotation was used to ex-

amine the extent to which these measures are interrelated and produced one

factor with total variance of 88.208% and an eigenvalue greater than one.

Cronbach alpha was 0.928, indicating that the measures are reliable.

ADAM S. MAIGA146

Table 1. Factor Loadings, Explained Variance and

Reliability Measures.

Factor

Loading

Explained

Variance

Reliability

Coefficient

(Cronbach Alpha)

Budget slack 88.208 0.928

To protect himself, a manager

submits a budget that can safely

be attained

0.764

In good business times, your

superior is willing to accept a

reasonable level of slack in the

budget

0.673

Slack in the budget is good to do

things that cannot be officially

approved

0.782

Budget participation 86.154 0.905

I am involved in setting all of my

budget

0.864

My superior clearly explains

budget revisions

0.793

I have frequent budget-related

discussions with my superior

0.914

I have a great deal of influence on

my final budget

0.696

My contribution to the budget is

very important

0.731

My superior initiates frequent

budget discussions when when

the budget is being prepared

0.805

Moral equity

Scenario A 72.519 0.873

Fair/unfair 0.859

Just/unjust 0.925

Morally right/

not morally right 0.903

Acceptable/unacceptable to

family

0.680

Scenario B 69.472 0.852

Fair/unfair 0.873

Just/unjust 0.920

Morally right/not morally right 0.885

Acceptable/unacceptable to

family

0.620

Scenario C 63.598 0.808


Budget Participation

Budget participation is measured using the Milani (1975) six-item measure.

The response scale is a 7-point Likert-type scale ranging from one (strongly

disagree) to seven (strongly agree). A principal component analysis with

varimax rotation produced one factor with total variance of 86.154% and

an eigenvalue greater than one. A reliability check for the measures pro-

duced a Cronbach alpha of 0.905, indicating that the measures are reliable.

Moral Equity

Four scenarios that the IMA Resources Center developed and that Flory

et al. (1992) used in a subsequent study are used in developing a measure of

moral equity because they portray substantially more involved, realistic

situations. Each scenario included an action statement to assure that all

respondents were reacting to the same stimulus. The action statement was

particularly necessary with the situations described in the present study.

Consequently, the four scenarios are used in this study.

Each scenario portrays a different sort of ethical dilemma. Scenarios A

and D describe actions that might not be perceived as explicitly ethical or

unethical, while scenarios B and C feature what most would label as def-

initely unethical behavior. Scenario A describes a superior who is making

questionable expenditures that he claims meet upper management’s ap-

proval. The manager, who may find himself in a marketing environment

different from his background, is faced with establishing the proper lines of

Table 1. (Continued )

Factor

Loading

Explained

Variance

Reliability

Coefficient

(Cronbach Alpha)

Budget slack 88.208 0.928

Fair/unfair 0.799

Just/unjust 0.886

Morally right/not morally right 0.774

/unacceptable to family 0.706

Scenario D 65.561 0.825

Fair/unfair 0.884

Just/unjust 0.851

Morally right/ not morally

right

0.799

Acceptable/ unacceptable to

family

0.686

ADAM S. MAIGA148

authority in connection with an issue that may not be a violation of com-

pany policy. Scenario B involves a controller who is asked to falsify external

financial statements for the purpose of procuring additional working cap-

ital. Although this may actually happen in some companies, managers typ-

ically agree that falsification of external statements is wrong. This is also

true of the specific violations of company policy shown in scenario C. A

difference in scenario C, besides the fact that it is an internal situation, is

that the manager had previously violated company policy, and now, in an

attempt to rectify a resulting failure, decides to violate the policy again.

Scenarios A, B, and C all implicitly involve a manager’s job security; but in

each situation, the individuals are seemingly concerned with their company’s

welfare. In contrast, scenario D emphasizes the manager’s personal prob-

lems. In this scenario, company policy is not clearly delineated, and there

could be some uncertainty whether the manager’s action is unethical. The

additional background information provided in scenario D allows the re-

spondent to empathize with the manager’s personal difficulties, although it

is unclear whether his personal situation has any bearing on his decision.

Respondents were asked to react to each scenario using a set of measures

developed by Reidenbach and Ronald (1991). The set focuses on the dy-

namics of decision making regarding a manager’s moral equity. It consists

of four bipolar seven-point Likert scales (Fair/Unfair, Just/Unjust, Morally

right/Not morally right, Acceptable to my family/Unacceptable to my

family).

Factor analysis with principal component analysis with varimax rotation

is used to examine the extent to which the moral equity measures under each

scenario are interrelated. One factor with eigenvalue 41 emerged from the

analysis for each scenario, with corresponding varimax rotation factor so-

lution retaining at least 67.31% of the total variance in the data. The

Cronbach alphas were 0.873, 0.852, 0.808, and 0.825, respectively, suggest-

ing that the measures are reliable.

To assess the content validity of the scales, each moral equity measure is

regressed on its corresponding ethical intention measure to test whether the

constructs in fact measure manager’s moral equity (Flory et al. 1992).

Manager’s ethical intention to each scenario is measured on a 7-point bi-

polar scale range from (1) ethical to (7) unethical. This is a common val-

idation procedure in the social sciences. A high covariation (R2) between

ethical intention and moral equity suggests that the moral equity captures

much of what the respondents mean by ‘‘ethical’’ (Flory et al. 1992). The

individual b values also help define the concept of ‘‘ethics’’ for the respond-

ents. The results which appear in Table 2 indicate that under the four


scenarios the moral equity measures explain 49.3–65.4% of the variance in

what the managers defined as ethical. Corresponding b values, ranging from

0.523 to 0.709, suggest that the ethical intention measures capture much of

what the respondents mean by ‘‘ethical.’’

Research Model and Testing Procedures

The average for the six responses for budget participation, the overall av-

erage for the four scenarios measuring moral equity, and the average for the

three responses for propensity to create slack were computed in order to test

the hypothesis. The hypothesis posits a moderating effect of manager’s

moral equity on the relationship between budget participation and propen-

sity to create slack. Based on this approach, the following regression model

is employed:

PCS ¼ a0 þ b1BPþ b2MEþ b3ðBP�MEÞ þ � (1)

Where PCS is the propensity to create slack; BP the budget participation;

ME the moral equity; a0 the intercept; b1; b2; and b3 are the regression

coefficients; and � the error term.

RESULTS

Descriptive Statistics

In Table 3, the mean and standard deviation values of the variables used to

answer the research question denote that many respondents indicated some

probability of engaging in the activity specified in the scenarios and their

level of budget participation. Additional information on respondents’ char-

acteristics is provided in Table 3. The respondents to the question regarding

Table 2. A Comparison of the Moral Equity Measure and the Ethical

Intention Measure.

Scenario Regression Results Moral Equity b1

R2

A 0.608 0.523

B 0.493 0.691

C 0.567 0.709

D 0.654 0.653

ADAM S. MAIGA150

number of years with the division have a mean of 9.14 in their current

position. To the number of years in management question, respondents

indicated a mean of 13.12 years. The results also show that the average

number of employees equals 241. For the 194 divisions that provided sales

figures, the mean was $5.4 million.

Table 4 presents the correlation matrix for the variables in the study. Both

budget participation and propensity to create slack negatively correlate with

the interaction term (r ¼ �0:195; p ¼ 0:002; r ¼ �0:498; p ¼ 0:000respectively).

Hypothesis Test

To test the hypothesis, standard scores8 are used for the independent var-

iables in order to provide a clearer basis to interpret signs of the interaction

Table 3. Descriptive Statistics.

Variables Number

of Items

Standard Mean Theoretical

Range

Actual Range

Propensity to create

slack

3 2.685 1.392 1–7 1.00–6.66

Budget participation 6 4.890 1.529 1–7 2.33–6.66

Moral equity 4 3.817a 0.722 1–7 1.66–4.66

Size (number of

employees)

N/A 240.841 148.897 N/A 32–764

Years at division N/A 9.143 8.999 N/A 4–16

Years in management

position

N/A 13.116 9.035 N/A 5–26

Net sales (million) N/A $5.524 $1.339 N/A 1.371–12.623

aOverall mean.

Table 4. Correlations among Variables.

BP ME BP�ME

BP 1 — —

ME 0.121 1 —

0.055 — —

BP�ME �0.195�� 0.012 1

0.002 0.844 —

PCS 0.104 0.046 �0.498**

0.100 0.471 0.000

��po0.001.


coefficient (Brownell & Hirst, 1986) and to minimize mulitcollinearity be-

tween main and cross-product effects (Cronbach, 1987). The interaction

term is constructed by multiplying the standardized scores of budget par-

ticipation and moral equity for each respondent. Tolerance greater than 0.10

is achieved. Variance inflation factor values from the regression analyses

conducted for all the variables are less than 2. Hence, mulitcollinearity does

not appear to be a problem.

To permit an acceptance of the hypothesis, the coefficient b3 in Eq. (1) is

required to be (a) significant, indicating an interaction between participation

and moral equity affecting manager’s propensity to create slack and

(b) negative to support the direction of the hypothesis. The results of the

regression analysis using Eq. (1) appear in Table 5 and indicate that the b3coefficient is statistically significant and negative (t ¼ �8:840; p ¼ 0:000). As

predicted by the hypothesis, budget participation and moral equity interact

to affect manager’s propensity to create slack.

To provide additional support for the hypothesis, the data are divided

into two groups on the basis of the overall mean score for moral equity –

high moral equity and low moral equity categories. Moral equity scores

below the mean are classified as low, while moral equity scores above the

mean are classified as high. Next, budget participation is regressed on pro-

pensity to create slack for each group. Results in Table 6 show that budget

participation is significant and positive with low moral equity group

(b ¼ 0:488; t ¼ 6:131; po0:001 with R2 ¼ 0:239; and adjusted R2 ¼ 0:232).For the high morality group, results show that budget participation is sig-

nificant and negative (b ¼ �0:355; t ¼ �4:279; po0:001 with R2 ¼ 0:126;and adjusted R2 ¼ 0:119). Therefore, the results provide additional supportfor the hypothesis.

Table 5. Regression Analysis.

Standardized

Coefficients b

T Significance Colinearity

Statistics

Tolerance

Variance

Inflation

Factor

Intercept — 55.595 0.000 — —

BP 0.002 0.041 0.968 0.948 1.055

ME 0.039 0.707 0.480 0.985 1.015

BP�ME �0.497 �8.840 0.000 0.962 1.040

R2 ¼ 0:249; n ¼ 251; F ð3; 247Þ ¼ 27:361; po0.0001.

ADAM S. MAIGA152

Table 6. Regression Results For Low And High Moral Equity Groups.

Regression Results for Low Moral Equity Regression Results for High Moral Equity

b S.E. t Significance b S.E. t Significance

Intercept 2.740 0.074 37.144 0.000 2.729 0.072 38.093 0.000

BP 0.419 0.068 6.131 0.000 �0.335 0.078 �4.279 0.000

R2 0.239 0.126

Adjusted-R2 0.232 0.119

F 37.586 (po0.000) 18.307 (po0.000)

N 122 129

Effect

ofManager’s

MoralEquity

153

A Further Analysis of the Interaction Effects

In order to facilitate understanding of the interaction effect, the interaction

term was investigated mathematically, as shown in Eqs. (2) and (3) below,

and then graphically presented in Fig. 2. The steps taken for this analysis

are: (a) taking the partial derivative of Eq. (1) over ME, (b) determining the

value of BP (the inflection point) at which this variable would have no

moderating effect on the relationship between ME and (ME�BP), and

(c) plotting the joint effect of the main and interaction terms (see Schoonh-

oven, 1981; Southwood, 1978).

The use of steps (a) and (b) above produced Eqs. (2) and (3) as

@PCS=@BP ¼ b2 þ b3ME (2)

and

ME ¼ �b2=b3 (3)

Substituting the values for b2 and b3 from Table 4 into Eq. (3) produced

ME ¼ 0:004 (the inflection point of the standardized equation). Again,

substituting the values for b2 and b3 from Table 5 and values for ME into

Eq. (2), the joint effect of the main and interaction terms was determined.

The results presented in Fig. 2 show the relationship between participa-

tion and budget slack due to the moderating effect of moral equity. For high

moral equity scores, i.e., moral equity values greater than 3.821 (3.817, the

mean+0.004, the standardized inflection point), budget participation is as-

sociated with low budget slack. For moral equity values o3.821, budget

participation is associated with high budget slack. For moral equity value

1 2 3 43.8214.968

6

7

4

3

∂B

-Sla

ck/∂

BP

Fig. 2. The Effect of Moral Equity on the Relationship between Budget

Participation and Propensity to Create Slack.

ADAM S. MAIGA154

C3.821, the association between budget participation and budget slack is

found to be neutral. Note that the value which is 3.821 for moral equity is

(a) within the range of values observed in this study and (b) close to the

mean value for moral equity (see Table 3 above). This value therefore rep-

resents the average moral equity in this study.

To investigate further the nature of the interaction, measures of moral

equity and budget participation are dichotomized (high, low) based upon

sample means. Table 7 indicates mean slack scores for the groups formed by

the process. A post ad hoc Scheffe test was carried out to compare the cell

means. The information reveals that when there is high budget participa-

tion, slack differs between levels of moral equity. Mean slack for the high

moral equity group, 2.434, is significantly less than mean slack for the low

group, 3.049 (p ¼ 0:001). Also, with low budget participation, slack differs

between the levels of moral equity. Mean slack for low moral equity, 2.329,

is significantly lower than mean slack for high group, 2.977 (p ¼ 0:000).However, there is no significant difference between low budget participa-

tion/high moral equity group and high budget participation/low moral eq-

uity group means (2.977 vs. 3.049; p ¼ 0:973), and low budget participation/

low moral equity group mean slack is not significantly different from that of

high budget participation/high moral equity (2.329 vs. 2.434, p ¼ 0:915).

CONCLUSION

This study extends our understanding of budgetary slack by investigating

the moderating effect of a manager’s moral equity on the relationship be-

tween budget participation and manager’s propensity to create slack. The

results suggest that the relationship between budgetary participation and

Table 7. Mean Scores for Propensity to Create Slack.

Low Budget Participation High Budget Participation

Low 2.329a 3.049

Moral 0.876b 0.827

Equity n ¼ 62 n ¼ 60

High 2.977 2.434

Moral 0.839 0.798

Equity n ¼ 60 n ¼ 69

aMean.bStandard deviation.


budgetary slack is moderated by moral equity. For managers with high

moral equity, the relationship between participation and slack is signif-

icantly negative while, for managers with low moral equity, the relationship

is significantly positive. However, as reported in Table 7, when budget par-

ticipation is low, managers with low moral equity create significantly less

slack than those with high moral equity. This analysis seems counterintu-

itive. Accordingly, different possible explanations are provided below:

(1) Given that slack creation was so low in the sample (overall mean for

slack of 2.685, which is well below the scale mid-point of 4) and the

means of the four cells in Table 7 were also below this mid-point, one

cannot rule out that the results are an artifact of this sample.

(2) As a construct, moral equity and slack may be seen from both a clinical

and an emotive sense. Also, slack creation may be situational and

therefore will affect the ethical decision-making process by altering in-

tention. That is, an individual may judge an action to be ethical, and yet

act in an unethical fashion due to a greater profit gained through the

unethical action (Hunt & Vitell 1986). Hence, managers’ questionnaire

responses may have been founded on an emotive platform and the re-

sults could be prejudiced.

(3) It may also be the case that cross-sectional survey methods are not

sufficient for the investigation of moral equity and propensity to create

slack, given the nature of the constructs, as well as the validity impli-

cations of examining relationships within a single industry, especially

when some non-response bias may still be present.

(4) Construction of the measures of budget participation, moral equity, and

propensity to create slack from the literature reviews and factor analyses

may have driven the results. In addition, it could be argued that the

findings obtained in this research are partly attributable to a poor

measure of moral equity and the dependent variable (propensity to cre-

ate slack).

(5) Also, as argued by Brownell (1982), the impact of budget participation

on budget slack might be contingent on other groups of variables such as

cultural, organizational, and interpersonal. Future research should focus

on those contingent variables in order to develop a comprehensive and

integrated model specifying the conditions under which budgetary par-

ticipation will produce favorable outcomes.

Future research could go well beyond the specific suggestions made here.

The treatment of ethical issue was limited to moral equity. The ethics lit-

erature suggests that a number of other ethical factors, such as relativism,

ADAM S. MAIGA156

idealism, and contractualism, may have significant effects on behavior. Al-

so, future research may need to investigate whether individuals have voice,

choice, or both in participative budgeting (Libby, 1999) and assess the im-

pact on budget slack within the ethical context. The implications of these

effects for budgeting remain to be investigated. Finally, field evidence of

these issues and well-designed experimental and case studies, and archival

tests are needed to distinguish among different explanations for observed

behavior to support the predictions being tested.

Despite these limitations, the current research is important. The current

study demonstrates that moral equity is an important variable in the re-

lationship between budget participation and budget slack. The findings,

based on the research question, suggest that organizational goals may take

precedence over self-interest for those managers with high moral equity.

Since this study is the first systematic investigation on the moderating effect

of moral equity on the relationship between budget participation and man-

ager’s propensity to create slack, it may enhance the richness of agency

research and provide additional insights to resolve some of the complex

relationships and issues in this traditional, but still important and

interesting, research area.

NOTES

1. Slack can also serve a positive purpose in the organization. For example, Cyertand March (1992) suggest that slack can be used to absorb fluctuations in an un-certain operating environment. Merchant (1989) suggests that superiors may allowslack in subordinates’ budgets to encourage coordination, motivation, and innova-tion. This contrasts with the agency theory perspective that budgetary slack is aninefficiency reflecting the effect of moving from an environment with perfect infor-mation to one with information asymmetry (e.g., Magee, 1980; Christensen, 1982;Baiman & Evans, 1983; Penno, 1984).2. Moral hazard is defined as an incentive to act in one’s self-interest in conflict

with the organization’s overall goals while being able to hide those actions throughprivately held information (Baiman, 1982).3. The term business unit is used to refer to a self-contained sub-unit (e.g., di-

vision) of a larger corporation.4. Research budget constraints did not allow larger sample size selection.5. The unusable returned questionnaires were either incomplete or did not meet

the two selection criteria.6. Because of contravening company policy, some preferred not to participate.7. Discriminant analysis was used to compare the early vs. late respondents

(Fowler, 1993). Results revealed that the two groups did not differ significantly ineither the level of the variables or in the relationship between the variables at the 0.05level. This suggests that non-response bias is not a problem.


8. The independent variables were transformed into new measurement variableswith a mean of 0 and standard deviation of 1.

ACKNOWLEDGMENTS

The author wishes to thank the editor, associate editor, and two anonymous

reviewers for the helpful comments. Additionally, the author acknowledges

the comments and suggestions provided by Fred A. Jacobs of Georgia State

University.

REFERENCES

Antle, R., & Eppen, G. D. (1985). Capital rationing and organizational slack in capital budg-

eting. Management Science, 31, 163–174.

Arrow, K. J. (1985). The economics of agency. In: J. W. Pratt & R. J. Zeckhauser (Eds),

Principals and agents: The structure of business (pp. 37–51). Boston, MA: Harvard

Business School Press.

Baiman, S. (1982). Agency research in managerial accounting: A survey. Journal of Accounting

Literature, (Spring), 154–213.

Baiman, S. (1990). Agency research in managerial accounting: A second look. Accounting,

Organizations and Society, 15(4), 341–371.

Baiman, S., & Evans, J. H., III (1983). Pre-decision information and participative management

control systems. Journal of Accounting Research, (Autumn), 21(2), 371–395.

Baiman, S., & Sivaramakrishnan, K. (1991). The value of private pre-decision information in a

principal-agent context. The Accounting Review, 66(4), 747–766.

Blanchard, G. A., & Chow, C. W. (1983). Allocating indirect costs for improved management

performance. Strategic Finance, 64(9), 38–41.

Bowie, N., & Duska, R. (1990). Business Ethics (2nd ed.). Englewood Cliffs, NJ: Prentice Hall.

Brownell, P. (1982). The role of accounting data in performance evaluation, budgetary par-

ticipation, and organizational effectiveness. Journal of Accounting Research, 20, 12–27.

Brownell, P., & Hirst, M. (1986). Reliance on accounting information, budgetary participation,

and task uncertainty. Journal of Accounting Research, 24(2), 241–249.

Cammann, C. (1976). Effects of the use of control systems. Accounting, Organizations and

Society, 1, 301–313.

Chow, C. W., Cooper, J. C., & Waller, W. S. (1988). Participative budgeting: Effects of a truth-

inducing pay scheme and information asymmetry on slack and performance. The Ac-

counting Review, 63, 111–122.

Christensen, J. (1982). The determination of performance standards and participation. Journal

of Accounting Research, 20, 589–603.

Collins, F. (1978). The interaction of budget characteristics and personality variables with

budgetary response attitudes. The Accounting Review, (April), 53(2), 324–335.

Coughlan, A. T., & Schmidt, R. M. (1985). Executive compensation, management turnover,

and firm performance: An empirical investigation. Journal of Accounting & Economics, 7,

43–66.

ADAM S. MAIGA158

Cronbach, L. J. (1987). Statistical tests for moderator variables: Flaws in analysis recently

proposed. Psychological Bulletin, 102(3), 414–417.

Cyert, F. L., & March, J. (1992). The Behavioral Theory of the Firm (2nd ed.). Cambridge, MA:

Blackwell Publishers.

DeGeorge, R. (1992). Agency theory and the ethics of agency. In: N. E. Bowie & R. E. Freeman

(Eds), Ethics and agency theory: An introduction (pp. 59–72). New York, NY: Oxford

University Press.

Douglas, P. C., Davidson, R. A., & Schwartz, B. N. (2001). The effect of organizational culture

and ethical orientation on accountants’ ethical judgments. Journal of Business Ethics,

34(2), 101–121.

Douglas, P. C., & Wier, B. (2000). Integrating ethical dimensions into a model of budgetary

slack creation. Journal of Business Ethics, 28, 267–277.

Dunk, A. (1993). The effect of budget emphasis and information asymmetry on the relation

between budgetary participation and slack. The Accounting Review, (April), 68(2),

400–410.

Evans, J. H., III, Hannan, R., Krishnan, R., & Moser, D. (2001). Honesty in managerial

reporting. The Accounting Review, 76, 537–559.

Flory, S., Phillips, T., Reidenbach, E., & Robin, D. (1992). A multidimensional analysis of

selected ethical issues in accounting. The Accounting Review, (April), 67(2), 284–302.

Forsyth, D. R. (1992). Judging the morality of business practices: The influence of personal

moral philosophies. Journal of Business Ethics, 11(5), 461–475.

Fowler, F. J., Jr. (1993). Survey Research Methods. Newbury Park, CA: Sage.

Gul, F. A. (1991). The effects of management accounting systems and environmental uncer-

tainty on small business manager’s performance. Accounting and Business Research, 22,

57–61.

Hughes, M. A., & Kwon, S. Y. (1990). An integrative framework for theory construction and

testing. Accounting, Organizations and Society, 15, 179–191.

Hunt, S. D., & Vitel, S. J. (1986). A general theory of marketing ethics. Journal of Macro-

marketing, 6(1), 5–16.

Jones, T. M. (1991). Ethical decision making by individuals in organizations: An issue-con-

tingent model. Academy of Management Review, 15, 366–392.

Kren, L. (1993). Control system effects on budget slack. Advances in Management Accounting,

2, 109–118.

Libby, T. (1999). The influence of voice and explanation on performance in a participative

budgeting setting. Accounting, Organizations and Society, 24(2), 125–137.

Luft, J. L. (1997). Fairness, ethics and the effect of management accounting on transaction

costs. Journal of Management Accounting Research, 9, 199–216.

Lukka, K. (1988). Budgetary biasing in organizations: Theoretical framework and empirical

evidence. Accounting, Organization and Society, 13, 281–301.

Magee, R. P. (1980). Equilibria in budget participation. Journal of Accounting Research, (Au-

tumn), 18(2), 551–573.

Merchant, K. A. (1981). The design of the corporate budgeting system: Influences on man-

agerial behavior and performance. The Accounting Review, 56, 813–829.

Merchant, K. A. (1985). Budgeting and the propensity to create budgetary slack. Accounting,

Organizations and Society, 10, 201–210.

Merchant, K. A. (1989). Rewarding Results: Motivating Profit Center Managers. Boston, MA:

Harvard Business School Press.


Merchant, K. A. (1995). Ethical Issues Related to ‘‘Results-Oriented’’ Management Control

Systems. In: Ethical issues in accounting: Proceedings of the professionalism and Ethics

seminar. Professionalism and Ethics Committee of the American Accounting Associa-

tion, St. Charles, IL, (June), 3(1), 2–4.

Milani, K. (1975). The relationship of participation in budget-setting to industrial supervisor

performance and attitudes: A field study. The Accounting Review, 50(2), 274–284.

Noreen, E. (1988). The economics of ethics: A new perspective on agency theory. Accounting,


Onsi, M. (1973). Factor analysis of behavioral variables affecting budgetary slack. The Ac-

counting Review, (July), 48(3), 535–548.

Penno, M. (1984). Asymmetry of pre-decision information and managerial accounting. Journal

of Accounting Research, 22, 177–191.

Pope, P. F. (1984). Information asymmetries in participative budgeting: A bargaining approach.

Journal of Business Finance & Accounting, 11, 41–59.

Reidenbach, R. E., & Ronald, D. P. (1991). An Application and extension of a multidimen-

sional ethics scale to selected marketing practices and marketing groups. Academy of

Marketing Science, 19(2), 83–92.

Robin, D. P., & Reidenbach, R. E. (1996). The perceived importance of an ethical issue as an

influence on the ethical decision-making of ad managers. Journal of Business Research,

5(1), 17–28.

Rutledge, R. W., & Karim, K. E. (1999). The influence of self-interest and ethical considerations

on manager’s evaluation judgments. Accounting, Organizations and Society, 24, 173–184.

Schoonhoven, C. B. (1981). Problems with contingency theory: Testing assumptions hidden

within the language of contingency theory. Administrative Science Quarterly, 26,

349–377.

Southwood, K. E. (1978). Substantive theory and statistical interaction: Five models. American

Journal of Sociology, 83, 1154–1203.

Stevens, D. E. (2002). The effects of reputation and ethics on budgetary slack. Journal of

Management Accounting Research, 14, 153–171.

Waller, W. S. (1988). Slack in participating budgeting: The joint effect of a truth-inducing pay

scheme and risk preferences. Accounting, Organizations and Society, 13, 87–98.

Young, S. M. (1985). Participative budgeting: The effects of risk aversion and asymmetric

information on budgetary slack. Journal of Accounting Research, (Autumn), 23(2),

829–842.

APPENDIX

PART I

SCENARIOS (Adopted from IMA Research Center)

The following four scenarios were used in this study. Each scenario ap-

peared on a separate page followed by brief instructions, a randomized

presentation of the scales, the univariate ethics measure, and the behavioral

intention measure as shown below.

ADAM S. MAIGA160

Scenario A. Tom Waterman is a young management accountant at a large,

diversified company. After some experience in accounting at headquarters,

he has been transferred to one of the company’s recently acquired divisions

run by its previous owner and president, Howard Heller. Howard has been

retained as vice president of this new division, and Tom is his accountant.

With a marketing background and a practice of calling his own shots, Ho-

ward seems to play by a different set of rules than those to which Tom is

accustomed. So far, it is working as earnings are up and sales projections are

high.

The main area of concern to Tom is Howard’s expense reports. Howard’s

boss, the division president, approves the expense reports without review,

and expects Tom to check the details and work out any discrepancies with

Howard. After a series of large and questionable expense reports, Tom

challenges Howard directly about charges to the company for typing that

Howard’s wife did at home. Although company policy prohibits such

charges. Howard’s boss again signed off on the expense. Tom feels uncom-

fortable with this and tells Howard that he is considering taking the matter

to the Board Audit Committee for review. Howard reacts sharply, remind-

ing Tom that ‘‘the Board will back me anyway’’ and that Tom’s position in

the company would be in jeopardy.

ACTION: Tom decides not to report the expense charge to the Audit

Committee. Please evaluate this action of Tom Waterman.

Fair___ ___ ___ ___ ___ ___ ___Unfair

Just___ ___ ___ ___ ___ ___ ___Unjust

Morally right___ ___ ___ ___ ___ ___ ___Not morally right

Acceptable to my family___ ___ ___ ___ ___ ___ ___Unacceptable to my

family

If you were responsible for making the decision described in the scenario,

how would you judge the decision?

Ethical ___ ___ ___ ___ ___ ___ ___ Unethical

Scenario B. Anne Devereaux, company controller, is told by the chief fi-

nancial officer that in an executive committee meeting the CEO told them

that the company ‘‘has to meet its earnings forecast, is in need of working

capital and that’s final.’’ Unfortunately, Anne does not see how additional

working capital can be raised even through increased borrowing, since


income is well below the forecast sent to the bank. Seth suggests that Anne

review bad debt expense for possible reduction and holding sales open

longer at the end of the month. He also brushes off the management letter

request from the outside auditors to write down the spare parts inventory to

reflect its ‘‘true value.’’

At home on the weekend, Anne discusses the situation with her husband,

Larry, a senior manager of another company in town. ‘‘They’re asking me

to manipulate the books,’’ she says. ‘‘On the one hand,’’ she complains, ‘‘I’m

supposed to be the conscience of the company and on the other, I’m sup-

posed to be absolutely loyal.’’ Larry tells her that companies do this all the

time, and when business picks up again she’ll be covered. He reminds her

how important her salary is to help maintain their comfortable lifestyle, and

that she should not do anything drastic that might cause her to lose her job.

ACTION: Anne decides to go along with the suggestions proposed by her

boss. Please evaluate this action of Anne Devereaux.

Fair___ ___ ___ ___ ___ ___ ___Unfair

Just___ ___ ___ ___ ___ ___ ___Unjust



family




Scenario C. Drew Isler, the plant’s chief accountant, is having a friendly

conversation with Leo Sullivan, operations manager and old college buddy,

and Fred LaPlante, the sales manager. Leo tells Drew that the plant needs a

new computer system to increase operating efficiency. Fred interjects that

with the increased efficiency and decreased late deliveries their plant will be

the top plant next year.

However, Leo wants to bypass the company policy which requires that

items greater than $5,000 receive prior Board approval and be capitalized.

Leo would prefer to generate purchase orders for each component part of

the system, each being under the $5,000 limit, and thereby avoid the ap-

proval ‘‘hassle.’’ Drew knows this is clearly wrong from a company and an

ADAM S. MAIGA162

accounting standpoint, and he says so. Nevertheless, he eventually says that

he will go along.

Six months later, the new computer system has not lived up to its ex-

pectations. Drew indicates to Fred that he is really worried about the prob-

lems with the computer, and the auditors will disclose how the purchase was

handled in the upcoming visit. Fred acknowledges the situation by saying

that production and sales are down and his sales representatives are also

upset. Leo wants to correct the problems by upgrading the system (and

increasing the expenses), and urges Drew to ‘‘hang in there.’’

ACTION: Feeling certain that the system will fail without the upgrade,

Drew agrees to approve the additional expense. Please evaluate this action

of Drew Isler.

Fair___ ___ ___ ___ ___ ___ ___Unfair

Just___ ___ ___ ___ ___ ___ ___Unjust



family




Scenario D. Paul Tate is the assistant controller at Stern Electronics, a me-

dium-sized manufacturer of electrical equipment. Paul is in his late 50’s and

plans to retire soon. His daughter has been accepted into medical school,

and financial concerns are weighing heavily on his mind. Paul’s boss is out

of the office recuperating from health problems, and in his absence Paul is

making all decisions for the department.

Paul receives a phone call from an old friend requesting a sizable amount

of equipment on credit for his new business. Paul is sympathetic but cog-

nizant of the risk of extending credit to a new company, especially under

Stern’s strict credit policy for such transactions. When Paul mentions this

conversation to Warren, the general manager, he is immediately interested.

Warren notes that the company needs an additional $250,000 in sales to

meet the quarterly budget and, thus, ensure bonuses for management,

including Paul.


ACTION: Paul decides to make the sale to his friend’s new business.

Please evaluate this action of Paul Tate.

Fair___ ___ ___ ___ ___ ___ ___Unfair

Just___ ___ ___ ___ ___ ___ ___Unjust



family




PART II

1. Is your division a profit center? ____Yes ______ No

2. Do you have a budget responsibility in your division? ______ Yes _____

No

PART III

If your answer to both 1 and 2 in Part IV is yes, please answer the remaining

parts of the questionnaire, otherwise stop at Part IV and return the ques-

tionnaire.

Participation

(response anchors: 1 ¼ stronglydisagree; 2 ¼ moderatelydisagree; 3 ¼

mildly disagree, 4 ¼ neutral; 5 ¼ mildly agree, 6 ¼ moderately agree, 7 ¼

strongly agree)

1. I am involved in setting all of my budget.

2. My superior clearly explains budget revisions.

3. I have frequent budget-related discussions with my superior.

ADAM S. MAIGA164

4. I have a great deal of influence on my final budget.

5. My contribution to the budget is very important.

6. My superior initiates frequent budget discussions when the budget is

being prepared.

Budget Slack

(response anchors: 1 ¼ strongly disagree, 2 ¼ moderately disagree, 3 ¼

mildly disagree, 4 ¼ neutral; 5 ¼ mildly agree, 6 ¼ moderately agree, 7 ¼

strongly agree)

1. To protect himself, a manager submits a budget that can safely be at-

tained.

2. In good business times, your superior is willing to accept a reasonable

level of slack in the budget.

3. Slack in the budget is good to do things that cannot be officially ap-

proved.

PART IV

Please answer the following:

1. What is the number of employees is your company? ___________

2. What is your approximate dollar volume of sales? _____________

3. Number of years is this position? ___________

4. Number of years in management? __________



166

ASYMMETRIC EFFECTS OF

ACTIVITY-BASED COSTING

SYSTEM COST REALLOCATION

M. G. Fennema, Jay S. Rich and Kip Krumwiede

ABSTRACT

Despite the many proposed benefits of activity-based costing (ABC),

many managers oppose implementing it. One important reason for this

resistance that has generally not been addressed in the literature is the

effect of cost reallocations on managers’ evaluations and compensation.

This study examines how the impact of installing an ABC system on

managers’ bonuses affects their support for ABC implementation. Be-

cause ABC systems usually allocate costs in different proportions than

traditional systems, some products may appear to be more profitable

while others may appear less so. This, in turn, causes the business units

responsible for the products to appear to be more or less profitable.

Based on prospect theory (Kahneman & Tversky (1979). Economet-

rica, 47, 263–291; Tversky & Kahneman (1992). Journal of Risk and

Uncertainty, 5, 297–323), we predict that the negative effect on man-

agers’ support for ABC in business units reporting less profit is greater

than the positive effect on managers’ support in the more profitable units.

The results of an experiment support this prediction. Since management

support is critical to successful system implementation, this asymmetric

effect has implications for cost management system changes.





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08007-X

167

INTRODUCTION

Increased global competition has motivated organizations to improve their

products and customer service. Such improvements typically include the

development of superior information systems that allow managers to track

customers, increase production efficiencies, and manage costs. One such

information system involves the calculation of product costs. Having ac-

curate product costs is critical in deciding whether the product can compete

against other companies’ products and also in the pricing of the item.

Determining product costs, however, is generally difficult to do. While

some costs are easily traced to specific products, many costs are expended to

produce a wide range of products. Assigning these ‘‘indirect’’ costs to in-

dividual products has always been difficult but has become increasingly so

as more automation is introduced into the production process. This auto-

mation has led to a higher percentage of indirect costs and therefore less

accurate product costs.

A highly publicized innovation developed over the last two decades to

produce better product cost information is activity-based costing (ABC).

When properly implemented and maintained, ABC systems assign indirect

costs based on resource consumption and generally provide more accurate

product costs than traditional cost allocation systems. However, replacing

old product cost information systems with newer ABC systems is not a

simple matter. ABC systems are generally complex and pose significant

implementation challenges (Kennedy & Affleck-Graves, 2001). Thus, in

spite of many proposed benefits of ABC, the majority of firms have not

implemented it (Kennedy & Affleck-Graves, 2001). Gosselin (1997) referred

to this as the ‘‘ABC paradox.’’

Because the information generated by the system affects many managers

in an organization, significant resistance to successful implementation can

arise. For example, Argyris and Kaplan (1994) describe situations in which

individuals engage defensive routines to block implementation when the

information from an ABC system will be highly embarrassing or potentially

threatening to them. An inability to overcome opposition can result in costly

failure of the implementation effort. One important reason for this resist-

ance that has generally not been addressed in the literature is how the

impact of cost reallocations on managers’ evaluations and compensation

affects their support for the new cost system. Because an ABC system re-

allocates indirect costs, some products may appear more profitable and

others may appear less so. The managers responsible for those products will

likewise be affected, since their divisions will appear to be more or less

M.G. FENNEMA ET AL.168

profitable. Thus, their support for the adoption of the new system may be

related to the profit effect on their division.

The major premise of this study is that although losses in profitability for

some products will be more or less mirrored by increases in profitability in

others, the reactions to prospective gains and losses by the managers re-

sponsible for the affected products will not be identical. Specifically, we

predict that the negative effect on managers’ support for ABC in business

units that will become less profitable is greater than the positive effect on

managers’ support in units that become more profitable. This prediction is

derived from the asymmetric value function in prospect theory (Kahneman

& Tversky, 1979; Tversky & Kahneman, 1992). The following sections de-

scribe the process of product costing, the proposed effect of ABC system

implementation on manager resistance, and the results of an experiment that

support this prediction.

BACKGROUND

Product Costing

The determination of a product’s cost requires an analysis of the resources

expended to create the product. Such costs are categorized as either direct or

indirect costs. Direct costs are those that are traceable to the final product

while indirect costs are those that are not. For example, the production of a

wooden desk involves several direct material costs including a certain

amount of wood and a set number of pieces of hardware. Determining the

cost of these inputs is relatively easy. In addition, human labor to cut the

wood or assemble the desk is also considered a direct cost since a standard

amount of time could be established for these operations. The calculation of

indirect costs is much more difficult than that of direct costs. In the desk

example, indirect costs might include items such as the depreciation (or rent)

expense for cutting equipment. Also included might be expenditures for

property taxes on the factory building and the salaries of factory supervi-

sors. Although all of these costs are incurred to produce the desks, they are

not traceable to specific desks. Therefore, these costs are more difficult to

assign to each desk.

Traditional cost accounting systems generally pool all indirect costs to-

gether and then spread them to products based on some volume-based

measure such as direct labor hours. For the desk example, the estimated

total annual amount of factory indirect costs might be divided by the

Asymmetric Effects of Activity-Based Costing System Cost Reallocation 169

estimated total annual direct labor hours to get a predetermined overhead

rate. This rate would then be applied to each desk based on the number of

direct labor hours required to produce it. This type of cost allocation system

yields reasonably accurate product costs as long as all the products made in

the factory consume similar percentages of indirect costs. Of course, if the

desk factory only made one type and size of desk, the traditional system

would be highly accurate.

However, if a factory makes diverse products, the traditional system is not

likely to provide accurate product costs. If the plant that makes desks also

makes tables, it is unlikely that a traditional system will allocate indirect

costs in the proportion that resources are consumed. For example, if tables

require less cutting of wood than do desks, the traditional system would

assign less indirect costs to each table based simply on the number of direct

labor hours consumed by each product. Even if the tables require more

setup and material handling costs than desks, they would still be allocated

lower overhead costs. The result would be distorted product costs for both

products. Such distortions caused by the limitations of traditional volume-

based allocation methods have been widely discussed in the literature (see

Kaplan & Cooper, 1998, Chapters 3 and 6, or Cokins, 2001, Chapter 1, for

overviews).

The goal of a cost system is to assign indirect costs to products in the

proportion that they are consumed by the product. Since traditional systems

often fail to achieve this goal, many companies have attempted to imple-

ment ABC. However, implementing ABC can be quite difficult. The next

section discusses ABC system implementation and the problems that can

arise.

ABC System Implementation and Manager Resistance

Organizational change has been the subject of much study (e.g., Goodman

& Dean, 1982; Beer, Eisenstat, & Spector, 1990; Huff, Huff, & Barr, 2001).

One example of organizational change is the implementation of a new

management accounting information system. Probably the most widely dis-

cussed innovation of the past two decades in management accounting in-

formation systems is ABC.

The promise of ABC is quite inviting. By focusing on the activities that a

product requires, ABC can provide a more precise allocation of indirect

costs, resulting in more accurate product costs. This increased accuracy

enables firms to make better critical decisions such as pricing and market


entry and exit decisions. Cooper and Kaplan (1991) also identify benefits

associated with lean production, such as reducing inventories, increasing

common components, increasing quality by minimizing total quality costs,

and assessing customer profitability. Ness and Cucuzza (1995) suggest that

once ABC is part of a firm’s critical management systems it becomes a

powerful tool, supporting continuous rethinking and improving not only

products and services but also process and market strategies. Ittner, Lanen,

and Larcker (2002) find some evidence that using ABC enhances manufac-

turing performance and some weak evidence that profits are enhanced when

ABC matches a plant’s operational characteristics. Kennedy and Affleck-

Graves (2001) find that firms adopting ABC outperform non-ABC firms by

up to 27% in both higher stock returns and accounting measures such as

return on equity and profit margin.

Although the potential benefits are significant, the introduction of any

new technology can be costly. ABC systems require significant investment in

model design and data-gathering costs. But since these systems generate

information that affects many individuals within a firm, behavioral and

organizational factors are of greater concern. In their models of cost man-

agement system implementation, Shields and Young (1989, 1995) identify

the linkage between the cost management system and performance evalu-

ation and compensation as one of seven behavioral and organizational var-

iables that are required for successful system implementation. Indeed, in an

empirical study of 143 firms, Shields (1995) finds this linkage to be empir-

ically correlated with success. Cokins (2001) notes that individuals involved

in successful ABC implementations spend 90% of their efforts on organ-

izational behavior issues and only 10% on technical issues. Unfortunately,

most organizations do just the opposite. The behavioral costs take the form

of barriers to implementation erected by individuals who feel that they may

be negatively affected by the new system.

Previous studies have addressed many reasons why managers might op-

pose the implementation of ABC. Cooper, Kaplan, Maisel, Morrissey, and

Oehm (1992) describe the process eight companies went through in trying to

implement ABC. Two of the company examples address behavioral issues

relating to the ABC implementation process. In Farrall, Inc. (Chapter 5),

representatives from the sales department committed only 5% of their time

to ABC implementation, compared with 10–100% committed from repre-

sentatives in accounting and information system departments. Sales and

marketing employees are often hesitant to help with ABC implementation

because they either do not value ABC or worry about what the results will

show. Williams Brothers Metals (Chapter 6) discusses the negative reaction


of product-line managers whose product margins are hurt by the ABC

analysis. One manager, whose two largest products now appeared unprof-

itable, believed the ABC information was ‘‘fundamentally flawed.’’ He

questioned the assumptions that the ABC model was built on and even

claimed they are ‘‘completely wrong.’’ Cooper et al. (1992) also found that

most firms involved with implementing ABC systems did not create the

critical link from the system to performance evaluation and compensation.

Argyris and Kaplan (1994) explore such barriers, identifying potential

embarrassing or threatening situations that can arise from an ABC system.

Examples include the identification of non-value-added activities (e.g., re-

work) or finding other production inefficiencies that had been hidden by the

old system. Faced with this situation, not only will affected managers not

support the ABC system, they will also respond defensively to block its

implementation. With barriers in place by affected stakeholders, successful

implementation is unlikely.

Indeed, successful implementation seems to be elusive in practice. Ittner et

al. (2002) find only 26% of U.S. manufacturing plants use ABC extensively.

Kennedy and Affleck-Graves (2001) find 20.1% of the top 1,000 firms in the

U.K. (ranked by sales) to be using ABC. Ness and Cucuzza (1995) observe

from their consulting experiences that employee resistance is the single big-

gest obstacle to successful ABC implementation. They find managers are

commonly nervous about revealing detailed information that could be used

to make them look bad or dramatically change the definitions of success and

failure. Nevertheless, the authors recommend that measurement and incen-

tive systems be tied to the ABC numbers.

Academic research relating to ABC implementation has likened it to an

innovation in a firm’s cost information system (Gosselin, 1997; Shields,

1995; Anderson, 1995). Studies have explained implementation of informa-

tion systems using a ‘‘process stage’’ model (Cooper & Zmud, 1990; Kwon

& Zmud, 1987). Anderson (1995) suggests integrating the system with man-

agement information and reward systems is critical to reaching the accept-

ance and usage stages. She states, ‘‘Implementation of ABC systems is

unlikely to succeed until these stakeholders’ concerns are addressed’’ (p. 9).

Krumwiede (1998) finds top management and nonaccounting support to be

critical to reaching the highest stages of implementation success (i.e., routine

and integrated).

Prior behavioral research on the resistance to ABC implementation has

focused on various organizational and individual factors but not on em-

ployee resistance to ABC implementation. This study uses an experimental

approach to explore the effect of changes in performance evaluation and


compensation resulting from ABC adoption on manager resistance. Clearly,

changes in compensation caused by a new ABC system can result in a

threatening situation for affected managers. However, while some studies

(e.g., see Mak & Roush, 1994) have examined the issue of performance

evaluation in an ABC environment, no study has investigated the potential

effect on managers’ support when compensation is based upon division per-

formance. In this situation, indirect costs will be reallocated, causing some

divisions to appear more profitable while others appear to be less profitable.

Consider the following example. A company has two divisions, one that

produces product A and another that makes an equal amount of product B.

Managers’ bonuses are based on a percentage of division profitability. The

current product costing system calculates the cost of A to be $5 per unit and

that of B to be $11 per unit. An ABC system calculates the unit costs to be

$7 for A and $9 for B. The same total costs exist, but they are applied to the

products in different ways. Since Division A would appear less profitable

and Division B would appear more so, a critical question arises: What effect

would such a shift of costs have on the two division managers’ support for

the new ABC system?

THEORY DEVELOPMENT AND HYPOTHESIS

Since the allocation of indirect costs to products is typically a zero-sum

game (i.e., any amount taken from one product is allocated to another), it

might be assumed that the effect on the support of managers responsible for

those products would be opposite but equal since the effect on their bonuses

would be opposite but equal. Such would be the predictions from standard

economic theory. But that theory has not been found to be particularly

descriptive of actual behavior in many situations (see Kahneman & Tversky,

1979). Prospect theory (Kahneman & Tversky, 1979; Tversky & Kahneman,

1992) has been proposed and widely tested as an alternative to standard

economic theory. This theory suggests asymmetrical reactions to expected

gains and losses. It has been applied in recent accounting studies relating to

budgetary slack (Lau & Eggleton, 2003), risk-taking tendencies in capital

budgeting decisions (Moreno, Kida, & Smith, 2002), framing effects in a

classic Asian disease-type business scenario (Chang, Yen, & Duh, 2002),

and loss aversion by traders in financial markets (Willman & Fenton-

O’Creevy, 2002).

There are two characteristics of prospect theory that are important to the

current situation (refer to Fig. 1 for an example of a prospect theory value


function). First, gains and losses are evaluated from some reference point. In

the above example, that reference point is most likely the current product

cost in each division (and thus the current bonus based on those product

costs earned by the managers). With the new ABC system, the manager of

the division producing product A would suffer a $2 per unit loss from the $5

reference point, and the manager of the division producing product B would

enjoy a $2 gain from the $11 reference point.

A second characteristic of prospect theory is a value function that is steeper

for losses than for gains. If the product cost changes caused the manager

of Division A to be penalized by the loss of $1,000 bonus and the manager of

Division B to be rewarded with an additional $1,000 bonus, the value func-

tion in Fig. 1 posits that the increase in value experienced by the Division B

manager would be significantly less than the decrease in value experienced by

the manager of Division A. Thus, the negative support from the first manager

would exceed the positive support of the second one. This leads to the fol-

lowing hypothesis:

H1. The loss in support from managers whose products increase in cost

due to the installation of an ABC system will exceed the increase in

support from managers whose products decrease in cost.

Gains Losses

Value

$1,000$1,000

Fig. 1. Prospect Theory Value Function.


The following experiment was designed to test this hypothesis using two

groups of participants. The first group included upper level college students

who have been trained to understand how ABC systems can improve man-

agerial decision making. The second involved practicing Certified Public

Accountants (CPAs). Since prior experience in a broad domain often results

in more defined knowledge structures about the decision process (Vera-

Munoz, Kinney, & Bonner, 2001), the experienced CPA participants were

chosen to validate the results of the student group.

METHODOLOGY

Participants/Task

Ninety-one university business students (junior and senior level accounting

majors) and 77 CPAs participated in a two-part exercise. The students were

enrolled in an accounting class where traditional and ABC product costing

systems had been taught. The two parts of their exercise were conducted two

days apart. The CPAs had an average of 14.8 years of professional expe-

rience and their data were collected during a one-day continuing profes-

sional education class, with the first part of the exercise completed in the

morning session and second part in the afternoon session.

In the first part of the exercise, participants were asked to examine a

company’s unit costs for a product under a current traditional costing sys-

tem and under a proposed ABC system. They were then asked to assume the

role of the chief financial officer (CFO) of the company and answer three

questions designed to measure the amount of their support for the proposed

system. The primary measure was the participant’s support for the new

system, on a 7-point Likert-type scale. The scale had verbal descriptions of

‘‘Would Strongly Support’’ and ‘‘Would Strongly Object to’’ at the end-

points and ‘‘Indifferent’’ at the midpoint. Participants were also asked to

assess the maximum the company should pay to install the new ABC system,

and to estimate the percentage of the cost of the new system that should be

allocated to each of two divisions. A copy of the instrument is included in

Appendix A.

In the second part (see Appendix B), participants were asked to play

the role of the manager of one of two divisions, either the Standard Division

or the Deluxe Division. In this part, the participants were shown the effect

of the proposed ABC system on the profitability of the divisions, and

therefore how it would affect their bonus. This effect was symmetric, with


the managers of the Standard Division receiving an extra $7,200 bonus and

the managers of the Deluxe Division losing the same amount. Participants

were then asked to answer the same three questions again.

Design

After assuming the CFO role in Part 1 of the experiment, participants

were randomly assigned to one of the between-participant conditions of

either the manager of the Standard Division or the Deluxe Division for Part

2. The dependent variables are the absolute value of the change (from the

CFO role to the division manager role) in the three variables that were used

to measure support for the new system.

RESULTS AND DISCUSSION

Tables 1 and 2 contain the descriptive results for the student participants

and CPA participants, respectively. All changes in support were calculated

and analyzed at the individual level. Panel A in each table contains the

results of the first part of the experiment for each group of participants.

Recall that in the first part, the participants were all assigned to the role of

the CFO of the firm. As such, it would be expected that there would be no

differences between the responses of those who would be in the two manager

roles. This is the case for both the students and the CPAs. It is interesting to

note that although these business students were taught about the benefits of

ABC systems, there was not overwhelming overall support for the proposed

system (i.e., 3.4 is nearly exactly in the middle of the 7-point scale). Also

note that for the CPAs the amounts to be allocated to the two divisions do

not add up to 100% because three participants allocated zero to both of the

divisions and one participant’s allocations did not add to 100%.

In Part 2 of the experiment, participants were assigned to one of the two

manager roles, and Panel B in Tables 1 and 2 contains the responses from

that part of the experiment. Panel C shows the increase or decrease in

support from that of the CFO role to that of the two division manager roles.

As expected, overall support for the system (as measured by the 7-point

scale), increased for the Standard Division managers and decreased for

those in the Deluxe Division. Similarly, the amount to be paid for the system

increased for those in the Standard Division and decreased for managers of

the Deluxe Division. Finally, the amount of the system’s cost to be allocated


to the manager’s division decreased for both divisions. Even though the

managers of the Standard Division would benefit significantly from the new

system, they were unwilling to accept more of the system’s cost than when

they were in the CFO role.

Tables 3 and 4 show the results of an ANOVA examining the differences

in responses over the two parts of the experiment, for the student group and

the CPA group respectively. Results for the primary measure of support

(support for the ABC system) are shown in panel A. For both students and

CPAs, there is a significant effect for the differential support of the new

ABC system. On the 7-point scale (1 ¼ Strongly Support, 7 ¼ Strongly

Object To), student participants in the Standard Division increased their

support by 1.1 points while those in the Deluxe Division reduced their

support by 2.0 points, or almost double that of the Standard Division. For

CPAs, Standard Division support increased by 1.9 points while Deluxe Di-

vision support decreased by 2.9 points. The difference in the absolute value

Table 1. Descriptive Results: Student Participants.

Standard Deluxe

Division Division

Role Role

(n ¼ 44) (n ¼ 47)

Panel A: Mean Responses – CFO Role (Part 1)a

Support for ABC systemb 3.4 3.4

Amount to be paid for ABC system ($) 322,501 314,161

Percent allocated to standard division 48 52

Percent allocated to deluxe division 52 48

Panel B: Mean Responses by Role (Part 2)



Percent allocated your division 42 34

Panel C: Increase (Decrease) in Mean Responses from CFO Role to Manager Role

Support for ABC systemb 1.1 (2.0)

Amount to be paid for ABC system ($) 7,476 (86,936)

Percent allocated to your division (5) (14)

aIn Part 1 of the experiment the participants were unaware of the role they would be asked to

assume in Part 2. Part 1 responses are presented consistent with their Part 2 roles to better

demonstrate how their judgments differed between the two parts.b1 ¼ Strongly support, 7 ¼ Strongly object to.


of these changes is significant for both the students (p ¼ 0:0088) and CPAs

(p ¼ 0:0337). This supports the hypothesis that the decrease in support from

those negatively affected by the ABC system will exceed the increase in

support from those positively affected by it.

In a secondary measure of support, the amount to be paid for the new

ABC system, results were mixed. Student participants in the Standard Di-

vision (whose costs went down and bonus went up) were willing to pay an

average of $7,476 more than when they were in the CFO role. However,

Deluxe Division managers were willing to pay $86,936 less than when they

were in the CFO role. The difference between the absolute value of these two

amounts is marginally significant at p ¼ 0:0578 (see panel B of Table 3). For

CPAs, the change in the amount to be paid for the new system did not differ.

Those in the Standard Division were willing to pay $157,263 more than

when they were in the CFO role while those in the Deluxe Division were

willing to pay $152,500 less. Panel B of Table 4 shows that the absolute

difference between these two responses is not significant.

Table 2. Descriptive Results: CPA Participants.

Standard Deluxe

Division Division

Role Role

(n ¼ 35) (n ¼ 42)

Panel A: Mean Responses – CFO Role (Part 1)a



Percent allocated to standard division 59 55

Percent allocated to deluxe division 34 43

Panel B: Mean Responses by Role (Part 2)



Percent allocated your division 51 21

Panel C: Increase (Decrease) in Mean Responses from CFO Role to Manager Role

Support for ABC systemb 1.9 (2.9)

Amount to be paid for ABC system ($) 157,263 (152,500)

Percent allocated to your division (8) (21)

aIn Part 1 of the experiment the participants were unaware of the role they would be asked to

assume in Part 2. Part 1 responses are presented consistent with their Part 2 roles to better

demonstrate how their judgments differed between the two parts.b1 ¼ Strongly support, 7 ¼ Strongly object to.


Panel C of Tables 3 and 4 contains the results of the other secondary

measure of support, the percent of cost managers would allocate to their

divisions. For both groups of subjects, both division managers wanted to bear

a smaller percent of the cost of the system (than they had allocated when they

Table 3. ANOVA Results of Hypothesis Test: Student Participants.

Source DF Sum of Squares Mean Square F Value p4F

Panel A: Support for ABC System

Manager role 1 22.49 22.49 7.17 0.0088

Error 89 278.99 3.13

Corrected total 90 301.48

Panel B: Amount to be Paid for ABC System

Manager role 1 143,482,968,757 143,482,968,757 3.69 0.0578

Error 89 3,456,371,618,501 38,835,636,163

Corrected total 90 3,599,854,587,257

Panel C: Percent Allocated to Your Division

Manager role 1 8,522 8,522 13.26 0.0005

Error 89 57,183 642

Corrected total 90 65,705

Table 4. ANOVA Results of Hypothesis Test: CPA Participants.

Source DF Sum of Squares Mean Square F Value p4F

Panel A: Support for ABC System

Manager role 1 18.69 18.69 4.68 0.0337

Error 75 200.45 3.99

Corrected total 76 318.14

Panel B: Amount to be Paid for ABC System

Manager role 1 433,073,610 433,073,610 0.02 0.9024

Error 75 2,143,359,521,714 28,578,126,956

Corrected total 76 2,143,792,595,325

Panel C: Percent Allocated to Your Division

Manager role 1 16,102 16,102 18.58 0.0001

Error 75 65,000 866

Corrected total 76 81,102


were in the CFO role), but Deluxe Division managers were even more un-

willing to bear the costs (p ¼ 0:0005 for students, p ¼ 0:0001 for CPAs).

These results also support the prospect theory-based hypothesis in this study.

CONCLUSIONS AND LIMITATIONS

This study documents an asymmetric effect of reallocating product costs on

managers who have an interest in ABC system information. Specifically,

these results provide evidence that the loss in support from managers whose

products increase in cost exceeds the increase in support from those whose

products decrease in cost. This was true over three possible measures of

support for the student participants and two for the CPA participants. This

simple but powerful finding has consequences for both theoretical models of

ABC implementation and practical applications.

Many prior studies have identified various inhibitors to ABC implemen-

tation success. In their models of cost management system implementation,

Shields and Young (1989, 1995) identify the link between the cost manage-

ment system and performance evaluation and compensation as one of seven

behavioral and organizational variables that are required for successful

system implementation. As noted earlier, Shields (1995) found this com-

pensation linkage to be correlated with success. Unfortunately, previous

studies (e.g., Cooper et al., 1992) have found that most firms involved with

implementing ABC systems do not create the critical link from the system to

performance evaluation and compensation. The current study underscores

the importance of this connection, and it also adds the finding that man-

agers will be asymmetrically affected by equal changes in that compensa-

tion. Identifying this effect helps develop a more complete theory of

individuals’ resistance to ABC implementation.

On a practical level, those attempting to install a new innovation such as

ABC need to anticipate individual resistance. Leonard-Barton (1987) likens

system implementation to an internal marketing campaign. By building on

system positives and countering negative ones, implementation managers can

speed the system change effort. Since cost reallocations can cause significant

loss of support from negatively affected managers, steps must be taken to

reduce that effect. Such measures might include temporary compensation

adjustments or permanent changes in the performance-based compensation

system, such as increased use of nonfinancial measures in bonus contracts

(Ittner, Larcker, & Rajan, 1997). In any case, the effect of such actions


should be well thought out to take into account the varying effects on the

managers.

As with all experimental research, the findings of this study need to be

considered in light of its limitations. The student participants in this study

lack experience in making decisions about the value of cost management

systems. However, the CPA group had much more experience and helps to

validate the results of the less experienced group. The design of this study

has limitations as well. The various pressures and incentives that would be

present for real-world decision makers are not present in our experiment. As

discussed in Lipe and Salterio (2000), experimental studies attempting to

model real-world phenomenon require simplifying assumptions that impair

realism. For example, the participants in this study took on both CFO and

Division Manager roles. In actual practice, these decision makers would

have very different backgrounds and priorities. Also, there may be other

company-wide incentives (e.g., stock plan) that affect a division manager’s

decision making. Another limitation of this study is that the division man-

agers may have various products, some of which show higher costs and

some showing lower costs using ABC. Finally, this study focuses on the

economic and behavioral impact of ABC on the division manager. It ignores

whether or not ABC is actually beneficial for the company as a whole.

This study adds to the limited body of research on manager resistance to

activity-based costing. Future studies might add to our knowledge by ad-

dressing ways to mitigate the asymmetric effect on managers’ support. Dif-

ferent compensation schemes might be examined in order to determine what

measures can be taken to increase support at the lowest cost. The goal of

these studies should be to increase our understanding of how managers

perceive changes in their environment. In this way, more accurate costing

can be achieved with increased support from affected managers.

REFERENCES

Anderson, S. W. (1995). A framework for assessing cost management system changes: The case

of activity based costing implementation at General Motors, 1986–1993. Journal of

Management Accounting Research, 7, 1–51.

Argyris, C., & Kaplan, R. (1994). Implementing new knowledge: The case of activity based

costing. Accounting Horizons, 8, 83–105.

Beer, M., Eisenstat, R., & Spector, B. (1990). Why change programs don’t produce change.

Harvard Business Review, 68, 158–166.

Chang, C. J., Yen, S., & Duh, R. (2002). An empirical examination of competing theories to

explain the framing effect in accounting-related decisions. Behavioral Research in Ac-

counting, 14, 35–64.


Cokins, G. (2001). Activity-based cost management. New York, NY: Wiley.

Cooper, R., & Kaplan, R. (1991). The design of cost management systems. Englewood Cliffs, NJ:

Prentice-Hall.

Cooper, R., Kaplan, R., Maisel, L., Morrissey, E., & Oehm, R. (1992). Implementing activity-

based cost management. Montvale, NJ: Institute of Management Accountants.

Cooper, R. B., & Zmud, R. W. (1990). Information technology implementation research: A

technological diffusion approach. Management Science, 36(2), 123–139.

Goodman, P., & Dean, J. (1982). Creating long-term organizational change. In: P. S. Goodman

and Associates (Ed.), Changes in organizations: New perspectives on theory, research, and

practice. San Francisco, CA: Jossey-Bass.

Gosselin, M. (1997). The effect of strategy and organizational structure on the adoption and

implementation of activity-based costing. Accounting, Organizations and Society, 22(2),

105–122.

Huff, A. S., Huff, J. O., & Barr, P. S. (2001). When firms change direction. New York, NY:

Oxford University Press.

Ittner, C. D., Lanen, W. N., & Larcker, D. F. (2002). The association between activity-based

costing and manufacturing performance. Journal of Accounting Research, 40(3), 711–726.

Ittner, C. D., Larcker, D. F., & Rajan, M. V. (1997). The choice of performance measures in

annual bonus contracts. The Accounting Review, 72(2), 231–255.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk.

Econometrica, 47, 263–291.

Kaplan, R. S., & Cooper, R. (1998). Cost & effect. Boston, MA: Harvard Business School Press.

Kennedy, T., & Affleck-Graves, J. (2001). The impact of activity-based costing techniques on

firm performance. Journal of Management Accounting Research, 13, 19–45.

Krumwiede, K. R. (1998). The implementation stages of activity-based costing and the impact

of contextual and organizational factors. Journal of Management Accounting Research,

10, 239–277.

Kwon, T. H., & Zmud, R. W. (1987). Unifying the fragmented models of information systems

implementation. In: R. J. Boland & R. Hirscheim (Eds), Critical issues in information

systems research. New York, NY: Wiley.

Lau, C. M., & Eggleton, I. R. C. (2003). The influence of information asymmetry and budget

emphasis on the relationship between participation and slack. Accounting & Business

Research, 33(2), 91–104.

Leonard-Barton, D. (1987). Implementing structured software methodologies: A case of inno-

vation in process technology. Interfaces, 17, 6–17.

Lipe, M. G., & Salterio, S. E. (2000). The balanced scorecard: Judgmental effects of common

and unique performance measures. The Accounting Review, 75(3), 283–298.

Mak, Y., & Roush, M. (1994). Flexible budgeting and variance analysis in an activity-based

costing environment. Accounting Horizons, 8, 93–103.

Moreno, K., Kida, T., & Smith, J. F. (2002). The impact of affective reactions on risky decision

making in accounting contexts. Journal of Accounting Research, 40(5), 1331–1349.

Ness, J., & Cucuzza, T. (1995). Tapping the full potential of ABC. Harvard Business Review, 73,

130–138.

Shields, M. (1995). An empirical analysis of firms’ implementation experiences with activity-

based costing. Journal of Management Accounting Research, 7, 148–166.

Shields, M., & Young, S. (1989). A behavioral model for implementing cost management

systems. Journal of Cost Management, 3, 17–27.


Shields, M., & Young, S. (1995). Behavioral and organizational issues. In: B. Brinker (Ed.),

Handbook of cost management. New York: Warren Gorham Lamont.

Tversky, A., & Kahneman, D. (1992). Advances in prospect theory: Cumulative representation

of uncertainty. Journal of Risk and Uncertainty, 5, 297–323.

Vera-Munoz, S. C., Kinney, W. R., & Bonner, S. E. (2001). The effects of domain experience

and task presentation format on accountants’ information relevance assurance. The

Accounting Review, 76(3), 405–429.

Willman, P., & Fenton-O’Creevy, M. (2002). Traders, managers and loss aversion in investment

banking: A field study. Accounting, Organizations & Society, 27(1, 2), 85–98.

APPENDIX A

Product Costing Case: Part 1

Micro Computer Industries makes memory boards for microcomputers.

They began operations in 1986 and had sales of about $10 million last year.

The production process involves the assembly of various memory board

circuits. Two types of memory boards are currently made, a Standard model

and a Deluxe model. Prices in this industry are very competitive, so prices

are based on competitors’ prices.

The company is organized into two divisions, one for each type of mem-

ory board. The senior officers (i.e., the chief executive officer, chief operating

officer, and chief financial officer) are salaried and are given stock options.

The managers of the two divisions are paid a salary and a substantial bonus,

based on the profitability of their division determined by the selling price

minus product costs and a portion of allocated administrative costs. Product

costs are determined by a traditional cost accounting system, where fixed

overhead is allocated based on direct labor hours. Exhibit A.1 shows the

unit costs for the two types of memory boards under the current traditional

system.

Recently, the CFO of the company has been attempting to improve the

accuracy of the product costing system. One attempt involves the possible

use of an ABC system. It is estimated that the cost of installing such a

system would be between $200,000 and $500,000, with the most likely cost

of $350,000. Four activities were identified and their rate per activity was

calculated. Exhibit A.2 shows the unit costs under the proposed ABC system

for the month of January, in which 1,137 memory boards were made in

the Standard Division and 288 memory boards were made in the Deluxe

Division.


Assume that you were the CHIEF FINANCIAL OFFICER. Give your

OPINION (there are no right or wrong answers) to the following questions.

What is the maximum that the company should pay to install the ABC

system? _________

What percentage of the cost of the new ABC system would you accept to

be allocated to each division?

Standard __________

Deluxe __________

How much would you support or object to the purchase and installation

of the new ABC system? (make a slash mark in the appropriate place along

the line)

|---------------|---------------|---------------|---------------|---------------|---------------|

Would

Strongly

Support

Would

Strongly

Object toIndifferent

Exhibit A.1. Unit Product Costs under Traditional System.

Memory Boards

Standard Deluxe

Direct materials 100 200

Direct labor 30 60

Manufacturing overhead 150 300

Total ($) 280 560

Exhibit A.2. Unit Product Costs under ABC System.

Memory Boards

Standard Deluxe


Direct labor 30 60


Total ($) 250 620


APPENDIX B

Product Costing Case: Part 2 (STANDARD condition)

Micro Computer Industries makes memory boards for microcomputers.

They began operations in 1986 and had sales of about $10 million last year.

The production process involves the assembly of various memory board

circuits. Two types of memory boards are currently made, a Standard model

and a Deluxe model. Prices in this industry are very competitive, so prices

are based on competitors’ prices.

The company is organized into two divisions, one for each type of mem-

ory board. The senior officers (i.e., the chief executive officer, chief operating

officer, and chief financial officer) are salaried and are given stock options.

The managers of the two divisions are paid a salary and a substantial bonus,

based on the profitability of their division which is determined by the selling

price minus product costs and a portion of allocated administrative costs.

Product costs are determined by a traditional cost accounting system, where

fixed overhead is allocated based on direct labor hours. Exhibit B.1 shows

the unit costs for the two types of memory boards under the current tra-

ditional system.

Recently, the CFO of the company has been attempting to improve the

accuracy of the product costing system. One attempt involves the possible

use of an ABC system. It is estimated that the cost of installing such a

system would be between $200,000 and $500,000, with the most likely cost

of $350,000. Four activities were identified and their rate per activity was

calculated. Exhibit B.2 shows the unit costs under the proposed ABC system

for the month of January, in which 1,137 memory boards were made in the

Exhibit B.1. Unit Product Costs under Traditional System.

Memory Boards

Standard Deluxe


Direct labor 30 60


Total ($) 280 560


Standard Division and 288 memory boards were made in the Deluxe

Division.

Remember that, as manager of the STANDARD division, your bonuses

are based on the profitability of that division. Exhibit B.3 shows the bonus

calculations for the current year.

If the ABC system had been in place in the current year, the bonus

calculation (ignoring the allocation of the system’s cost) would have been as

in Exhibit B.4.

Assume that you are the MANAGER of the STANDARD Division.

Give your OPINION (there are no right or wrong answers) to the following

questions.

What is the maximum that the company should pay to install the ABC

system? _________

What percentage of the cost of the new ABC system would you accept to

be allocated to your division? ________________

Exhibit B.2. Unit Product Costs under ABC System.

Memory Boards

Standard Deluxe


Direct labor 30 60


Total ($) 250 620

Exhibit B.3. Current Year Actual Bonus Calculation.

Division

Standard Deluxe

Unit sales 12,000 6,000

Selling price per unit ($) 450 950

Cost per unit ($) 280 560

Gross margin ($) 2,040,000 2,340,000

Allocated administration costs ($) 900,000 1,100,000

Profit ($) 1,140,000 1,240,000

Bonus (2% of profit) ($) 22,800 24,800


How much would you support or object to the purchase and installation

of the new ABC system? (make a slash mark in the appropriate place along

the line)

|---------------|---------------|---------------|---------------|---------------|---------------|

Would

Strongly

Support

Would

Strongly

Object toIndifferent

Exhibit B.4. Current Year Bonus Calculation if ABC System had been

in Place.

Division

Standard Deluxe

Unit sales 12,000 6,000

Selling price per unit ($) 450 950

Cost per unit ($) 250 620

Gross margin ($) 2,400,000 1,980,000

Allocated administration costs ($) 900,000 1,100,000

Profit ($) 1,500,000 880,000

Bonus (2% of profit) ($) 30,000 17,600



188

EXAMINING THE ROLE OF

CULTURE AND ACCULTURATION

IN INFORMATION SHARING

Stephen B. Salter and Axel K.-D. Schulz

ABSTRACT

In the current environment, an important firm asset is the employee

knowledge base, which in a large part depends on employee willingness to

share information. Yet prior research has noted that while employees are

delighted to reveal success they are often reluctant to reveal errors. While

there are many factors affecting managers’ reluctance to reveal errors,

this study focuses on cultural differences between Chinese migrants and

Anglo residents as well as the role of acculturation. This is particularly

relevant given the very significant foreign direct investment into China,

and migration of managers and high-end technical staff from portions of

Greater China to the management and higher technical classes of the

Anglo world. Prior studies including Chow, Harrison, McKinnon, and Wu

(1999a). Accounting, Organizations and Society, 24, 561–582, Chow,

Deng, and Ho (2000). Journal of Management Accounting Research, 12,

65–95, and Tinsley and Pillutla (1998). Journal of International Business

Studies, 29(4), 711–728, provide conflicting views and evidence for dif-

ferences in information sharing between Chinese and Anglo managers,

and there is no accounting or management literature that deals with

changes in information sharing behavior in the migration process.





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08008-1

189

This study employs an experiment to test for differences in individuals’

willingness to share information about a prior costing error. Using a

sample of students from two different nationalities drawn from a major

Australian university (Australian and Hong Kong SAR, China), this

study finds that migrant Chinese share less information than Anglo-

Australians. This study further provides empirical evidence that the rel-

ative change in willingness to share this information when the supervisor is

removed from the decision context is lower for the migrant Chinese than

for the Anglo-Australians. Finally, this study finds evidence for accultur-

ation as the willingness of migrant Chinese managers changes with the

length of their stay in the new society. Acculturation occurs relatively

quickly and highly acculturated Chinese information-sharing behavior is

not significantly different from the Australian-born subjects.

INTRODUCTION

In the current knowledge-based environment, a large portion of a firm’s

assets are in its information and knowledge. Often, the source of this

knowledge is the database of best practices and past failures that accumu-

lates among employees.1 While there are many aspects of information shar-

ing in organizations, this study focuses on the impact of cultural differences

as well as the effect of changes in cultural differences over time (accultur-

ation) in relation to individuals revealing past errors. In particular, this

study centers on differences in culture between Anglo-Australian residents

and Chinese migrants.2

How to encourage the release of information through control systems has

been widely debated in the Anglo world (see, e.g., Chenhall, 1992; Nanni,

Dixon, & Vollman, 1992; Peters, 1994; Smith, 1994; Levinthal & March,

1993). Periodically removing the supervisor, who has the ability to punish

subordinates based on the information revealed, has been suggested as one

possible solution to reduce the risk of information sharing (Peters, 1994;

Chow et al., 1999a). In addition to Anglo studies, several authors have tried

to extend the research on information sharing to other cultures, primarily

Greater China,3 only to find that the results are inconsistent. Chow et al.

(1999a) and Chow, Schulz, and Wu (1999b) find no statistical differences in

base propensity to share information between Australian and Chinese

(Taiwanese) managers. In contrast, Chow et al. (2000) report significantly

more sharing by Peoples Republic of China (PRC) managers than American

STEPHEN B. SALTER AND AXEL K.-D. SCHULZ190

managers. Tinsley and Pillutla (1998), in a related negotiation study, found

Hong Kong Chinese to be less willing to share information than American

subjects.

Further, Chow et al. (1999a, b) examined inter-cultural differences in

information sharing before and after the removal of the supervisor from the

decision context. Both find that Australian and Chinese managers are more

likely to share information when a supervisor is absent although neither

compares the relative magnitude of the change between cultures. Content

analysis presented by both Chow et al. (1999a) and Chow et al. (2000) point

to differences in factors underlying subjects’ information sharing decisions

between the Anglo-(Australian and American) and Chinese (Taiwanese and

PRC) cultures.

This literature can be synthesized by the view that Anglo subjects operate

primarily in an individualistic/self interest paradigm compatible with agency

theory. Chinese subjects, on the other hand, are subject to opposing cultural

forces; collectivism and face. Collectivism is the desire to serve the needs of

the many, if necessary, at the expense of the one. Collectivism encourages

information sharing. Face, the need to protect the individual’s reputation

among those who judge him/her, creates a situation similar to agency where

the protection of one’s reputation may require hiding or not revealing

damaging information. This face effect can be further broken down by the

constituencies who judge an individual: (1) face before the group (of peers)

and (2) face before one’s superior/supervisor. The latter effect is magnified

in high-power distance cultures. The results from Chow et al. (1999a, b)

indicate that Taiwanese and possibly other members of the overseas Chinese

diaspora may be harder to predict and may even be unstable in their re-

sponses to information sharing stimuli. The three previous studies are based

on comparing Anglo managers, who live and work in an Anglo society

(U.S.A. or Australia), with Chinese managers, who live and work in a

Chinese society (PRC, Taiwan or Hong Kong).

The last quarter of the 20th century and the first years of the 21st century

have seen not only significant global flow of capital, but also a greatly

increased global flow of labor. These are largely not the ‘‘huddled masses

yearning to be free’’ of the 19th century, but rather knowledge workers who

enter positions of responsibility in the new host countries.4 This migration

first leads to the immigrants’ culture potentially clashing with that of his or

her new home. Cross-cultural studies in general, and the first part of our

study, provide some answers to this.

The second and perhaps more interesting issue of migration is that the

values and behavior of migrants are not static. Over time, immigrants

Examining the Role of Culture and Acculturation in Information Sharing 191

change their tastes, advertising and media usage, voting patterns, consump-

tion patterns, and behavior and values. This process is commonly referred to

as acculturation (see Kim, La Roche, & Tomiuk, 2004; Yammarino & Jung,

1998; Ueltschy & Krampf, 1997; Berry, 1997 for summaries). Cross-cultural

studies are insufficient to address this change. They tell only half the story

(i.e., how did the migrant’s values differ from his new host country prior to

migrating). The second part of our study investigates the effect of accul-

turation crucial to managers (i.e., changing attitudes to information shar-

ing). Our question is how behavior changes over time, if at all. In order to be

as relevant as possible, this study examines a group of the overseas Chinese

diaspora that is far more likely to immigrate to Anglo-culture nations

(i.e., citizens of Hong Kong SAR).

The results indicate that when there are disincentives to share information,

Chinese participants, with little or no acculturation, were significantly less

likely to disclose information than Anglo participants. The behavior is similar

to that described by Tinsley and Pillutla (1998). With the removal of the

supervisor, both groups disclosed more, but the Anglos exhibited significantly

larger growth in willingness to share information. Finally, the behavior of the

immigrants change over time; and acculturation brings decision patterns of

these immigrants closer to, but not completely in line with, that of the host

society. The pace of acculturation appears to be rapid with a clear break at 5

years. These results indicate that managers in immigrant receiving countries

need to manage not only the cross-cultural differences that immigrants bring

with them, but also the process of change as immigrant workers adjust or fail

to adjust their behavior to the norms of their adopted country.

This paper now continues with a brief review of the literature followed by

hypotheses, methodology, results, and conclusion.

LITERATURE REVIEW

Explanations of information sharing have taken a number of different

directions. A substantial body of literature has examined agency-based expla-

nations and proposed solutions. An alternative stream of research has exam-

ined the impact of culture on behavior patterns. Two of these studies (Chow

et al., 1999a; Tinsley & Pillutla, 1998) have attempted to link cultural tenets

and agency explanations. Given trends in cross-national movements of per-

sons, there also appears to be a need to examine the impact of acculturation on

information sharing behavior. This section of the paper reviews relevant lit-

erature for each of these streams and proposes three hypotheses.


Why Are Managers Reluctant to Release Information?

Examining the Basic Literature

Persuading employees to be open is not always easy (see, e.g., Chow et al.,

1999b). Why are managers reluctant to release information? The extant

literature appears to argue that economic consequences and perception of

personal gain or loss drives information sharing behavior. As Chow et al.

(1999b, p. 440) explain:

Despite the large number of students who had indicated a concern with the company’s

welfare and ethics, concern for their own job security and prospects had deterred many

of them from fully sharing information about their mistakes. This deterrent effect, in

turn, had depended on the expected reaction of the company and superior to the in-

formation disclosure.

This response appears to be tied closely to prior U.S. work in escalation of

commitment using an agency framework.5 In the agency studies, fear of loss

cause subordinates not to report mistakes if supervisors appear to have no

way of finding out. Kanodia, Bushman, and Dickhaut (1989), for example,

developed an equilibrium model in which deliberately failing to correct

errors and, in fact, escalating them is rational in an agency framework.

While this behavior may be irrational from the firm’s perspective, it may be

rational to the decision maker when the decision maker’s reputation is at

stake and information asymmetries are present. Harrell and Harrison (1994)

present evidence that respondents not only cover up errors, but also escalate

commitment to decisions when incentive to shirk and asymmetric informa-

tion are both present, as predicted by agency theory.

In the information-sharing literature, one proposed method of enhancing

the likelihood of information sharing is by withdrawing supervisors. Chow

et al. (1999a) proposed that doing so will cause a related reduction in threat

to the participants and hence an increase in the willingness to share infor-

mation. Such a reduced threat can have the equivalent valence of many

other types of truth-inducing incentives.

Why Are Managers Reluctant to Release Information?

Examining the Cross-Cultural Literature

The extant literature suggests that there are certain universal information-

sharing behaviors by managers. However, a growing body of research,

including Chow, Kato, and Merchant (1996), Chow, Kato, and Shields

(1994), Birnberg and Snodgrass (1988), Harrison, McKinnon,


Panchapakesan, and Leung (1994), Merchant, Chow, and Wu (1995), and

O’Connor (1995) suggests that people in different nations often vary in how

they react to given job-related conditions or decisions.

In the case of the impact of culture on information sharing, there are a

number of competing studies and theories. Tinsley and Pillutla (1998) and

Chow et al. (1999a) provide conflicting cultural models of information

sharing but the same results. Both papers find residents of greater China

relatively unwilling to share information. In the case of Chow et al. (1999a),

the original expectation based on Hofstede (1980) was that the Chinese were

more collectivist than Anglo-Australians and would share more information

for the benefit of all. The results showed this not to be the case. On the basis

of open questions, the results were ascribed to the Chinese sense of face – a

desire not to be shamed in front of their peers. Tinsley and Pillutla (1998),

using the Schwartz (1992) value study, hypothesize that Hong Kong res-

idents balance competing dimensions in decision making. While they value

self-transcendence (collectivism), they are also conservative and hence re-

sistant to change. An integral part of this conservatism is a desire to avoid

the disruption caused by information sharing which is associated with the

need of the individual to maintain face. Tinsley and Pillutla (1998) find that

conservatism (face) dominates the decision process, and Hong Kong sub-

jects share less information than Americans. Chow et al. (2000), on the other

hand, studying subjects in the PRC argue and find the opposite to Tinsley

and Pillutla (1998). They posit and find that a person in a collectivist society

will be more inclined to share information that is beneficial to the collective.

Chow et al. (2000), however, does not involve the disclosure of information

before a supervisor but simply within groups of equals.

Contrasting Chow et al. (2000) with the Tinsley and Pillutla (1998) model,

both acknowledge the intuitive link between collectivism and information

sharing in the Chinese society; however, the difference arises in the impor-

tance of the counterforce of face. Chow et al. (2000) see face as a byproduct

of and dominated by collectivism, with the citizens of Greater China being

collectivist and information sharing. Tinsley and Pillutla (1998) start with

the Schwartz (1987) view of collectivism being overshadowed by face as part

of Chinese conservatism (a desire to avoid change or suggest that something

is not actually what it appears to be).

Tinsley and Pillutla’s (1998) position is very much in line with the views of

Ho (1976) and Redding and Wong (1986). Redding and Wong (1986) argue

that one of the features that distinguish the importance of face in Chinese

cultures is the sheer degree of concern with it. Ho (1976, p. 867), for

example, explains the dynamic of face as follows: ‘‘Face is lost when the


individual fails to meet essential requirements placed upon him by virtue of

the social position he occupies.’’ Thus an individual who fails to achieve a

promised commitment, whatever the reason, suffers a loss of face and hence

a loss of psychic income in addition to any personal monetary consequences.

Persons disappointing either a superior or a group may lose face. In both

cases, perceived failure leads to a loss of face from a superior or from a

group as a whole. Whether one loses face before a superior or subordinate,

the perception of that loss is measured with regard to external constituen-

cies. Thus, ‘‘yface is assessed in terms of what others think of him’’ (Ho,

1976, p. 871). It is possible that Chow, Deng and Ho (2000) removes part of

face by removing the supervisor but the net effect of the papers cited in the

preceding paragraphs is that combined effect of collectivism and face on

information sharing remains an open question and one worth reexamining.

Given the opposite views in the prior literature, our initial starting point

or anchor is that Chinese managers who have migrated to an Anglo country

will have a different propensity to reveal private information than Anglo

managers residing in the same country. As such, we hypothesize

H1. Chinese immigrant managers will have a different propensity to re-

veal private information than born and raised Anglo-resident managers.

Drawing on the cultural model of Hofstede (1980, 1991) supplemented by

the work of Ho (1976) on face, Chow et al. (1999a) also proposed an in-

teraction between the cultural origin and the effect of the removal of the

supervisor on information-sharing behavior. While Chow et al. (1999a) did

not find statistical support for this hypothesis, as both types of managers

responded similarly to the removal of the supervisor, further content anal-

ysis suggested that Anglo-Australian managers used different decision mod-

els and criteria than Taiwanese-Chinese (hereinafter referred to as Chinese)

managers in their decision to reveal information.

Australian subjects were presented in the Chow et al. (1999a) study as

fairly simple and calculative. The role of a supervisor is to reward success

and punish failure. Thus, the presence of a supervisor provides the major

barrier to information disclosure and, as such, provides the incentive for

subordinates not to disclose information about prior mistakes that could

harm the individual’s prospects. The removal of a supervisor results in the

removal of this barrier to information disclosure. The Chinese were pre-

sented as facing a more complex series of factors in making the decision to

reveal information. Chinese managers face three barriers to information

sharing: (1) a desire to avoid a loss of face before the group; (2) a desire to

avoid a loss of face before the supervisor; and (3) the perception of the


superior as the greater authority and the legitimate and obliged decision

maker (power distance).

While two of these barriers are removed when the supervisor is with-

drawn, loss of face before the group remains. Thus, one might anticipate a

smaller increase in the propensity to share information for members of the

Chinese society vis-a-vis the Anglo-Australians whose major significant

barrier to sharing, fear of punishment, has been removed. Based on these

cultural perspectives, we therefore propose that Chinese managers who have

migrated to an Anglo country will exhibit a relatively lower level of will-

ingness to release private information about an error in response to the

removal of a supervisor than Anglo managers who are residing in the same

country. Our second hypothesis is

H2. The change in willingness to release private information about a

prior mistake, as a result of the removal of the supervisor, would be lower

for Chinese immigrant managers than born and raised Anglo-resident

managers.

Acculturation and Information Sharing

If one believes the extant literature, cross-national differences in values can

and do affect behavior. Even if a manager never plans to immigrate, she/he

is still likely to encounter cross-cultural control issues. Large-scale immi-

gration6 makes it increasingly unclear from whence an American, Canadian,

or Australian manager has drawn their values and how they may react to

particular decision-making situations.

While non-native-born individuals may either immediately or never ab-

sorb the new culture, considerable evidence from marketing and other social

science literature suggests that the values and behaviors of immigrants

change over time as they interact with the existing workforce (see Ueltschy

& Krampf (1997) and Yammarino & Jung (1998) for summaries), a process

referred to as acculturation. In each of these studies, immigrant group

members are found to hold norms and values somewhere between those of

the culture of origin and the host society. The more acculturated the in-

dividual, the greater the progression toward the attitudes and values of the

host society (Faber, O’Guinn, & Meyer, 1987). Studies of Latino immi-

grants have, for example, noted differences in media preferences and ad-

vertising effectiveness (Hayes-Bautista, Schinck, Chapa, & Soto, 1984;

Adelson, 1989; Ueltschy & Krampf, 1997). The marketing literature


contains a similar stream of literature on Chinese immigrants which is

summarized in Lee and Tse (1994).

The process of acculturation is not linear and takes place in stages

(Ueltschy & Krampf, 1997). Studying Mexican immigrants to the U.S.,

Ueltschy and Krampf (1997) identified three statistically viable clusters of

behaviors with members of each cluster varying in terms of level of accul-

turation to media.

1. A low acculturation cluster of persons who have lived in the U.S. between

1 and 5 years

2. A bicultural cluster of persons born in Mexico who have lived in the U.S.

from 10 to 15 years

3. A high acculturation cluster of persons born in the U.S. of non-U.S.

parents and grand parents; this group is not particularly dissimilar from

the Anglo residents around them

These differences between the groups in the sample translate into significant

differences in advertising preferences. Lee and Tse (1994) use a four-part a

priori classification of Hong Kong and Canadian subjects based on immi-

gration laws and status, including English-speaking Caucasian Canadians

and three levels of Chinese immigrants. Like Ueltschy and Krampf (1997),

they find significantly different media usage patterns in each group but use a

breaking point for immigrants of 7 years based on the elapsed time since a

major change in immigration law.

In summary, the extant literature seems to indicate that immigrants grow

closer to the host country in tastes and behavior over time. The previous

literature (Chow et al., 1999a, 2000) finds that Australian subjects are ex-

pected to begin at a lower level of information sharing than Chinese and

increase their likelihood of sharing information (about a prior mistake) to a

greater extent than Chinese subjects in response to the removal of the su-

pervisor. Therefore, as the migrant Chinese group becomes more accultur-

ated, they will also show a much greater increase in likelihood of sharing

information in response to the removal of the supervisor than their

un-acculturated peers. Thus our third hypothesis is

H3. There is a positive relationship between the change in the willingness

to release private information about a prior mistake when the supervisor

is removed and the degree of acculturation.


RESEARCH METHODS

Overview of Design

An experiment was designed to focus on private information disclosure of

Chinese immigrants and Australian residents and the effect of acculturation

on Chinese immigrants’ disclosure behavior. A 2� (2) repeated factorial de-

sign was utilized to test hypotheses 1–3. The first factor represented the type

of culture and was assessed in terms of ethnic origin and country of birth,

with subjects being classified into two categories of Australian and Chinese

(more detail about the categories is provided in subsequent discussion). The

second factor represented the removal of the supervisor, which was manip-

ulated within subjects. Participants were required to make two decisions with

the removal of the supervisor occurring between the first and second deci-

sion. To test H3, we split the Chinese group into a high- and low-acculturated

subgroup on the basis of tenure in Australia, which resulted in a 3� (2)

repeated factorial design. After the split, the first factor represented the level

of acculturation ranging from low-acculturated Chinese to high-acculturated

Chinese to Australian groups. The second factor remained the removal of the

supervisor. The dependent variable was the subjects’ degree of willingness to

disclose private information revealing prior personal mistake.

Decision Task

The decision task was adapted from the Chow et al. (1999a) instruments.

Each subject received an experimental booklet containing instructions and

the experimental material. Participants were asked to read the experimental

material carefully and assume the role of a department manager of a hy-

pothetical plant. The instructions also contained very explicit statements to

the effect that there was no ‘‘correct’’ or ‘‘incorrect’’ response to the ex-

periment, but rather that the best response was one that most closely re-

flected their true feelings and belief.

The experimental material consisted of a background note and two ques-

tions. For the background note, subjects were provided with information

about their recent promotion to department manager and a decision they

had taken to endorse a new technology shortly after their promotion. Sub-

jects were told that after endorsing the new technology, they found out that

they had underestimated the variable costs of the new technology. In

hindsight, if they had correctly estimated the variable cost, the best decision


would have been to stay with the old technology. Subjects were further

informed that the organization monitors performance very closely but does

not track actual costs and revenues from individual projects. This allowed

them to work very hard to squeeze cost savings and revenues from other

projects to make up for the higher than planned costs of the new technology.

Subjects were further told that they had just been invited to attend a meeting

with another department manager in another plant, who was also consid-

ering adopting a new technology. Subjects were informed that the technology

was very similar to the one they had adopted and that the other manager had

also underestimated the variable costs associated with the technology.

Following the background note, subjects had to make two decisions. For

both decisions, the participants had to indicate the degree to which they would

reveal their mistake in underestimating the variable cost of the new technology

in order that people in the other plant would make a better decision. For the

first decision, subjects were told that their supervisor would be attending the

same meeting with the other department manager. For the second decision,

subjects were informed that their supervisor was now not attending the meeting

with the other department manager and that there was little chance of an-

ything that was said would find its way back to the supervisor.

Further, subjects were also asked to indicate the extent to which their two

decisions truly reflected how they would think and act if they were in the

manager’s place. Finally, subjects were asked to indicate their ethnic origin

along with their country of birth and information about the tenure in

Australia or any other country in which they have lived. Demographic in-

formation about their gender was also collected.

Dependent Variable

The dependent variable measured the degree to which the individual would

reveal their private information about prior personal failure. Subjects were

asked to indicate to what extent they would definitely reveal their own

mistake of underestimating. They recorded their answer on a 9-point scale

ranging from 1 (definitely not reveal) to 9 (definitely reveal).

Independent Variables

The independent variables were culture/degree of acculturation and the

presence or absence of the supervisor. In measuring culture/acculturation,


the prior research was examined for precedent. Researchers have used a

variety of measures to operationalize culture/acculturation. These include

the following:

1. Self-identification. This approach has been deemed most appropriate by

cross-cultural behavioral researchers, especially those in cultural anthro-

pology and social psychology (Cohen, 1978, Minor, 1992; Kara & Kara,

1996).

2. Place of birth. Padilla (1980) and Valencia (1985) use place of birth (U.S.

vs. foreign born) to proxy for acculturation.

3. Place of birth/time in country groupings. Both Lee and Tse (1994) and

Ueltschy and Krampf (1997) use time-bounded clusters, which then form

the basis of testing the main hypotheses of their study.7

We divide our sample initially into Australian born and non-Australian

born. All of the Australian-born participants identify themselves as

Caucasian and the Chinese sample was of Chinese ethnic origin. This re-

moves the possibility of Australian-born Chinese ethnic origin subjects and

means that place of birth and self-identified origin collapse into one. The

two broad categories can be described as Australian born and Chinese born.

Within the Chinese-born sample, an initial regression indicates that time in

country affects the difference in likelihood that the individual is going to

reveal their prior error between when the supervisor is present and when the

supervisor is absent. We are then faced with choosing a viable break point to

group our Chinese-born sample. Fortunately, the immigration-based ap-

proaches taken by Lee and Tse (1994) and the rules of thumb of Ueltschy

and Krampf (1997) can be synthesized. This study uses a categorization

consistent with Ueltschy and Krampf (1997) of less than 5 years and 5 or

more years living in the new country. This categorization is also consistent

with Australian immigration rules specifically requiring immigration appli-

cations to be resubmitted at the end of each phase of education. Our second

year students, who are resident for more than 5 years, would have to have

completed at least two immigration reviews before the Australian Govern-

ment (high school and college entry) or have achieved permanent residence

by migrating themselves or been part of a family migration.

The two groups, Australian born and Chinese born, were used to test for

initial cultural differences in H1 and H2 and three groups Australian Born

Culture (ABC), Chinese Born High Acculturation (CHA) and Chinese Born

Low Acculturation (CLA) were used for the purpose of testing the impact of

acculturation H3. The ABC group consists of participants born in Australia,

essentially of Anglo ethnic origin, born in a country with a dominant Anglo


culture. The CHA group consists of Chinese participants, born in

Hong Kong, with 5 or more years of tenure in Australia. While members

of this group were not born in a country with a dominant Anglo culture,

their tenure for 5 or more years in Australia provided them with a high level

of opportunity to become acculturated with the Anglo culture. The CLA

group consists of Chinese participants, born in Hong Kong, with less than 5

years of tenure in Australia. Members of this group were born in a country

that does not have a dominant Anglo culture and their tenure in Australia

has been minimal, providing them with only a low level of opportunity to

become acculturated with the Anglo culture.

The role of the supervisor in preventing or enhancing information disclo-

sure was manipulated within subjects by the removal of the supervisor in the

second decision. For the first decision, subjects were told that the supervisor

was attending the meeting between themselves and the other plant managers

(‘‘supervisor present’’ condition), while for the second decision the supervisor

was not attending the meeting (‘‘supervisor absent’’ condition).

Manipulation Checks

A question testing for the respondents’ willingness to answer the questions

truthfully was included in the experimental instrument.8 Participants were

asked to respond on a 9-point scale ranging from 1 (‘‘not at all’’) to 9

(‘‘totally’’). Five participants were omitted for failing to answer the ques-

tion. This resulted in 115 usable responses. The mean response of the 115

participants was 7.157 (significantly greater than the midpoint of the scale,

t ¼ 18:403; p ¼ 0:000). Of the 115 participants, 7 responded below the mid-

point of the scale. The 115 usable responses did not significantly differ

between treatment groups on this question. These results provide comfort

that participants not only responded truthfully, but also did not system-

atically interact with the treatment groups. As such, all 115 responses were

retained for the main analysis.9

Participants

All participants were enrolled in the same second-year management ac-

counting subject at the time of the experiment.10 Demographic data was

obtained for gender. While the distribution of the 57 male and 58 female


subjects was not consistent across all treatments, a post hoc comparison

between gender and the dependent variable was not significant.

RESULTS

Analysis

To test the hypotheses, a combination of two repeated ANOVA tests (for H1

and H2) as well as a linear regression and Bonferroni contrast tests (for H3)

were employed. Bonferroni contrast tests were used to test the very specific a

priori relationships of interest in H3. All directional hypotheses were tested

using 1-tailed tests. All of the analysis was conducted using SYSTAT 10.

Descriptive Statistics

To test the first two hypotheses, the sample was split on the basis of country

of birth. A few general trends emerged from the descriptive statistics re-

ported in Table 1. In all situations and for all sample subgroups, the will-

ingness to reveal information was greater than the midpoint of the scale

indicating a propensity to reveal private information about a prior mistake.

On average, the Chinese migrant sample had a lower propensity to reveal

information than the Anglo-Australian sample. This is concomitant with

Tinsley and Pillutla (1998) findings that American managers show a greater

willingness to share information than Chinese managers in negotiation

strategies. In the supervisor absent condition, there was an increase for both

the Chinese sample and Australian sample in the willingness to reveal pri-

vate information about a prior mistake.

To address the issue of acculturation, the complete Chinese sample was

split into two subgroups resulting in three groups: ABC, CHA, and CLA, as

discussed previously. This sample breakdown reveals some general trends

between the CHA and CLA groups reported in the descriptive statistics

(refer Table 2). On average, the CLA group was less willing to reveal in-

formation than the CHA group. Although the CHA group increased their

willingness to disclose information once the supervisor was removed, the

CLA group did not. Finally, examining the responses in Table 2, the results

show a positive trend between the difference in willingness to reveal the

information (supervisor absent vs. supervisor present) and the degree of

acculturation from CLA (�0.066)11 to CHA (+1.100) to ABC (+1.300).


Hypothesis Testing

H1 postulates that migrant Chinese managers working in an Australian

culture will have a different propensity to reveal information about a prior

mistake than resident Australian managers. From the data shown in Table 1

and the results in Table 3, H1 was supported. The results show that Chinese

participants were significantly (F ¼ 4:916; p ¼ 0:029; two-tailed) less likelyto disclose information (mean ¼ 6:138) than Australian participants

(mean ¼ 6:850). These results are consistent with those previously found

by Tinsley and Pillutla (1998).

H2 proposed that the removal of the supervisor would result in a smaller

increase in information sharing for migrant Chinese than Australian res-

ident subjects. The descriptive statistics reported in Table 1 and the results

reported in Table 3 show that the tendency to release information when the

supervisor is absent is marginally significantly (F ¼ 2:286; p ¼ 0:068;12 one-tailed) greater for Australian subjects (difference in means ¼ 1.300) than

Chinese subjects (difference in means ¼ 0.831). As the difference is only

marginally significant, H2 was not supported. One of the potential expla-

nations for the nonsignificant difference could be that subjects highly ac-

culturated to the Australian context show behavior consistent with the

Australian subjects. This explanation is tested more specifically when ex-

amining H3.

Table 1. Means (Standard Deviations) for Australian Born vs. Chinese

Born Participants.

Supervisor Culture Partial Means

Australian Chinese

Supervisor present

Mean 6.200 5.723 5.930

S.D. (2.070) (1.972) (2.021)

Supervisor absent

Mean 7.500 6.554 6.965

S.D. (1.632) (1.924) (1.821)

Partial means 6.850 6.138

(1.632) (1.956)

N 60 65 115

Note: Higher values indicate a greater willingness to definitely reveal mistakes.


H3 proposed a positive relationship between the change in willingness to

release private information about a prior mistake when the supervisor is

absent and the degree of acculturation. The descriptive statistics reported in

Table 2 and the results reported in Table 4 do indeed show a significant

positive relationship (F ¼ 4:206; p ¼ 0:009; one-tailed) between the propen-

sity to disclose the information (supervisor absent vs. supervisor present)

Table 2. Mean (Standard Deviation) for Australian vs. Chinese High/

Low Acculturation.

Culture Partial Means

ABC CHA CLA

Supervisor present

Mean 6.200 5.660 5.933 5.930

S.D. (2.070) (1.923) (2.187) (2.021)

Supervisor absent

Mean 7.500 6.760 5.867 6.965

S.D. (1.632) (1.779) (2.031) (1.821)

Partial means 6.850 6.210 5.900

(1.966) (1.924) (2.074)

N 50 50 15 115

Note: Higher values indicate a greater willingness to definitely reveal mistakes. ABC, Austral-

ian-Born Culture Group; CHA, Chinese High Acculturation Group; CLA, Chinese Low

Acculturation Group.

Table 3. Repeated ANOVA for Australian vs. Chinese Born

Participants.

Analysis of Variance

Source SS DF MS F p (2-tailed)

Between subjects

Culture 28.616 1 28.616 4.916 0.029

Error 657.758 113 5.821

Within subjects

Supervisor 64.155 1 64.155 47.130 0.000

Supervisor�Culture 3.111 1 3.111 2.286 0.133

Error 153.819 113 1.361


and the degree of acculturation CLA (�0.066) to CHA (+1.100) to ABC

(+1.300).

A follow up reported in Table 5 shows that the difference in propensity to

release information between the two Chinese groups (CHA vs. CLA) was

significant (c1; F ¼ 7:083; p ¼ 0:004; one-tailed) while the difference in pro-

pensity between the high-acculturated Chinese subjects and the Australian

group was not significant (c2; F ¼ 0:045; p ¼ 0:416; one-tailed). Therefore,H3 was partially supported.

The sub-sample of subjects born in China was further examined in order

to explore the different information disclosure behavior resulting from ac-

culturation. A regression comparing the difference in information disclosure

(supervisor present vs. supervisor absent) to the number of years these sub-

jects have lived in Australia indicates that the coefficient is both positive

(0.134) and significant (t ¼ 2:738; p ¼ 0:008; two-tailed). This provides

further evidence of acculturation; with the increased number of years the

Chinese migrants spent in Australia, the more their behavior resembled the

high-acculturated Chinese group, which was previously shown to be insig-

nificant from the Australian group (Table 6).

In summary, Chinese migrant subjects on average were significantly less

likely to reveal information. In addition, the relative difference in willingness

to reveal information as the situation shifts, from supervisor present to one

where the supervisor is absent, is lower for Chinese subjects than Australian

subjects although the difference is only marginally significant. However, the

small difference can be explained in terms of acculturation. Our results show

that while the high-acculturated Chinese group did not differ significantly

from the Australian subjects, the low-acculturated Chinese group did.

Table 4. Repeated ANOVA for Australian vs. Chinese

High/Low Acculturation.

Analysis of Variance

Source SS DF MS F p (2-tailed)

Between subjects

Acculturation 30.834 2 15.417 2.634 0.076

Error 655.540 112 5.853

Within subjects

Supervisor 25.521 1 25.521 19.582 0.000

Supervisor�Acculturation 10.964 2 5.482 4.206 0.017

Error 145.967 112 1.303


CONCLUSION AND LIMITATIONS

This study examined the impact of the multiple forces that may affect the

decision to release information about a prior mistake in a cross-cultural

setting. It finds that while the absence of a supervisor can enhance infor-

mation sharing, this effect is not constant across cultures. Further and more

important, as managers migrate, a process of acculturation takes place that

shifts response patterns for the migrant to a point closer to but not com-

pletely the same as persons born in his/her host culture.

The results support Tinsley and Pillutla’s (1998) view that for Chinese

subjects, conservatism (face) dominates their base collectivism. Thus, the

willingness to share information that logically should exist in persons from a

collectivist culture is not found. This raises questions as to why Chow et al.

(2000) find the opposite. It is possible that the in-group portion of Chow

et al.’s (2000) sample reinforces cultural collectivism that the overseas Chinese

(Taiwanese and Hong Kong residents of Chinese origin) used in this sample.

Tinsley and Pillutla (1998) and Chow et al. (1999a) represent a variant on

Table 5. Bonferroni Contrasts for Acculturation.

Contrast SS DF MS F p (2-tailed)

c1 15.705 1 15.705 7.083 0.009

c2 0.099 1 0.099 0.045 0.833

Note: c1 ¼ (CHA vs. CLA) vs. (supervisor present vs. supervisor absent). c2 ¼ (ABC vs.

CHA) vs. (supervisor present vs. supervisor absent).

Table 6. Regression of Change in Willingness to Report and

Acculturation.

Effect Coefficient Std. Error Std. Coefficient Tolerance t-stat Probability (2-tailed)

Constant �0.176 0.399 0.000 �0.441 0.661

LIA 0.134 0.049 0.326 1.000 2.738 0.008

Source Sum of Squares DF Mean Square F-ratio p

Regression 11.609 1 11.609 7.499 0.008

Residual 97.529 63 1.548

Note: N ¼ 65; R2 ¼ 0:106; LIA ¼ years lived in Australia, Durbin Watson D Statistic ¼ 2.443,

First-order Autocorrelation ¼ �0.222.


those in the PRC or that the managers in Chow et al. (2000) represent one

view of Chinese society.

Further, the behavior of the immigrant appears to change over time and

acculturation brings decision patterns of these immigrants closer, but not

completely in line with that of the host society. This supports the extant

marketing literature that indicates that changes in consumer behavior can be

tied to acculturation. The pace of acculturation appears to be rapid with a

clear break at 5 years.13

One of the limitations of this study lies in its use of students as surrogates

for managers. While we believe that student subjects are acceptable

surrogates for managers, as discussed previously in this study, it does raise

an issue of the potential moderating role of general corporate culture

(particularly for foreign-owned companies), which is outside the scope of our

study.

A second limitation lies in this study’s reliance on established cultural

differences (such as differences in face) based on different nationalities or-

igins backgrounds. Future research may strengthen this by measuring this

difference more directly. Along with this limitation, this study raises ques-

tions that might well be answered by further research. Further work needs to

be done to test the pace of acculturation in non-Anglo societies that accept

immigrants such as France or Italy. Acculturation should also be studied

using participants who are not attending university in the host country.14

Also, given the variance of responses to information sharing between

Hong Kong, PRC, and Chinese members of the Chinese diaspora observed

in the literature, there appears to be a need to test for differences within the

Chinese culture including differences between regions of mainland China. A

different direction of interest could lead to the examination of whether the

cultural background of the supervisor has an affect on the willingness to

disclose information about prior mistakes when they are not present. An-

other interesting direction is to explore the potential moderating role of

general corporate culture on the relationships found in our study. Partic-

ularly where general corporate culture is at odds with the national culture in

which the organization operates, managers may face competing cultural

pressures in their information-sharing decisions. Other possible variables of

interest may include size of business and ethnicity of fellow employees.

While this study exclusively examined the sharing behavior concerning neg-

ative information, part of the information and knowledge assets also include

positive information. While substantial bodies of literature examine the

sharing of positive information, the authors are not aware of any accul-

turation work in this area.


This study triangulates three views of culture in Schwartz (1992), Hofstede

(1980) and Ho (1976) in preparing its hypotheses. However, as Harrison and

McKinnon (1999) would argue, since each of these descriptions of culture

focuses on the values of each society, our study may be ‘‘limited in its ability

to examine and understand the dynamic process of management control sys-

tems and their cultural interplays.’’ Other lenses, particularly those emerging

from anthropology, sociology, and history literatures, may provide valuable

additional insights. In addition, this study is limited by the potential of a

laboratory experiment involving a simplified information-sharing task.

Finally, while this study is intended to build the academic knowledge

base, there are certain implications for managers and management control

system designers. If companies want to utilize the talents of the world’s

brightest and best, they must realize that this comes at a price. That price is

constant vigilance and understanding of how people from different cultures

respond to and change their response to their work context over time.

ACKNOWLEDGMENTS

The authors would like to acknowledge the helpful comments made by

Margaret Abernethy, Andrea Drake, Chee Chow, Graeme Harrison, Jane

Hronsky, and the participants of the 2001 AAANZ Conference and faculty

at workshops at the University of Melbourne, University of Wisconsin and

Bowling Green State University. U.S. Department of Education’s BIE pro-

gram and the University of Melbourne Faculty of Economics and Commerce

and Visiting Scholar Award provided financial support for this project.

NOTES

1. As Macintosh (1994) noted, the manner in which these employees gather, storeand move information should be of vital interest to accounting and informationsystem managers.2. We are using a very general definition of migrants as ‘‘individuals who moved

from their country of origin to a new country.’’3. Greater China includes the Peoples Republic of China and other areas where

persons of Chinese origin dominate the business culture. This group of countriestypically includes Singapore, Hong Kong and Taiwan.4. See for example ‘‘After the Flood,’’ The Economist Newspaper, September 7,

2000 and Bogumil and Lawrence (2000).


5. Agency theory has arguably had more influence on (Western) financial eco-nomics and accounting than any other theory in the last 20 years (Baiman, 1982,1990; Watts & Zimmerman, 1990). The key feature that distinguishes it from classicaleconomics is the assumption that there may be a divergence of goals between themanager (agent) and the firm’s owners (principal). Managers act in their self-interestrather than the firm’s interest (i.e., shirk) when two conditions are simultaneouslysatisfied: first, there is an incentive for them to do so (they stand to gain personally bytaking a particular action), and second, information asymmetry exists between themand principals, who have less information than the agents.6. If there is a defining demographic trend in these ‘‘immigrant Anglo countries’’

of the world, it has been the return of significant waves of immigrants from culturesthat have little or no relation to existing cultures. In Australia, for example, over20% of the population is composed of immigrants and 40% of these immigrants areChinese (SBS Television, 2000). In the 1996 Canadian census, approximately 25% ofthe population is first generation immigrants and of those immigrants approximately20% are from South and East Asia. Although the U.S. sees itself as a nation ofimmigrants, it has a smaller immigrant population with immigrants contributingapproximately 8% of the population in 1990. However, the role of Chinese immi-grants in the U.S. has been a topic of increasing importance as the percentage ofChinese immigrants among total U.S. immigrants has increased from 9.1% in 1960to 37% in 1992 (Min, 1995).7. Lee and Tse (1994) (Hong Kong and Canadian subjects) use a four-part a priori

classification viz:(1) English-speaking Caucasian Canadians (those who had lived inCanada for more than 10 years); (2) Long-time Hong Kong immigrants (those whoemigrated to Canada more than 7 years ago); (3) New Hong Kong immigrants (thosewho emigrated to Canada less than 7 years ago); and (4) Hong Kong residents (HK,ethnic Chinese who had lived in Hong Kong for more than 10 years). Hong Kongresidents and Caucasian Canadians were included as the anchoring points for com-parison purposes. Ueltschy and Krampf (1997), based on clustering of responses to amultivariate scale, define a three-part classification specifically: (1) Cluster 1 a lowacculturation group that have lived in the U.S. between 1 and 5 years; (2) Cluster 2 abicultural group born in Mexico but having lived in the U.S. from 10 to 15 years; and(3) Cluster 3 a high acculturation group born in the U.S. as were their parents andgrand parents but not particularly dissimilar from the Anglo residents around them.8. This question was included to test for any systematic differences across subject

groups in terms of how truthful subjects were in responding to the questions in theinstrument.9. To gain further comfort all of the analyses reported in this study were also

conducted with a covariate controlling for the response on the manipulation checkquestion. The covariate was coded 1 for a response below the midpoint of the scaleand 2 for a response above the midpoint. Results obtained were consistent with thosereported in this study.10. We believe that student subjects are acceptable surrogates for managers in our

study, as these subjects were studying (i.e., working) in an environment culturallyrepresentative of their future working environment insofar as national culture.11. CLA/Supervisor absent 5.867 minus CLA supervisor present 5:933 ¼ �0:066

in Table 2.


12. Table 3 provides the results for a two-tailed test. The one-tail p-value is ob-tained by dividing the two-tail p-value by 2.13. We tested a further break at 10 years and found no evidence of significant

further acculturation.14. It is interesting to note that the Australian government seems to recognize the

value of university training as a source of acculturation by their policy that foreignerseducated at Australian universities need to obtain only 90% of the points required ofother potential candidates for permanent residence.

REFERENCES

Adelson, A. (1989). English role in Hispanic marketing, The New York Times, (January 26),

D19, col 1.

Baiman, S. (1982). Agency research in management accounting: A survey. Journal of Accounting

Literature, 1, 154–213.

Baiman, S. (1990). Agency research in management accounting: A second look. Accounting,

Organizations and Society, 15, 341–371.

Berry, J. W. (1997). Immigration, acculturation, and adaptation. Applied Psychology: An In-

ternational Review, 46, 5–34.

Birnberg, J., & Snodgrass, C. (1988). Culture and control: A field study. Accounting, Organ-

izations and Society, 13, 447–464.

Bogumil, A., & Lawrence, W. (2000). Foreign knowledge workers as a strategic staffing option.

Academy of Management Executive, 14(4), 71–84.

Chenhall, R. (1992). Contemporary performance measurement. Accounting Communique

(Vol. 39). Australian Society of CPAs.

Chow, C., Deng, J., & Ho, J. (2000). The openness of knowledge sharing within organizations:

A cross-cultural investigation. Journal of Management Accounting Research, 12, 65–95.

Chow, C., Harrison, G., McKinnon, J., & Wu, A. (1999a). Cultural influences on information

sharing in Chinese and Anglo-American organizations: An exploratory study. Account-

ing, Organizations and Society, 24, 561–582.

Chow, C., Kato, Y., & Merchant, K. (1996). The use of organizational controls and their effects

on data manipulation and management myopia: A Japan vs. U.S. comparison. Ac-

counting, Organizations and Society, 21, 175–192.

Chow, C., Kato, Y., & Shields, M. (1994). National culture and the preference for management

controls: An exploratory study of the firm-labor market interface. Accounting, Organ-

izations and Society, 19, 381–400.

Chow, C., Schulz, A., & Wu, A. (1999b). Roz Inc.: A case for extending management ac-

counting coverage to managers’ information acquisition process. Journal of Accounting

Education, 17, 429–442.

Cohen, R. (1978). Ethnicity: Problem and focus in anthropology. Annual Review of Anthro-

pology, 7, 379–403.

Faber, R., O’Guinn, T., & Meyer, T. (1987). Television portrayals of Hispanics: A comparison

of ethnic perceptions. International Journal of Intercultural Relations, 11, 155–169.

Harrell, A., & Harrison, P. (1994). An incentive to shirk, privately held information and man-

agers’ project evaluation decisions. Accounting, Organizations and Society, 19, 569–577.


Harrison, G., & McKinnon, J. (1999). Cross-cultural research in management control systems

design: A review of the current state. Accounting, Organizations and Society, 24, 483–509.

Harrison, G., McKinnon, J., Panchapakesan, S., & Leung, M. (1994). The influence of culture

on organizational design and planning and control in Australia and the United States

compared with Singapore and Hong Kong. Journal of International Financial Manage-

ment and Accounting, 5, 242–261.

Hayes-Bautista, D., Schinck, W., Chapa, J., & Soto, S. (1984). Hispanic shoppers. Washington

DC: Food Marketing Institute.

Ho, D. Y. (1976). On the concept of ‘Face’. American Journal of Sociology, 81, 867–884.

Hofstede, G. (1980). Culture’s consequences: International differences in work-related values.

Beverly Hills, CA: Sage.

Hofstede, G. (1991). Cultures and organizations: Software of the mind. Berkshire, UK: McGraw-

Hill.

Kanodia, C., Bushman, R., & Dickhaut, J. (1989). Escalation errors and the sunk cost effect:

An explanation based on reputation and information asymmetry. Journal of Accounting

Research, 27(1), 59–77.

Kara, A., & Kara, N. (1996). Ethnicity and consumer choice: A study of Hispanic decision

processes across acculturated levels. Journal of Applied Business Research, 12(2), 22–36.

Kim, C. Laroche., La Roche, M., & Tomiuk, M. (2004). The Chinese in Canada: A study in

ethnic change with emphasis on gender roles. Journal of Social Psychology, 144(1), 5–30.

Lee, W., & Tse, D. (1994). Changing media consumption in a new home: Acculturation patterns

among Hong Kong immigrants to Canada. Journal of Advertising, 23(1), 57.

Levinthal, D., & March, J. (1993). The myopia of learning. Strategic Management Journal, 14,

95–112.

Macintosh, N. (1994). Management accounting and control systems. Chichester: Wiley.

Merchant, K., Chow, C., & Wu, A. (1995). Measurement, evaluation and reward of profit

centre managers: A cross-cultural field study. Accounting, Organizations and Society, 20,

619–638.

Min, P. (1995). Chinese-Americans: Contemporary trends and issues. Thousand Oaks, CA: Sage.

Minor, M. (1992). Comparing the Hispanic and Non Hispanic markets: How different are they?

Journal of Service Marketing, 6(2), 29–32.

Nanni, A., Jr., Dixon, J., & Vollman, T. (1992). Integrated performance measurement: Man-

agement accounting to support the new manufacturing realities. Journal of Management

Accounting Research, 4, 1–19.

O’Connor, N. (1995). The influence of organizational culture on the usefulness of budget

participation by Singaporean-Chinese managers. Accounting, Organizations and Society,

20, 383–403.

Padilla, A. (1980). The role of cultural and ethnic loyalty in acculturation. In: A. Padilla (Ed.),

Acculturation: Theory, models and some new findings (pp. 47–84). Boulder, CO: Westview

Press.

Peters, T. (1994). The Tom Peters seminar: Crazy times call for crazy organizations. London:

Macmillan.

Redding, G., & Wong, G. Y. Y. (1986). The psychology of Chinese organizational behavior. In:

M. H. Bond (Ed.), The psychology of the Chinese people (pp. 267–295). Hong Kong:

Oxford University Press.

Schwartz, S. (1987). Toward a universal psychological structure of human values. Journal of

Personality and Social Psychology, 53, 550–562.


Schwartz, S. (1992). Universals in the content and structure of values: Theoretical advances and

empirical tests in 20 countries. In: M. P. Zana (Ed.), Advances in experimental social

psychology, (Vol. 25, pp. 1–65). New York, NY: Academic Press.

Smith, M. (1994). New tools for management accountants. Melbourne: Longman Professional

Publishing.

Ueltschy, L., & Krampf, R. (1997). The influence of acculturation on advertising effects to the

Hispanic market. Journal of Applied Business Research, 13(2), 87.

Valencia, H. (1985). Hispanic values and cultural research. Journal of Academy of Marketing

Science, 17, 23–28.

Watts, R., & Zimmerman, J. (1990). Positive accounting theory: A ten-year retrospective. The


Yammarino, J., & Jung, D. (1998). Chinese Americans and leadership: A levels of analysis

perspective. The Journal of Applied Behavioral Science, March, 34(1), 47–67.


THE EFFECTS OF VALUE

ATTAINMENT AND COGNITIVE

ROLES OF BUDGETARY

PARTICIPATION ON JOB

PERFORMANCE

Vincent K. Chong, Ian R. C. Eggleton and

Michele K. C. Leong

ABSTRACT

This chapter examines the effects of the value attainment and cognitive

roles of budgetary participation on job performance. A structural model

consisting of variables such as budgetary participation, job-relevant in-

formation, job satisfaction, and job performance is proposed and tested

using a survey questionnaire on 70 senior managers, drawn from a cross-

section of the financial services sector. Their responses are analyzed using

a structural equation modeling (SEM) technique. The results reveal that

budgetary participation is positively associated with job-relevant infor-

mation. These results lend support to the cognitive effect of budgetary

participation, which suggests that subordinates participate in the budget

setting process to share information. In addition, the results suggest that

budgetary participation is positively associated with job satisfaction.





ISSN: 1475-1488/doi:10.1016/S1475-1488(04)08009-3

213

These results support the value attainment role of budgetary participa-

tion, which increases subordinates’ levels of job satisfaction. Furthermore,

the results reveal that there are positive relationships between job-relevant

information and job satisfaction, job-relevant information and job per-

formance, and job satisfaction and job performance.

INTRODUCTION

Since the pioneering work of Argyris (1952) concerning the behavioral as-

pects of budgeting, a sizable body of research has developed in management

accounting examining the effect of budgetary participation on job

performance (see Covaleski, Evans, Luft, & Shields, 2003, for a compre-

hensive review). To date, the results of empirical evidence on the cognitive

role of budgetary participation on job performance have produced consist-

ent outcomes (Chenhall & Brownell, 1988; Kren, 1992; Chong & Chong,

2002).1 Kren (1992) developed a research model, which explicitly

examined the cognitive function of budgetary participation on job per-

formance. He argues that budgetary participation can facilitate the acqui-

sition and use of job-relevant information. Job-relevant information, in

turn, can improve performance. Specifically, Kren finds that budgetary

participation affects job performance indirectly through job-relevant

information.

A review of the literature on participative budgeting indicates that value

attainment is another role of budgetary participation (Locke & Schweiger,

1979; Locke & Latham, 1990; Shields & Shields, 1998). Indeed, numerous

studies have recognized the value attainment role of budgetary participation

(see e.g. Chenhall, 1986; Chenhall & Brownell, 1988; Chong & Bateman,

2000). However, no studies have explicitly tested its impact on subordinates’

job performance, and its potential influence within a cognitive model of

budgetary participation. The value attainment role of budgetary participa-

tion theoretically affects subordinates’ levels of job satisfaction (Shields &

Shields, 1998). Specifically, the value attainment effect of budgetary par-

ticipation suggests that allowing subordinates to participate in the budget-

setting process will increase the likelihood that they will feel satisfied with

their values (French, Israel, & As, 1960; Strauss, 1963; Lowin, 1968; Locke

& Schweiger, 1979). Subordinates’ values may include: (1) the opportunity

to express their views, (2) the feeling of being treated equally, and (3) the

desire for respect or dignity (Argyris, 1955; Davis, 1957). To date, the

VINCENT K. CHONG ET AL.214

existing literature has not explicitly tested the value attainment role of

budgetary participation within the cognitive model of budgetary participa-

tion developed by Kren (1992). This gap in the accounting literature,

which remains unexplored, constitutes the primary motivation for our

study.

This study extends Kren’s cognitive model by incorporating the variable

of job satisfaction in our theoretical model (see Fig. 1). We argue that

participation in the budget-setting process will help managers attain values,

and that, subsequently, such value attainment of budgetary participation

will manifest itself as higher job satisfaction, which in turn, enhances job

performance. We posit that the cognitive role of budgetary participation

enhances the gathering of job-relevant information (Link 1, Fig. 1), and that

the value attainment effect of budgetary participation increases subordi-

nates’ levels of job satisfaction (Link 2, Fig. 1). In addition, we argue that

the availability and use of job-relevant information enhances job satisfac-

tion and job performance (Links 3 and 4, respectively, Fig. 1). Finally, we

propose that subordinates with higher levels of job satisfaction will be more

likely to perform better in their job (Link 5, Fig. 1). In summary, the cog-

nitive and value attainment roles of budgetary participation should initially

Cognitive Effect Performance Effect

Budgetary Participation

Link 2

Link 1 Job-RelevantInformation

Job Satisfaction

Job Performance

Link 5

Link 4

Link 3

Value Attainment Effect Performance Effect

Fig. 1. Theoretical Model.

Roles of Budgetary Participation on Job Performance 215

enhance job-relevant information and job satisfaction, which in turn, lead to

higher job performance.

The remainder of the chapter is organized as follows: In the next section,

the relevant literature is reviewed and hypotheses underlying the study are

developed. Subsequent sections present the research method, results, and

conclusion and limitations of the study.

RESEARCH MODEL AND THEORETICAL

DEVELOPMENT

The Cognitive Effect Hypothesis: The Relationship between Budgetary

Participation and Job-Relevant Information

The first hypothesis is concerned with the relationship between budgetary

participation and job-relevant information (Link 1 of Fig. 1).2 Prior ac-

counting literature suggests that budgetary participation provides an op-

portunity for subordinates to gather job-relevant information to facilitate

their decision-making process (Kren & Liao, 1988; Chenhall & Brownell,

1988; Kren, 1992; Magner, Welker, & Campbell, 1996). For example, Kren

and Liao (1988) suggest that participation provides cognitive benefits, which

enable the subordinates to clarify and comprehend the means by which

objectives can be fulfilled. Chenhall and Brownell (1988), on the other hand,

claim that budgetary participation allows managers to acquire job-relevant

information that assists and clarifies their role expectations, the methods

used in fulfilling their role expectations, or the consequences of role per-

formance. Kren (1992) and Magner et al. (1996) find that budgetary par-

ticipation is positively associated with job-relevant information. The

psychological literature (see Latham & Saari, 1979; Campbell & Gingrich,

1986) also indicates that budgetary participation has a positive and direct

effect on job-relevant information.

In summary, the above literature review and empirical evidence suggest

that budgetary participation serves as a cognitive function by enabling

managers to obtain, exchange, and disseminate job-relevant information.

Hence, this study postulates that the cognitive role of budgetary participa-

tion is positively associated with job-relevant information. The following

hypothesis is tested:

H1. Budgetary participation is positively associated with job-relevant

information.


The Value Attainment Effect Hypothesis: The Relationship between

Budgetary Participation and Job Satisfaction

The second hypothesis is concerned with the relationship between budg-

etary participation and job satisfaction (Link 2 of Fig. 1). The notion of the

value attainment role of budgetary participation suggests that allowing

subordinates to participate in the budget-setting process will increase the

likelihood that they will feel satisfied with their values (French et al., 1960;

Strauss, 1963; Lowin, 1968; Locke & Schweiger, 1979).3 Prior studies (e.g.

Chenhall & Brownell, 1988; Lau & Tan, 2003) support the value attainment

role of budgetary participation. For example, Chenhall and Brownell

(1988, p. 231) conclude that ‘‘yparticipation is most helpful in decreasing

managers’ role ambiguity, and that decreased role ambiguity improves

job satisfaction.’’ Chenhall and Brownell attribute their findings to the

managers’ opportunity to participate in a budget-setting process, which

allows them to attain their values (i.e. to reduce the level of role ambiguity).

Consequently, these managers felt highly satisfied with their job because

of such value attainment. Similarly, Lau and Tan (2003) find that bud-

getary participation is likely to improve subordinates’ levels of job satis-

faction because allowing them to get involved in the budget-setting

process enhances their ability to meet their budget targets (i.e. to meet their

values).

In addition, subordinates who participate in the budget-setting process

may experience feelings of dignity and self-respect (Cherrington, 1980;

Shields & Shields, 1998). Shields and Shields (1998) suggest that budgetary

participation helps to increase the subordinates’ self-esteem, morale, and to

enhance their job satisfaction. They claim that the positive relationship be-

tween budgetary participation and job satisfaction is due to the fact that

‘‘ythe act of participation allows a subordinate to experience self-respect

and feelings of equality arising from the opportunity to express his or her

values’’ (Shields & Shields, 1998, p. 59). Furthermore, the opportunity to

participate in the budget-setting process may cause subordinates to feel that

their jobs are more fulfilling and induce them to exert greater work-related

effort (see Deci & Ryan, 1985). Prior research suggests that the exertion of

effort in the job itself provides fulfillment of peoples’ intrinsic needs to be

competent and effective, hence contributing to job satisfaction (Aronson &

Mills, 1959; Cardozo, 1965; Emmons, 1986).

In summary, the above discussion suggests that the value attainment role

of budgetary participation is expected to increase subordinates’ levels of job

satisfaction. Thus, the formal hypothesis is as follows:


H2. Budgetary participation is positively associated with job satisfaction.

The Relationship between Job-Relevant Information and Job Satisfaction

The third hypothesis is concerned with the relationship between job-relevant

information and job satisfaction (Link 3 of Fig. 1). A review of the literature

suggests that job-relevant information is related to subordinates’ job sat-

isfaction (see O’Reilly & Caldwell, 1979; White & Mitchell, 1979; Griffin,

1983; Lau & Tan, 2003). Lau and Tan (2003, p. 23), for example, find that

‘‘yjob-relevant information, which promotes feeling of success through

successful task completion, was associated with improved subordinates’ job

satisfaction.’’ Lau and Tan’s argument was based on the needs-satisfaction

model proposed by Salancik and Pfeffer (1977, 1978), which suggests that

the expectation of feeling success (failure) is associated with feelings of sat-

isfaction (dissatisfaction). In addition, the needs-satisfaction model suggests

that needs fulfillment leads to increased job satisfaction. Specifically, it is

argued that ‘‘yjobs which fulfill a person’s needs are satisfying; those that

do not are not satisfying’’ (Salancik & Pfeffer, 1977, p. 428).

In summary, the more job-relevant information that subordinates have

about how to perform their job, the more job satisfaction they will have.

Thus, the following hypothesis is tested:

H3. Job-relevant information is positively associated with job satisfaction.

The Relationship between Job-Relevant Information and Job Performance

The fourth hypothesis is concerned with the relationship between job-

relevant information and job performance (Link 4 of Fig. 1). Job-relevant

information gathered through a participative process could enhance an in-

dividual’s ability to perform (Beehr & Love, 1983). The existing literature

suggests that the use of job-relevant information enhances job performance

(see e.g. Campbell & Gingrich, 1986; Kren, 1992; Chong & Chong, 2002).

Kren (1992), for example, finds that job-relevant information is positively

associated with job performance. He attributes his results to the fact that

job-relevant information helped subordinates to improve their action choic-

es through better-informed effort, and consequently, improved perform-

ance. Kren (1992, p. 512) summarizes the usefulness of job-relevant

information for decision making as follows:


yJob-relevant information can improve performance because it allows more accurate

predictions of environmental states and thus allows more effective selection of appro-

priate courses of action.

Chong and Chong (2002) conclude that the use of job-relevant information

improves subordinates’ job performance because job-relevant information

allows them to improve their decision choices. In summary, the above dis-

cussion suggests that the higher the level of job-relevant information, the

higher the level of subordinates’ job performance. Hence, the following

hypothesis is tested:

H4. Job-relevant information is positively associated with job performance.

The Relationship between Job Satisfaction and Job Performance

The fifth hypothesis is concerned with the relationship between job satis-

faction and job performance (Link 5 of Fig. 1). Recall that a primary ob-

jective of this study is to demonstrate the value attainment role of budgetary

participation on subordinates’ job performance. Numerous prior account-

ing studies (see e.g. Choo & Tan, 1997; Poznanski & Bline, 1997) have

demonstrated that job satisfaction is an antecedent to job performance. For

example, Choo and Tan (1997) find that job satisfaction mediates the re-

lationship between disagreement in budgetary performance evaluation style

and job performance. Specifically, they find that when subordinates’ pre-

ferred budgetary performance evaluation styles differed from their superi-

ors’ preferred budgetary performance evaluation styles, this disagreement

leads to lower levels of job satisfaction and poorer job performance amongst

subordinates. A dissatisfied subordinate may be more likely than a highly

satisfied subordinate to decide simply not to perform in his or her job (see

Franken, 1982). For example, Franken (1982, p. 451) claims that:

Job dissatisfaction is an important issue because it has been linked toythe decision

simply to not perform. ydissatisfaction is likely to lead to ysimply poor performance.

In addition, a subordinate who is satisfied with his or her job is assumed to

perform better (Locke, 1986; Katzell, Thompson, & Guzzo, 1992). Katzell

et al. (1992, p. 198) suggest that job satisfaction should be regarded as an

attitudinal state of arousal that disposes one to exert effort. These studies

attribute such findings (i.e. that there is a positive relationship between job

satisfaction and job performance), to the fact that highly satisfied subor-

dinates are more likely to exert additional effort to perform. Effort refers to

‘‘the amount of energy spent on [an] act per unit of time’’ (Naylor,


Pritchard, & Ilgen, 1980, p. 6). It represents the force, energy, or activity by

which work is accomplished (see Naylor et al., 1980; Ilgen & Klein, 1988).

Brown and Peterson (1994) found that increased effort leads to improved

job performance.4

In summary, the above discussion suggests that the higher the levels of job

satisfaction, the higher the levels of subordinates’ job performance. Thus,

the following hypothesis is tested:

H5. Job satisfaction is positively associated with job performance.

RESEARCH METHOD

The data collection method used in this study was a mailed survey ques-

tionnaire. This method was employed as it enables a large random sample to

be tested, which enhances external validity. A total of 141 senior-level man-

agers from firms in the financial services sector were randomly drawn from

the Kompass Australia (1999) business directory. Telephone calls were made

to ensure that each of the senior-level managers selected would receive the

questionnaire, and would be the person involved in answering the ques-

tionnaire. Most importantly, these managers were called to ensure that they

held budget responsibilities in their respective firms. Respondents were in-

volved in the preparation of budgets for planning and control purposes.

Since all of the respondents occupied positions of high responsibility and

accountability, their role of budgetary participation in our sample is not one

of pseudo-participation.5

Each participant was sent a survey questionnaire with a covering letter

explaining the objective of the study and a reply paid self-addressed enve-

lope. To enhance our response rate, each respondent was promised a gift

voucher of 15 Australian dollars for returning the completed questionnaire.

Each questionnaire was pre-coded to enable non-respondents to be traced

and follow-up to be executed. A follow-up letter and another copy of the

questionnaire were sent to those people who had not responded after four

weeks. The response rate to the mail-out was 54%.6 Of the 77 question-

naires, seven were excluded from the study because they were considered

outliers.7 This resulted in 70 usable responses for the final data analysis. The

organizations surveyed have an average of 908 employees. Each manager,

on average, was responsible for 98 employees. The average age of each

participant was 42 years. The average length of time spent in the position


was nearly 4 years and each manager had, on average, 10 years experience in

his or her current area of responsibility.

Measurement of Variables

Budgetary participation was measured by a three-item scale, similar to those

used by Kren (1992) and Magner et al. (1996), which was originally devel-

oped by Milani (1975). These three similar items were chosen to insure that

the budgetary participation scale would be comparable with those of Kren

(1992) and Magner et al. (1996). The three items were: (1) ‘‘To what extent

are you involved in setting your budget?’’, (2) ‘‘How much influence do you

feel you have on the final budget?’’, and (3) ‘‘How do you view your con-

tribution to the budget?’’. Each item was measured on a 7-point Likert-type

scale (scored 1–7), with higher values on the item scale indicating higher

budgetary participation. Cronbach’s alpha coefficient (Cronbach, 1951) ob-

tained for this scale was 0.89, which indicates high internal reliability for the

scale (Nunnally, 1967).

Job-relevant information was measured using Kren’s (1992) three-item, 7-

point Likert-type scale. The objective of this measure is to assess the extent

to which managers’ perceived information availability is necessary for the

purposes of evaluating important decision alternatives and making effective

job-related decisions. The scale ranges from 1 (strongly disagree) to 7

(strongly agree). The three items were: (1) ‘‘I am always clear about what is

necessary to perform well on my job’’, (2) ‘‘I have adequate information to

make optimal decisions to accomplish my performance objectives’’, and (3)

‘‘I am able to obtain the strategic information necessary to evaluate im-

portant decision alternatives’’. Cronbach’s alpha coefficient obtained for

this scale was 0.77, which indicates satisfactory internal reliability for the

scale.

Job satisfaction was measured by a two-item, 7-point Likert-type scale

developed by Dewar and Werbel (1979). The two items were: (1) ‘‘All in all,

I am satisfied with my job’’, and (2) ‘‘In general, I like working in this

company’’. Cronbach’s alpha coefficient obtained for this scale was 0.91,

which indicates high internal reliability for the scale.

A single-item, 7-point Likert-type scale was used to measure job per-

formance as it was consistent with numerous prior accounting studies (e.g.

Merchant, 1981, 1984; Mia & Chenhall, 1994; Dunk, 1995). Respondents

were asked to rate their overall performance from ‘‘well below average’’ to

‘‘well above average’’ on a fully anchored 7-point Likert-type scale.


RESULTS

Table 1 shows the descriptive statistics and Pearson correlation matrix for

the variables used in the study.

To test the theoretical model (Fig. 1) and associated hypotheses, a struc-

tural equation modeling (SEM) technique was used. The SEM technique

allows us to simultaneously test the entire budgetary participation–per-

formance structural model. It also allows the specification of more complex

models and makes allowances for errors in measurement. The SEM tech-

nique used in this study was based on the EQS structural equation computer

program (Bentler, 1995). An approach recommended by Anderson and

Gerbing (1988) was utilized, which involves two steps. First, the measure-

ment model was evaluated by confirmatory factor analysis. Second, on the

basis of the results of the measurement model analysis, only those items

reflecting a common construct were aggregated to derive unidimensional

composite scales for the structural model tests (Anderson & Gerbing, 1988).

Analysis of the Measurement Model

The first step of the analysis is to develop a measurement model. The

measurement model consists of three factors (budgetary participation, job-

relevant information, and job satisfaction). The w2 statistic and fit indices are

summarized in Table 2, panel A. Although, the w2 statistic has been the most

Table 1. Descriptive Statistics and Pearson Correlation Matrix

ðn ¼ 70Þ.

Variable Actual (Theoretical) Range Mean S.D. 1 2 3 4

1. Budgetary participation 2.00–7.00 5.16 1.58 1.00

(1.00–7.00)

2. Job satisfaction 1.00–7.00 5.53 1.22 0.39�� 1.00

(1.00–7.00)

3. Job-relevant information 2.33–7.00 5.39 1.06 0.24� 0.32�� 1.00

(1.00–7.00)

4. Job performance 4.00–7.00 6.08 0.81 0.25� 0.36�� 0.44�� 1.00

(1.00–7.00)

��Significant at the 0.01 level (2-tailed).�Significant at the 0.05 level (2-tailed).


Table 2. Results of Confirmatory Factor Analysis.

Panel A: w2 Statistics and Fit Indices

Fit Measures Recommended Values Result of This study

w2 NA 36.58

Df NA 17

w2 (p-value) 40.05 0.003

w2/df p3.00 2.15

Fit indices

CFI X0.90 0.94

NFI X0.90 0.89

NNFI X0.90 0.90

GFI X0.90 0.89

Residual analysis

RMSEA p0.10 0.13

AOSR p0.05 0.06

Panel B: Standardized Loading, Composite Reliability, Variance Extracted Estimate and

Cronbach’s Alpha

Variable Standardized

Loading

Composite

Reliability

Variance

Extracted

Estimate

Cronbach’s

Alpha

Budgetary

participation

0.89 0.73 0.89

BP1 0.78�

BP2 0.93�

BP3 0.85�

Job-relevant

information

0.87 0.66 0.77

JRI1 0.86�

JRI2 0.95�

JRI3 0.59�

Job satisfaction 0.91 0.84 0.91

JS1 0.85�

JS2 0.98�

CFI ¼ Comparative Fit Index, higher values indicate better fit.

NFI ¼ Normed Fit Index, higher values indicate better fit.

NNFI ¼ Non-Normed Fit Index, higher values indicate better fit.

GFI ¼ Goodness-of-Fit Index, higher values indicate better fit.

RMSEA ¼ Root Mean Square of Approximate, lower values indicate better fit.

AOSR ¼ Average Off-Diagonal Standardized Residual, lower values indicate better fit.�All standardized loadings are significant at po0.05.


often used indicator for model fit, it is not as useful as other fit indicators

(Browne & Mels, 1994). Bollen and Long (1993) suggest the use of multiple

model fit indicators. The results shown in Table 2, panel A reveal that the w2

statistic for our measurement model does not fit the data (w2 ¼ 36:58;po0.003). However, the value of the w2 divided by the degree of freedom is

2.15, which is below the cutoff value of 3 and is indicative of model fit

(Hartwick & Barki, 1994; Segars & Grover, 1993). Further, the fit measures

such as comparative fit index (CFI) and non-normed fit index (NNFI) are

0.94 and 0.90, respectively, which meets or exceeds the 0.90minimum

threshold (Bentler & Bonett, 1980; Anderson & Gerbing, 1988; Hair et al.,

1998). The normed fit index ðNFI ¼ 0:89Þ and goodness-of-fit index

(GFI ¼ 0.89) fall marginally short of the desired 0.90 criteria. The root

mean square error of approximate (RMSEA) is 0.13, indicating a reasonable

fit, although this value is somewhat higher than desired (Segars & Grover,

1993; Hartwick & Barki, 1994; Fogarty, Singh, Rhoads, & Moore, 2000).

Finally, the average off-diagonal standardized residual is 0.06, suggesting a

marginal fit. Taken together, the fit indices seem to be adequate and re-

specification is not necessary. Furthermore, the tests for convergent validity

and discriminant validity provide further support for this decision.

Convergent Validity Test

The convergent validity of the scale is assessed by three measures: stand-

ardized loading, composite reliability, and variance extracted estimate

(Fornell & Larcker, 1981). The results shown in Table 2, panel B, reveal that

the standardized loadings for all the items of the three scales are highly

significant (po0.05). The composite reliability reflects the internal consist-

ency of the indicators measuring a given factor (Fornell & Larcker, 1981). It

is analogous to Cronbach’s (1951) alpha coefficient for measuring the re-

liability of a multiple-item scale. The composite reliability is 0.87 or greater

for each scale, indicating that the items comprising each scale are highly

correlated. Variance extracted estimates assess the amount of variance that

is captured by an underlying factor in relation to the amount of variance due

to measurement error (see Fornell & Larcker, 1981). The variance extracted

estimates for three of the factor scales exceeds the minimum level of 0.50

recommended by Fornell and Larcker (1981). Taken together, these results

provide support for the convergent validity of the three scales.

Discriminant Validity Test

Discriminant validity is inferred when measures of each construct converge

on their respective true scores, which are different from the scores of other


constructs (see Churchill, 1979). The discriminant validity of the three scales

is assessed using the procedure outlined by Bagozzi, Yi, and Phillips (1991).

Specifically, a model is estimated in which the correlation between (1)

budgetary participation and job-relevant information; (2) budgetary par-

ticipation and job satisfaction; (3) job-relevant information and job satis-

faction, respectively, are restricted to unity (i.e. the correlation is fixed at

1.0). The fits of the constrained models are then compared with those of the

original unconstrained models. The results of the discriminant validity

analysis for the three scales are shown in Table 3.

As shown in Table 3, panel A, the w2 difference tests are significant for the

models, which estimate the correlations between the budgetary participation

and job-relevant information scales (w2ð1Þ ¼ 63:45; po0.01), the budgetary

participation and job satisfaction scales (w2ð1Þ ¼ 12:26; po0.05), and job-

relevant information and job satisfaction scales (w2ð1Þ ¼ 8:16; po0.05).

These results suggest that the above-mentioned scales exhibit very strong

properties of discriminant validity. A further test for the discriminant va-

lidity of the three scales is conducted by comparing the variance extracted

estimates with the squared of the correlations between the latent constructs

(Fornell & Larcker, 1981). As shown in Table 3, panel B, the variance

extracted estimates for all constructs exceed the squared of the correlations.

These results provide strong support for the discriminant validity of the

three scales.

Analysis of the Structural Model

The hypotheses are tested by relying on the standardized parameter esti-

mates for the theoretical model as shown in Fig. 2. As expected, the results

reveal that H1, which states that budgetary participation is positively as-

sociated with job-relevant information, is statistically significant (standard-

ized path coefficient ¼ 0:24; po0.05). Thus, these results support H1 and

lend support to the cognitive effect of budgetary participation, which sug-

gests that subordinates participate in the budget-setting process to share

information. In addition, the results shown in Fig. 2 reveal that budgetary

participation is positive and statistically significantly associated with job

satisfaction (standardized path coefficient ¼ 0:34; po0.05). These results

support H2. Support is also found for H3, H4, and H5, since significantly

positive relationships are demonstrated between job-relevant information

and job satisfaction (H3), job-relevant information and job performance

(H4), and job satisfaction and job performance (H5) (standardized path


coefficients ¼ 0.24, 0.36, and 0.24, respectively po0.05). Overall, the model

accounts for 24% of the variance in job performance (R2 ¼ 0:24; see Fig. 2).In order to understand better the impact of the linkage between budgetary

participation and job performance, a model that excluded the value attain-

ment effect of budgetary participation (i.e. job satisfaction) is tested for

comparison purposes. As expected, budgetary participation is positive and

statistically significantly associated with job-relevant information (stand-

ardized path coefficient ¼ 0:36; po0.05), and job-relevant information is

also positive and significantly associated with job performance (standard-

ized path coefficient ¼ 0:45; po0.05). The cognitive effect model accounts

for 21% of the variance in job performance ðR2 ¼ 0:21Þ: Recall that the

value attainment-cognitive effect model accounts for 24% of the variance in

job performance, while the cognitive effect model alone accounts for only

21% of the variance in the job performance. Taken together, these results

reveal that the introduction of the value attainment effect into the cognitive

effect model result in a statistically significant (F change ¼ 3:06; po0.043,

1-tailed) increase in R2, suggesting that the combined value attainment and

cognitive effect improve the predictive ability of our model.

CONCLUSION AND LIMITATIONS

The main objective of this study is to test the impact of the value attainment

role of budgetary participation on job performance, and its influence within

the cognitive model. This study contributes to the participative budgeting

literature in a number of ways. First, it introduces the value attainment role

of budgetary participation to the accounting literature, and explicitly ex-

amines this role of budgetary participation on job performance. Second, it

provides empirical support to the value attainment role of budgetary par-

ticipation, which was theorized to increase subordinates’ job satisfaction.

This result is consistent with our value attainment effect hypothesis and

prior studies (e.g. Chenhall & Brownell, 1988; Chong & Bateman, 2000).

In addition, this study extends prior studies by incorporating the value

attainment role of budgetary participation into the cognitive model. The

results of this study reveal that the joint effects of the value attainment and

cognitive roles of budgetary participation significantly improved subordi-

nates’ job performance. Furthermore, the results of this study provide ad-

ditional empirical evidence to support the robustness of the findings of prior

studies that examined the cognitive role of budgetary participation (e.g.

Kren, 1992; Magner et al., 1996).


Table 3. Results of Discriminant Validity Tests.

Panel A: The w2 Difference Test

Model w2 df Dw2

Constrained model: BP-JRI 13.02 8

Unconstrained model: BP-JRI 76.47 9 63.45��

Constrained model: BP-JS 6.86 5

Unconstrained model: BP-JS 19.12 4 12.26�

Constrained model: JRI-JS 11.31 5

Unconstrained model: JRI-JS 3.15 4 8.16�

Panel B: Variance Extracted Estimate Test

Intercorrelation Squared Intercorrelation Variance Extracted Estimate

BP-JRI 0.41 0.17�� 0.73–0.66

BP-JS 0.43 0.18�� 0.73–0.84

JRI-JS 0.35 0.12�� 0.66–0.84

�po0.05.��po0.01.��The squared of the correlation is less than both variance extracted estimates.

Budgetary Participation

0.34*

0.24* Job-RelevantInformation

Job Satisfaction

Job Performance

0.24*

0.36*

0.24*

R2 = 0.06 R

2 = 0.24

R2 = 0.21

Fig. 2. Standardized Path Coefficients. (*Significant at 0.05 level. Model w2 ¼ 0:51;d.f. ¼ 1 (po0.48); Bentler-Bonnet Normed Fit Index (NFI) ¼ 0.99; Bentler-Bonnet

Nonnormed Fit Index (NNFI) ¼ 1.09; Comparative Fit Index (CFI) ¼ 1.00; Good-

ness-of-Fit Index (GFI) ¼ 0.99; Root Mean Square Error of Approximate

(RMSEA) ¼ 0.00; Average Off-Diagonal Standardized Residual (AOSR) ¼ 0.01.)


Several limitations of this study need to be noted. First, sample was se-

lected from the financial services sector. Hence, in generalizing the results to

other industries, caution should be exercised. Further studies that compare

two industries such as the manufacturing and financial services sectors

would be worthwhile. A related issue to the sample is the use of relatively

small sample size ðn ¼ 70Þ in this study. Prior studies (e.g. Bentler & Bonnet,

1980; Zimmerman, Eason, & Gowan, 1999) have criticized the problems

associated with the use of small sample size for structural equation mode-

ling. An alternative approach, path analysis, could be used as it would likely

yield similar results to structural equation modeling regarding the signif-

icance of the relations between variables. Second, this study uses a self-

rating scale to measure job performance which is likely to have resulted in

higher leniency and lower variability errors in this measure (Prien & Liske,

1962; Thornton, 1968). Thus, care should be taken in interpretation of the

results. Future studies could employ different research methods (e.g. lon-

gitudinal field studies) to investigate systematically the theoretical relation-

ships proposed in this study. In addition, future study may also consider

employing objective measures of performance (e.g. return-on-investment or

return-on-assets to measure performance).

Third, this study focuses on an examination of the value attainment and

cognitive roles of budgetary participation without considering the potential

motivational function of budgetary participation on job performance (Nouri

& Parker, 1998; Chong & Chong, 2002; Wentzel, 2002). An attempt to test

the three roles (i.e. motivational, cognitive, and value attainment) of budg-

etary participation in a single study would provide more insight into the

process as to how budgetary participation really affects job performance.

Finally, while this study tested a recursive model, a non-recursive model

might be more applicable to the situation. In other words, there could be

simultaneous links between (1) budgetary participation and job-relevant in-

formation; (2) job-relevant information and both job satisfaction and per-

formance, and (3) job satisfaction and job performance. Applying the test of a

non-recursive model was not possible in this study due to identification

problems. Future research may attempt to test for a non-recursive model of

participative budgeting.

NOTES

1. The cognitive mechanism suggests that the process of participation improvessubordinate’s performance by increasing the quality of decisions as a result of the


subordinate sharing job-relevant information with the superior (Kren, 1992; Shields& Shields, 1998, p. 59).2. Job-relevant information refers to information that assists job-related decision-

making (Kren, 1992). It is also known as ex-ante information (see Baiman, 1982),decision-facilitating information (see Demski & Feltham, 1976; Magee, 1986; Hilton,1994), and task-relevant knowledge (Murray, 1990; Wier, 1993).3. Pinder (1984, p. 95) suggests that ‘‘yvalues are those things that a person

believes are conducive to his/her welfare’’, while Locke (1983, p. 1034) claims that‘‘ya value is what a person consciously or subconsciously desires, wants, or seeks toattain’’. As noted earlier, subordinates’ values may include: (1) the opportunity toexpress their views, (2) the feeling of being treated equally, or (3) the desire forrespect or dignity (Argyris, 1955; Davis, 1957). It is suggested that values play animportant role in determining job satisfaction (Katzell, 1964).4. In general, it is suggested that increased effort can either lead to immediate

performance increases if it is directed toward current performance, or lead to delayedperformance increases if it is directed toward learning (see Bonner & Sprinkle, 2002).The focus of this study is to investigate the increased effort, which is directed towardcurrent performance (i.e. immediate performance), rather than directed towardlearning (i.e. delayed performance).5. Pseudo-participation refers to a budget-setting process in which subordinates

are involved, but the superior makes the final decision. It is a consultative-typebudgeting process in which the subordinate’s input to the budget is being ignored(Argyris, 1952; see also Pasewark & Welker, 1990). Pasewark and Welker (1990)suggest pseudo-participation can have a de-motivating effect on subordinates.6. We tested for non-response bias using the approach suggested by Oppenheim

(1966, p. 34). No statistically significant differences in the mean scores between theearly and late responses were found.7. The seven responses considered as outliers were from individuals in companies

that employed substantially more people than the other firms. These seven firmsemployed a range of 10,000–140,000 employees. We conducted a univariate assess-ment of the values of the standardized scores, which revealed that all these sevenresponses exceeded the recommended threshold standardized values (Z scores) rangefrom 73 to 74 (Hair, Anderson, Tatham, & Black. 1998, p. 65). The structuralequation modeling (SEM) analyses were repeated before the exclusion of the sevenoutliers. The results revealed that there were no differences between SEM resultsbased on 77 (before exclusion of the seven responses) and those based on 70 re-sponses. This implies that our results are relatively robust to variations in the size(i.e. number of employees) of the sample.

ACKNOWLEDGEMENTS

The authors appreciate the helpful comments and suggestions of

Vicky Arnold (Editor), the Associate Editor, two anonymous reviewers,

and seminar participants at York University, Toronto on the earlier drafts

of this chapter. An earlier version of this paper was presented at the 2001


Asian-Pacific Conference on International Accounting Issues, Rio de

Janeiro, Brazil.

REFERENCES

Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: a review

and recommended two-step approach. Psychological Bulletin, 103(3), 411–423.

Argyris, C. (1952). The impact of people on budgets. New York: Controllership Foundation.

Argyris, C. (1955). Organizational leadership and participative management. Journal of Busi-

ness, 28, 1–7.

Aronson, E., Mills, J. (1959). The effect of severity of initiation on liking for a group. Journal of

Abnormal and Social Psychology, (September), 59, 177–181.

Bagozzi, R. P., Yi, Y., & Phillips, L. W. (1991). Assessing construct validity in organizational

research. Administrative Science Quarterly, 36, 421–458.

Baiman, S. (1982). Agency research in managerial accounting. Journal of Accounting Literature,

1, 154–213.

Beehr, T. A., & Love, K. G. (1983). A meta-model of the effects of goal characteristics, feed-

back, and role characteristics in human organizations. Human Relations, 36(2), 151–166.

Bentler, P. M. (1995). EQS structural equation program manual. Encino, CA: Multivariate

Software, Inc.

Bentler, P. M., & Bonnet, D. G. (1980). Significance tests and goodness-of-fit in the analysis of

covariance structure. Psychological Bulletin, 88(3), 588–600.

Bollen, K., & Long, I. S. (1993). Tests for structural equation models. Newbury Park, CA: Sage

Publication.

Bonner, S. E., & Sprinkle, G. B. (2002). The effects of monetary incentives on efforts and task

performance: theories, evidence, and a framework for research. Accounting, Organiza-

tions and Society, 27, 303–345.

Brown, S. P., & Peterson, R. A. (1994). The effect of effort on sales performance and job

satisfaction. Journal of Marketing, 70, 74–80.

Browne, M., & Mels, G. (1994). RAMONA users’ guide. Columbus, OH: The Ohio State

University.

Campbell, D. J., & Gingrich, K. F. (1986). The interactive effects of task complexity and

participation on task performance: a field experiment. Organizational Behavior and Hu-

man Decision Processes, 38(2), 162–180.

Cardozo, R. N. (1965). An experimental study of consumer effort, expectation, and satisfaction.

Journal of Marketing Research, 2, 244–249.

Chenhall, R. H. (1986). Authoritarianism and participative budgeting: a dyadic analysis. The


Chenhall, R. H., & Brownell, P. (1988). The effect of participative budgeting on job satisfaction

and performance: role ambiguity as an intervening variable. Accounting, Organizations

and Society, 13(3), 225–233.

Cherrington, D. J. (1980). The work ethic: working values and values that work. New York:

Amacom.

Chong, V. K., & Bateman, D. (2000). The effects of role stress on budgetary participation and

job satisfaction-performance linkages: a test of two different models. Advances in Ac-

counting Behavioral Research, 3, 91–118.


Chong, V. K., & Chong, K. M. (2002). Budget goal commitment and informational effects of

budget participation on performance: a structural equation modeling approach. Be-

havioral Research in Accounting, 14, 65–86.

Choo, F., & Tan, K. B. (1997). A study of the relations among disagreement in budgetary

performance evaluation style, job-related tension, job satisfaction and performance.

Behavioral Research in Accounting, 9, 199–218.

Churchill, G. A. (1979). A paradigm for developing better measures of marketing constructs.

Journal of Marketing Research, 16, 64–73.

Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16,

297–334.

Covaleski, M. A., Evans, J. H., Luft, J. L., & Shields, M. D. (2003). Budgeting research: three

theoretical perspectives and criteria for selective integration. Journal of Management

Accounting Research, 15, 3–49.

Davis, K. (1957). Management by participation: its place in today’s business world. Manage-

ment Review, 46, 69–79.

Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self determination in human be-

haviour. New York: Plenum.

Demski, J. S., & Feltham, G. A. (1976). Cost determination: a conceptual approach. Ames: Iowa

State University Press.

Dewar, R., & Werbel, J. (1979). Universalistic and contingency predictions of employee sat-

isfaction and conflict. Administrative Science Quarterly, 24(3), 426–448.

Dunk, A. S. (1995). The joint effects of budgetary slack and task uncertainty on subunit

performance. Accounting and Finance, 35(2), 61–75.

Emmons, R. A. (1986). Personal strivings: an approach to personality and subjective well-being.

Journal of Personality and Social Psychology, 51, 1058–1068.

Fogarty, T. J., Singh, J., Rhoads, G. K., & Moore, R. K. (2000). Antecedents and consequences

of burnout in accounting: beyond the role stress model. Behavioral Research in Ac-

counting, 12, 31–67.

Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation models with observable

variables and measurement error. Journal of Marketing Research, 18, 39–50.

Franken, R. E. (1982). Human motivation. Monterey, CA: Brooke/Cole Publishing Company.

French, J. R. P., Jr., Israel, J., & As, D. (1960). An experiment on participation on Norwegian

factory: Interpersonal dimensions of decision-making. Human Relations, 13, 3–19.

Griffin, R. W. (1983). Objective and social sources of information in task redesign: a field

experiment. Administrative Science Quarterly, 28, 184–200.

Hair, J. F., Jr., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data

analysis with readings. Englewood Cliffs, NJ: Prentice-Hall International, Inc.

Hartwick, J., & Barki, H. (1994). Explaining the role of user participation in information system

use. Management Science, 40(4), 440–465.

Hilton, R. W. (1994). Managerial accounting. New York: McGraw-Hill, Inc.

Ilgen, D. R., & Klein, H. J. (1988). Individual motivation and performance: Cognitive influ-

ences on effort and choice. In: J. P. Campbell & R. J. Campbell (Eds), Productivity in

organizations (pp. 143–176). San Francisco: Jossey-Bass.

Katzell, R. A. (1964). Personal values, job satisfaction, and job behavior. In: H. Borow (Ed.),

Man in a world at work. Boston: Houghton Mifflin.

Katzell, R. A., Thompson, D. E., & Guzzo, R. A. (1992). How job satisfaction and job per-

formance are and are not linked. In: C. J. Cranny, P. C. Smith & E. F. Stone (Eds), Job


Satisfaction: How people feel about their jobs and how it affects their performance.

New York: Lexington Books.

Kompass Australia. (1999). Prahran: Peter Isaacson.

Kren, L. (1992). Budgetary participation and managerial performance: The impact of infor-

mation and environmental volatility. The Accounting Review, 67(3), 511–526.

Kren, L., & Liao, W. M. (1988). The role of accounting information in the control of organ-

izations: a review of the evidence. Journal of Accounting Literature, 7, 280–309.

Latham, G. P., & Saari, L. M. (1979). Importance of supportive relationships in goal setting.

Journal of Applied Psychology, 64, 151–156.

Lau, C. M., & Tan, S. L. C. (2003). The effects of participation and job-relevant information on

the relationship between evaluative style and job satisfaction. Review of Quantitative

Finance and Accounting, 21(17), 17–34.

Locke, E. A. (1983). The nature and causes of job satisfaction. In: M. D. Dunnette (Ed.),

Handbook of industrial and organizational psychology. New York: Wiley.

Locke, E. A. (1986). Job attitudes in historical perspective. In: D. Wren (Ed.), Paper dedicated

to the development of modern management. Briarcliff Manor, NY: Academy of Man-

agement.

Locke, E. A., & Latham, G. P. (1990). A theory of goal setting and task performance. Englewood

Cliffs, NJ: Prentice-Hall.

Locke, E. A., & Schweiger, D. M. (1979). Participation in decision-making: one more look.

Research in Organizational Behavior, 1, 265–339.

Lowin, A. (1968). Participative decision making: A model, literature critique, and prescriptions

for research. Organizational Behavior and Human Performance, 3, 68–106.

Magee, R. P. (1986). Advanced management accounting. New York: Harper & Row.

Magner, N., Welker, R. B., & Campbell, T. L. (1996). Testing a model of cognitive budgetary

participation processes in a latent variable structural equations framework. Accounting

and Business Research, 27(1), 41–50.

Merchant, K. A. (1981). The design of the corporate budgeting system: influences on man-

agerial behavior and performance. The Accounting Review, 56(4), 813–829.

Merchant, K. A. (1984). Influences on departmental budgeting: an empirical examination of a

contingency model. Accounting, Organizations and Society, 9(3/4), 291–307.

Mia, L., & Chenhall, R. H. (1994). The usefulness of management accounting systems, func-

tional differentiation and managerial effectiveness. Accounting, Organizations and So-

ciety, 19(1), 1–13.

Milani, K. (1975). The relationship of participation in budget setting to industrial

supervisor performance and attitudes: a field study. The Accounting Review, 50(2),

274–284.

Murray, D. (1990). The performance effects of participative budgeting: An integration of in-

tervening and moderating variables. Behavioral Research in Accounting, 2, 104–123.

Naylor, J. C., Pritchard, R. D., & Ilgen, D. R. (1980). A theory of behavior in organizations.

New York: Academic Press.

Nouri, H., & Parker, R. J. (1998). The relationship between budgetary participation and job

performance: The roles of budget adequacy and organizational commitment. Account-

ing, Organizations and Society, 23(5/6), 467–483.

Nunnally, J. C. (1967). Psychometric theory. New York: McGraw-Hill.

Oppenheim, A. N. (1966). Questionnaire design and attitudes measurement. London: Heinemann

Educational Books, Ltd.


O’Reilly, C. A., & Caldwell, D. F. (1979). Informational influence as a determinant of perceived

task characteristics and job satisfaction. Journal of Applied Psychology, 64, 157–165.

Pasewark, W., & Welker, R. (1990). A Vroom-Yetton evaluation of subordinate participation

in budgetary decision making. Journal of Management Accounting Research, 2, 113–126.

Pinder, C. C. (1984). Work motivation. Glenview, IL: Scott Foresman.

Prien, E. P., & Liske, R. E. (1962). Assessment of higher-level personnel: III. Rating criteria:

A comparative analysis of supervision ratings and incumbent self-rating of job per-

formance. Personnel Psychology, 15, 187–194.

Poznanski, P. J., & Bline, D. M. (1997). Using structural equation modeling to investigate the

causal ordering of job satisfaction and organizational commitment among staff ac-

countants. Behavioral Research in Accounting, 9, 154–171.

Salancik, G. R., & Pfeffer, J. (1977). An examination of need-satisfaction model of job design.

Administrative Science Quarterly, 22, 427–456.

Salancik, G. R., & Pfeffer, J. (1978). A social information process approach to job attitudes and

task design. Administrative Science Quarterly, 23, 224–253.

Segars, A. H., & Grover, V. (1993). Re-examining perceived ease of use and usefulness:

A confirmatory factor analysis. MIS Quarterly, 17(4), 517–525.

Shields, J. F., & Shields, M. D. (1998). Antecedents of participative budgeting. Accounting,


Strauss, G. (1963). Some notes on power-equalization. In: H. Leavitt (Ed.), The social science of

organizations: Four perspectives. Englewood Cliffs, NJ: Prentice-Hall, Inc.

Thornton, G. C. (1968). The relationship between supervisory and self-appraisals of executive

performance. Personnel Psychology, 21, 441–456.

Wentzel, K. (2002). The influence of fairness perceptions and goal commitment on managers’

performance in a budget setting. Behavioral Research in Accounting, 14, 247–271.

White, S. E., & Mitchell, T. M. (1979). Job enrichment versus social cues: A comparison and

competitive test. Journal of Applied Psychology, 64, 1–9.

Wier, B. (1993). The effects of task relevant knowledge, goal level, goal commitment, and

motivation on the participation-performance linkage: an empirical examination.

Unpublished Ph.D. Thesis, Texas Tech University.

Zimmerman, R. A., Eason, P., & Gowan, M. (1999). Taxpayer performance between income

tax and consumption tax: Behavioral life cycle effects. Behavioral Research in Account-

ing, 11, 111–142.



234

advances in accounting behavioral research

Documents

Transcript of advances in accounting behavioral research