3,474 1,393 1MB
Pages 249 Page size 431 x 649 pts Year 2007
ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH
i
ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH Series Editor: Vicky Arnold Recent Volumes: Volumes 1–4: Series Editor: James E. Hunton Volumes 5–8: Series Editor: Vicky Arnold
ii
ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH
VOLUME 9
ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH EDITED BY
VICKY ARNOLD Dixon School of Accounting, University of Central Florida, 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 JAI Press is an imprint of Elsevier
iii
JAI Press is an imprint of Elsevier The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands 525 B Street, Suite 1900, San Diego, CA 92101-4495, USA First edition 2006 Copyright r 2006 Elsevier Ltd. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN-13: 978-0-7623-1353-2 ISBN-10: 0-7623-1353-6 ISSN: 1475-1488 (Series) For information on all JAI Press publications visit our website at books.elsevier.com Printed and bound in The Netherlands 06 07 08 09 10 10 9 8 7 6 5 4 3 2 1
iv
CONTENTS LIST OF CONTRIBUTORS
vii
REVIEWER ACKNOWLEDGEMENTS
ix
EDITORIAL POLICY AND SUBMISSION GUIDELINES
xi
THE IMPACT OF ACCOUNTABILITY ON THE PROCESSING OF NONDIAGNOSTIC EVIDENCE Michael Favere-Marchesi and Karen V. Pincus
1
AUDITORS’ MEMORY OF INTERNAL CONTROL INFORMATION: THE EFFECT OF DOCUMENTATION PREPARATION VERSUS REVIEW Lori S. Kopp and James L. Bierstaker
27
INTERNAL AUDITOR BURNOUT: AN EXAMINATION OF BEHAVIORAL CONSEQUENCES Timothy J. Fogarty and Lawrence P. Kalbers
51
FAIRNESS, BUDGET SATISFACTION, AND BUDGET PERFORMANCE: A PATH ANALYTIC MODEL OF THEIR RELATIONSHIPS Adam S. Maiga
87
v
vi
CONTENTS
UNDERSTANDING INVESTMENT EXPERTISE AND FACTORS THAT INFLUENCE THE INFORMATION PROCESSING AND PERFORMANCE OF INVESTMENT EXPERTS Kinsun Tam, James L. Bierstaker and Inshik Seol
113
THE INFLUENCE OF OUTCOME KNOWLEDGE ON JUDGES AND JURORS’ EVALUATIONS OF AUDITOR DECISIONS: A REVIEW AND SYNTHESIS OF PRIOR RESEARCH D. Jordan Lowe and Philip M. J. Reckers
157
WHY YOU SHOULD CONSIDER SEM: A GUIDE TO GETTING STARTED Cindy Blanthorne, L. Allison Jones-Farmer and Elizabeth Dreike Almer A POSTMODERN STAKEHOLDER ANALYSIS OF TELEWORK Anita Reed, James E. Hunton and Carolyn Strand Norman
179
209
LIST OF CONTRIBUTORS Elizabeth Dreike Almer
School of Business Administration, Portland State University, USA
James L. Bierstaker
Department of Accountancy, Villanova University, USA
Cindy Blanthorne
Department of Accounting, University of North Carolina at Charlotte, USA
Michael Favere-Marchesi
Faculty of Business Administration, Simon Fraser University, Canada
Timothy J. Fogarty
Department of Accountancy, Case Western Reserve University, USA
James E. Hunton
Department of Accounting, Bentley College, USA
L. Allison Jones-Farmer
Department of Management, Auburn University, USA
Lawrence P. Kalbers
Department of Accountancy, Loyola Marymount University, USA
Lori S. Kopp
Faculty of Management, University of Lethbridge, Canada
D. Jordan Lowe
School of Global Management and Leadership, Arizona State University, USA
Adam S. Maiga
Department of Accounting, University of Wisconsin-Milwaukee, USA
Carolyn Strand Norman
Department of Accounting, Virginia Commonwealth University, USA vii
viii
LIST OF CONTRIBUTORS
Karen V. Pincus
Sam Walton School of Business, University of Arkansas, USA
Philip M. J. Reckers
W.P. Carey School of Business, Arizona State University, USA
Anita Reed
School of Accountancy, University of South Florida, USA
Inshik Seol
Graduate School of Management, Clark University, USA
Kinsun Tam
Department of Accounting, University at Albany, State University of New York, 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.
Mohammed Abdolmohammadi Bentley College, USA
Janne Chung York University, Canada
Jillian Alderman University of Central Florida, USA
Peggy Dwyer University of Central Florida, USA
Elizabeth Dreike Almer Portland State University, USA
Dann Fisher Kansas State University, USA
Richard Baker Adelphi University, USA
Timothy J. Fogarty Case Western Reserve University, USA
Philip Beaulieu University of Calgary, Canada
Clark Hampton University of Central Florida, USA
Dennis M. Bline Bryant College, USA
Julia Higgs Florida Atlantic University, USA
Wray Bradley University of Tulsa, USA
Karen L. Hooks Florida Atlantic University, USA
Gary Braun University of Texas El Paso, USA
Kathy Hurtt Baylor University, USA
Vincent Chong The University of Western Australia, Australia
Stacy Kovar Kansas State University, USA ix
x
REVIEWER ACKNOWLEDGEMENTS
Ganesh Krishnamoorthy Northeastern University, USA
Ed O’Donnell University of Kansas, USA
Tanya Lee University of North Texas, USA
Will Quilliam University of South Florida, USA
Theresa Libby Wilfred Laurier University, Canada
Randall Rentfro Florida Atlantic University, USA
Tim Lindquist The University of Northern Iowa, USA
Steve Salter University of Cincinnati, USA
Adam S. Maiga University of Wisconsin-Milwaukee, USA Mario Maletta Northeastern University, USA Maureen Mascha Marquette University, USA James Maroney Northeastern University, USA
Scott Summers Brigham Young University, USA Steve Sutton University of Central Florida, USA Sally Webber Northern Illinois University, USA Patrick Wheeler University of Missouri, USA
Robyn Moroney Monash University, Australia
Bernard Wong-On-Wing Washington State University, USA
Rob Nieschwietz University of Colorado Denver, USA
Alex Yen Suffolk University, USA
EDITORIAL POLICY AND SUBMISSION GUIDELINES Advances in Accounting Behavioral Research (AABR) publishes articles encompassing all areas of accounting that incorporate theory from and contribute 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 revisions, 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 document 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 Kenneth G. Dixon School of Accounting University of Central Florida P. O. Box 161400 Orlando, FL 32816-1400, USA xi
xii
EDITORIAL POLICY AND SUBMISSION GUIDELINES
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 Hampton (2005).’’ In the text, use the form McCall et al. (2006) 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. (Bobek, 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 manuscript. Only articles referenced in the text should be included in the reference 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 selectionsocialization 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 (Eds), Judgment and decision making in accounting and auditing (pp. 207–230). New York: Cambridge University Press.
THE IMPACT OF ACCOUNTABILITY ON THE PROCESSING OF NONDIAGNOSTIC EVIDENCE Michael Favere-Marchesi and Karen V. Pincus ABSTRACT Previous research on auditors’ processing of nondiagnostic evidence demonstrates the existence of a dilution effect – the tendency to underreact to diagnostic information when accompanied by nondiagnostic information. Prior audit studies find that accountability, a prominent feature in audit settings, does not affect the magnitude of the dilution effect exhibited by auditors. Based on more recent theories about accountability, this line of research is extended by exploring whether (1) the dilution effect previously identified is a robust phenomenon that can be replicated, (2) accountability has an impact on both the frequency and magnitude of dilution effect, and (3) the impact of accountability on both the frequency and magnitude of dilution effect is conditional on the degree of accountability experienced by the participants through various reporting levels. The experimental results from a sample of internal auditors provide evidence supporting the first two propositions; however, the results related to reporting levels are not significant. A discussion of the implications of these findings for audit research and practice follows.
Advances in Accounting Behavioral Research, Volume 9, 1–25 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09001-6
1
2
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
INTRODUCTION Over 30 years of research in social psychology has examined the role of accountability on judgmental biases. Accountability refers to the implicit or explicit expectation that individuals may be called on to justify their judgments and decisions to others (Scott & Lyman, 1968; Semin & Manstead, 1983; Tetlock, 1992). In experimental settings, accountability can be removed by instructions that ensure participants’ confidentiality. However, accountability is a prominent feature in various business settings, including the audit environment. As a normal part of the audit process, individual auditors’ judgments and decisions are subject to review at multiple levels before audit conclusions are reported. Parties outside of the audit team also monitor auditors’ performance (e.g., boards of directors and regulators). Thus, the role of accountability is an important topic in studies of audit judgment and decisionmaking. The social contingency model of judgment and choice (Tetlock, 1992) is built on the premise that decision-makers function in social settings where they feel accountable to others, in varying degrees and ways. The social contingency model posits that, given certain conditions, accountability will motivate decision makers to become more complex, nuanced, and differentiated thinkers as they anticipate objections and engage in pre-emptive self-criticism (Tetlock, 1983a; Tetlock & Boettger, 1989); similar observations are made in accounting (see Messier & Quilliam, 1992) and other business settings (e.g., Fandt, 1993). A more complex processing induced by accountability has debiasing benefits, such as reducing order effects (Kennedy, 1993; Tetlock, 1983b), sunk cost effects (Simonson & Nye, 1992), and overconfidence (Tetlock & Kim, 1987). However, accountability also has biasing costs, including increasing the dilution effect (Tetlock & Boettger, 1989; Tetlock, Lerner, & Boettger, 1996). The dilution effect refers to ‘‘the tendency for nondiagnostic information to dilute the extremity of predictions that people make when presented with only diagnostic information’’ (Tetlock & Boettger, 1989, p. 389). In other words, the dilution effect captures the tendency to lose confidence in the predictive power of diagnostic information when nondiagnostic information is also present. Waller and Zimbelman (2003) provide indirect evidence that the dilution effect occurs in audit practice, a field setting where accountability cannot be manipulated. They use archival data to demonstrate that auditors implicitly fail to fully consider prior period misstatements, relative to a regression model, when assessing risk.
The Impact of Accountability
3
This important bias is worth considering in studies of fraud risk assessment because such tasks typically require auditors to pay attention to a few pieces of diagnostic evidence mixed among a vast quantity of nondiagnostic information. Many ‘‘red flag’’ indicators of potential fraud, such as a domineering chief executive, may be present in both fraud and non-fraud cases. Moreover, since fraud is intentional, perpetrators often exert effort to conceal evidence from auditors. As Bologna and Lindquist (1995, p. 133) point out, ‘‘An accumulation of small differences is often the very essence of a sophisticated large fraud.’’ In fraud risk assessments, the presence of a dilution effect poses the threat that auditors will render a judgment of minimal risk when risk is beyond an acceptable threshold, a particularly costly error if the diluted risk assessments prevent the timely detection of fraud. Companies and their auditors share fraud risk assessment and detection responsibility (e.g., CICA Handbook Section 5135, ISA No. 240, and SAS No. 99). The primary responsibility rests with management, boards of directors, and audit committees (National Association of Corporate Directors (NACD), 1998; Public Oversight Board (POB), 2000). While professional standards require external auditors to assess the risk of material fraud, then plan and perform the audit to provide reasonable assurance that the financial statements are free of material misstatements caused by error or fraud. SAS No. 99 recognizes that internal auditors are the first line of defense against fraud and thus are effective partners in the prevention and detection of fraud. Internal auditors who understand the various types of fraud and their relative rates of occurrence are more likely to recognize red flags and be better prepared to fight the high organizational cost of corruption (Institute of Internal Auditors (IIA), 1999; Independent Commission Against Corruption (ICAC), 1994). Several prior research studies have examined the dilution effect using external auditors in experimental settings. Hackenbrack (1992) demonstrates the existence of dilution effects in fraud risk assessments. Hoffman and Patton (1997) find not only that auditors’ fraud risk judgments become more conservative when they are accountable to superiors, but also that participants, anticipating the views of their evaluators, exhibit a dilution effect of similar magnitude whether or not they are held accountable. Glover (1997) also finds that accountability does not affect the magnitude of the dilution effect exhibited by auditors assessing the risk of a material misstatement in accounts receivable, which includes the possibilities of both error and fraud. Finally, Shelton (1999), while not directly examining the influence of accountability on dilution, reports that senior auditors, but not partners, exhibit the dilution effect in a going-concern task. While past research has shown the impact of the dilution effect on various decision-making
4
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
processes, this study enhances the knowledge base by including the realistic, multi-faceted component of accountability, and integrating more recent theories of accountability from psychology into practical audit settings. Over time, the social psychology literature has accumulated evidence that led to a modification of the social contingency model (Lerner & Tetlock, 1999) to recognize the more complex and multidimensional role of accountability in judgment and decision-making. Following this modification, the study reexamines the relationship between accountability and the dilution effect for a fraud risk assessment task. This study presents evidence from a sample of internal auditors that demonstrates a more complex role of accountability than previously understood. Evidence is provided that accountability can decrease the occurrence rate of dilution effect – a judgment quality benefit. However, results also show that, when accountable participants exhibit a dilution effect, this effect is magnified – a judgment quality cost. A major contribution of this study is to refine the findings of prior studies by showing that accountability may not only attenuate dilution effect on some dimension (frequency of occurrence) but can also amplify it on other dimension (magnitude of dilution) as predicted by the flexible contingency model proposed by Lerner and Tetlock (1999). The reduced frequency of occurrence implies accountability may reduce the number of costly decision errors, providing evidence that the accountability process can add value to the audit. Concluding comments discuss future research ideas to examine the more problematic cases where the dilution effect is magnified by accountability.
THEORETICAL BACKGROUND AND HYPOTHESES One pervasive finding of the literature on judgment and choice is that decision makers often fail to use optimal strategies (Slovic, Fischhoff, & Lichtenstein, 1977). Human decision-making is subject to a number of biases, including the tendency to underreact to diagnostic information when it is accompanied by nondiagnostic information – a ‘‘dilution effect.’’ Nondiagnostic evidence can be defined as information with ‘‘little or no value for predicting the outcome’’ (Nisbett, Zukier, & Lemley, 1981, p. 249), yet social judgment research has shown that decision makers are often influenced by nondiagnostic information (Nisbett & Ross, 1980; Nisbett et al., 1981; Zukier, 1982). When presented with a mix of diagnostic and nondiagnostic evidence, people tend to make less extreme judgments than when presented with only diagnostic evidence.
The Impact of Accountability
5
The potential for dilution effect in audit settings exists because auditors usually face a mix of diagnostic and nondiagnostic evidence. The dilution effect is most often explained as the result of the representativeness heuristic, whereby auditors judge the probability of the event of interest (e.g., the risk of fraudulent financial reporting) by comparing client information (evidence) to their mental model of conditions that produce the event (Tversky & Kahneman, 1982; Frederick & Libby, 1986). The similarity between client information and the event of interest is a positive function of the salience, weight, and number of distinctive features. Nondiagnostic pieces of evidence are distinctive features because they are client characteristics that have no association with or predictive ability for the event of interest. Hence, nondiagnostic evidence reduces (dilutes) the similarity between the target (client) and the event (risk of fraudulent financial reporting) suggested by the diagnostic evidence. The first hypothesis of this study is to replicate the dilution effect observed in prior studies of fraud risk assessment (Hackenbrack, 1992; Hoffman & Patton, 1997; Glover, 1997). H1. The fraud-risk assessment based on diagnostic evidence together with nondiagnostic evidence will be lower than the assessment based on diagnostic evidence alone. Studies of suboptimal decision-making led to the development of contingent decision-making models (Beach & Mitchell, 1978; Payne, 1982). These models recognize that decision makers, as they encounter new and changing task conditions, change their behavior to comply with the demands of the environment. The choice of a decision strategy depends on task demands – a function of the decision problem, decision maker, and environment. Attributes of the decision environment include the degree to which decision makers are personally accountable for their decisions. Suedfeld and Tetlock (1991, p. 55) describe human beings as cognitive managers who ‘‘react to specific challenges and opportunities’’ by adapting, ‘‘the complexity of their information processing in response to variables such as the importance of the decision, [and] the need to justify one’s views to an audience or constituency.’’ Accountability in audit research was first addressed by Emby and Gibbins (1988) who describe the factors contributing to good judgment in public accounting. They state that, ‘‘justification procedures occupy much of the time taken by the professional judgment process and continue after the action has been taken’’ (p. 288). Auditors are made personally accountable through working papers documenting the rationale of their decisions. This justification process is supplemented by supervision and review, required by
6
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
quality control standards for external auditors (SQCS No. 2) and standards for the practice of internal auditing (IIA, 1995). Many studies have been conducted to determine why accountability has an impact on the judgment process (McAllister, Mitchell, & Beach, 1979; Tetlock, 1983a; Curley, Yates, & Abrams, 1986; Tetlock & Kim, 1987; Tetlock, Skitka, & Boettger, 1989; Cloyd, 1997; Tetlock, 1999). Those studies show that, when people do not know the views of the evaluative audience, they are motivated to think in more complex, analytic ways. Several studies show that accountability can improve decision quality (Hagafors & Brehmer, 1983; Weldon & Gargano, 1985; Ashton, 1990, 1992; Johnson & Kaplan, 1991), and even reduce or eliminate judgmental biases (Tetlock, 1983b, 1985; Simonson & Staw, 1992; Simonson & Nye, 1992; Kennedy, 1993). Other studies demonstrate that accountability may worsen the quality of the decision-making process. Ashton (1990) notes that decision makers, knowing their decisions would be evaluated, might experience increased anxiety and become more preoccupied with good justification of performance than good performance itself. Tetlock and Boettger (1989) report that accountability made their participants use more information, but did not make them more discriminating judges of the usefulness of that information. Participants, having a desire to be favorably evaluated and not knowing the views of the person to whom they felt accountable, process as much information as possible, including nondiagnostic cues, in an attempt to identify the most acceptable position. Accountability in this case becomes ‘‘a social magnifier of the dilution effect.’’ Recently, Lerner and Tetlock (1999, p. 255) conclude that it is ‘‘a mistake – and a rather common one – to view accountability as a unitary phenomenon.’’ They point out that even the simplest manipulation of accountability involves several distinguishable aspects, including the mere presence of another as an observer, the identifiability of participants to the observer, and the evaluation of the participants by the observer. They suggest that the effect of accountability is complex and propose a flexible contingency model as an explanation. This model fits within a developing body of literature on dual-process theories of reasoning (Green, Visser, & Tetlock, 2000; Stanovich & West, 2000 for a review) that focus on when and why individuals move from an automatic associative system of heuristic reasoning to a controlled rule-based system of analytic processing. Mindful integrative complexity can improve decision quality as accountable decision makers use more analytic decision strategies, paying closer attention to the information set, for example by discounting weak or suspicious
The Impact of Accountability
7
evidence or separating the relevant information from the irrelevant information (McAllister et al., 1979; Tetlock & Kim, 1987). In particular, Lerner and Tetlock (1999, p. 263) argue that accountability can attenuate bias on tasks ‘‘to the extent that (a) suboptimal performance results from lack of selfcritical attention to the judgment process and (b) improvement requires no special training in formal decision rules, only greater attention to the information provided.’’ Hence, predecisional accountability improves judgment because, when participants expect to justify their judgment, they want to avoid appearing foolish in front of the audience and thus prepare themselves by engaging in a self-critical search for reasons to justify their actions (Lerner & Tetlock, 1994). This search leads participants to pay greater attention to the cues they use and gain greater awareness of their cognitive processes by regularly monitoring the cues that are allowed to influence judgment and choice. This type of integrative complexity implies that accountability should reduce the occurrence rate of the dilution effect. Accountability may lead auditors to pay more attention to the task, including heeding the warnings of the professional literature that fraud cues are often buried in large quantities of non-fraud cues. If this occurs, accountability will reduce the frequency of dilution effect. H2. The occurrence of a dilution effect will be less frequent when internal auditors are held accountable for their judgments, than when there is no accountability. However, integrative complexity can also be dysfunctional if decision makers shift into a ‘‘relatively mindless’’ process where all evidence is treated with elevated importance, whether diagnostic or nondiagnostic (Tetlock & Boettger, 1989). The same overarching motive underlies amplification in both judgment and choice tasks: a desire to avoid appearing foolish in front of the audience. Lerner and Tetlock (1999, p. 264) argue that, in judgment tasks, predecisional accountability will amplify bias ‘‘to the extent that a given bias results from naive use of normatively (but not obviously) irrelevant cues.’’ When a bias results from a lack of awareness that certain cues are proscribed, the desire to avoid appearing foolish in front of an audience only makes matters worse: it heightens the use of all the cues, even irrelevant ones. For a fraud risk assessment task, Lerner and Tetlock’s model implies that, if an accountable auditor unknowingly pays attention to the irrelevant cues, the dilution effect will be magnified. H3. When a dilution effect occurs, the magnitude of the dilution will be greater for accountable than for nonaccountable internal auditors.
8
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
The previous two hypotheses aim to show that accountability may have both benefits and costs. These predictions are very much in line with the flexible contingency model of Lerner and Tetlock (1999). Predecisional accountability will attenuate biases that arise from lack of self-critical attention to one’s decision processes, but is also likely to amplify bias to the extent that a given judgment bias results from using normatively (but not obviously) proscribed information. Thus, bias attenuation is expected to reduce the occurrence rate in dilution effect for accountable auditors, while bias amplification is expected to translate into an increased magnitude of the dilution effect for accountable auditors that exhibit such bias. This dual prediction is supported by the notion that ‘‘accountability is a logically complex construct that interacts with characteristics of decision makers and properties of the task environment to produce an array of effects – only some of which are beneficial’’ (Lerner & Tetlock, 1999, p. 270). Most previous experiments have included only two accountability levels (i.e., accountable and non-accountable, or low accountable and high accountable) and thus did not provide insight about sensitivity to various levels of accountability. In fact, while ‘‘explicit accountability’’ may be related to an intended evaluative audience, internal auditors also have ‘‘implicit accountability’’ to other evaluative audiences of their work (e.g., their immediate supervisor, other company employees, external auditors, regulatory authorities, the general public, and even themselves). This implicit accountability may also lead internal auditors to be more conservative in their judgments, since such conservatism is expected by the evaluative audiences and is thus a ‘‘safe’’ approach. Nevertheless, following recommendations by Messier and Quilliam (1992), participants in this study were instructed that they would be required to justify/defend their decisions to one specific evaluative audience. Internal auditors operate in environments where organizational lines of reporting vary from company to company. For example, some internal audit departments report to a management level of the organization (e.g., Controller), while others report directly to the audit committee of the Board of Directors. The 2002 Healthcare Internal Auditing Survey, a survey of 1,200 members of the Association of Healthcare Internal Auditors, found that 38% of the respondents had internal audit functions reporting directly to the board’s audit committee, while 22% reported to the company’s CFO, 18% to the CEO, 8% to the board as a whole, and 15% elsewhere. A direct reporting relationship to the audit committee provides stronger internal audit independence from management, which is important because internal auditors must be able to review the conduct of management
9
The Impact of Accountability
(BRC, 1999) and remain independent of undue management influence. Recent U.S. legislation (the Sarbanes–Oxley Act) and listing requirement changes by the New York Stock Exchange and NASDAQ demand a direct line of reporting from the internal audit function to the board’s audit committee as a means to strengthen corporate governance. A reporting relationship to the board’s audit committee establishes a higher level of reporting for the internal auditor, which may or may not translate into a different degree of accountability. These various reporting levels might create different amounts of pressure, increasing with higher levels of reporting. In the absence of research findings establishing a direct link between level of reporting and degree of accountability, the last two hypotheses are exploratory in nature and predict a non-directional difference in the frequency and magnitude of dilution effect between internal auditors who are accountable to top management and those accountable to the audit committee. H4. The frequency of occurrence of the dilution effect will be different between internal auditors reporting to top management and those reporting to the audit committee of the Board of Directors. H5. When a dilution effect occurs, the magnitude of dilution will be different between internal auditors reporting to top management and those reporting to the audit committee of the Board of Directors.
RESEARCH METHOD Participants Since the primary responsibility for fraud risk assessment and detection rests with companies making internal auditors the first line of defense, this study uses internal auditors as participants. With the assistance of the IIA, several firms on the western coast of the United States were identified as having an internal audit department comprised of 10 staff or more and an audit committee. One of the researchers personally contacted the head of the internal audit department of each company to enlist its participation in the study. A total of 192 internal auditors from 38 firms participated in the experiment. One of the researchers personally administered the experiment on site at the headquarters of each firm under controlled conditions, and participants were selected by the firms based on their schedule availability since all participants at each firm were administered the research instrument concurrently. Three of the 38 firms also had one member of their audit committee and one
10
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
member of their top management participate in the study in order to operationalize the accountability conditions discussed below. Experimental Tasks and Procedures A fraud risk assessment task from previous research allows a comparison between internal auditor and external auditor results. The task includes the increasing risk case with one diagnostic and four positive (favorable to the client) non-diagnostic pieces of evidence from Hackenbrack (1992), with minor wording variations necessary to adapt the setting from external to internal auditor-participants. Hackenbrack case materials present a narrative description and working papers prepared by the prior year’s audit team; his participants are placed in the role of current year external auditors. The primary wording variation is to note that the company described in the case is being considered as an acquisition target by participants’ management; audit working papers are presented as being prepared by the acquisition target’s external auditors and made available for pre-acquisition review; participants are placed in the role of internal auditors doing a pre-acquisition review, in which fraud risk assessment would be a natural task. Pilot testing was done with two groups of internal auditors who did not participate in the experiment. The experimental materials and instructions were clearly understood, pilot test participants generally felt the case was realistic and adequately performed the required tasks with proper use of the response scales. Similar to Hackenbrack (1992), participants are asked to complete two tasks: a case-study task and a rating task. For the case-study task, participants are initially instructed to read a description about the background of a company being considered for acquisition by their management. This description, covering the company’s prior activities, establishes a baseline against which to evaluate working paper excerpts from a recent interim review performed by the company’s external auditors. As in Hackenbrack (1992), the diagnostic evidence is a memo about the change in company compensation policy placing greater weight on bonuses tied to each responsibility center’s financial performance. Other working paper excerpts that normatively have no predictive value in assessing fraud risk represent the nondiagnostic evidence (see Fig. 1 and Hackenbrack, 1992 for details). After reading the excerpts, participants are asked to rate how much this information has changed their impression of the company’s exposure to fraudulent reporting relative to the initial description. The rating is made on Hackenbrack’s 21-point scale, anchored at the lower endpoint 0 (labeled
Rating Task • Read descriptions of five increasing fraud-risk situations: 1. Changes in management compensation package with increased emphasis on achieving budgeted targets and bonuses tied to responsibility-center financial performance. 2. Potential breach of restrictive debt covenants, which could cause a significant amount of longterm debt to become current with no waiver from lender. 3. Delays in accounting activities due to cutbacks in resources, poor staffing levels, and increases in activity volume, with management unsupportive of accounting department. 4. Hands-off top management style, focusing on “big picture,” with no thorough review of financial information. 5. Strong competition and lower margins, with management determined to retain market share by introducing new services, with initial cost lowering company’s profits. • Rate situations on a 101-point scale as to how much each would increase fraud risk.
The Impact of Accountability
Case-Study Task • Read about company targeted for acquisition and five working papers: o Memo reflecting fraud-risk situation #1 below (diagnostic evidence) o Analysis of interim financial performance and position (non-diagnostic evidence) o Permanent file’s description of cash collections cycle (non-diagnostic evidence) o Memo on accounts receivable and related allowance (non-diagnostic evidence) o Memo on preparation of time budgets (non-diagnostic evidence) • Rate change in fraud risk after reading working papers, based on 21-point scale, where “0” means no change and “10” means change in fraud risk for another company with same initial risk level, for which auditors just learned of fraud-risk situation #2 below.
Experimental Manipulations Non-accountable participants read, “Your responses will be confidential and your company’s management will not have access to this information.” Participants accountable to top management read, “A member of your company’s top management may be reviewing your decisions and the justifications you gave for those decisions, and could call upon you to discuss this case further.” Participants accountable to the Board of Directors read, “A member of your company’s audit committee of the Board of Directors might be reviewing your decisions and the justifications you gave for those decisions, and could call upon you to discuss this case further. However, your company’s management will not have access to this information.”
11
Fig. 1.
Summary Features of the Experiment (adapted from Hackenbrack, 1992).
12
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
‘‘no change in fraud-risk assessment’’) and at the midpoint 10 (labeled with reference to a fraud-related situation used in the rating task), with an unlabeled upper endpoint (20). For the rating task, auditors first read (in random order) descriptions of five increasing fraud-risk situations. Participants rank the situations from high to low according to how much each situation would increase a company’s exposure to fraudulent financial reporting. The situation that would increase exposure most is assigned to the upper endpoint (100) of Hackenbrack’s 101-point scale, with a lower endpoint (0) labeled ‘‘does not increase your risk assessment.’’ Participants then rate the remaining four situations relative to the two endpoints. One of the five situations is the same memo about bonus plan changes used in the case-study task. Hence, participants evaluate this same fraud-related situation twice, first in combination with nondiagnostic evidence (case-study task) and second without nondiagnostic evidence (rating task). Fig. 1 illustrates the tasks and procedures, which are identical to those used by Hackenbrack (1992). The influence of the nondiagnostic evidence is measured by comparing the ratings participants made in the two tasks. The computation of the dependent variables accounts for the difference in scales. A questionnaire is used to collect data on participants’ experience (both general and fraudrelated), perform a manipulation check on the accountability condition, and assess participants’ impressions of task realism. Most participants (97%) rated the case and related task as realistic. Independent Variable: Accountability Participants are randomly assigned to one of the three conditions (nonaccountable, accountable to top management, and accountable to the audit committee). Participants in the ‘‘non-accountable group’’ are told that their responses are confidential and that only the researcher has access to this information. Participants in the ‘‘accountable groups’’ are told that this study is administered to internal auditors of several companies, and that, for participants selected at random, a member of their top management or audit committee will review their decisions and call upon them to further discuss their response. To complete the accountability manipulation, after the experiment was administered to all 192 participants, the researcher attended meetings with six participants who explained their decisions to a member of their employer’s top management or audit committee. All of the experimental materials were made available to the participating members of top management
13
The Impact of Accountability
or audit committee so that they could engage the participants in discussions related to their decisions. A post-experiment debriefing questionnaire asks participants to report the group to which they felt most accountable for their case responses. Of the 192 participants, 11 participants who did not correctly perceive the accountability condition to which they had been assigned and one participant who did not properly use the rating scale were dropped from the analysis, yielding a total of 180 useable responses. Dependent Variables: Frequency and Magnitude of Dilution effect The dependent variables measure the influence of nondiagnostic evidence on the assessment of the company’s exposure to fraudulent reporting. The within-subject experiment allows for a repeated measure. Participants evaluate the same fraud-related situation twice, first in combination with nondiagnostic evidence (case-study task) and second without nondiagnostic evidence (rating task). The influence of the nondiagnostic evidence is measured by comparing the participants’ ratings in the two separate tasks. The scale used in the casestudy task has two points in common with the scale used in the rating task (a) the lower endpoint 0 on each scale represents ‘‘no change’’ in fraud risk, and (b) the description of the anchor value 10 on the case-study scale is one of the fraud-related situations used in the rating task. These two sets of points define a common unit of measure between the two scales. Because the scales have different upper endpoints, rescaling follows the procedure described by Hackenbrack (1992). For each participant, the same three measures used in Hackenbrack (1992) are available: (a) a case-study rating (X1) in the presence of nondiagnostic evidence, (b) the rating of the same fraud-related situation absent in the nondiagnostic evidence in the rating task (X2), and (c) a common unit of measurement between the case-study and the rating scales (10/X3, where 10 is the anchor value on the case-study scale and X3 represents the rating of its description on the rating scale). The influence of the nondiagnostic evidence on each participant’s case-study assessment, measuring the magnitude of the dilution effect, is evaluated by calculating D ¼ (X1) [X2 (10/X3)], in case-study units. D is negative when the nondiagnostic evidence reduces a participant’s case-study assessment (dilution effect). Design ¯ between the ratings on the case-study and the A mean of the differences (D) rating tasks is computed for each treatment condition and on an overall
14
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
basis. If the rating on the case-study task is lower than the rating on the rating task, there is a dilution effect (negative mean difference). w2 tests are used to compare frequency of occurrence between groups. Differences between the means are evaluated with both parametric and nonparametric tests.
RESULTS Descriptive Statistics On average, participants had worked as external auditors for 2 years, then as internal auditors for five and a half years, giving them a total of seven and a half years of general audit experience. Participants reported an average number of frauds detected of 0.5 while working as an external auditor, and of 1.5 while working as an internal auditor. The increase in the number of frauds detected when working as an external auditor as opposed to working as an internal auditor is expected given the fact that internal auditors, contrary to external auditors, traditionally had a strong mandate to search for fraud. Hence, participants had sufficient experience for the experimental task. When clustered by accountability condition, there were no significant differences between the participant groups with respect to general experience measures (experience as external auditor, p ¼ 0:95; experience as internal auditor, p ¼ 0:37) and task-specific experience measures (fraud detection in public accounting, p ¼ 0:13; fraud detection in internal audit, p ¼ 0:71), as shown in Table 1. Table 1.
Participants’ Experience Measures Between Accountability Conditions.
Accountability Conditions
TM
AC
NA
Panel A: Means of participants’ general experience (in months) As an external auditor 22 20 24 As an internal auditor 62 75 62
Ha
Sig.
0.10 2.01
0.95 0.37
Panel B: Means of participants’ task-specific experience (no. of engagements/reviews with detected fraud) As an external auditor 0.96 0.42 0.16 4.08 0.13 As an internal auditor 1.52 1.79 1.18 0.69 0.71 Note: TM Participants accountable to top management; AC Participants accountable to audit committee; NA Non-accountable participants. a H is the test statistic of the Kruskall–Wallis test, which has approximately a w2 distribution.
15
The Impact of Accountability
Hypotheses Tests H1: Dilution Effect Lilliefors tests for each condition and the combined data indicate a departure from normality. Thus, both t-tests and Wilcoxon signed-rank tests are used to determine whether a dilution effect occurred. Further, since the distributions are asymmetric, sign tests are also conducted.1 Summary statistics and test results are presented in Table 2. The nondiagnostic evidence has a clear dilutive impact on the participants’ judgments, with a combined average mean shift of 10.37.2 The observed significance levels of the t-tests and the nonparametric tests provided support for H1. H2 and H3: Impact of Accountability on Frequency and Magnitude of Dilution Effect Table 3 (Panel A) shows the frequency of occurrence of the dilution effect. The Pearson w2 statistic is 6.59 (p-value ¼ 0.00), indicating that the proportion of participants who exhibit a dilution effect is smaller for the accountable group (68%) than for the non-accountable group (86%).3 Hence, as expected, accountability reduces the frequency of the dilution effect, supporting H2. Table 2.
Dilution Effect (H1): Summary Statistics and Hypothesis Test Results.
Panel A: Parametric tests Accountability
No. of Participants
Audit committee Top management Non-accountable TOTAL
62 56 62 180
Mean
Median
15.39 8.81 6.75 10.37
Std. Dev.
t
Sig.
36.4 18.8 11.9 24.9
5.33 3.51 4.47 5.58
0.00 0.00 0.00 0.00
4.25 3.72 4.50 4.39
Panel B: Nonparametric tests Accountability Audit committee Top management Non-accountable TOTAL a
No. of Participants
Wa
Sig.b
S
Sig.
62 56 62 180
5.4 4.3 5.5 9.1
0.00 0.00 0.00 0.00
4.3 2.6 5.8 7.5
0.00 0.01 0.00 0.00
Wilcoxon signed-rank test statistic. n test statistic.
b
16
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
Table 3.
Accountability and Dilution Effect.
Panel A (H2): Frequency (all participants, n ¼ 180)
Accountable Non-accountable
Dilution
No Dilution
80 (68%) 53 (86%)
38 (32%) 9 (14%)
Panel B (H3): Magnitude (participants who exhibited a dilution effect, n ¼ 133) Accountability Accountable Non-accountable
No. of Participants Mean Dilution Effects Std. Dev. 80 53
18.75 8.57
33.5 11.8
t
Sig. M–Wa Sig.
2.47 0.01 2.40 0.02
Note: Pearson w2 statistic: 6.59 (p-value ¼ 0.00). a Standard normal deviates of the Mann–Whitney scores.
The question still remains as to whether accountability also influences the magnitude of dilution, when a dilution effect occurs. Table 3 (Panel B) shows the summary statistics and test results for the participants who exhibit a dilution effect. The magnitude of the dilution effect is significantly greater for accountable than for non-accountable participants (pvalue ¼ 0.01), supporting H3. The Mann–Whitney test corroborates this conclusion. H4 and H5: Degree of Accountability on Frequency and Magnitude of Dilution Effect Table 4 (Panel A) shows the frequency of the dilution effect for participants accountable to top management and to the audit committee. The Pearson w2 statistic is 1.37 (p-value ¼ 0.24), indicating that, for participants whose judgments reflect a dilution effect; the proportion of participants accountable to top management (63%) is not significantly different from that of participants accountable to the audit committee (73%), failing to support H4. A t-test and a Mann–Whitney test are conducted to test the impact of reporting level on the magnitude of the dilution effect. Table 4 (Panel B), shows the magnitude of the dilution effect for accountable participants who exhibit a dilution effect. The magnitude of the dilution effect is not significantly different for participants accountable to top management and for those accountable to the audit committee (p-value ¼ 0.19), failing to support H5. The Mann–Whitney test corroborates this conclusion.
17
The Impact of Accountability
Table 4.
Degree of Accountability and Dilution Effect.
Panel A (H4): Frequency (accountable participants, n ¼ 118)
Top management Audit committee
Dilution
No Dilution
35 (63%) 45 (73%)
21 (37%) 17 (27%)
Panel B (H5): Magnitude (accountable participants who exhibited a dilution effect, n ¼ 80) Accountability Top management Audit committee
No. of Participants Mean Dilution Effects Std. Dev. 35 45
14.97 21.69
21.4 40.4
t
Sig. M–Wa Sig.
0.88 0.19 0.28 0.78
Note: Pearson w2 statistic is 1.37 (p-value ¼ 0.24). a Standard normal deviates of the Mann–Whitney scores.
Additional Analysis Shelton (1999) in a going-concern task finds that, while audit seniors exhibit a dilution effect, audit managers and partners with significantly more experience on this task do not fall prey to the dilution effect and are able to ignore irrelevant information. Internal auditors’ judgments based on experience levels are examined to see whether Shelton’s experience finding can be replicated. General experience is first clustered along several dimensions: experience as external auditors, experience as internal auditors, and total experience (public accounting, internal audit, and general business combined). Upper and lower one-third extremities of the various distributions are then selected to conduct statistical tests. The frequency of dilution effect along those general experience dimensions are displayed in Table 5 (Panel A). The significance level of the Fisher’s exact test for general experience as external auditor is 0.44, indicating that the proportion of participants with 3 years or more of public experience who exhibit a dilution effect (70%) is not significantly different than the proportion of participants with no public accounting experience whose judgments reflect a dilution effect (76%). Similarly, the significance level of the Fisher’s exact test for general experience as internal auditor is 0.64, indicating that the proportion of participants with 7 years or more of internal audit experience who exhibit a dilution effect (78%) is not significantly different than the proportion of participants with 2 years or less of internal audit experience, whose judgments reflect a dilution effect (72%). Finally, the significance level of the Fisher’s exact test for combined general experience is 0.50, indicating that
18
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
Table 5. Experience and Frequency of Dilution Effect. Panel A: General experience Type of experience External auditor Internal auditor Combined experience
Dilution No experience 3 years or more 2 years or less 7 years or more 7 years or less 17 years or more
72 37 33 38 36 37
(76%) (70%) (72%) (78%) (77%) (79%)
No dilution 23 16 13 11 11 10
(24%) (30%) (28%) (22%) (23%) (21%)
Panel B: Task-specific experience Engagements/Reviews with detected fraud Public accounting Internal audit
No Yes No Yes
Dilution 102 (74%) 31 (72%) 80 (71%) 53 (79%)
No Dilution 35 12 33 14
(26%) (28%) (29%) (21%)
the proportion of participants with 17 years or more of combined experience who exhibit a dilution effect (79%) is not significantly different than the proportion of participants with 7 years or less of combined experience whose judgments reflect a dilution effect (77%). Cognizant of the fact that general experience whether as an external auditor or an internal auditor, or even combined general experience, may not be a good proxy for participants’ level of familiarity with fraudulent financial reporting, participants were also clustered along two task-specific experience dimensions, reflected by the number of engagements or reviews where fraud has been detected either in a public accounting or an internal audit environment. Again, the upper and lower extremities of the various distributions were chosen to conduct statistical tests. The frequency of dilution effect along those task-specific experience dimensions are displayed in Table 5 (Panel B). The significance level of the Fisher’s exact test for task-specific experience in public accounting is 0.84, indicating that the proportion of participants with such experience who exhibit a dilution effect (72%) is not significantly different than the proportion of participants with no such experience whose judgments reflect a dilution effect (74%). Similarly, the significance level of the Fisher’s exact test for task-specific experience in internal audit is 0.29, indicating that the proportion of participants with such experience who exhibit a dilution effect (79%) is not significantly different than the proportion of participants without such experience, whose judgments reflect a dilution
19
The Impact of Accountability
effect (71%). Hence, whether using dimensions of general or task-specific experience, no reduction in the frequency of dilution effect related to increased levels of experience is observed to the extent of the reduction found in the presence of accountability. Tests using those same clusters also reveal that, when dilution effect occurs, experience neither exacerbates nor attenuates the magnitude of the dilution effect. The mean dilution effect ( 14.39) of participants with task-specific experience in public accounting who exhibit a dilution effect was not significantly different from that ( 14.78) of participants without such experience, whose judgments reflect a dilution effect (p-value ¼ 0.95). Similarly, the mean dilution effect ( 15.80) of participants with task-specific experience in internal audit who exhibit a dilution effect was not significantly different from that ( 13.86) of participants without such experience, whose judgments reflect a dilution effect (p-value ¼ 0.71). Experienced auditors in public accounting have well-developed knowledge structures for going-concern assessment that is performed in each and every audit engagement (Ricchiute, 1992), and general experience can eliminate the dilution effect in such task (Shelton, 1999). Going-concern assessment is a fairly uniform task benefiting from tools, such as analytical procedures to identify conditions indicative of possible substantial doubt about a company’s ability to continue as a going concern, with established links to bankruptcy prediction models (Hopwood, McKeown, & Mutchler, 1994; Altman, 1993; Koh, 1991; Mckee, 1989; Ohlson, 1980). Fraud-risk assessment, on the contrary, is a task not characterized by a uniform environment, but is specific to each audit engagement or review. Frequently, auditors rely heavily on their perceptions of management’s attitude or character when assessing the risk of financial-statement fraud (Wilks & Zimbelman, 2004; Heiman-Hoffman, Morgan, & Patton, 1996), since existing quantitative fraud prediction models tend to overestimate the likelihood of fraud (Hansen, McDonald, Messier, & Bell, 1996; Nieschwietz, Schultz, & Zimbelman, 2000). Hence, in fraud-risk assessment tasks, accountability seems to be a much more effective debiaser of the dilution effect than experience.
CONCLUSION Limitations There are limitations to this study. First, the data are collected using a case developed by Hackenbrack (1992). Hence, the study inherits the biases and limitations of the case itself. However, a trade-off was necessary to achieve
20
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
comparability in the dilution effect results of this study and that of Hackenbrack. The replicated finding of a dilution effect with non-accountable internal auditors strengthens earlier findings. This study also adds to the knowledge base by examining accountability to different evaluative audiences and measuring dilution effect along dimensions of frequency and magnitude that are reflected in a more recent theory of accountability (Lerner & Tetlock, 1999). Second, generalizing from experimental studies to real audit situations is always problematic. For example, given that time constraints limit the size of information sets presented to participants, more than one piece of diagnostic evidence would likely be presented and the proportion of nondiagnostic evidence to diagnostic evidence in real audit situations would be greater than in experiments. This limitation, however, serves to bias against an experimental finding of a dilution effect. Research Findings: Accountability and the Dilution Effect Consistent with prior research in social psychology and audit judgment, this study finds that nondiagnostic evidence can moderate the impact of evidence considered useful for fraud-risk assessment – a dilution effect. On average, participants’ risk assessments based on diagnostic evidence together with nondiagnostic evidence are significantly lower than the assessments based on diagnostic evidence alone. This study extends the understanding of accountability in audit settings by demonstrating that accountability influences the dilution effect in two ways. Accountability decreases the frequency of occurrence – a judgment quality benefit, but increases the magnitude of the effect when dilution occurs – a judgment quality cost. These findings fit within the evolving theoretical work on the complexity of the dilution effect and the developing body of literature on dual-process theories of reasoning. The impact of accountability on the dilution effect may conceivably be influenced by other environmental pressures, such as deadlines, incentives, and feedback (Ashton, 1990). Tan, Ng, and Mak (2002) suggest a three-way interaction between accountability, knowledge, and task complexity that provides a possible framework to further explore this relationship. Another consideration is suggested by the work of Kadous, Kennedy, and Peecher (2003) – participants with stronger commitment to a goal of building ‘‘a justifiable case’’ for a client’s accounting method – are more likely to accept a client’s aggressive accounting method in the presence of quality assessment than in its absence. Such goal commitment may also be an indication of firm culture.
The Impact of Accountability
21
Contrary to expectations, the reporting level does not significantly influence the frequency or the magnitude of the dilution effect. Participants accountable to top management exhibit a dilution effect as often and in the same magnitude as participants accountable to the audit committee. One potential explanation is that the manipulation of accountability is not strong enough: no statements about the characteristics of the audit committee were made (e.g., number of independent directors, number of members with financial expertise, and number of committee meetings per year). Another potential explanation is that internal auditors may implicitly not perceive a different amount of pressure between accountability to top management and accountability to the audit committee. Perhaps the fact that the board’s audit committee outranks management is counterbalanced by top management’s day-to-day involvement in company operations, equalizing the accountability pressures. Audit committees may not necessarily be viewed as higher levels of accountability if top management can still exercise considerable pressures on internal auditors through employment and compensation decisions. Further research is needed to fully assess the relationship between reporting relationships and accountability to various evaluative audiences. With respect to audit practice, the observed reduction of judgment error frequency supports the establishment of accountability measures, such as documentation and review requirements, for quality assurance. At the same time, while accountability significantly reduces the frequency of the dilution effect, the judgment of over half of accountable participants still reflects this bias. Practitioners should be concerned that this error in judgment is so prevalent and that accountability can exacerbate the dilution effect. Thus, a key question for further exploration in an audit setting is whether the benefits of error frequency reduction exceed the costs of dilution effect magnitude amplification. Given that many audit decisions are categorical (e.g., fairly presented or not fairly presented), are the observed magnitude differences in risk assessments sufficient to change ultimate decision categories? Further research that includes a progression from risk assessment to audit test planning to final decision could provide additional insight. Another key question is whether additional quality control measures could reduce the magnitude amplification of the dilution effect in audit practice. Since accountability and documentation typically take place after a decision has been made, the magnitude of judgment errors could potentially be decreased by debiasers introduced earlier in the process. For example, Ashton and Kennedy (2002) demonstrate that self-review can be an effective debiaser for mitigating recency effects; similarly, debiasing interventions might mitigate the magnitude of the dilution effect.
22
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
NOTES 1. The sign test is a less powerful alternative to the Wilcoxon signed-rank test, but does not assume that the population probability distribution is symmetric. 2. The mean shift for the internal auditors is stronger than Hackenbrack’s external auditor sample, which had an average mean shift of 4.30 for the increasing fraud risk case with favorable nondiagnostic evidence used in this study. 3. In Hackenbrack (1992), 32 of 39 participants evaluating increasing fraud risk cases (82%) exhibited a dilution effect, which is comparable to the nonaccountable participants in this study; there was no experimental accountability in Hackenbrack’s study.
ACKNOWLEDGMENTS The authors thank the companies and internal auditors, who participated in the study and acknowledge the financial support of the Deloitte Foundation.
REFERENCES Altman, E. I. (1993). Corporate financial distress and bankruptcy. New York, NY: Wiley. American Institute of Certified Public Accountants (AICPA). (2002). Statement on auditing standards no. 99: Consideration of fraud in a financial audit. New York, NY: AICPA. American Institute of Certified Public Accountants (AICPA). (1996). Statement on quality control standards (SQCS) no. 2: System of quality control for a CPA firms accounting and auditing practice. New York, NY: AICPA. Ashton, R. H. (1990). Pressure and performance in accounting decision settings: Paradoxical effects of incentives, feedback, and justification. Journal of Accounting Research, 28(Suppl.), 148–180. Ashton, R. H. (1992). Effects of justification and a mechanical aid on judgment performance. Organizational Behavior and Human Decision Processes, 52(2), 292–306. Ashton, R. H., & Kennedy, J. (2002). Eliminating recency with self-review: The case of auditors’ ‘going concern’ judgments. Journal of Behavioral Decision Making, 15(3), 221–231. Beach, L. R., & Mitchell, T. R. (1978). A contingency model for the selection of decision strategies. Academy of Management Review, 3, 439–449. Blue Ribbon Committee on Improving the Effectiveness of Corporate Audit Committees (BRC). (1999). Report and recommendations of the Blue Ribbon Committee on improving the effectiveness of corporate audit committees. New York, NY: New York Stock Exchange and National Association of Securities Dealers. Bologna, G. J., & Lindquist, R. J. (1995). Fraud auditing and forensic accounting: New tools and techniques (2nd ed.). New York, NY: Wiley. Canadian Institute of Chartered Accountants (CICA). The Auditor’s responsibility to consider fraud and error in an audit of financial statements: CICA handbook section 5135. Toronto, ON: CICA.
The Impact of Accountability
23
Cloyd, C. B. (1997). Performance in tax research tasks: The joint effects of knowledge and accountability. The Accounting Review, 72(1), 111–131. Curley, S. P., Yates, J. F., & Abrams, R. A. (1986). Psychological sources of ambiguity avoidance. Organizational Behavior and Human Decision Processes, 38, 230–256. Emby, C., & Gibbins, M. (1988). Good judgment in public accounting: Quality and justification. Contemporary Accounting Research, 4(2), 287–313. Fandt, P. M. (1993). Encouraging decision vigilance through increased accountability. Journal of Applied Business Research, 9(3), 129–135. Frederick, D. M., & Libby, R. (1986). Expertise and auditors’ judgments of conjunctive events. Journal of Accounting Research, 24(2), 270–290. Glover, S. M. (1997). The influence of time pressure and accountability on auditors’ processing of nondiagnostic information. Journal of Accounting Research, 35(2), 213–226. Green, M. C., Visser, P. S., & Tetlock, P. E. (2000). Coping with accountability cross-pressures: Low-effort evasive tactics and high-effort quests for complex compromises. Personality and Social Psychology Bulletin, 26, 1380–1391. Hackenbrack, K. (1992). Implications of seemingly irrelevant evidence in audit judgment. Journal of Accounting Research, 30(1), 126–136. Hagafors, R., & Brehmer, B. (1983). Does having to justify one’s judgments change the nature of the judgment process? Organizational Behavior and Human Performance, 31, 223–232. Hansen, J. V., McDonald, J. B., Messier, W. F., & Bell, T. B. (1996). A generalized qualitative – response model and the analysis of management fraud. Management Science, 42, 1022– 1032. Heiman-Hoffman, V. B., Morgan, K. P., & Patton, J. M. (1996). What do auditors believe are the best ways to spot fraud? The warning signs of fraudulent financial reporting. Journal of Accountancy, 189(10), 75–81. Hoffman, V. B., & Patton, J. M. (1997). Accountability, the dilution effect, and conservatism in auditors’ fraud judgments. Journal of Accounting Research, 35(2), 227–237. Hopwood, W., McKeown, J. C., & Mutchler, J. F. (1994). A reexamination of auditor versus model accuracy within the context of the going-concern opinion decision. Contemporary Accounting Research, 10(2), 409–431. Independent Commission Against Corruption (ICAC). (1994). Unraveling corruption – a public sector perspective. Sydney, New South Wales: ICAC. Institute of Internal Auditors (IIA). (1995). Standards for the professional practice of internal auditing. Altamonte Springs, FL: IIA. Institute of Internal Auditors (IIA). (1999). Business fraud survey, jointly conducted with the Institute of Management and Administration (IOMA). Altamonte Springs, FL: IIA. International Auditing and Assurance Standards Board (IAASB). (2004). International standards on auditing 240 (revised): The auditor’s responsibility to consider fraud in an audit of financial statements. New York, NY: IAASB. Johnson, V. E., & Kaplan, S. E. (1991). Experimental evidence on the effects of accountability on auditor judgments. Auditing: A Journal of Practice and Theory, 10(Suppl.), 96–107. Kadous, K., Kennedy, S. J., & Peecher, M. E. (2003). The effect of quality assessment and directional goal commitment on auditors’ acceptance of client-preferred accounting methods. The Accounting Review, 78(3), 759–778. Kennedy, J. (1993). Debiasing audit judgment with accountability: A framework and experimental results. Journal of Accounting Research, 31(2), 231–245.
24
MICHAEL FAVERE-MARCHESI AND KAREN V. PINCUS
Koh, H. C. (1991). Model predictions and auditor assessments of going concern status. Accounting and Business Research, 21, 331–338. Lerner, J. S., & Tetlock, P. E. (1994). Accountability and social cognition. In: V. S. Ramachandra (Ed.), Encyclopedia of human behavior (Vol. 1, pp. 3098–3121). San Diego, CA: Academic Press. Lerner, J. S., & Tetlock, P. E. (1999). Accounting for the effects of accountability. Psychological Bulletin, 125(2), 255–275. McAllister, D. W., Mitchell, T. R., & Beach, L. R. (1979). The contingency model for the selection of decision strategies: An empirical test of the effects of significance, accountability, and reversibility. Organizational Behavior and Human Performance, 24, 228–244. Mckee, T. E. (1989). Modern Analytical Auditing. Westport, CT: Greenwood Press, Inc. Messier, W. F., & Quilliam, W. C. (1992). The effect of accountability on judgment: Development of hypotheses for auditing. Auditing: A Journal of Practice and Theory, 11(Supplement), 123–151. National Association of Corporate Directors (NACD). (1998). Report of the NACD best practices council: Coping with fraud and other illegal activity, a guide for directors, CEOS, and senior managers. Washington, DC: NACD. Nieschwietz, R., Schultz, J., & Zimbelman, M. (2000). Empirical research on external auditors’ detection of financial statement fraud. Journal of Accounting Literature, 19, 190–246. Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice-Hall. Nisbett, R. E., Zukier, H., & Lemley, R. E. (1981). The dilution effect: Nondiagnostic information weakens the implications of diagnostic information. Cognitive Psychology, 13, 248–277. Ohlson, J. A. (1980). Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research, 18(1), 109–131. Payne, J. W. (1982). Contingent decision behavior. Psychological Bulletin, 92, 382–402. Public Oversight Board (POB). (2000). The panel on audit effectiveness report and recommendations. Stamford, CT: POB. Ricchiute, D. N. (1992). Working-paper order effects and auditors going-concern decisions. The Accounting Review, 67(1), 46–48. Scott, M., & Lyman, S. (1968). Accounts. American Sociological Reviews, 33, 46–62. Semin, G. R., & Manstead, A. S. R. (1983). The accountability of conduct: A social-psychological analysis. New York, NY: Academic Press. Shelton, S. W. (1999). The effect of experience on the use of irrelevant evidence in auditor judgment. The Accounting Review, 74(2), 217–224. Simonson, I., & Nye, P. (1992). The effect of accountability on susceptibility to decision errors. Organizational Behavior and Human Decision Processes, 51(3), 416–446. Simonson, I., & Staw, B. M. (1992). Deescalation strategies: A comparison of techniques for reducing commitment to losing courses of action. Journal of Applied Psychology, 77(4), 419–426. Slovic, P., Fischhoff, B., & Lichtenstein, S. (1977). Behavioral decision theory. Annual Review of Psychology, 28, 1–39. Stanovich, K. E., & West, R. F. (2000). Individual differences in reasoning: Implications for the rationality debate. Behavioral and Brain Sciences, 23, 645–726. Suedfeld, P., & Tetlock, P. E. (1991). Psychological advice about political decision making: Heuristics, biases, and cognitive defects. In: P. Suedfeld & P. E. Tetlock (Eds), Psychology and social policy. Washington, DC: Hemisphere.
The Impact of Accountability
25
Tan, H.-T., Ng, T. B., & Mak, B. W. (2002). The effects of task complexity on auditors’ performance: The impact of accountability and knowledge. Auditing: A Journal of Practice and Theory, 21(2), 81–95. Tetlock, P. E. (1983a). Accountability and complexity of thought. Journal of Personality and Social Psychology, 45(1), 74–83. Tetlock, P. E. (1983b). Accountability and the perseverance of first impressions. Social Psychology Quarterly, 46(4), 285–292. Tetlock, P. E. (1985). Accountability: A social check on the fundamental attribution error. Social Psychology Quarterly, 48(3), 227–236. Tetlock, P. E. (1992). The impact of accountability on judgment and choice: Toward a social contingency model. Advances in Experimental Social Psychology, 25, 331–376. Tetlock, P. E. (1999). Accountability theory: Mixing properties of human agents with properties of social systems. In: L. L. Thompson & J. M. Levine (Eds), Shared cognition in organizations: The management of knowledge. LEA’s organization and management series (pp. 117–137). Mahwah, NJ: Lawrence Erlbaum Associates. Tetlock, P. E., & Boettger, R. (1989). Accountability: A social magnifier of the dilution effect. Journal of Personality and Social Psychology, 57(3), 388–398. Tetlock, P. E., & Kim, J. I. (1987). Accountability and judgment processes in a personality prediction task. Journal of Personality and Social Psychology, 52(4), 700–709. Tetlock, P. E., Lerner, J. S., & Boettger, R. (1996). The dilution effect: Judgmental bias, conversational convention, or a bit of both? European Journal of Social Psychology, 26(6), 915–934. Tetlock, P. E., 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. Tversky, A., & Kahneman, D. (1982). Judgments of and by representativeness. In: D. Kahneman, P. Slovic & A. Tversky (Eds), Judgment under uncertainty: Heuristics and biases (pp. 84–98). Cambridge, England: Cambridge University Press. Waller, W. S., & Zimbelman, M. F. (2003). A cognitive footprint in archival data: Generalizing the dilution effect from laboratory to field settings. Organizational Behavior and Human Decision Processes, 91, 254–268. Weldon, E., & Gargano, G. M. (1985). Cognitive effort in additive task groups: The effects of shared responsibility on the quality of multiattribute judgments. Organizational Behavior and Human Decision Processes, 36, 348–361. Wilks, T. J., & Zimbelman, M. F. (2004). Decomposition of fraud risk assessments and auditors’ sensitivity to fraud cues. Contemporary Accounting Research, 21(3), 719–745. Zukier, H. (1982). The dilution effect: The role of the correlation and the dispersion of predictor variables in the use of nondiagnostic information. Journal of Personality and Social Psychology, 43(6), 1163–11114.
This page intentionally left blank
26
AUDITORS’ MEMORY OF INTERNAL CONTROL INFORMATION: THE EFFECT OF DOCUMENTATION PREPARATION VERSUS REVIEW Lori S. Kopp and James L. Bierstaker ABSTRACT This study contributes to the cognitive processes and expertise research in judgment and decision-making in auditing. It uses the levels-of-processing theory (Craik & Lockhart, 1972) to investigate the amount of auditor attention given to information during internal control documentation procedures, and the effect of this attention on internal control information acquisition and risk assessment. Based on levels-of-processing, the attention required to complete an internal control questionnaire (ICQ) is predicted to result in the acquisition of more internal control information than when a completed ICQ is reviewed. In addition, auditors who complete an ICQ should assess control risk more like experts’ than auditors, who review an ICQ completed by another individual. Results suggest that the audit seniors who completed an ICQ retained significantly more internal control information than audit seniors who reviewed an ICQ completed by another individual. This result held when separately examining the internal control strengths and weaknesses. In addition, audit seniors Advances in Accounting Behavioral Research, Volume 9, 27–50 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09002-8
27
28
LORI S. KOPP AND JAMES L. BIERSTAKER
who completed an ICQ-assessed control risk at a level comparable to the control risk assessments of audit managers in the same firm.
INTRODUCTION Auditors of publicly traded companies must now express an opinion on the effectiveness of the management’s internal control over financial reporting (PCAOB, 2004). Although auditors have always been required to understand internal control to properly plan the audit (American Institute of Certified Public Accountant (AICPA), 1995), they now have the additional responsibility of identifying and documenting each material of internal control weakness. However, the PCAOB does not mandate the form of internal control documentation that auditors choose. Auditors may document internal control in the form of flowcharts, questionnaires, and/or narratives (AICPA, 1988; Canadian Institute of Chartered Accountants (CICA), 2003b, International Federation of Accountants (IFA), 2003).1 Given the recently elevated importance of internal control evaluation to the audit process, understanding how auditors’ documentation procedures affect their evaluation of controls is important. Consequently, this study examines how auditors’ documentation procedures affect their memory of internal control information and control risk assessments.2 Specifically, this study examines whether audit seniors from a Big 4 audit firm retain more internal control information when they complete an internal control questionnaire (ICQ) than when they review an ICQ completed by another individual. Based on the levels-of-processing framework (Craik & Lockhart, 1972), the attention required to complete an ICQ is predicted to result in the acquisition of more internal control information. This attention should also result in control risk assessments comparable to experts’ control risk assessments. To test these predictions, 76 audit seniors from one Big 4 firm evaluated internal control information using one of two documentation procedures. In one condition, participants reviewed a previously completed internal control narrative and then completed an ICQ. In the other, participants reviewed a previously completed narrative and ICQ. Participants then completed a test to examine the amount and type of internal control knowledge acquired and provided an internal control risk assessment. The results of the study indicate that audit seniors who completed an ICQ retained significantly more internal control information than audit seniors who reviewed an ICQ completed by another individual. Results further suggest that seniors who completed an ICQ made control risk assessments
Auditors’ Memory of Internal Control Information
29
that were comparable to risk assessments, which were made by managers of the same Big 4 firm after they reviewed the narrative and ICQ. However, the risk assessments of audit seniors who reviewed an ICQ were significantly different than managers’ risk assessments. These findings imply a potential advantage of completing (rather than reviewing) an ICQ during internal control evaluation, which may be particularly important given the heightened attention paid to internal controls under Sarbanes–Oxley (PricewaterhouseCooper (PWC), 2003, p. 13). This paper adds to the cognitive processes audit judgment and decisionmaking research by using the psychological theory of levels-of-processing to examine the effect of the depth of mental processing on memory of internal control information and control risk assessments. This study also contributes to audit judgment and decision-making research by examining performancebased measures of audit expertise. Performance-based measures have principally been used in studies aimed at assessing the quality of expert decisions (Bouwman & Bradley, 1997). The performance-based measure in this study is the comparison of the risk assessments of audit seniors and managers. The remainder of this study is organized as follows: the next section discusses theory and development of the hypotheses; the third section discusses the methodology to be used; the fourth section presents the results; and the final section summarizes and concludes the paper.
THEORY AND HYPOTHESIS DEVELOPMENT Documentation and Evaluation of Internal Control Information The Committee of Sponsoring Organizations of the Treadway Commission (COSO) and the Canadian Institute of Chartered Accountants Criteria of Control Board (COCO) define what constitutes internal control and set a standard against which the effectiveness of a firm’s internal controls can be evaluated (AICPA, 1992; CICA, 1995). The Auditing Standards Board amended SAS No. 55 to incorporate the COSO framework. This amendment, SAS No. 78, requires that auditors obtain ‘‘a sufficient understanding of internal control to plan the audit and to determine the nature, timing, and extent of tests to be performed’’ (AICPA, 1995).3 It also requires auditors to document in the audit working papers their understanding of an entity’s control activities and assessments of control risk. In addition, the Securities and Exchange Commission, making specific reference to COSO, requires all publicly held companies to use ‘‘a recognized internal control framework
30
LORI S. KOPP AND JAMES L. BIERSTAKER
that is established by a body or group that has followed due-process procedures’’ to evaluate internal control (IT Governance Institute, 2003). Auditors evaluate a client’s internal control system to determine the extent to which it is capable of preventing or detecting material misstatements, whether caused by error or fraud. Based on their evaluation of internal controls, auditors assess the risk that the control system will not prevent or detect material misstatements (i.e., control risk). Control risk assessments are directly associated with audit investment through their effect on the nature, timing, and extent of substantive testing (ISA 400; IFA, 2003). In addition, all publicly traded organizations in the United States are required to document, evaluate, monitor, and report on internal control, and auditors must issue a report on control effectiveness (PCAOB, 2004). Auditors have a choice of formats to use for documenting internal control. Two commonly used formats are narratives and questionnaires (Bierstaker, 1999a). Prior research has shown that the internal control documentation format influences an auditor’s data collection activities (Purvis, 1989), and may affect auditors’ encoding (Plumlee, 1985), recall (Bierstaker, 2003), and recognition (Plumlee, Tuttle, & Moeckel, 2002) of internal control information. Therefore, procedures used to document internal controls may influence auditors’ information processing activities, perhaps because of the manner in which auditors process internal control information into memory when completing and reviewing narratives and ICQs. Typically staff auditors prepare internal control documentation, and senior auditors evaluate controls (Abdolmohammadi & Usoff, 2001). In addition, internal control documentation may be obtained from internal auditors to enhance audit efficiency (Louwers, Ramsay, Sinason, & Strawser, 2005). Efficiency concerns also can cause auditors to rely on long-term memory, rather than performing a thorough re-reading of the working papers (Libby & Trotman, 1993). Less time spent re-reading working papers and greater reliance on long-term memory would mean less total time spent on audit procedures, and greater audit efficiency. Moreover, Moeckel and Plumlee (1989) and Harding, Hughes, and Trotman (2005) suggest that auditors rely confidently on items in memory, even incorrect ones. Thus, an increase in the number of accurate items in memory would improve audit effectiveness. Finally, it is not always feasible for auditors to take the time to refer back to the working papers. Therefore, they often must rely on memory for evidence encountered during the review process (Tan, 1995). The next section uses the levels-of-processing approach (Craik & Lockhart, 1972) as a framework for examining the relationship between attention and memory of internal control information.
Auditors’ Memory of Internal Control Information
31
LEVELS-OF-PROCESSING APPROACH Craik and Lockhart (1972) proposed the levels-of-processing framework, an approach that has been one of the many theories used to study human memory.4 In the levels-of-processing approach, information to be learned (i.e., stimuli), receives some type of mental processing. The by-product of this processing is a memory trace. Deeper processing of stimuli results in longer lasting and stronger traces. More information will be remembered when the memory trace is stronger. In the levels-of-processing framework, the amount of attention devoted to a stimulus will determine the depth to which information is processed. That is, more attention results in greater retention. In an audit setting, after reviewing a narrative describing the presence or absence of fraud indicators for a hypothetical audit client, auditors were asked to document this evidence using one of two documentation methods (Plumlee et al., 2002).5 Consistent with levels-of-processing theory, the auditors who used notes documented a higher proportion of fraud-related items than those who used checklists. That is, the auditors taking written notes likely used deeper processing and devoted more attention to the information in the narrative than the auditors using a checklist. In summary, the amount of information acquired is a function of the depth of processing, where greater attention increases the depth in which this information is processed. The next section discusses how elaborations of textual information and questions can be used to increase the amount of attention given to information. Elaborations of Textual Information and Questioning Techniques Results of non-accounting research examining elaborations of textual information can be used as a basis for considering auditors’ review of narrative information during internal control evaluation (Palmere, Benton, Glover, & Ronning, 1983). Self-generated elaborations of textual information resulted in more information stored in memory than experimenterprovided elaborations (Bobrow & Bower, 1969; Slamecka & Graf, 1978; Stein & Bransford, 1979). The PQ4R method is a technique intended to improve memory for textual material (Thomas & Robinson, 1972). It derives its name from the six phases of the method: preview, question, read, reflect, recite, and review. One of the key stages of this technique, read, involves question answering. During this stage, individuals answer the questions made up in the question stage. This
32
LORI S. KOPP AND JAMES L. BIERSTAKER
question-answering feature encourages deeper processing of textual information (Anderson, 1990), and has resulted in greater information retention. For example, either generating or answering questions while reading text resulted in retention of more information than just studying the textual material without having to generate or answer questions (Frase, 1975; Frase & Schwartz, 1975). In addition, focusing attention on text segments containing information related to the questions resulted in greater information retention (Reynolds & Anderson, 1982). Having to generate answers to questions in the ICQ is expected to cause auditors to focus more of their attention on information in the internal control narrative. This increased attention should result in deeper processing and greater attention to internal control information than reviewing an ICQ in which the answers are already provided. In summary, the question answering required when completing an ICQ should result in more attention to, and greater retention of information in an internal control narrative than reviewing an ICQ. Based on the role of attention in the level-of-processing framework the first hypothesis is as follows: H1. Auditors’ memory for internal control information will be greater when they complete an ICQ rather than when they review an ICQ completed by another individual. Control Risk Assessments A performance-based measure is one technique that can be used to assess audit expertise (Bouwman & Bradley, 1997). Control risk assessments are an example of a performance-based measure. During internal control evaluation initial control risk assessments made by audit seniors are subsequently reviewed and adjusted (if necessary) by audit managers. Previous auditing research suggests that relevant task specific audit experience is likely to impact information acquisition and subsequent task performance (Bonner, 1990). For example, Mock and Turner (1981) and Biggs, Mock, and Watkins (1988) found a positive relationship between experience and the quantity of data collected. Biggs and Mock (1983) and Biggs, Messier, and Hansen (1987) linked experience with typical accounting and control systems to the completeness of information search. In other words, more experienced auditors acquired more relevant information than less ones. In addition, these studies suggest that task-specific experience can influence the quality of auditor decision processes. Similarly, Davis (1996) and Purvis (1989) found that task-specific experience improves auditors’
Auditors’ Memory of Internal Control Information
33
acquisition of internal control information. These findings suggest that the information acquisition strategy of more experienced auditors is influenced by their knowledge of the important features of a well-designed internal control system (Bonner & Pennington, 1991). Since less experienced auditors do not possess this knowledge, they may not acquire internal control information as thoroughly as audit managers, and their task performance is likely to suffer. However, if completing an ICQ leads to more complete information acquisition, and deeper levels of information processing for audit seniors, their performance may rise to the level of audit managers. Based on the level-of-processing framework the second hypothesis is: H2. Audit seniors who complete an ICQ will assess control risk more like managers’ assessments than audit seniors who review an ICQ completed by another individual.
METHOD Participants To test the hypotheses, 76 audit seniors from one Big 4 accounting firm completed an internal control evaluation case. The participants averaged 2.6 years (standard deviation ¼ 1.1 years) of experience in public accounting, their average age was 26.1 years (standard deviation ¼ 3.5 years), 53% were CPAs, and 53% were male. To ensure that the participants had experience documenting internal controls and assessing control risk, two questions were asked, both of which were measured on 11-point scales (0 ¼ ‘‘Never’’; 10 ¼ ‘‘Very often’’). The questions were as follows: ‘‘In the course of conducting audits, how often do you document internal controls?’’ and ‘‘In the course of conducting audits, how often do you assess control risk?’’ Participant responses for the documentation (mean ¼ 8.04; standard deviation ¼ 2.10) and control risk assessment (mean ¼ 7.89; standard deviation ¼ 2.46) questions indicate that the participants had considerable experience documenting internal controls and assessing control risk. In addition, expert data were collected from ten audit managers from the same Big 4 firm at the same training sessions where data were collected from audit seniors. The audit managers averaged 8.7 years (standard deviation 3.1 years) of experience in public accounting. Other research that included audit managers in their expert panel includes Libby and Libby (1989), Bedard, Biggs, and DiPietro (1998), and Bierstaker (2003).
34
LORI S. KOPP AND JAMES L. BIERSTAKER
Manipulated and Measured Variables The ICQ documentation procedure was manipulated at two levels. In the first condition, participants reviewed a narrative and completed an ICQ (COMPLETE). In the second condition participants reviewed a narrative and reviewed an ICQ previously completed by another person (REVIEW). The ICQ includes 21 internal controls for the purchasing cycle. These internal controls include the 20 internal controls taken from a list developed by Frederick (1991) as well as an additional control (vouchers are pre-numbered) that consistently appears in the auditing literature (e.g., Yost, 1997).6 The completed ICQ and narrative appear in the appendix. Three measured variables were used to test the hypotheses: (1) memory of internal control strengths, (2) memory of internal control weaknesses, and (3) internal control risk assessments. A recognition test was used to measure memory of internal control strengths and weaknesses. Recognition and recall tests are two methods used to measure memory (Ashcraft, 1994). A recognition test was used because memory is facilitated when the encoding context (e.g., the list of internal controls in the ICQ) is similar to the retrieval context (e.g., the list of internal controls in a recognition test).7 Recognition tests have been used in prior audit judgment research (Moeckel & Plumlee, 1989; Moeckel, 1990; Sprinkle & Tubbs, 1998, Ricchiute, 1999, Plumlee et al., 2002). Auditors make internal control risk assessments using qualitative (e.g., words) or quantitative scales (CICA, 2003b). Based on previous research (e.g., Choo & Trotman, 1991; Asare, 1992; Houston, Peters, & Pratt, 1999) an 11-point scale with linguistic descriptors (0 ¼ ‘‘Minimum’’; 5 ¼ ‘‘Moderate’’; 10 ¼ ‘‘Maximum’’) was used to elicit auditors’ control risk assessments. In recognition tests, individuals are shown two types of items: target items and distractor items (Ashcraft, 1994). The target (old) items are items originally reviewed or studied whereas distractor (new) items are those not originally reviewed or studied. In this study, the target items were the 21 internal controls listed in the ICQ. The distractor items were ten internal controls not originally reviewed. Yes/no questions are commonly used to measure memory on a recognition test. Table 1 shows the four possible outcomes in a recognition test. These outcomes are defined as follows: a hit is a ‘‘yes’’ response to an old item, a false alarm is a ‘‘yes’’ response to a new item, a correct rejection is a ‘‘no’’ response to a new item, and a miss is a ‘‘no’’ response to an old item. The percentage of hits for the internal control strengths and weaknesses was used to measure memory.8
35
Auditors’ Memory of Internal Control Information
Table 1.
Possible Outcomes in a Recognition Test. Participant’s Response
Test item
Olda Newb
Old
New
Hitc False alarm
Miss Correct rejection
a
Internal control strengths and weaknesses originally reviewed. Internal controls not originally reviewed. c The number of hits for the internal control strengths and weakness was used to measure memory of internal control information. b
Experimental Procedures Prior to formal experimentation, steps were taken to ensure that the experimental materials were externally valid. The research instrument was developed with assistance from ten auditors representing all Big 4 accounting firms, who provided insights about their firms’ internal control evaluation process and reviewed preliminary versions of the materials. The case also was pilot tested with 54 graduate accounting students. The experiment was conducted with two classes attending a Big 4 firm’s national training program. One of the researchers was present during the administration of the experiment that took place during a 45-minute session. Participants in the first class session were asked not to discuss the experiment with any other individuals participating in the training program. Participants in the second session communicated that they did not have any prior knowledge of the experiment. There were 44 (32) participants in the first (second) class session. Both documentation manipulations were presented during each of the experimental sessions. This was done to assure that there were an equal number of participants in the two documentation conditions. Pilot testing showed there was no difference in the performance time of individuals in the two documentation conditions. The sequence of experimental tasks is shown in Fig. 1. The case materials consisted of three parts. Part 1 consisted of an introduction and the documentation procedures. In the introduction the participants were told they would be examining internal control evaluation. After the introduction they were given a narrative to review and an ICQ to either review or complete. Because of time constraints, only internal controls for the purchasing cycle were examined. An evaluation at the cycle level is appropriate because control risk can be assessed at the account balance or class-of-transactions
36
LORI S. KOPP AND JAMES L. BIERSTAKER Part 1 1. Introduction 2. Documentation procedures a. COMPLETE condition: Reviewed a narrative and completed an ICQ b. REVIEW condition: Reviewed a narrative and an ICQ Part 2 3. Cognitive ability assessment 4. Demographic questions Part 3 5. Control risk assessment 6. Recognition test
Fig. 1.
Sequence of Experimental Tasks
level (AICPA, 1983; CICA, 2003b). The purchasing cycle was chosen because there is a higher likelihood of finding both internal control strengths and weaknesses during internal control evaluation for this cycle (Reckers & Taylor, 1979). The narrative used in the experiment was organized in the order in which documents flow through the accounting system. Based on interviews with Big 4 auditors, this organization is consistent with practice. In Part 2 of the case, presented after completing the documentation procedure, an assessment of cognitive ability and demographic questions were used as distractor tasks to clear short-term memory before the auditors were asked to make the control risk assessment and perform the recognition test. The Wonderlic Personnel Test was used to measure cognitive ability (Wonderlic, 1999).9 Part 3 of the case materials consisted of the control risk assessment and the recognition test. The participants did not know ahead of time that they would be performing a recognition test. This is consistent with prior research (Moeckel & Plumlee, 1989; Moeckel, 1990; Sprinkle & Tubbs, 1998, Ricchiute, 1999; Plumlee et al., 2002; Harding et al., 2005). To ensure that the auditors relied on memory, and not documentation, the narrative and ICQ were not made available when the risk assessments were made. This is consistent with prior research where auditors were not able to refer back to previous information before taking a memory test (Moeckel & Plumlee, 1989; Moeckel, 1990; Christ, 1993; Sprinkle & Tubbs, 1998; Plumlee et al., 2002). To prevent the list of internal controls from acting as cues during the risk assessment process, the participants took the recognition test after the risk assessment. The participants were not allowed to adjust their control risk assessments after they completed the recognition test.
Auditors’ Memory of Internal Control Information
37
RESULTS Preliminary Analyses Audit seniors were asked to document internal controls and then assess control risk in this study. As an additional external validity check, participants were asked how often they documented internal controls before assessing control risk when conducting audits. An 11-point scale (0 ¼ ‘‘Never’’; 10 ¼ ‘‘Very often’’) was used to answer this question. A mean response of 7.22 (standard deviation ¼ 2.49) showed that the participants had adequate experience performing this activity. Ten managers from the same Big 4 firm were also asked how often the same person completes the documentation and assesses control risk. A mean response of 6.86 (standard deviation ¼ 2.04) on an 11-point scale (0 ¼ ‘‘Never’’; 5 ¼ Often; 10 ¼ ‘‘Very often’’) provides additional evidence that auditors often document internal controls before making control risk assessments. Hypothesis Testing H1 predicts that memory for internal control information will be greater when participants review a narrative and complete an ICQ (COMPLETE) than when they review both a narrative and a completed ICQ (REVIEW).10 Memory of internal control information was measured in two ways: the percentage of the 13 internal control strengths identified (PER_CS), and the percentage of the eight internal control weaknesses identified (PER_CW). Standard deviation measures for the two memory variables demonstrate that participants in the REVIEW condition had greater variability in their memory of internal control information, than participants in the COMPLETE condition. The results reveal that the effect of documentation procedure on memory of internal control information is in the predicted direction. Specifically, participants in the COMPLETE condition retained more internal control information than those in the REVIEW condition. Table 2 provides the means, medians, and standard deviations for the two memory variables for each documentation condition. Results of t-tests demonstrate that participants in the COMPLETE condition had significantly better memory of internal control information than participants in the REVIEW condition for PER_CS (t ¼ 1.93; p ¼ 0.029) and PER_CW (t ¼ 2.43; p ¼ 0.009), supporting H1. Predictions are directional; therefore, p-values are one-tailed. As Levene’s tests suggest differences in the variances for the two memory
38
Table 2.
LORI S. KOPP AND JAMES L. BIERSTAKER
Memory of Internal Control Information by Documentation Condition Means [Medians] (Standard Deviations).
Documentation Condition
Memory Variables PER_CS
PER_CW
COMPLETE N ¼ 38
87.65% [92.31] (12.84%)
74.67% [75.00] (19.82%)
REVIEW N ¼ 38
81.17% [84.62] (16.26%)
61.84% [62.50] (25.82%)
Notes: PER_CS ¼ Percentage of the 13 internal control strengths identified; PER_CW ¼ Percentage of the 8 internal control weaknesses identified; COMPLETE ¼ Auditors reviewed a narrative and completed an internal control questionnaire; REVIEW ¼ Auditors reviewed a narrative and internal control questionnaire.
variables are significant (PER_CS: F ¼ 5.10, p ¼ 0.03; PER_CW: F ¼ 4.70, p ¼ 0.03), H1 was also tested using nonparametric robust rank-order tests.11 Similar results are obtained using this procedure. Additional analyses were performed to ensure results found for H1 were not attributable to other variables. The possible effects of personal characteristics (i.e., age, gender, public accounting experience, cognitive ability) and internal control documentation and control risk assessment experience on memory of internal control information were examined. Age, gender, months of public accounting experience, internal control documentation experience, and control risk assessment experience were not significantly related to any of the memory variables. The results of the t-test demonstrate that there was not a significant difference (t ¼ 0.411; p ¼ 0.682) in the cognitive ability of the audit seniors in the two documentation conditions. Spearman rank-order correlations found that cognitive ability was significantly related to PER_CW (r ¼ 0.232; p ¼ 0.044), consistent with prior research (Bierstaker & Wright, 2001). However, including cognitive ability as a control variable did not affect the results. Signal detection theory indices, A0 and B00 , are used to measure the quality of the participants’ memory performance (Snodgrass & Corwin, 1988; Neath, 1998; Sprinkle & Tubbs, 1998). Hit and false alarm rates are used to calculate these indices.12 As Levene’s tests suggest differences in the variances for the hit and false alarm rates, nonparametric analyses were performed to calculate A0 and B00 (Snodgrass & Corwin, 1988; Neath, 1998).
39
Auditors’ Memory of Internal Control Information
A0 measures the degree to which participants were able to discriminate between old and distractor (new) items.13 It ranges from 0 to 1 with 0.5 reflecting chance performance. Numbers greater than 0.5 demonstrate the ability to better discriminate old items from the distractor items. B00 is an index that measures the response bias of individuals taking a recognition test. This measure ranges from 1 to 1. Zero represents no bias. A positive number represents conservative bias, defined as a tendency to respond ‘‘new’’ more often than ‘‘old’’. Negative numbers represent liberal bias, defined as a tendency to respond ‘‘old’’ more often than ‘‘new’’. Table 3 provides the means, medians, and standard deviations for A0 and 00 B for each documentation condition. Kolmogorow–Smirnov analysis suggests that the A0 measure is not normally distributed. As a result, Mann– Whitney U tests are used to examine differences in A0 for the two documentation groups. Mann–Whitney U/t-tests demonstrate that there was not a significant difference in the A0 /B00 measures of participants in the two documentation conditions.14 This was expected based on the random assignment of participants. Consequently, discussion of A0 and B00 is limited to the overall group of participants. A mean of 0.75 for A0 indicates that the Table 3. Memory of Internal Control Information by Documentation Condition Utilizing Signal Detection Theory Indices Means [Medians] (Standard Deviations). Documentation Condition
Signal Detection Theory Indices A0
B00
COMPLETE N ¼ 38
0.78 [0.89] (1.67)
0.06 [ 0.05] (0.49)
REVIEW N ¼ 38
0.72 [0.81] (0.25)
0.01 [ 1.1] (0.53)
OVERALL N ¼ 76
0.75 [0.81] (0.23)
0.02 [ 0.09] (0.51)
Notes: A0 measures the degree to which participants were able to discriminate between old and distractor (new) items in a recognition test. A0 ranges from 0 to 1. Numbers greater than 0.5 demonstrate the ability to better discriminate old items from distractor items. B00 measures the response bias of individuals taking a recognition test. B00 ranges from 1 to 1. Zero represents no bias. Positive/negative numbers represent conservative/liberal bias.
40
LORI S. KOPP AND JAMES L. BIERSTAKER
participants were able to discriminate old items from the distractor items relatively well. A mean B00 measure of 0.02 suggests that the participants exhibited almost no response bias. Therefore, the quality of the memory performance of the participants was high. H2 predicts that audit seniors who completed an ICQ will assess control risk more like audit managers’ assessments than audit seniors who review an ICQ completed by another individual. Testing compared audit seniors control risk assessments with risk assessments of ten audit managers in the same Big 4 firm. Like the audit seniors in the REVIEW condition, the audit managers made a control risk assessment after reviewing the narrative and completed ICQ. As Levene’s tests suggest differences in the variance for the risk assessments (F ¼ 4.34, p ¼ 0.016), testing is performed using nonparametric robust rank-order tests. Audit managers had significantly higher (U` ¼ 2.216; p ¼ 0.027) risk assessments (median ¼ 5.50; mean ¼ 6.00; standard deviation ¼ 2.40) than audit seniors in the REVIEW (median ¼ 4.50, mean ¼ 4.45; standard deviation ¼ 1.69) condition. There was no significant difference (U` ¼ 1.27; p ¼ 0.206) in the risk assessments of the managers and audit seniors in the COMPLETE (median ¼ 5.00; mean ¼ 5.00; standard deviation ¼ 2.43) condition.15 Overall, these results support H2 and suggest completing an ICQ aided auditors’ risk assessment performance.16
SUMMARY AND CONCLUSIONS This study extends cognitive process audit judgment and decision-making research by using the level-of-processing framework to examine how documentation procedure affects memory of internal control information and internal control risk assessments. More specifically, it investigates the amount of attention given to internal control information, and the effect of this attention on internal control information acquisition. Results indicate that the audit seniors who completed an ICQ retained significantly more internal control information than audit seniors who reviewed an ICQ completed by another individual, suggesting that preparing an ICQ may enhance auditors’ knowledge of an auditee’s internal control. This result held while separately examining the internal control strengths and weaknesses. Identification of internal control weaknesses is required by Section 404 of Sarbanes–Oxley and is important for error and fraud detection as well as avoidance of audit failure. Identification of internal control strengths is important to avoid unnecessary substantive testing.
Auditors’ Memory of Internal Control Information
41
Results also indicate that participants in the REVIEW condition had greater variability in their memory of internal control information. One of the goals of the audit profession is to have consistency across engagements (Bell & Wright, 1995). The greater variability of the memory of auditors in the REVIEW documentation condition could lead to more inconsistencies during internal control evaluation. The user involvement in internal control documentation performed by auditors in the COMPLETE condition could increase the level of consistency, an important consideration given the increased responsibilities for internal control evaluation imposed on management and auditors by Sarbanes–Oxley (PWC, 2003). The results of this study have implications for improving audit effectiveness. Audit seniors in the COMPLETE condition made risk assessments that were comparable to subsequent risk assessments made by managers of the same firm. That is, better memory of internal control information led to more expert control risk assessments because more attributes of the control system were considered when the assessment was made. Libby and Libby (1989) demonstrated how auditors’ use of a mechanical decision aid provided control reliance decisions that were more like the decisions made by a group of firm experts. Similarly, this study contributes to internal control judgment research and expertise research. Studies aimed at assessing the quality of expert decisions typically use performance-based measures (Bouwman & Bradley, 1997). The performance-based measure in this study is the comparison of the risk assessments of audit seniors and managers. By demonstrating that the completion of an ICQ leads to control risk assessments more like experts’ (i.e., audit managers) control risk assessments, the results of this study suggest questionnaires may elevate the performance of auditors to be more like experts. In addition, the results of this study suggest that auditors who review a narrative and prepare an ICQ acquire more information than auditors who review a narrative and ICQ. Taken together, these results imply questionnaires enhance auditor expertise by improving their depth of information processing. In the REVIEW condition the auditors reviewed documentation prepared by someone else. Audit staff members often prepare this documentation (Abdolmohammadi & Usoff, (2001). In an audit setting, competitive pressures may lead auditors to place greater reliance on audit staff for internal control information because it is cheaper to do so. Based on the results of this study, reliance on audit staff for ICQ documentation would lead to less effective internal control evaluation. Thus, there is efficiency versus effectiveness trade-offs involved in having staff versus senior auditors prepare ICQs.
42
LORI S. KOPP AND JAMES L. BIERSTAKER
Like all research, this study is subject to limitations. The first limitation was the procedure used to compare the risk assessments of the audit seniors and managers. The managers used the same documentation condition as participants in the REVIEW condition. If the managers completed the ICQ, there might have been significant differences in the control risk assessments of participants in the COMPLETE condition and the audit managers. While there might have been a significant difference in the performance of these two groups, audit managers would typically rely on seniors or staff to complete an ICQ on actual audit engagements. A second limitation was the type of experimental materials used in the study. The documentation about the client’s internal control system was limited to written information. Auditors work in a richer information environment when evaluating internal control. Although prior or current contact with the client can influence auditors’ evaluation of internal controls during audit planning; it was not part of the experimental design. Future research can manipulate or control for client involvement. The third limitation was the duration of the distractor tasks in this study. Consistent with the psychology literature, the distractor tasks used in this study to clear short-term memory were relatively brief. The study did not examine the effects when the time interval between examining information and recognition of information is extended, as can be the case in some audit situations. Future research could extend the duration of this time interval to determine if a longer-time period would lead to similar results. The study’s results suggest other possibilities for further research. This study demonstrated that more attention during internal control documentation led to better memory of internal control information and more accurate control risk assessments. Future research could examine how knowledge embedded in questionnaires may enhance auditor performance when evaluating internal controls (Bierstaker & Thibodeau, 2006) and how the review or completion of an ICQ may influence an auditor’s planned extent and nature of testing. Further, future research can examine whether attention during documentation of going-concern and fraud risk assessment tasks provides similar results. It can also examine what additional procedures besides user involvement in documentation cause auditors to devote more attention to internal control or other audit information. Results show that auditors were able to recognize a higher percentage of internal control strengths than weaknesses. Future research can investigate what procedures could better improve memory of internal control weaknesses during internal control evaluation. Finally, this study examined the relationship between individual
Auditors’ Memory of Internal Control Information
43
task involvement, memory and control risk assessments. The influence of group task involvement on memory or decision-making also could be examined.
NOTES 1. Narratives, internal control questionnaires, and flowcharts have been used in practice and experimental research to document an understanding of internal control (e.g., Ashton, 1974; Bierstaker, 1999a). Since narratives and questionnaires are used most frequently to document internal control, and are often used in combination (Bierstaker, 1999a), this study investigates these documentation formats. 2. In this study, memory is defined in terms of the amount and types (i.e., internal control strengths and weaknesses) of internal control knowledge. 3. This is also required by Canadian and International Standards (CICA, 2003a; International Federation of Accountants, 2003) 4. Examples of other theories used in audit memory research are the split-attention effect during internal control evaluation (Bierstaker, 1999b), the use of distributive memory theories to examine memory conjunction errors in a multiple client audit environment (Lindberg & Maletta, 2003) and the use of impression formation to demonstrate that the integration of inconsistent items lead to deeper processing and better memory of items suggesting firm failure (Libby & Trotman, 1993). 5. The first method was a 12-item yes/no checklist of items that indicated a high or low risk of management fraud. In the second method, the auditors were instructed to take notes that they might use later to write a memorandum documenting evidence about the likelihood of fraud. Auditors in the notes condition were cued with questions that prompted them to address the same twelve items on the checklist. Three of the items on the 12-item checklist were indicators of fraud. 6. Thirteen of the internal controls serve as internal control strengths, while eight serve as weaknesses. The proportion of internal control strengths and weaknesses is similar to the proportion of positive (six) and negative (12) indicators of fraud used in a study of fraud risk assessment (Plumlee et al., 2002). 7. This is known as the encoding specificity principle (Tulving & Thomson, 1973). 8. Participants’ ability to discriminate between old and new items also was examined. Indices suggested by signal detection theory were used to measure the discriminative ability and response bias of the participants. The hit and false alarm rates were used to calculate these indices. 9. The Wonderlic Personnel Test is a 12-minute short-form test of cognitive ability. It has been administered to more than 100 million individuals. It is correlated highly with longer tests of cognitive ability, such as the Wechler Adult Intelligence Scale – Revised and the Otis–Lennon Ability test. Reliabilities range between 0.88 and 0.94 when measured for internal consistency (Wonderlic Personnel Test, 1998). 10. The ICQ was completed with a mean accuracy rate of 86% (standard deviation ¼ 9%). ICQ accuracy was significantly correlated with both CS and CW. 11. The robust rank-order test makes less stringent distributional assumptions about the sampled population than the Mann–Whitney U test and may be more appropriate where significant variation in sample variances are detected (Siegel & Castellan 1988).
44
LORI S. KOPP AND JAMES L. BIERSTAKER
12. Memory accuracy (i.e., a hit rate) of 90% and a false alarm rate of 90% tell us that the participant responded ‘‘yes’’ to 90% of the target and distractor items. This individual was not able to discriminate between target and distractor items. Caution should be used when interpreting memory accuracy results if participants do not demonstrate the ability to discriminate between target and distractor items and they respond ‘‘old’’ more often than ‘‘new’’ (i.e., demonstrate liberal response bias) to items presented to them. 13. See Snodgrass and Corwin (1988) for descriptions of how to calculate the A0 and B00 indices. 14. Similar results are obtained if A0 /B00 analyses are performed using parametric t-tests. 15. Similar results are obtained if H2 was tested using parametric multiple comparison least significant difference (LSD) tests. 16. Additional analyses were performed to ensure results found for H2 were not attributable to other variables. For the audit seniors age, internal control documentation experience, cognitive ability, public accounting experience, gender, and control risk assessment experience were not significantly related to control risk assessment. Therefore, these variables were not included in H2 testing.
ACKNOWLEDGMENTS The authors thank the auditors from a Big 4 firm for participating in the study. We gratefully acknowledge the valuable comments of Vicky Arnold (the editor), the associate editor, two anonymous reviewers, and participants at the 2001 American Accounting Audit Midyear Meeting and the 2002 Canadian Academic Accounting Association Conference.
REFERENCES Abdolmohammadi, M. J., & Usoff, C. A. (2001). The assessment of task structure, knowledge base, and decision aids for a comprehensive inventory of audit tasks. Westport, CT: Quorum Books. American Institute of Certified Public Accountants (AICPA). (1983). Statement on auditing standards no. 47: Audit risk and materiality in conducting an audit. New York, NY: American Institute of Certified Public Accountants, Inc. American Institute of Certified Public Accountants (AICPA). (1988). Statement on auditing standards no. 55: Consideration of the internal control structure in a financial statement audit. New York, NY: American Institute of Certified Public Accountants, Inc. American Institute of Certified Public Accountants (AICPA). (1992). Internal control – integrated framework. Jersey City, NJ: AICPA. American Institute of Certified Public Accountants (AICPA). (1995). Statement on auditing standards no. 78: Consideration of internal control in a financial statement audit: An amendment to SAS no. 55. New York: American Institute of Certified Public Accountants, Inc.
Auditors’ Memory of Internal Control Information
45
Anderson, J. R. (1990). Cognitive psychology and its implications (3rd ed.). New York: W.H. Freeman and Company. Asare, S. K. (1992). The auditor’s going-concern decision: Interaction of task variables and the sequential processing of evidence. The Accounting Review, 67(2), 379–383. Ashcraft, J. H. (1994). Human memory and cognition (2nd ed.). New York: HarperCollins College Publishers. Ashton, R. H. (1974). An experimental study of internal control judgments. Journal of Accounting Research, 12(1), 143–157. Bedard, J., Biggs, S., & DiPietro, J. (1998). The effects of hypothesis quality, client management explanations, and industry experience on audit planning decisions. Advances in Accounting, (16), 49–73. Bell, T. B., & Wright, A. M. (1995). Auditing practice, research, and education: A productive collaboration. New York: American Institute of Certified Public Accountants. Bierstaker, J. L. (1999a). Performance in internal control evaluation: Internal control documentation: Which format is preferred? The Auditor’s Report, 22(2), 12–13. Bierstaker, J. L. (1999b). A test of the split-attention effect in a professional context. Journal of Business and Behavioral Sciences, 6(2), 177–189. Bierstaker, J. L. (2003). Auditor recall and evaluation of internal control information: Does taskspecific knowledge mitigate part-list interference? Managerial Auditing Journal, (18), 90–99. Bierstaker, J. L., & Thibodeau, J. (2006). Implementing Section 404 of the Sarbanes–Oxley Act of 2002: The effect of documentation format and task-specific experience on auditor internal control evaluation. Managerial Auditing Journal, forthcoming. Bierstaker, J. L., & Wright, S. (2001). A research note concerning practical problem-solving ability as a predictor of performance in suditing tasks. Behavioral Research in Accounting, (13), 49–62. Biggs, S. F., & Mock, T. J. (1983). An investigation of auditor decision processes in the evaluation of internal controls and audit scope decisions. Journal of Accounting Research, 63(1), 234–255. Biggs, S. F., Messier, W. F., Jr., & Hansen, J. V. (1987). A descriptive analysis of computer audit specialists’ decision-making behavior in advanced computer environments. Auditing: A Journal of Practice and Theory, 6(2), 1–21. Biggs, S. F., Mock, T. J., & Watkins, P. R. (1988). Auditors’ use of analytical review in audit program design. Accounting Review, 63(1), 148–161. Bobrow, S. A., & Bower, G. H. (1969). Comprehension and recall of sentences. Journal of Experimental Psychology, 80(3), 455–461. Bouwman, M. J., & Bradley, W. E. (1997). Judgment and decision making, part II: Expertise, consensus, and accuracy. In: V. Arnold & S. G. Sutton (Eds), Behavioral Accounting Research: Foundations and frontiers. Sarasota: American Accounting Association. Bonner, S. E. (1990). Experience effects in auditing: The role of task-specific knowledge. The Accounting Review, 65(1), 72–92. Bonner, S. E., & Pennington, N. (1991). Cognitive processes and knowledge as determinants of auditor expertise. Journal of Accounting Literature, 10, 1–50. Canadian Institute of Chartered Accountants (CICA). (2003a). CICA handbook assurance: Section 5205 internal control in the context of an audit, understanding internal control for audit planning purposes. Toronto: Canadian Institute of Chartered Accountants. Canadian Institute of Chartered Accountants (CICA). (2003b). CICA handbook assurance: Section 5210 assessing control risk/documenting control risk assessments. Toronto: Canadian Institute of Chartered Accountants.
46
LORI S. KOPP AND JAMES L. BIERSTAKER
Canadian Institute of Chartered Accountants (CICA). (1995). Canadian Institute of Chartered Accountants Criteria of Control Board (COCO): Guidance on control. Toronto: Canadian Institute of Chartered Accountants. Choo, F., & Trotman, K. T. (1991). The relationship between knowledge structure and judgments for experienced and inexperienced auditors. The Accounting Review, 6(3), 464–485. Christ, M. (1993). Evidence on the nature of audit planning problem representations: An examination of auditor free recalls. The Accounting Review, 68(April), 304–322. Craik, F. I. M., & Lockhart, R. S. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11(December), 671–684. Davis, J. T. (1996). Experience and auditors’ selection of relevant information for preliminary control risk assessments. Auditing: A Journal of Practice and Theory, 15(1), 16–37. Frase, L. T. (1975). Prose processing. In: G. H. Bower (Ed.), Psychology of learning and motivation, Vol. 9. New York: Academic Press. Frase, L. T., & Schwartz, B. J. (1975). Effect of question production and answering on prose recall. Journal of Educational Psychology, 67(October), 628–635. Frederick, D. M. (1991). Auditors’ representation and retrieval of internal control knowledge. The Accounting Review, 66(1), 240–258. Harding, N.S., Hughes, S., & Trotman, K.T. (2005). Auditor calibration in the review process. Advances in Accounting Behavioral Research, (8), 41–57. Houston, R. W., Peters, M. F., & Pratt, J. H. (1999). The audit risk model, business risk and audit-planning decisions. The Accounting Review, 74(3), 281–298. International Federtation of Accountants (IFA). (2003). Handbook of international auditing, assurance and ethics pronouncements. New York, NY: IFA. IT Governance Institute. (2003). IT control objectives for Sarbanes–Oxley. Rolling Meadows, IL: IT Governance Institute. Libby, R., & Libby, P. A. (1989). Expert measurement and mechanical combination in control reliance decisions. The Accounting Review, 64(4), 729–747. 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. Lindberg, D. L., & Maletta, M. M. (2003). An examination of memory conjunction errors in multiple client audit environments. Auditing: A Journal of Practice and Theory, 22(1), 126–141. Louwers, T. J., Ramsay, R. J., Sinason, D. H., & Strawser, J. R. (2005). Auditing and assurance services. New York, NY: McGraw-Hill/Irwin. Mock, T., & Turner, J. L. (1981). Internal accounting control evaluation and auditor judgment Audit Research Monograph no. 3. New York: AICPA. Moeckel, C. (1990). The effect of experience on auditors’ memory errors. Journal of Accounting Research, 28(Autumn), 368–387. Moeckel, C., & Plumlee, R. D. (1989). Auditors confidence on recognition of audit evidence. The Accounting Review, 64(4), 653–666. Neath, I. (1998). Human memory: An introduction to research, data, and theory. Pacific, Grove, CA: Brooks/Cole. Palmere, M., Benton, S. L., Glover, J. A., & Ronning, R. R. (1983). Elaboration and recall of main ideas in prose. Journal of Educational Psychology, 76(December), 898–907. PCAOB. (2004). An audit of internal control over financial reporting performed in conjunction with an audit of financial statements. Release No. 2004-001 – March 9, 2004. Plumlee, R. D. (1985). The Standard of objectivity for internal auditors: Memory and bias effects. Journal of Accounting Research, 23(2), 683–699.
Auditors’ Memory of Internal Control Information
47
Plumlee, R.D., Tuttle, B., & Moeckel, C.L. (2002). Auditors’ memory for documented evidence. Advances in Accounting Behavioral Research, (5), 51–75. PricewaterhouseCoopers (PWC). (2003). Stand and be counted. New York, NY: Advertorial Campaign. Purvis, S. E. C. (1989). The effect of audit documentation format on data collection. Accounting, Organizations and Society, 14(5/6), 551–563. Reckers, P. M., & Taylor, M. E. (1979). Consistency in auditors’ evaluations of internal accounting controls. Journal of Accounting and Finance, 3(1), 42–55. Reynolds, R. E., & Anderson, R. C. (1982). Influence of questions on the allocation of attention during reading. Journal of Experimental Psychology, 74(October), 623–632. Ricchiute, D. N. (1999). The effect of audit seniors’ decisions on working paper documentation on partners’ decisions. Accounting, Organizations and Society, 24(2), 155–171. Slamecka, N. J., & Graf, P. (1978). The generation effect: Delineation of a phenomenon. Journal of Experimental Psychology: Human Learning and Memory, 4(Nov), 592–604. Snodgrass, J. G., & Corwin, J. (1988). Pragmatics of measuring recognition memory: Applications to dementia amnesia. Journal of Experimental Psychology: General, 117(March), 34–50. Sprinkle, G. B., & Tubbs, R. M. (1998). The effects of audit risk and information importance on auditor memory during working paper review. The Accounting Review, 73(4), 475–502. Stein, B. S., & Bransford, J. D. (1979). Constraints on effective elaboration: Effects of precision and subject generation. Journal of Verbal Learning and Verbal Behavior, 18(December), 769–777. Tan, J. (1995). Effects of expectations, prior involvement, and review awareness on memory for audit evidence and judgment. Journal of Accounting Research, 33(1), 113–135. Thomas, E. L., & Robinson, H. A. (1972). Improving reading in everyday class: A sourcebook for teachers. Boston, MA: Allyn and Bacon. Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352–373. Wonderlic, C. F. (1999). Wonderlic personnel test – form II. Libertyville: IL: Wonderlic Personnel Test, Inc. Wonderlic Personnel, Test Inc.. (1998). Wonderlic personnel test and scholastic level exam – user’s manual. Libertyville: IL: Wonderlic Personnel Test, Inc.. Yost, G. C. (1997). The audit its environment and application: An experiential approach. Upper Saddle River, NJ: Prentice-Hall.
APPENDIX: CASE MATERIALS Internal Control Narrative: Wittim Medical Supplies, Inc. Wittim Medical Supplies, Inc. manufactures a variety of medical supplies, including test tubes, thermometers, and disposable surgical garments. The following narrative describes the accounting system and related internal controls for Wittim’s purchasing cycle. If an internal control is not being described in this narrative, it is not addressed within Wittim’s control system. The supplies manager initiates the purchase and maintains the inventory records. The records include reorder points for all regularly used items. The
48
LORI S. KOPP AND JAMES L. BIERSTAKER
supplies manager prepares a purchase requisition on a two-part pre-numbered form. After signing the requisition, he files one copy by requisition number and sends the other copy to the purchasing department. Requisitions for items that will cost over $100 must be approved by the production manager before being sent to the purchasing department. Purchasing Department The purchasing department checks a purchase requisition for proper approval and selects a vendor. A five-part pre-numbered purchase order is prepared. Copies are sent to the vendor, receiving department, accounts payable, and the supplies manager. The purchasing department records the current purchase and files its purchase order and requisition copies by purchase order number in the open order file. The receiving department files its copy in a file by purchase order number. The supplies manager files his copy with its corresponding purchase requisition. Receiving Department An authorized receiving department employee counts the goods to verify that they were received, compares the count to the packing slip, and prepares a four-part receiving report. Copies of the receiving report are sent to the supplies manager, the purchasing department, and accounts payable. The receiving department files the copy of the receiving report and the packing slip with its copy of the purchase order. The supplies manager updates the inventory records when he receives the receiving report and then files the purchase order, purchase requisition, and receiving report by purchase order number. The purchasing department files its copy with the order in the open order file. The purchasing department receives two-part invoices from the vendors. One copy of the approved invoice is sent to accounts payable. The purchase order, purchase requisition, invoice, and receiving report are then filed in the closed order file by purchase order number. Accounts Payable Department The accounts payable department receives and matches purchase orders and approved invoices from the purchasing department with receiving reports from the receiving department. The invoices are checked for prices, quantities, and mathematical accuracy. A clerk initials them if they are accurate.
49
Auditors’ Memory of Internal Control Information
When all the documents are received, a clerk posts the payable amount to the particular vendor’s payable account, prepares a pre-numbered disbursement voucher, and attaches it to the purchase order, receiving report, and invoice. The package is then given to the accounts payable manager for review and approval for payment. The manager gives the approved disbursement vouchers to the second clerk. The clerk batches and totals the approved vouchers and prepares a batch summary. The batch summary is sent to the accounting department. A third clerk completes a two-part, prenumbered check for each disbursement voucher. The check and the disbursement vouchers are sent to the cashier. Cash Disbursements Department The cashier totals the checks and compares the total to the batch summary. She then signs the checks with the treasurer’s signature using a check-signing machine. She is authorized by the board of directors to disburse company funds. She then places the first copies of the check/remittances in envelopes, and sends them to the vendors. The second copy is sent to the accounting department. The checks for aborted payments are marked to prevent their payment. Internal Control Questionnaire Purchasing Cycle (Review Condition) Please review the following internal control questionnaire. Internal controls Authorization objective (1) Purchase orders are authorized. (2) There are authorization procedures and limits on buying power (3) There is an approved list of suppliers (4) Access to receiving areas is authorized and controlled (5) Cash disbursers are authorized by the board of directors (6) Ability to record goods received in the inventory records is restricted to those authorized Accuracy objective (7) Certain limits or ranges are put into place so that only correct information can be entered into the purchase order (8) Invoices are checked for prices, quantities, and mathematical accuracy
Yes/No Y Y N N Y N
N Y
50
LORI S. KOPP AND JAMES L. BIERSTAKER
(9) Batch totals of invoices entered are compared to totals of invoices recorded (10) Cash disbursement records are reconciled to monthly bank statements Completeness objective (11) Purchase orders are pre-numbered (12) Receiving reports are pre-numbered (13) Receiving reports are matched to purchase orders (14) Invoices are matched to purchase orders (15) Vouchers are pre-numbered (16) Payments are pre-numbered Validity objective (17) Invoices are matched to receiving reports (18) Goods received are physically verified (19) Supporting documentation is reviewed before authorizing payments (20) Paid invoices are effectively cancelled (21) Aborted payments are marked to prevent their being treated as valid payments
N N
Y N Y Y Y Y Y Y Y N Y
INTERNAL AUDITOR BURNOUT: AN EXAMINATION OF BEHAVIORAL CONSEQUENCES Timothy J. Fogarty and Lawrence P. Kalbers ABSTRACT The burnout condition of employees – characterized by three interrelated symptoms of emotional exhaustion, reduced personal accomplishment and depersonalization – is a well-known phenomenon in psychology and several applied business disciplines. Following persistent recognition in the practice community, academic recognition of this topic has begun to appear in the accounting literature. Using a measure of burnout developed for boundaryspanning positions, this paper shows that burnout among internal auditors is a serious concern. Results offer evidence that the burnout condition is directly related to several of the important behavioral and attitudinal outcomes in internal accounting practice. In order to provide greater clarity for future research, this study offers a separate treatment of the three dimensions of burnout, two very different organizational commitment constructs and two turnover directions. Implications for the management of human resources in this area are included.
INTRODUCTION Over the last 20 years, the literature on accounting practice has included a persistence of articles that draw attention to the existence and consequence Advances in Accounting Behavioral Research, Volume 9, 51–86 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09003-X
51
52
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
of work-related stress. The importance of this work has been somewhat undercut by those practitioners that report that they thrive upon a sense of stress and urgency. Thus, stress itself, is not always a negative condition. The literature has responded by focusing upon a darker side of stress. Burnout is believed to have only dysfunctional consequences for the behavioral performance and psychological well-being of accounting professionals. Burnout symptoms have been reported anecdotally to occur in public accounting (Rose, 1983; Sanders, 1998), internal auditing (Kusel & Deyoub, 1983) and management accounting (Journal of Accountancy, 1984; Figler, 1980). Academic treatments of the consequences of role stress have forged connections between stress, life enjoyment, career success, health, and withdrawal from organizational participation (Senatra, 1980; Collins & Killough, 1992; Haskins, Baglioni, & Cooper, 1990). However, these studies have produced a body of conflicting results. This could be because the source of truly negative results, job burnout, has not been adequately theorized and incorporated in this work. For many years, only two primarily descriptive unpublished papers studied accountant burnout on an explicit basis (Bokemeier, Bokemeier, & Tipgos, 1990; Bokemeier, Lorentzen, Bokemeier, & Tipgos, 1995). These studies produced mixed evidence associating burnout with gender, family, and work concerns. More recently, burnout has enjoyed a broader recognition in the published literature. Fogarty, Singh, Rhoads, and Moore (2000) produced evidence that burnout should be considered as a mediating variable between role stress and traditional behavioral outcomes. Sweeney and Summers (2002) show that the burnout experienced in public accounting spikes during busy season. Kalbers and Fogarty (2005) illustrate how burnout feelings may be grounded in perceptions of inadequate skills and insufficient control in an era of job insecurity. Notwithstanding the recent increase in academic attention, there still remains a set of important questions surrounding the importance of the topic for the accounting profession. Consequently, our ability to develop cumulative knowledge, so that we could draw comparisons with other professions (e.g., law, medicine, and nursing) and within the accounting discipline (e.g., external auditors, internal auditors) has been slow. This paper aims to fill some of the gaps in this literature. Joining the nascent accounting literature on this topic, we draw upon studies in fields such as occupational health and applied psychology to argue for a place for the burnout construct that is distinctive from other aspects of job/role stress. In order to suggest that the effects of burnout cannot be ignored, we develop a theoretical model that places burnout as an exogenous condition for behavioral and
Internal Auditor Burnout
53
attitudinal consequences. From this model, testable hypotheses are drawn. This improves upon previous work that used similar outcomes as if they were unrelated and independent. This study also more effectively controls the sample to eliminate variation in job content. Using data collected from internal auditors, results provide evidence that the individual dimensions of burnout are relevant for this type of accountant, both in terms of its magnitude and its adverse affects on outcomes valued by employer and employee.
THE BURNOUT CONSTRUCT Burnout describes a specific psychological condition in which people suffer emotional exhaustion, experience a lack of personal accomplishment and tend to depersonalize others (Freudenberger, 1974). The occupational health and applied psychology literature reveals an extensive body of research that has established a domain for the burnout construct, validated its measurement and legitimated the study of its causes and consequences (Lee & Ashforth, 1996). The consensus position in this literature is that burnout tendencies involve a psychological condition characterized by three dimensions (Maslach & Jackson, 1996) that are interrelated (see Shirom, 1989). First, emotional exhaustion is evidenced by feelings of depleted energy and related sensations as a result of excessive psycho-emotional demands. These excessive demands stem from work tasks that require innovative and creative solutions, and produce high levels of arousal such as would be expected to occur when accounting personnel work for a clientele under time pressure or pertaining to matters that involve great consequence (Jackson, Schwab, & Schuler, 1986). Second, reduced personal accomplishment entails attributions of inefficacy, low motivation and reduced self-esteem. Often these conditions are associated with the belief that future efforts will not be worthwhile because past efforts have repetitively failed to produce desired results (Abramson, Seligman, & Teasdale, 1978). Depersonalization, the third dimension of burnout, is the tendency to dehumanize others, often through a cynical, callous and uncaring attitude toward them. Treating others as if they were objects occurs as a part of this burnout condition. Burnout is caused by conditions referred to as role stressors. Despite its connection to stress, burnout is not a stressor per se. Instead, burnout is the result of stressors, that when present in particular degrees and combinations, overwhelm the coping resources of the individual (Hunsaker, 1986; Feldman & Weitz, 1988). In other words, the notion of burnout accepts the premise that different role stressors may not be excessive individually (Toppinen-Tanner,
54
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
Kalimo, & Mutaneu, 2002). However, when taken together, the cumulative effect of role stressors may be overwhelming. When one or more role stressors are excessive, burnout symptoms are likely to emerge. Furthermore, burnout is a situation that is invariably adverse whenever present. Contrariwise, role stress is believed to be productive and functional (often referred to as ‘‘eustress’’) up to some point (Seyle, 1976). Beyond these levels, the impact of role stress on important work outcomes turns dysfunctional. Since burnout has not been ‘‘on the radar screen’’ of accounting researchers until recently, the literature on the impact of role stressors on job outcomes in the discipline may exhibit specification bias. Such bias could occur by omitting an important mediator variable such as burnout. For example, while several studies show that low levels of job satisfaction and high levels of turnover intentions are correlated with high levels of role conflict and role ambiguity (e.g., Slavin, 1980; Strawser, Kelly, & Hise, 1982; Heian, 1988), there is equivocal support for these relationships (e.g., Bartunek & Reynolds, 1983; Collins & Killough, 1992; Rebele & Michaels, 1990; Aranya & Ferris, 1983; Wood & Wilson, 1987; Reed & Kratchman, 1985). For example, Senatra (1980) found no relationship between role conflict and either turnover intentions or job satisfaction. The impact of other role stressors on turnover intentions has produced particularly mixed results (e.g., Hellriegal & White, 1973; Kelley & Seiler, 1982; Heian, 1988). Even the most consistent result, the link between higher role ambiguity and higher turnover intentions, has not always been found (e.g., Senatra, 1980). This wide variability in the results may be because burnout is a more proximate construct than role stressors for job outcomes. Although burned out individuals tend to exhibit three highly specific symptoms, some people may have more profound and consequential manifestations of only one or two burnout dimensions. Each burnout symptom is a distinct psychological condition with its own unique pattern of antecedent relationships with role stressor and consequential influences on job outcomes. Therefore it is necessary to ‘‘unpack’’ the burnout construct and analyze consequences on a more precise level consistent with the meaningfulness of these symptoms for an accounting professional’s quality of work life.
EFFECTS OF BURNOUT TENDENCIES ON JOB OUTCOMES: A MODEL A large body of literature outside accounting has documented the consequences of burnout. In reviews and meta-analysis of this literature, Lee and Ashforth
Internal Auditor Burnout
55
(1993, 1996) and Cordes and Dougherty (1993) conclude that burnout has often been associated with diminished levels of performance and with a general withdrawal of the individual from a meaningful psychological involvement with their organization. Burned out individuals stop taking the usual degree of care in their work and, as a result, lower levels of quality are produced. The psychological effects span a wide spectrum, ranging from the person’s reaction to the work to his/her attachment to the organization. Longitudinal work has also supported the continuation of the burnout problem as a social one (Kalimo, Pahkin, Mutaneu, & Toppinen-Tanner, 2003). Consistent with this work, a simple yet powerful theoretical model can be considered for individuals employed by large organizations. Establishing burnout as an exogeneous condition that may emerge from employment allows one to posit two levels of consequences. The first level constitutes the impairment of the emotional conditions and most practical outcomes of employment. This can be broadly classified into reduced psychological connection with the job and the organization, and diminished performance. In a second level, one can expect the departure of the individual from employment. Once burnout has eroded performance and distanced the individual from the community of the organization, second level predicts action to escape from the irritating circumstances (Moore, 2000a). The expectation is that second-level consequences will be affected by first-level consequences and exogeneous conditions. Thus, the final organizational behavior (turnover) will be the result of the exogeneous precipiant condition (burnout) as well as the initial endogeneous results of burnout (poor performance, reduced involvement). Fig. 1 depicts these theoretical partitions.
HYPOTHESIS DEVELOPMENT This section elaborates the expectations that stem from a more detailed consideration of the model shown in Fig. 1. Importantly, burnout is expanded into its three constituent elements: emotional exhaustion, reduced personal accomplishment, and depersonalization. Collectively these can be referred to as burnout tendencies. More precision also is needed about the individual’s involvement with the organization. This theoretical construct is divided into the more operational variables of job satisfaction, affective commitment, and continuance commitment. Finally, two types of turnover are offered (inter-organizational, intra-organizational) as the second-level consequences. Fig. 2 summarizes these elaborations, and the hypotheses to be discussed below.
56
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS First-Level Consequences
Second-Level Consequences
Reduced Performance Quality
Turnover
Burnout
Diminished Psychological Involvement
Exogenous Conditions
Fig. 1.
Job Satisfaction
a H2 ( Burnout Tendencies: a. Emotional Exhaustion b. Reduced Pers. Accomp. c. Depersonalization
H3 H4
H11
,b, c)
H9 H5 (a,b ,c)
(a ,b
(a , b
Organizational Commitment: Affective
H8
H1
H7
) b ,c (a ,
External Turnover Intentions
H 6 (a ,b ,c) ,c)
Organizational Commitment: Continuance
,c)
Positive relationship
0 H1
Internal Turnover Intentions
Job Performance
Negative relationship
Fig. 2.
Empirical Model of Burnout Consequences to be Tested
Internal Auditor Burnout
57
The notion that job satisfaction is an important consequence of burnout is firmly established in the literature (Maslach, 1982; Wolpin & Greenglass, 1991). Psychological burnout is the result of an appraisal process by which an individual evaluates the balance between job demands and personal resources. The outcome of this appraisal should affect an individual’s psychological well-being on the job (Halbeslebin & Buckley, 2004). Since job satisfaction is also an affective response to the overall nature of the work, it is hypothesized that burnout feelings ought to be inversely related to job satisfaction. H1. High levels of the three burnout tendencies will be associated with low levels of job satisfaction. Accountants caught in a burnout syndrome generally view the organization in adversarial terms and tend to withdraw psychologically from it. Emotionally exhausted accounting professionals who view others in a detached and callous manner, and who do not feel their job provides them with a meaningful sense of accomplishment, tend to withdraw from the organization (Maslach, 1982). Initially, this withdrawal may take the form of absenteeism, physical isolation, and extended breaks, as the employee avoids contact with organizational members and clients (Hellriegal & White, 1973; Gaertner, Hemmeter, & Pitman, 1987). This contrasts dramatically with employees that have a high level of organizational commitment. Conventionally, organizational commitment has been understood as the willingness of the individual to ‘‘go the extra mile’’ on behalf of the organization (Smith, Organ, & Near, 1984). Accordingly, the expectation can be stated that those that are burned out will be unable or unwilling to exhibit this form of affective organizational commitment. H2. High levels of the three burnout tendencies will be associated with low levels of affective organizational commitment. Kalbers and Fogarty (1995) argue for a more differentiated understanding of commitment as this construct applies to internal auditors. For a variety of reasons centering around the lack of an ‘‘up or out’’ convention in this occupation, internal auditors may experience what has been called continuance commitment. Internal auditors may be destined to share a future with their organization by virtue of the transaction costs that they would suffer if they were to leave. Unlike affective commitment, burnout is not expected to decrease the perceived magnitude of these inadvertent and progressive forms of commitment. Instead, burnout tendencies may magnify
58
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
the awareness of the individual that this attachment exists and that alternatives may not be viable. H3. High levels of the three burnout tendencies will be associated with high levels of continuance commitment. Burnout affects behavioral outcomes such as job performance, because it reduces the energy at the disposal of the individual and leads to reduced efforts at work. Burnout feelings also entrap accounting professionals in a vicious spiral where they are less prone to seek and obtain support, react well to constructive feedback and, as a result, continue to perform ineffectively. Burnout affects job performance directly since the individual perceives little or no control over the job situation, and his/her confidence in tackling work-related problems declines. H4. High levels of the three burnout tendencies will be associated with low levels of job performance. If burnout persists, the worker will likely seek permanent avoidance by leaving the position, the firm, or even the career. Therefore, an association between burnout and turnover intentions can be expected. However, the particular nature of internal auditor positions requires a retreat from the usual monolithic construct of turnover. Leaving the internal audit function can be accomplished in two different ways. New jobs can be found both inside and outside the organization. Although the relief offered by a new position might be stronger for the burned out individual that desires to make the more drastic external move, the intention to pursue an internal turnover strategy would also be a reasonable response to the burnout syndrome. Therefore, burnout should be positively associated with both internal and external turnover intentions. H5. High levels of the three burnout tendencies will be associated with high levels of external turnover intentions. H6. High levels of the three burnout tendencies will be associated with high levels of internal turnover intentions. The combination of these six hypotheses can be conceived as an expectation that burnout will have a pervasive impact upon a set of job outcomes. However, an integrative test of these effects as a model requires that some expectations be formulated about the interrelationships among the outcome variables. Although the focus of this paper pertains to burnout, a true
Internal Auditor Burnout
59
picture of burnout’s impact cannot ignore the likelihood that the outcomes are not independent. Because of the large body of work in this area, combined with the need to focus on the impact of burnout tendencies, these additional relationships are meant to serve only as environmental conditions for the operation of burnout and will not be developed at length. Nonetheless, their inclusion completes the model and prevents the overstatement of burnout effects inherent in a more piecemeal hypothesis testing approach. Based on considerable work published both in the accounting literature and in the more general scholarship on business organizations, the link of job satisfaction to turnover intentions is reasonably well established (e.g., Reed, Kratchman, & Strawser, 1994; Gregson, 1992b). An individual that does not find some degree of intrinsic reward in performing the tasks required by a job is likely to seek another. This leads to two research expectations because this may involve either a transfer to another position within the organization, or a departure to another organization. H7. Low levels of job satisfaction will be associated with high levels of external turnover intentions. H8. Low levels of job satisfaction will be associated with high levels of internal turnover intentions. These general notions also support an expectation that lower affective commitment will be associated with higher turnover intentions. In this instance, the individual focuses more upon feelings about the organization than upon the job that needs to be performed. Therefore, no relationship should exist with internal turnover intentions. H9. Low levels of affective organizational commitment will be associated with high levels of external turnover intentions. Continuance commitment in many ways acts as the mirror image of affective commitment. As a result of previous investments in an organization, an individual feels compelled to continue within it. Whereas the feeling of being ‘‘stuck’’ should work as a damper on the inclination to leave the organization’s boundaries, it should not be related to the tendency to seek transfers within the entity. H10. High levels of continuance organizational commitment will be associated with low levels of external turnover intentions.
60
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
No model or outcomes would be complete without anticipating that job satisfaction and organizational commitment are positively related. In this model, this effect is specified as a relationship between affective organizational commitment and job satisfaction (Poznanski & Bline, 1997). In other words, there should be a psychological link between feelings about the organization that provides the work and the work itself. H11. High levels of affective organizational commitment will be associated with high levels of job satisfaction. No expectation about the relationship between job performance and turnover intentions was formed. This choice is the result of the opposing possibilities that poor performers may quit in advance of being dismissed and that high performers may be more demanded in other capacities, both inside and outside the organization. In sum, two sets of hypotheses are offered. The first pertains to the direct consequences of the three burnout dimensions. Hypotheses 1–6 expect an impact of the separate burnout elements upon each of six behavioral and attitudinal outcomes. These represent the results from the exogeneous conditions of Fig. 1 on both first- and second-level consequences. Hypotheses 7–11 complete the model by linking first- and second-level consequences. The test of these effects, in addition to preventing the overstatement of the direct effects of exogeneous conditions in their absence, allow for the creation of indirect effects. Therefore, burnout may affect turnover through the effect it has on diminished psychological involvement and performance, if those constructs are themselves linked to turnover.
THE STUDY Sample To reflect indications of burnout among accounting personnel employed in a specific accounting engagement, a sample of internal auditors was attained through the cooperation of 23 organizations in a diverse number of industries with offices in the Midwest United States during April 2000. This work was facilitated with the cooperation of the local chapter of the Institute of Internal Auditors. This organization sanctioned the study and encouraged cooperation with the researchers. For these purposes, firms that agreed to participate were provided with questionnaires, which they distributed to their internal auditor staff. The number of questionnaires was arranged
61
Internal Auditor Burnout
according to the size of the staff and their availability. Participating individuals were sent questionnaires along with a cover letter from the researchers encouraging participation. Completed questionnaires were mailed by respondents directly to the researchers. Confidentiality was extended as a condition of participation. This study suggests that internal auditors are a valuable focus for a burnout study for several reasons. First, the very existence of internal auditing within organizations creates an intrinsically conflictual potentiality. Internal auditors are simultaneously expected to be members of ‘‘the team’’ and independent from it. Second, internal auditing is often not a primary destination for college graduates. Since many internal auditors have had previous employment, they may exhibit a seasoned attitude toward their work. Third, the role stress of internal auditing is less likely than public accounting to be distorted by seasonality (see Sweeney & Summers, 2002).
Measurement Burnout historically has been measured from the Maslach Burnout Inventory (Shirom, 1989; Rafferty, Lemkau, Purdy, & Rudisill, 1986). The original scale contained 22 items that were divided across three-subclasses corresponding to the burnout dimensions. Golembiewski, Hilles, and Dally (1983) revised this instrument by altering the scale headings. However, the restricted applicability of the MBI (slanted toward health care and human services) gradually caused many researchers to develop and customize their own instruments. Among these, Singh, Goolsby, and Rhoads (1994), in applying burnout to customer service personnel, discovered that the modified scale failed to capture the complexity of the role-set of boundary spanning personnel. These authors, building on the theoretical propositions of Leiter and Maslach (1988) and others, proposed a multidimensional rolespecific burnout (MROB) measure that uses 24 items to span the three conceptual dimensions of burnout and four target role senders (i.e., immediate supervisor, top management, coworkers, and customers). This scale allows burnout feelings unique to particular members of the role set to be expressed, thereby elaborating the capabilities of burnout measurement. This measure also avoids confusing burnout with more generic forms of depression and anxiety (see Shirom & Ezrachi, 2003). Internal auditors fill critical boundary spanning roles within their organizations. They interact with and evaluate managers across the organization and report results to upper management and the audit committee. Based upon the acceptable
62
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
psychometric properties for the MROB that were offered in Singh et al. (1994), and more recently for accounting professionals (Fogarty et al., 2000) and account managers in different industries (Demerouti, Verbeke, & Bakker, 2005), the current study utilized the MROB scale. The instrument allows internal auditors to report their potentially variable distress regarding supervisors, coworkers, top management, and clients (auditee managers). The scale was slightly modified to be relevant to internal auditing professionals. Job satisfaction was measured with a scale developed by Brayfield and Rothe (1951). This scale has been used extensively in many of the business fields, including studies that pertain to burnout. This scale was selected over others because of its superior consideration of a set of more specific aspects of jobs. A seven-item scale used by Fogarty et al. (2000) to study accountants was used to measure performance. This scale, by asking subjects to evaluate their performance relative to others, minimizes the leniency tendencies of self-reported absolute measures in this area. Organizational commitment was measured in two ways. Because commitment can be predicated as an affective response to the organization, or as an instrumental reaction (see Meyer, Paunonel, Gellatyl, Goffin, & Jackson, 1989), a 13-item scale that made this distinction was used. Instrumentation from Meyer and Allen (1991, 1997) was used for these two variables, hereinafter designated as affective commitment and continuance commitment. The last outcome variable, turnover intentions, was modified from Kalbers and Fogarty (1995). This measure was deemed particularly applicable because it differentiated intra-organizational (internal turnover intentions) and interorganizational (external turnover intentions) mobility. These authors argue that since internal auditors are likely to move into non-auditing managerial capacities, a more robust measurement of mobility intentions is needed. All items were Likert scaled, using seven-point scaling. An appendix to this paper contains the items seen by the respondents.
Data Analysis Methods The advantages of the structure of the hypotheses are that they form a unified model of effects. This lends itself to latent variable structural equation (LVSE) analysis techniques. This method has gained considerable popularity in the social sciences literatures (Bentler & Dudgeon, 1996) and has recently attracted the attention of accounting researchers (Gregson, 1992a, 1992b; Poznanski & Bline, 1997). The LVSE method offers the ability of correcting
63
Internal Auditor Burnout
for measurement error and simultaneously estimating the modeled path coefficients. This produces coefficients with unbiased and minimal variance and contextualizes the test of specific hypotheses within a model that can itself be evaluated using a variety of fit statistics. AMOS 4.0, one of several specific software packages that offers LVSE analysis, was used. In order to reduce measurement error and to improve measurement fit, two-item composite measures were formed for most of the latent variables (see West, Finch, & Curran, 1995; Landis, Beal, & Tesluk, 2000). This was done by randomly sorting the measured items into two groups (the minimum needed by LVSE methods) to form the composite items used in the analysis. This technique facilitates the identification of the model by reducing the number of parameters to be estimated. Because the focus in this paper is on the relationships among the constructs, the departure from the unique measured items imposes very small losses in precision. This was not done for the turnover scales, because the number of measured items were either at or close to the minimum. LVSE was first used to test the hypothesized relationships as a unified model of effects. LVSE may also be used to explore new models or modify hypothesized models (Jo¨reskog & So¨rbom, 1993) through modification indices that indicate possible ways to improve the fit of the overall model. After testing the hypothesized model, the model was then modified through iterations of deleting non-significant paths and adding significant paths. Results of such an exercise must be viewed with caution because fitting data with only one sample may involve chance. Researchers must consider whether changes are meaningful within the context of theory and common sense (Arbuckle & Worthke, 1999; MacCallum, 1995; Daft, 1983). Therefore, results for the hypothesized model and the modified model are described and compared.
RESULTS Response A total of 298 questionnaires was received representing an 81.6% response rate. Four were found incomplete or otherwise unusable, and were deleted from consideration. The response rate that was achieved was much higher than recent accounting burnout studies, including the 20% reported by Fogarty et al. (2000), and other studies that pertain to internal auditors (e.g., Harrell, Taylor, & Chewning, 1989; Pei & Davis, 1989). No systematic response bias was detected, using tests recommended by Oppenheim (1966).
64
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
Descriptive Statistics Demographic information about the sample is contained in Table 1. The sample was slightly more female (59%) than male. Almost two-thirds of the sample had an earned bachelors degree as their highest educational credential. The average time in the practice of internal auditing was 7.1 years, and nearly 30% of the sample had more than 10 years in this organizational function. An approximately equal division of the sample according to rank within the internal audit group was achieved, with close to one-third of the sample in each of the three levels that the questionnaire allowed the subjects to select. Forty-two percent of the sample had an accounting license or other professional credential (i.e., CPA, CIA, or CMA). Other demographics not shown on this table included the fact that the respondents were predominantly married (64%). The mean age of respondents was 36.6. The average tenure with the organization was 7.5 years. Table 2 summarizes the key descriptive statistics for the constructs under study. For these purposes, we report both the original items (see the appendix) and the composite items. Table 2 shows the number of measured items, the Alpha levels, means and dispersion statistics. In addition to the separate commitment and turnover intention variables that we found necessary in the Table 1.
Demographic Profile of the Respondents. Percent
Professional Certification
Percent
Sex Male Female
41 59
Certified Public Accountant Certified Internal Auditor Certified IS Auditor One or more (CPA, CIA, CISA)
25 10 11 42
Educational attainment High school Bachelors degree Masters degreea
11 66 23
Internal audit experience o1 year 1–2 years
15 24
3–5 years 6–10 years 410 years
18 13 30
Responsibility level Entry level/staff Experienced/senior Supervisor/manager a
One person had a Ph.D.
25 39 36
65
Internal Auditor Burnout
Table 2. Measurement Characteristics for Study Constructs. Measure
Emotional exhaustion Reduced personal accomplishment Depersonalization Job performance Job satisfaction Organizational commitment: affective Organiztional commitment: continuance External turnover intentions Internal turnover intentions
Number of Items
Reliability
Mean
Standard Deviation
Original
Composite
Original
Composite
Original
Composite
Original
Composite
8
2
0.79
0.83
2.64
2.65
0.94
0.94
8
2
0.86
0.92
2.29
2.29
0.74
0.74
8 7 7 6
2 2 2 2
0.81 0.83 0.92 0.86
0.92 0.84 0.91 0.89
2.61 5.77 5.15 4.42
2.61 5.78 5.14 4.42
1.02 0.76 1.23 1.32
1.02 0.76 1.22 1.32
7
2
0.81
0.88
3.54
3.55
1.25
1.27
3
NA
0.83
NA
3.32
NA
1.65
NA
2
NA
0.50
NA
3.38
NA
1.46
NA
internal auditor context (see above), Table 2 also reports the three separate burnout dimensions. As shown by the close match between the original and the composite columns, this procedure has done little to alter the data that will be analyzed for hypothesis testing. Table 2 also includes the mean values and standard deviations for all the variables in the study. Burnout tendencies appear to be relatively low, compared to the other variables. However, because burnout is a serious condition but one not experienced by most people, even relatively low median scores are indicative of an occupational problem. The other descriptive information attests to the typicality of the internal auditor sample. The pattern of higher emotional exhaustion and depersonalization and lower reduced personal accomplishment is a familiar one (Brookings, Bolton, Brown, & McEvoy, 1985). This similarity underscores the importance of burnout tendencies for internal auditing professionals. Each of the measures utilized in the study has acceptable reliability and nomological validity. Cronbach’s alpha reliability, an index of internal consistency of a measure, exceeds 0.80 for each measure utilized except internal turnover intentions. This indicates that, for the most part, over 80% of the variance in the measures utilized can be attributed to a systematic source (i.e., the underlying construct). This level exceeds the acceptable standard for the reliability of measures (Nunnally, 1967).
66
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
Table 3.
Exhaustion Reduced Deperson Job performance Affective commitment Continuance commitment Job satisfaction Internal turnover
Reduced
Deperson
0.416
0.629 0.502
Correlations.
Job Affective Continuance Job Performance Commitment Commitment Satisfaction 0.237 0.549 0.290
0.385 0.377 0.476 0.260
Internal Turnover
External Turnover
0.124 0.118 0.113 0.107
0.421 0.454 0.439 0.226
0.129 0.039 0.098 0.013
0.407 0.278 0.436 0.153
0.134
0.628
0.233
0.662
0.013
0.151
0.136
0.166
0.613 0.168
Significant at 0.05. Significant at 0.01.
Correlations among the latent variables are shown in Table 3. The burnout dimensions correlate positively among themselves (values ranging from 0.42 to 0.63), all significant at po0.01. Burnout dimensions correlate negatively with job satisfaction (values ranging from 0.42 to 0.45), job performance (values ranging from 0.24 to 0.55), and affective commitment (values ranging from 0.38 to 0.48), again universally significant at po0.01. Burnout dimensions are also significantly (po0.01) related to external turnover intentions (values ranging from 0.28 to 0.44). Some dimensions of burnout are related to continuance commitment and internal turnover intentions at po0.05. Taken together, this suggests that burnout and other measures captured in this study are a reasonable foundation for studying the relevance and significance of burnout tendencies among these accounting professionals.
Model Evaluation Results from the hypothesized and modified models are presented in Tables 4 and 5. The modified model results from an iterative process of deleting nonsignificant hypothesized paths and adding significant paths not hypothesized or paths that turned significant when other paths had been eliminated. The overall fit statistics in Table 4 reveal that both models, including burnout tendencies as an essential exogenous variable, fit the data from internal auditing professionals reasonably well. The fit statistics for the modified model are an improvement over those of the hypothesized model. The w2 test statistic associated with the null hypothesis that the proposed model can effectively reproduce the observed
67
Internal Auditor Burnout
Table 4.
Standardized Parameter Estimates and Fit Statistics Empirically Derived Model.
Emotional exhaustion 1 Emotional exhaustion 2 Reduced personal accomplishment 1 Reduced personal accomplishment 2 Depersonalization 1 Depersonalization 2 Affective commitment 1 Affective commitment 2 Continuance commitment 1 Continuance commitment 2 Job performance 1 Job performance 2 Job satisfaction 1 Job satisfaction 2 External turnover 1 External turnover 2 External turnover 3 Internal turnover 1 Internal turnover 2 Goodness-of-fit index Adjusted goodness-of-fit index Normed fit index w2 ratio Root mean square residual Root mean square error approximation
Hypothesized Standardized Parameter Estimates
Modified Standardized Parameter Estimates
0.903 0.794 0.908 0.946 0.903 0.936 0.919 0.864 0.828 0.970 0.836 0.861 0.913 0.920 0.785 0.799 0.774 0.999 0.332
0.898 0.797 0.907 0.947 0.904 0.934 0.919 0.864 0.921 0.872 0.835 0.862 0.914 0.920 0.780 0.811 0.769 0.893 0.371
Fit statistics 0.924 0.887 0.935 1.85 0.115 0.54
Fit statistics 0.931 0.903 0.942 1.56 0.091 0.044
covariances resulted in a ratio of 1.85 and 1.56, respectively. Good fitting models evidence ratios of 5.0 or less (Wheaton, Muthen, Alwin, & Summers, 1977). The various measures of relative and absolute fit index (ranging from 0 to 1, with 0 implying poor fit and 1 indicating perfect fit) including the goodness-of-fit (GFI) and the normed fit (NFI) and the comparative fit (CFI) indices exceed 0.90 for both models, the usual benchmark of a well-fitting model (Bentler, 1990). This reflects the relatively large number of parameters in the model. Noting that different fit indices have different strengths and weaknesses, this consistent evidence of exceeding the target value of 0.90 for good fitting models is encouraging.
68
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
Table 5. Hypothesis
1
2
3
4
5
6
7
8
9
10
11
Relationship (Hypothesized Direction)
Burnout-Job satisfaction ( )
BurnoutAffective organizational commitment ( ) BurnoutContinuance organizational commitment (+) Burnout-Job performance ( )
BurnoutExternal turnover intentions (+) BurnoutInternal turnover intentions (+) Job satisfactionExternal turnover intentions ( ) Job satisfactionInternal turnover intentions ( ) Affective commitmentExternal turnover intentions ( ) Continuance commitmentExternal turnover intentions ( ) Affective commitmentJob satisfaction (+)
Summary of Results.
Burnout Component
Hypothesized Model Standard coefficient
Emotional exhaustion Reduced personal accomplishment Depersonalization Emotional exhaustion Reduced personal accomplishment Depersonalization Emotional exhaustion Reduced personal accomplishment Depersonalization Emotional exhaustion Reduced personal accomplishment Depersonalization Emotional exhaustion Reduced personal accomplishment Depersonalization Emotional exhaustion Reduced personal accomplishment Depersonalization
Sign
Sign
Modified Model Standard coefficient
Sign
Sign
0.167
0.05
0.139
0.05
0.206
0.01
0.196
0.01
NS NS
NA NA
0.05
0.186
0.060 0.097
NS NS
0.173
NS NS
NS NS 0.01
0.365 0.101
NS
0.01 NS
0.428 0.226
+
0.01 0.01
0.106
NS
NS
0.154
+
0.05
0.044 0.026
NS NS
NS NS
NA NA
NS NS
NS NS
0.01
0.618
NS NS
NA 0.172
0.05
0.142
0.618 0.024 0.133
NS NS
0.140
0.01 NS +
NS 0.01 0.05
0.088 0.060
NS NS
NS NS
NA NA
NS NS
NS NS
0.096
NS
NS
NA
NS
NS
0.125 0.349
NS
NS 0.01
NA 0.335
NS
NS 0.01
0.197
+
0.01
NA
NS
NS
0.475
0.01
0.532
0.118
0.05
NA
NS
NS
0.01
0.556
+
0.01
0.573
+
0.01
69
Internal Auditor Burnout
Table 5. (Continued ) Hypothesis
NA
NA
NA
Relationship (Hypothesized Direction)
Burnout Component
Hypothesized Model Standard coefficient
Sign
Sign
Affective commitmentInternal turnover intentions Continuance commitment Internal turnover intentions Affective commitmentContinuance commitment
Modified Model Standard coefficient 0.351
Sign
Sign
+
0.01
0.257
0.318
0.01
+
0.01
NA ¼ not applicable; NS ¼ not significant. Significant at 0.10
Table 4 also indicates that the difference between reproduced and observed covariances is rather small as evidenced by the root mean square residual of 0.115 and 0.091 and the root mean square of approximation of 0.054 and 0.044, respectively. For these two measures, the target upper bound of 0.10 was achieved for both in the modified model, but was slightly higher for the root mean square residual for the hypothesized model. In sum, the hypothesized and modified models of burnout and its consequences in Fig. 1 are acceptable and reasonable portrayals of the data and serve as a sound basis for interpreting the specific hypotheses. The estimated maximum likelihood parameters of the measurement part of the model of burnout and its consequences are consistently large and significant. Each of the estimated loadings is significant at po0.05. In this model, as is typical in confirmatory analysis, the cross-loadings are constrained to be zero and the measures are allowed to load only on their hypothesized constructs. The deletion and addition of paths between latent variables were the only changes made to the modified model.
Test of Hypotheses related to Burnout The hypothesized model is used to test the hypotheses as stated. The modified model is used to further consider the hypotheses as well as additional
70
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
results. Table 5 provides a summary of the standardized coefficients, signs, and significance for each hypothesis for both models. As discussed in the modified model results section, three paths not hypothesized were added to the modified model. Consistent with Hypothesis 1, two of the burnout tendencies have a significant negative influence on job satisfaction. Internal auditors who evidence more emotional exhaustion and reduced personal accomplishment tend to have lower levels of job satisfaction. Hypotheses 1a and 1b are supported, with regards to those dimensions of burnout. The absence of a path from depersonalization to job satisfaction indicates the failure of this relationship to be statistically significant. Thus, no support for Hypothesis 1c exists. Hypothesis 2 suggested a negative relationship between the burnout tendencies and the affective form of organizational commitment. As expected, reduced personal achievement and depersonalization are negatively associated with affective commitment. These relationships support Hypotheses 2b and 2c. Again, one of the expected effects is not supported. Hypothesis 2a, pertaining to the influence of emotional exhaustion, is not significant. Hypothesis 3 relates to continuance commitment. In the hypothesized model, which includes all hypothesized paths, significant or not, none of the paths from the three burnout measures to continuance commitment are significant. However, two of the three paths are in the hypothesized direction. Based on this model, there is no support for a relationship between burnout and continuance commitment. The expected negative relationship between burnout and job performance is the subject matter of Hypothesis 4. Consistent with the theoretical Fig. 1, one path indicates a significant negative association of a burnout dimension with job performance. Specifically, internal auditors who sense a reduced sense of personal accomplishment have lower job performance. Thus, Hypothesis 4b is supported. The lack of evidence in support of Hypotheses 4a and 4c indicate a certain behavioral resilience to the burnout emotions of emotional exhaustion and depersonalization. Hypothesis 5 is not supported by the results. Reduced personal accomplishment is significantly (po0.05) related to external turnover, but negatively, which is the opposite sign of that hypothesized. The relationship between emotional exhaustion and external turnover is positive (but significant only at the higher p level of 0.10). There is no support for Hypothesis 6. Apparently internal auditors who are burned out are not more, or less, likely than others to seek to move to other positions within the organization. In sum, selective support exists for the hypotheses pertaining to the individual components of burnout. Collectively, the burnout dimensions form
71
Internal Auditor Burnout
powerful antecedents of the psychological and behavioral outcomes for internal auditors. In total, six of a possible eighteen effects (three burnout dimensions to six outcomes) were significant. Because the literature has not progressed to the point where more precise expectations could be formed, this level of realization of the a priori idea is considerable. The different impact of the burnout dimensions on the two types of commitment and the two types of turnover intentions illustrate the importance of these conceptual distinctions. Hypotheses among Outcome Variables Fig. 1 suggests that the outcomes of the burnout dimensions would occur in two levels. Because this implies that a full model requires the specification of relationships among these outcomes, Hypotheses 7–11 have been constructed and tested. In addition, these effects are necessary so that the influence of burnout is not overstated. Hypothesis 7 specified a negative relationship between job satisfaction and external turnover intentions. As shown in Table 5, job satisfaction is negatively associated with external turnover. At po0.01, Hypothesis 7 is therefore supported. In this regard, internal auditing is similar to most occupations. Relevant to Hypothesis 8, the relationship between job satisfaction and internal turnover is significant, but in the opposite direction from that hypothesized. Evidently, job satisfaction within internal auditing may lead to a willingness to serve the organization in other areas. Because of the opposite sign, Hypothesis 8 is not supported. The subject of both Hypotheses 9 and 10 is the relationship between commitment and external turnover intentions. Hypothesis 9 anticipates the inverse relationship between affective commitment and turnover intentions. The results indicate support for this relationship at po0.01. Table 5 also shows that the relationship between continuance commitment and external turnover intentions is significant at the po0.05 level. Therefore, Hypothesis 10 is supported. The affective organizational commitment–job satisfaction relationship appears strongly significant and in the expected positive direction. Hypothesis 11 is confirmed. Modified Model Results Results from the modified model are shown with the hypothesized model in Table 5 and are graphically displayed in Fig. 3. The modified model represents the optimal fit of the data, wherein all depicted relationships are
72
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
.05 -.139 (
Job Satisfaction
)
Burnout: Emotional Exhaustion
Burnout: Reduced Personal Accomp.
Burnout: Depersonalization
.556 96 -.1 .2 26 6 8 -.1
.154 (
28 -.4
Organizational Commitment: Affective
.318 .05) -.6 18
Organizational Commitment: Continuance
Job Performance
-.33 5
.172
-.532 -.142 (.05) .35 1
-.257
External Turnover Intentions
Internal Turnover Intentions
Positive relationship Negative relationship
All arrows are significant at .01 except as shown parenthetically.
Fig. 3.
Empirically Derived Modified Model
significant at po0.01 except as noted as po0.05. In Table 5, it can be observed that of the total 23 hypotheses (three forms of each of the first six hypotheses and four additional hypotheses), the results in the modified model are consistent with 18 of the hypotheses in the hypothesized model. For the other five, three have turned from not significant to significant, and two have turned from significant to not significant. This can occur as paths are deleted and added, because even paths that are not significant influence the relationships among the latent variables. Three hypotheses related to the burnout measures changed from not significant to significant, and in the hypothesized direction, in the modified model. Emotional exhaustion and reduced personal accomplishment became positively and significantly related to continuance commitment. This result demonstrates the importance of distinguishing this outcome from affective commitment. Apparently, those who are more emotionally exhausted by their jobs perceive fewer other reasonable choices than staying associated with their organization. Along similar lines, internal auditors
Internal Auditor Burnout
73
who see themselves unable to achieve as much as they once did are more dependent upon their continuing relationship with the organization. Therefore, the modified model provides potential support for Hypotheses 3a and 3b. Also, emotional exhaustion became positively and significantly (po0.01) related to external turnover intentions (po0.10 in hypothesized model), thus suggesting some support for Hypothesis 5a. Two hypotheses that were supported in the hypothesized model received no support in the modified model. Those internal auditors more satisfied with their job were not more likely to transfer within the organization (Hypothesis 8) and those with higher levels of continuance commitment were not more likely to want to leave the organization (Hypothesis 10). The three paths added in the modified model that were not hypothesized provide some insights about possible relationships that caused the two hypothesized relationships above to go from significant to not significant. Affective commitment was found to be positively and significantly associated with internal turnover intentions and continuance commitment. Evidently, those that feel a stronger bond with the organization are more likely to want to leave the internal audit function but to stay with the company. However, the continuance commitment–internal turnover relationship is negative, and significant at po0.01. Apparently, higher levels of continuance commitment are associated with a lower desire, or ability, for internal auditors to move on to other jobs, even within the organization. The modified model also reveals the lack of any significant relationship between job performance and either internal or external turnover intentions. Internal auditors, like several other occupational groups (see Abramis, 1994), do not follow the conventional wisdom that better workers are more likely to stay in their current positions. In sum, the modified model is largely consistent with the results of the hypothesized model. The modified model provides the potential for further understanding of the relationships of the variables under consideration. Consistent with Fig. 1, Fig. 3 illustrates that only one of the outcome variables does not have a direct link with at least one of the burnout dimensions and that the three dimensions of burnout operate in quite different ways. Together, these dimensions produce a much more specific set of effects than would be possible with a unified burnout construct or with a set of discrete but unconnected outcomes. The models constructed in this research allows for the calculation of the magnitude of indirect effects. The strength of any two linear paths can be estimated by multiplying the two coefficients. Thus, from Fig. 3, indirect burnout effects flow through job satisfaction, affective commitment and
74
Table 6.
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
Summary of Estimated Standardized Total Effects of Burnout Components – Modified Model. Emotional Exhaustion
Affective commitment Continuance commitment Job performance Job satisfaction External turnover intentions Internal turnover intentions
Reduced Personal Accomplishment
0.000 0.226 0.000 0.139 0.219 0.058
0.186 0.095 0.618 0.300 0.058 0.090
Depersonalization
0.428 0.136 0.000 0.238 0.308 0.115
continuance commitment. Combining indirect effects and direct effects allows for the computation of total effects of any of the burnout dimensions on any of the outcomes. Table 6 contains this matrix.
DISCUSSION The aim of this study has been to demonstrate the importance of burnout tendencies to the accounting profession by showing evidence for the relevance of this construct for important attitudinal and behavioral consequences. To accomplish this, the study offered and tested a model of burnout’s influence on traditional outcome variables such as job satisfaction and job performance. Furthermore, the organizational commitment and turnover intentions constructs have been elaborated to refine our appreciation for the influence of burnout tendencies in the internal auditing setting. Finally, the hypothesized model was modified to explore potential relationships beyond those specified and to improve model fit.
The Burnout Construct and its Relevance More than 15 years ago, Kusel and Deyoub (1983) drew the attention of the accounting literature to the phenomenon of burnout among internal auditors. Since that time, the accounting literature has included a sustained interest in role stress and its consequences. This has included a handful of studies that empirically addressed the burnout phenomenon. The more recent of these studies has provided a clear discussion of the conceptualization, operationalization, and discriminant validity of burnout tendencies.
Internal Auditor Burnout
75
However, the published literature does not include a burnout study specifically focused on its consequences in the internal auditing context. The task of this study has been to build upon the methodological advances and to apply them to a singular occupational context. The results provide initial support for the overarching conclusion that burnout tendencies have a consistent, significant, and dysfunctional influence on the traditional outcomes of job satisfaction, job performance, affective organizational commitment, and external turnover intentions. This study provides support for the hypothesis that burnout is a key influence on job outcomes for internal auditors. Thus, there appears to be enough evidence to suggest that future researchers interested in role stress and its consequences risk misspecification of underlying phenomenon if they disregard the burnout variable.
Specific Insights into the Direct Influence of Burnout Dimensions When the specific tendencies of burnout are examined, a strong network of consequences from emotional exhaustion, depersonalization, and reduced personal achievement to the traditional outcomes of interest is evidenced in the results of the hypothesized and modified models. Specifically, it is apparent from Fig. 3 that the role/job environment of internal auditors is such that these people are susceptible to harsh consequences from all three dimensions. This departs from the conclusion that one might draw from the high levels of emotional exhaustion and depersonalization alone. In fact, one could argue that reduced personal accomplishment is the key burnout dimension for this group. Once these accounting professionals begin to show symptoms of a diminishing feeling of a personal contribution to the work environment, a large number of behavioral and psychological consequences are triggered. Emotional exhaustion has modest impact in the hypothesized model, reducing only job satisfaction in a significant way. In the modified model, a more systematic impact is demonstrated, where those with higher levels of emotional exhaustion are more likely to want to leave the organization, but also to have higher levels of continuance commitment. The lack of association with affective commitment suggests that internal auditors who do not have the emotional wherewithal for the work blame it on their job rather than on the organization that provides them that job. Reduced personal accomplishment has a direct influence on five of the six outcome variables. It is negatively associated with job satisfaction, affective commitment, job performance, and external turnover intentions. It is positively
76
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
associated with continuance commitment. Those internal auditors who no longer feel they contribute to the work environment have decreased job performance and diminished affection for their job and organization. It should be no surprise that those who perceive their accomplishments facility diminishing also rate their ex-post performance lower. Confidence in one’s ability to successfully execute appears to be important to the realization of a job well done. The failure of the other dimensions of burnout to have similar direct effects upon job performance suggests that internal auditors are able to bravely soldier on in the face of severe emotional conditions. Performance is the last thing to go for a group whose work can be somewhat severed from its excessive human components. Depersonalization, in both models, has the fewest consequences of the three burnout dimensions. It is significant in the development of lower affective commitment. Once the internal auditor begins to treat others as objects, the problem generalizes to the organization as a whole. Depersonalization implies a level of cynicism and negativity inconsistent with shared organizational empathy. Continuance commitment is less familiar to behavioral researchers. Occupations that lack an ‘‘up or out’’ norm might attract people who prefer career stability. However, this situation requires a recognition that people are committed to organizations just because of ‘‘sunk costs’’ or the difficulty of persuing alternative work. In the modified model, two of the burnout dimensions increased this form of commitment. Increased continuance commitment tends to be formed among those who report more emotional exhaustion and reduced personal accomplishment. The hopelessness of emotional detachment and the diminishment of job task abilities appears to translate into a despair that one might not be employed elsewhere. In sum, each hypothesis that expected direct effects from burnout has a unique story to tell about internal auditors. No burnout dimension is without importance nor does any one dominate across the set of six outcomes considered. Five of six outcome variables show some direct connection to burnout. The results are consistent with previous evidence that shows that burnout sufferers will not all exhibit the same consequences (Moore, 2000b).
Relationships not involving Burnout The relationships not involving burnout add to a long tradition of behavioral work on accounting professionals. Therefore, they will receive a rather abbreviated explanation.
77
Internal Auditor Burnout
Hypothesis 7 suggests that low job satisfaction will be associated with higher external turnover intentions. The fact that this is supported for internal auditors is not surprising. That the strength of this association is modest (po0.05) suggests that the presence of the burnout direct effects in the model may have extracted some of the variance from this relationship. Nonetheless, those that are unsatisfied with their work tend to seek alternative employers. Job satisfaction is not so powerful a variable to trigger internal turnover, however. Hypothesis 8 expects such an effect, and was supported in the hypothesized model. However, in the modified model, the relationship is not significant, suggesting that when the non-significant paths are eliminated that it is feelings for the organization, not for the internal auditing job, that positively influence the desire to transfer within the organization. Accordingly, satisfaction may not be the comprehensive pathway for the translation of the first level outcomes into the second level outcomes of Fig. 1. Hypotheses 9 and 10 mirror the previous two hypotheses, substituting commitment variables for satisfaction. The first relationship, relating affective commitment with external turnover intentions was supported. Considerable previous evidence for this association has been found for many groups, both inside accounting and outside this field. Those that report sharing the concerns and normative perspectives of the organization are less likely to consider leaving. Hypothesis 10 suggested that high levels of continuance commitment would be inversely associated with lower levels of external turnover intentions. That this effect was found in the hypothesized model, but not the modified model, may be explained by the direct effects of two burnout dimensions on external turnover intentions, one of which was added in the modified model. Thus, whether or not internal auditors consider themselves to have a store of irreversible investments with the organization does not change their inclination to leave its employ. Burnout elements need to be taken into consideration. The final Hypothesis (H11) covers the familiar scholarship ground that relates job satisfaction to affective organizational commitment. Those internal auditors who like the work of internal auditing also have warm feelings for the organization that provides them the opportunity to perform this work.
Indirect Effects Although the size of most of these indirect effects are small relative to the direct effects, they do merit some attention. Of particular note is the tendency of depersonalization to reduce affective commitment and then change turnover
78
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
intentions. Interestingly, this indirect effect is positive regarding external turnover and negative as it pertains to internal turnover. For internal auditors, depersonalization works through change in concerns about the organization to drive people away from their employers. This callous negativity does not work as a valuable currency in an organization’s internal labor market. These indirect effects compensate for the lack of a direct effect from depersonalization.
CONCLUSIONS Behavioral accounting research has recognized role stress in accounting positions in a large number of studies that have been conducted over the last two decades. Throughout that time, judicious applications of psychological theory have enriched the work in this field. Rather inexplicably, the burnout construct has been largely ignored. This has impeded our ability to theorize and to measure. This paper has been an attempt to continue to bridge the gap between the accounting literature and its source disciplines on this point. By using internal auditors only, a clear benchmark has been offered for an important occupational group. The results of this work on burnout have considerable practical implication for accounting careers and for accounting organizations. Any organization that cares about the psychological welfare of its workers should be on guard for the onset of burnout. Burnout is more likely a statement about the work environment than it is about the workers (Leiter & Maslach, 2001). Academic accountants that seek to describe the important behavioral dimensions of stress in the workplace should not ignore the role played by burnout. Those that offer interventions such as flexible work arrangements have grounded their suggestions in terms of burnout levels (Almer & Kaplan, 2002). This paper invites researchers to think in more sophisticated ways about direct and indirect effects in models that have this sort of practical consequence. The recognition of burnout as an important part of the work environment leads to other questions that are not answered in this research. For example, we have considered a single functional area, internal auditing. It would also be useful to consider whether burnout varies across work organizations in its consequences (e.g., Lachman & Aranya, 1986). These extensions should also be conducted with attention to the individual components of the burnout construct, as was done in the present research. Perhaps even more importantly, we need a more finely tuned idea of internal auditor performance. Following Witt, Andrews, and Carlson (2004), the quality and quantity of performance may have different burnout antecedents.
Internal Auditor Burnout
79
The opening of the traditional role stress model to the more general idea of the mediation of environmental characteristics provides new ways of looking at the three main role stressors. Although we did not look at role conflict, role ambiguity, and role overload in this paper, the fact that burnout clearly deserves a ‘‘place at the table’’ bears upon future specifications of a larger model that would reach back to antecedents of burnout (see Kalbers & Fogarty, 2005). Ceteris paribus, the lack of constructs between these environmental variables and outcomes of interest (e.g., performance, satisfaction, commitment, turnover intentions) tends to overstate the power of the former. This paper’s use of burnout can be taken as representative of the need for more analytically complex specifications. The present paper also stands as an exemplar of the need to use methods capable of testing these more complex models in their entirety. Beyond its academic value, the idea of burnout is one surfeit with the heavy toll often extracted from internal auditors by virtue of their employment. The human cost in terms of lost productivity, negative carryover to non-work life, and a failure to enjoy life in many senses exists in burnout in a much clearer sense than it does in notions of stress. Recognizing a dark side of internal auditing careers may have constructive value. As put by Maslach (1982, p. 40), ‘‘the promise inherent in understanding burnout is the possibility of doing something about it.’’ Investigating the balance between job demands and job resources may be a fruitful avenue (see Bakker, Demerouti, & Verbeke, 2004). Inadequate team learning may be a critical resource deficiency that induces burnout among accounting professionals (Kleinman et al., 2002). The direct consequences of burnout on job outcomes suggest that researchers and practitioners need to focus on role stressors (and their management) to better understand when and how they result in dysfunctional outcomes. Future research must carry out this next logical step. However, support for our hypotheses of burnout as a source of effects suggests that attention must be directed toward the development of burnout tendencies. This would entail the search for effective coping mechanisms that thwart the emergence of burnout tendencies and help insulate the individual from a stressful role environment.
ACKNOWLEDGMENTS The authors would like to acknowledge the support and cooperation of the Northeast Ohio Chapter of the Institute of Internal Auditors and the 23 organizations that participated in the study.
80
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
REFERENCES Abramis, D. (1994). Work role ambiguity, job satisfaction and job performance: Meta-analysis and review. Psychological Reports, 75(3), 1411–1430. Abramson, L., Seligman, M., & Teasdale, J. (1978). Learned helplessness in humans: Critique and reformulation. Journal of Abnormal Psychology, 87, 49–74. Almer, E., & Kaplan, S. (2002). The effects of flexible work arrangements on stressors, burnout, and behavioral job outcomes in public accounting. Behavioral Research in Accounting, 14, 1–34. Aranya, N., & Ferris, K. (1983). Organizational-professional conflict among U.S. and Israeli professional accountants. Journal of Social Psychology, 53(1), 153–161. Arbuckle, J., & Worthke, W. (1999). AMOS 4.0 user’s guide. Chicago, IL: SmallWaters Corporation. Bakker, A., Demerouti, E., & Verbeke, W. (2004). Using the job-demands-resources model to predict burnout and performance. Human Resource Management, 43(1), 83–104. Bartunek, J., & Reynolds, C. (1983). Boundary spanning and public accounting role stress. Journal of Social Psychology, 121(1), 65–72. Bentler, P. (1990). Comparative Fit indexes in structural models. Psychological Bulletin, 107, 588–606. Bentler, P., & Dudgeon, P. (1996). Covariance structure analysis: Statistical practice, theory and directions. Annual Review of Psychology, 47, 588–606. Bokemeier, L., Bokemeier, J., & Tipgos, M. (1990). Burnout among men and women CPAs. Paper presented at the Annual Meetings of the American Accounting Association. Bokemeier, L., Lorentzen, J., Bokemeier, J., & Tipgos, M. (1995). Work and family factors associated with burnout among women CPAs. Paper presented at the Midwest Regional Meetings of the American Accounting Association. Brayfield, A., & Rothe, H. (1951). An index of job satisfaction. Journal of Applied Psychology, 35, 307–311. Brookings, J., Bolton, B., Brown, C., & McEvoy, A. (1985). Self-reported, job burnout among human service professionals. Journal of Occupational Behavior, 8, 143–150. Collins, K., & Killough, L. (1992). An empirical examination of stress in public accounting. Accounting, Organizations and Society, 17(6), 535–547. Cordes, C., & Dougherty, T. (1993). A review and integration of research on job burnout. Academy of Management Review, 18(4), 621–656. Daft, R. L. (1983). Learning the craft of organizational research. The Academy of Management Review(October), 539–546. Demerouti, E., Verbeke, W. J. M. I., & Bakker, A. B. (2005). Exploring the relationship between a multidimensional and multifaceted burnout concept and self-rated performance. Journal of Management, 31(2 April), 186–209. Feldman, D., & Weitz, B. (1988). Career plateaus in the salesforce: Understanding and removing blockages to employee growth. Journal of Personal Selling and Sales Management, 8(1), 23–32. Figler, H. (1980). Managing stress. Management Accounting, 62(2), 22–28. Fogarty, T., Singh, J., Rhoads, G., & Moore, R. (2000). Antecedents and consequences of burnout in accounting: Beyond the role stress model. Behavioral Research in Accounting, 12, 31–67.
Internal Auditor Burnout
81
Freudenberger, H. (1974). Staff burn-out. Journal of Social Issues, 30(1), 159–165. Gaertner, J., Hemmeter, P., & Pitman, M. (1987). Employee turnover in public accounting: A new perspective. CPA Journal, 58(8), 30–37. Golembiewski, R., Hilles, R., & Dally, R. (1983). Phases of progressive burnout and their work site covariants: Critical issues in OD research and praxis. Journal of Applied Behavioral Science, 19(3), 461–481. Gregson, T. (1992a). The advantages of LISREL for accounting researchers. Accounting Horizons, 6(4), 42–48. Gregson, T. (1992b). An investigation of the causal ordering of job satisfaction and organizational commitment in turnover models in accounting. Behavioral Research in Accounting, 4, 80–95. Halbeslebin, J., & Buckley, R. (2004). Burnout in organizational life. Journal of Management, 30(6), 859–880. Harrell, A., Taylor, M., & Chewning, E. (1989). An examination of management’s ability to bias the professional objectivity of internal auditors. Accounting, Organizations and Society, 14(3), 259–269. Haskins, M., Baglioni, A., & Cooper, C. (1990). An investigation of the sources, moderators and psychological symptoms of stress among audit seniors. Contemporary Accounting Research, 6(2), 361–385. Heian, J. (1988). A study of stress in public accounting at the middle management level. Unpublished paper, The Pennsylvania State University. Hellriegal, D., & White, G. (1973). Turnover of professionals in public accounting: A comparative analysis. Personnel Psychology, 26(2), 239–246. Hunsaker, J. (1986). Burnout: The culmination of long term stress. Industrial Management, 28(November–December), 24–26. Jackson, S., Schwab, R., & Schuler, R. (1986). Toward an understanding of the burnout phenomenon. Journal of Applied Psychology, 71(4), 630–640. Jo¨reskog, K. G., & So¨rbom, D. (1993). LISREL 8: Structural equation modeling with the SIMPLIS command language I. Hillsdale, NJ: Earlbaum. Journal of Accountancy. (1984). Management burnout: What the CPA can do. 157(January), 123. Kalbers, L., & Fogarty, T. (1995). Professionalism and its consequences a study of internal auditors. Auditing: A Journal of Practice and Theory, 14(1), 64–86. Kalbers, L., & Fogarty, T. (2005). Antecedents to internal auditor burnout. Journal of Managerial Issues, 17, 134–161. Kalimo, R., Pahkin, K., Mutaneu, P., & Toppinen-Tanner, S. (2003). Staying well or burning out at work: Work characteristics and personal resources as long-term predictors. Work and Stress, 17(2), 109–122. Kelley, T., & Seiler, R. (1982). Auditor stress and time budgets. CPA Journal, 52(7), 24–36. Kleinman, G., Siegel, P., & Eckstein, C. (2002). Teams as a learning forum for accounting professionals. Journal of Management Development, 21(6), 427–460. Kusel, J., & Deyoub, N. (1983). Internal auditor burnout. Internal Auditor, 40(5), 22–25. Lachman, R., & Aranya, N. (1986). Job attitudes and turnover intentions among professionals in different work settings. Organizational Studies, 7, 279–294. Landis, R., Beal, D., & Tesluk, P. (2000). A comparison of approaches to forming composite measures in structural equation models. Organizational Research Methods, 3, 186–207.
82
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
Lee, R., & Ashforth, B. (1993). A longitudinal study of burnout among supervisors and managers: Comparisons between Leiter and Maslach (1988) and Golembiewski et al. (1986) models. Organizational Behavior and Human Decision Processes, 54, 369–398. Lee, R., & Ashforth, B. (1996). A meta-analytic examination of the correlates of the three dimensions of job burnout. Journal of Applied Psychology, 81(2), 123–133. Leiter, M., & Maslach, C. (1988). The impact of interpersonal environment on burnout and organizational commitment. Journal of Organizational Behavior, 9(3), 297–308. Leiter, M., & Maslach, C. (2001). Burnout and quality in a sped-up world. Journal for Quality and Participation, 24(2), 48–51. MacCallum, R. C. (1995). Model specification: Procedures, strategies, and related issues. In: H. H. Rick (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 16–36). Thousand Oaks, CA: Sage. Maslach, C. (1982). Understanding burnout: Definitional issues in analyzing a complex phenomenon. In: W. Paine (Ed.), Job stress and burnout: Research, theory, and intervention perspectives (pp. 29–40). Beverly Hills, CA: Sage. Maslach, C., & Jackson, S. (1996). Maslach burnout inventory: Manual (3rd ed.). Palo Alto, CA: Consulting Psychologists. Meyer, J., & Allen, N. J. (1991). A three-component conceptualization of organizational commitment. Human Resource Management Review, 1, 61–89. Meyer, J., & Allen, N. J. (1997). Commitment in the workplace: Theory, research, and application. Thousand Oaks, CA: Sage. Meyer, J., Paunonel, F., Gellatyl, I., Goffin, R., & Jackson, D. (1989). Organizational commitment and job performance: It’s the type of commitment that matters. Journal of Applied Psychology, 74, 152–156. Moore, J. (2000a). One road to turnover: An examination of work exhaustion in technology professionals. MIS Quarterly, 24(1), 141–168. Moore, J. (2000b). Why is this happening? A causal attribution approach to work exhaustion consequences. Academy of Management Review, 25(2), 335–349. Nunnally, J. (1967). Psychometric theory. New York: McGraw-Hill. Oppenheim, A. (1966). Questionnaire design and attitude measurement. New York: Basic Books. Pei, B., & Davis, F. (1989). The impact of organizational structure on internal auditor organizational–professional conflict and role stress. Auditing: A Journal of Practice and Theory, 8(1), 101–115. Poznanski, P., & Bline, D. (1997). Using structural equations modeling to investigate the causal ordering of job satisfaction and organizational commitment among staff accountants. Behavioral Research in Accounting, 9, 154–171. Rafferty, J., Lemkau, J. P., Purdy, R. P., & Rudisill, J. (1986). Validity of Maslach burnout inventory for family practice physicians. Journal of Clinical Psychology, 42(3), 488–492. Rebele, J., & Michaels, R. (1990). Independent auditors’ role stress: Antecedent, outcome and moderating variables. Behavioral Research in Accounting, 2, 124–153. Reed, S., & Kratchman, S. (1985). The differential impact of role overload, inter-role conflict and coping behavior upon the job satisfaction of male and female accountants. Unpublished paper, Texas A&M University. Reed, S., Kratchman, S., & Strawser, R. (1994). Job satisfaction, organizational commitment, and turnover intentions of United States accountants: The impact of locus of control and gender. Accounting, Auditing and Accountability, 7, 31–58.
83
Internal Auditor Burnout
Rose, R. (1983). Burnout. Journal of Accountancy, 156(4), 22–25. Sanders, S. (1998). How to keep burnout from becoming one more side effect of busy season. Accounting Office Management and Administration Report, 98(1), 2–3. Senatra, P. (1980). Role conflict, role ambiguity and organizational climate in a public accounting firm. The Accounting Review, 55(4), 594–603. Seyle, H. (1976). The stress of life. New York: McGraw-Hill. Shirom, A. (1989). Burnout in work organizations. In: G. Cooper & I. Robertons (Eds), International review of industrial and organizational psychology (pp. 25–48). New York: Wiley. Shirom, A., & Ezrachi, Y. (2003). On the discriminant validity of burnout, depression and anxiety: A reexamination of the burnout measure. Anxiety, Stress and Coping, 16(1), 83–97. Singh, J., Goolsby, J., & Rhoads, G. (1994). Behavioral and psychological consequences of boundary spanning burnout for customer service representatives. Journal of Marketing Research, 31, 558–569. Slavin, N. (1980). Role ambiguity and conflict in an alien audit environment. Unpublished doctoral dissertation, City University of New York. Smith, C., Organ, D., & Near, J. (1984). Organizational citizenship behavior: Its nature and antecedents. Journal of Applied Psychology, 68(4), 653–663. Strawser, R., Kelly, J., & Hise, R. (1982). What causes stress for management accountants? Management Accounting, 64(3), 32–35. Sweeney, J., & Summers, S. (2002). The effects of busy season workload on public accountants’ job burnout. Behavioral Research in Accounting, 14, 223–246. Toppinen-Tanner, S., Kalimo, R., & Mutaneu, P. (2002). The process of burnout in white-collar and blue-collar jobs: Eight-year prospective study of exhaustion. Journal of Organizational Behavior, 23(5), 555–670. West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with nonnormal variables: Problems and remedies. In: H. H. Rick (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 56–75). Thousand Oaks, CA: Sage. Wheaton, B., Muthen, B., Alwin, D., & Summers, G. (1977). Assessing reliability and stability in panel models. In: D. Heise (Ed.), Sociological methodology 1977 (pp. 1–35). San Francisco: , CA. Witt, L., Andrews, M., & Carlson, D. (2004). When conscientiousness isn’t enough: Emotional exhaustion and performance among call center customer service representation. Journal of Management, 30(1), 149–160. Wolpin, J., & Greenglass, E. (1991). Is job satisfaction an antecedent or a consequence of psychological burnout? Human Relations, 44(2), 193–208. Wood, D., & Wilson, J. (1987). The auditors speak: An empirical study of behavioral auditing. Unpublished manuscript.
APPENDIX Job satisfaction It seems that my friends are more interested in their jobs than I am. I feel fairly well satisfied with my present job. I definitely dislike my work.
84
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
I feel that I am happier in my work than most other people. Most days I am enthusiastic about work. I like my job better than the average worker does. I find real enjoyment in my work. Affective organizational commitment
I would be very happy to spend the rest of my career in this organization. I really feel as if this organization’s problems are my own. I do not feel like ‘‘part of the family’’ at my organization. I do not feel ‘‘emotionally attached’’ to this organization. This organization has a great deal of personal meaning for me. I do not feel a strong sense of belonging to my organization. Continuance organizational commitment
It would be very hard for me to leave my organization right now, even if I wanted to. Too much of my life would be disrupted if I decided I wanted to leave my organization right now. Right now, staying with my organization is a matter of necessity as much as desire. I believe that I have too few options to consider leaving this organization. One of the few negative consequences of leaving this organization would be scarcity of available alternatives. One of the major reasons I continue to work for this organization is that leaving would require considerable personal sacrifice; another organization may not match the overall benefits I have here. If I had not already put so much of myself into this organization, I might consider working elsewhere. Job performance How do you rate yourself in terms of the quantity of work you accomplish? How do you rate yourself in terms of your ability to reach your goals? How do you rate yourself in terms of the evaluations you have received from your supervisor(s)? How do you rate yourself in terms of the quality of your relations with those that you audit? How do you rate yourself in terms of your ability to manage time and expenses? How do you rate yourself in terms of the respect you receive from others for your job performance?
Internal Auditor Burnout
85
How do you rate yourself in terms of the quality of your performance with regard to the use of appropriate audit procedures in the appropriate circumstances? External turnover intentions I plan to actively look for a job with another company within the next year. I often think about quitting my job. I will voluntarily leave this company within the next three years. Internal turnover intentions I expect internal auditing to be a good ‘‘stepping stone’’ to better positions within the company for me. In the foreseeable future, the company plans to rotate me out of internal auditing to another area of the company. Burnout: Emotional exhaustion
Working with clients is really a strain for me. I feel that I am working too hard for my clients. Working with my boss directly puts too much stress on me. I feel emotionally drained by the pressure my boss puts on me. I feel frustrated because of working directly with coworkers. I feel I work too hard trying to satisfy coworkers. I feel dismayed by the actions of top management. I feel burned out from trying to meet top management’s expectations. Burnout: Reduced personal achievement
I I I I I I I I
feel feel feel feel feel feel feel feel
I perform effectively to meet the needs of my clients. effective in solving the problems of my clients. I am an important asset to my supervisor. my supervisor values my contribution to the firm. my coworkers truly value my assistance. I am a positive influence on my coworkers. I satisfy many of the demands set by top management. I make a positive contribution toward top management goals.
Burnout: Depersonalization I feel I treat some clients as if they were impersonal ‘‘objects.’’ I feel indifferent toward some of my clients. I feel a lack of personal concern for my supervisor.
86
TIMOTHY J. FOGARTY AND LAWRENCE P. KALBERS
I I I I I
feel feel feel feel feel
I am becoming more hardened toward my supervisor. I have become callous toward my coworkers. insensitive toward my coworkers. I am becoming less sympathetic toward top management. alienated from top management.
FAIRNESS, BUDGET SATISFACTION, AND BUDGET PERFORMANCE: A PATH ANALYTIC MODEL OF THEIR RELATIONSHIPS Adam S. Maiga ABSTRACT This study uses a path analytic model to investigate the influence of fairness (i.e., procedural fairness, distributive fairness, and interactional fairness) on managers’ budget satisfaction and the influence of managers’ budget satisfaction on budget performance. To this end, data from 92 U.S. individual managers are used for the study. The results show that fairness perceptions have a significant positive impact on budget satisfaction which, in turn, positively affects budget performance. Further analyses indicate that budget satisfaction mediates the relationship between fairness measures and budget performance. The implications, limitations, and directions for future research are discussed.
INTRODUCTION Budgeting is one of the fundamental decision-making processes in organizations. During budget formulation, officials determine the portion of the Advances in Accounting Behavioral Research, Volume 9, 87–111 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09004-1
87
88
ADAM S. MAIGA
organization’s resources that the manager of each unit will be authorized to spend. The perceived fairness of budgeting is likely to be related to managers’ attitudes and behavior as it provides a condition in which norms of entitlement or propriety are fulfilled. Therefore, fair treatment of managers sets into motion a social exchange process by which the supervisory and organizational efforts to make fair decisions engender an obligation to reciprocate on the part of managers (Lind & Tyler, 1988). Consequently, much research has examined the relationship between organizational fairness – an organization’s fair treatment of its employees (managers) and work attitudes and behaviors (Colquitt, Conlon, Wesson, Porter, & Ng, 2001; Cropanzono & Greenberg, 1997). The main effect approach that characterized the first wave of organizational fairness research (Alexander & Ruderman, 1987; Folger & Konovsky, 1989; McFarlin & Sweeney, 1992) has since been complemented by research that seeks to explain the mechanisms that underpin the reported relationship between organizational fairness and managers’ work outcomes (Konovsky & Pugh, 1994; Moorman, Blakely, & Niehoff, 1998). Likewise, interest in fairness research has also grown in managerial accounting during the last 10 years. However, most of the accounting research to date has involved experimental methods in a budget (Lindquist, 1995; Libby, 1999, 2001) or transfer pricing decision setting (Kachelmeier & Towry, 2002). This research has generally supported the contention that consideration of fairness issues is relevant to the design of the process being studied. Two recent exceptions are Wentzel (2002) and Lau and Lim (2002), who used survey methodology in an organizational setting. Wentzel (2002) conducted her study in a downsized corporate environment and found that participation in the budget process was positively associated with perceived distributive fairness, which in turn was associated with goal commitment. Lau and Lim (2002) found that procedural fairness has an indirect effect on performance through participation. Although this stream of research has contributed substantially to explaining the social exchange basis of employee attitudes and behaviors, it is not without limitations. First, in spite of the tripartite conceptualization of organizational fairness (distributive, procedural, and interactional), much of this research has not examined all three dimensions of fairness simultaneously (Konovsky & Pugh, 1994; Kumar, Scheer, & Steenkamp, 1995; Livingstone, Roberts, & Chonko, 1995; Netemeyer, Boles, McKee, & McMurrian, 1997; Organ, 1988). For example, Masterson, Lewis, Goldman, and Taylor (2000) examined procedural and interactional fairness while Moorman et al. (1998) examined procedural fairness. A fuller understanding of the social exchange basis of employee work attitudes and behaviors may require a simultaneous
Fairness, Budget Satisfaction, and Budget Performance
89
examination of all three dimensions of organizational fairness. Second, while some studies in the organizational behavior domain demonstrate a positive association between fairness and performance (Brockner et al., 1994; Brockner & Siegel, 1995; Earley & Lind, 1987; Lind, Kanfer, & Earley, 1990), Renn (1998) finds no evidence that procedural fairness perceptions are directly related to task performance. The mixed results of recent accounting studies also suggest that the impact of fairness perceptions on performance may not be direct, but rather indirect. For example, Libby (1999) suggests that prior studies were unable to link increased perception of fairness to increased performance because it is unclear whether fairness leads to performance directly or through some intervening or moderator variables. She concluded that ‘‘further work allowing for causal analysis of data collected in field settings is required in order to further explore the process by which perceptions of fairness translate into improved performance.’’ Recent studies in social exchange research suggest that satisfaction mediates the influence of fairness dimensions on performance (e.g., see Netemeyer et al., 1997). Unfortunately, empirical studies in budget settings have largely failed to study the presence and importance of such mediating processes. This study, therefore, attempts to fill this gap in the literature. It contributes to the budgeting literature by investigating, through budget satisfaction, both the direct and indirect impact of perceptions of fairness on performance. Addressing this gap will enable integration of processes that heretofore have been examined independently. Also, inclusion of all three fairness dimensions avoids misspecification bias due to omitted constructs, and an understanding of the mediating pathways is likely to yield theoretical payoffs and more concrete managerial guidelines for enhancing fairness–outcome relationships. The remainder of this paper is organized as follows. In the next section, definitions are provided, the relevant literature is reviewed, and the hypotheses are developed. The research methodology and statistical results are discussed in the third and fourth sections, respectively. The paper concludes with a discussion of the findings and suggestions for future research.
DEFINITIONS, LITERATURE REVIEW, AND HYPOTHESES DEVELOPMENT Definitions Procedural fairness is concerned with the procedure used in allocating resources (Thibaut & Walker, 1975), the impact of the fairness of
90
ADAM S. MAIGA
decision-making procedures used to determine those procedural outcomes (Levanthal, 1980), and the attitudes and behavior of the people involved in and affected by those decisions (Levanthal, 1980; Lind & Tyler, 1988). Distributive fairness describes the perceived fairness of the outcomes employees receive (Adams, 1963, 1965; Bies & Moag, 1986; Blau, 1964; Colquitt et al., 2001; Cropanzano & Greenberg, 1997; Homans, 1961). In the budgeting literature, the concept of distributive fairness is related to the notion of ‘‘fair share.’’ A fair share is an expectation concerning the size of the resource distribution that a manager should receive relative to other managers. It reflects the ‘‘base’’ distribution that managers receive in the previous budget period adjusted for a proportion of any increase (decrease) in the firm’s total distribution. Interactional fairness is concerned with the quality of treatment received from decision-makers and the extent to which formal decision-making procedures are properly enacted (Bies & Moag, 1986; Tyler & Bies, 1990). In this study, interactional fairness is defined as the extent to which managers felt they have been treated fairly regarding personal interaction with supervisors throughout the budget-setting process. Budget satisfaction is one of the most frequently studied attitudinal variables in the supervisory literature that pertains to the use of budget and performance. Satisfaction is a pleasurable or positive emotional state that results from self-appraisal of experiences (Livingstone et al., 1995). Lofquist and Dawis (1969) define satisfaction as ‘‘the pleasurable state resulting from the appraisal of the extent to which the work environment fulfills an individual’s requirement.’’ Solly and Hohenshil (1986) state that ‘‘satisfaction is an attitude individuals hold about their work consisting of a general or global factor of satisfaction as well as a collection of specific factors related to sources of work reinforcement.’’ According to Spector (1997, p. 2), ‘‘satisfaction is simply how people feel about their jobs and different aspects of their jobs.’’ Satisfaction is typically conceptualized as a multidimensional construct, including satisfaction with one’s job, supervisor, coworkers, payment conditions, promotional programs, company policy, and feelings of job security (Churchill, Ford, & Walker, 1979). In this study, the concept of satisfaction involves collaboration with the supervisor, support from the supervisor, and overall satisfaction with the budget, where overall satisfaction refers to the manager’s cumulative satisfaction with all prior budgetsetting processes as well as satisfaction received from the most recent budget-setting process. Budget performance is the degree to which managers perceive they have met budgetary targets.
91
Fairness, Budget Satisfaction, and Budget Performance
Literature Review and Hypotheses Development This study proposes a model (see Fig. 1) in which budget satisfaction mediates the relations between fairness perceptions and budget performance. Specifically, the model contends that more fairness during budgeting fosters a higher sense of budget satisfaction, which, in turn, increases managers’ budget performance. The model also investigates the direct relationship between fairness and budget performance. Procedural Fairness and Budget Satisfaction Procedural fairness proposes that people consider the fairness of the formal organizational procedures that result in decisions. Procedural fairness is important to employees because it offers some control over the process and outcomes of decisions, thereby reassuring them about the likely fairness of their outcomes (Thibaut & Walker, 1975). In a budget setting, managers may view the fair enactment of budgetary procedures by their supervisor as a necessary condition for overall procedural fairness in budgeting. If this necessary condition is not met, managers may care little about the fairness of the formal budgetary procedures. If, on the other hand, the supervisor enacts budgetary procedures fairly and thereby fulfills the necessary condition for overall procedural fairness in budgeting, managers may look beyond how the procedures were enacted to focus on the fairness of the formal procedures themselves. Thibaut and Walker (1975) contend that allowing individual disputants a fair procedure leads to enhanced satisfaction. This leads to the following hypothesis: H1. The greater the procedural fairness, the greater the manager’s budget satisfaction. Procedural Fairness Budget Satisfaction
Distributive Fairness
Interactional Fairness
Fig. 1.
Theoretical Framework.
Budget Performance
92
ADAM S. MAIGA
Distributive Fairness and Budget Satisfaction When employees observe that reasonable standards are not applied consistently across all employees, distributive fairness judgments are likely to be affected (Kumar et al., 1995). In a limited-budget scenario, employees know that inconsistent application of standards could upset the input–outcome relationship by providing a greater allocation to some and a lesser allocation to others (Sashkin & Williams, 1990). This notion is in accordance with the equity theory which suggests that individuals may perceive distributive fairness as the ratio of their outputs to inputs, i.e., people compare the ratio of their own inputs and outcomes with those of relevant others and judge outcome fairness according to the match between the inputs and outcomes of each party (Adams, 1965). Therefore, H2. Distributive fairness positively relates to manager’s budget satisfaction. Interactional Fairness and Budget Satisfaction Interactional fairness theory holds that an individual’s reaction to an organization is dependent on the individual’s interpersonal treatment during the allocative decision process, the decision makers’ provision of adequate explanation for the decision, and the treatment of employees with respect when implementing the decision (Brockner & Wiesenfeld, 1996; Moorman, 1991; Skarlicki & Folger, 1977). Supervisors promote interactional fairness when they allow the employees to participate in setting budgetary procedural fairness. When supervisors help managers develop a plan to improve budget performance and communicate clearly that the organization is concerned for their well-being, interactional fairness is likely enhanced. Therefore, managers’ perceptions of interactional fairness may be associated with how they perceive supervisors’ valuation of their contribution, thereby affecting satisfaction (Moorman, 1991). Although similar value judgments can be communicated through formal procedures, the quality interactions with the supervisor in budgetary decision-making provide compelling evidence of an individual manager’s worth (Klieman, Quinn, & Harris, 2000). Thus, the following hypothesis is stated: H3. Interactional fairness positively influences a manager’s budget satisfaction. Budget Satisfaction and Budget Performance Fox (1974) argues that the key link with performance is to get employees not just to do their job but to act beyond contract to go over and above what they are formally required to do. Organ (1977) and Petty, McGee, and Cavender (1984) also argue that a satisfied employee is also a productive employee.
Fairness, Budget Satisfaction, and Budget Performance
93
Hence, one of the key ways to improve performance is to improve the level of satisfaction. However, research linking satisfaction to performance has not been conclusive (Brown & Peterson, 1993). For example, Judge, Bono, Thoresen, and Patton (2001) update previous findings and note that though the satisfaction–performance relationship is weak (correlation ¼ .30), it is positive and significant. Ostroff (1992) and Ostroff and Schmitt (1993) have found reliable relations between satisfaction and performance at the organizational level. Also, Harter, Schmidt, and Creglow (1998) have linked overall satisfaction to various indicators of the performance of a variety of business units. On the other hand, Iaffaldano and Muchinsky (1985) suggest that satisfaction and performance form only ‘‘an illusory correlation between two variables that we logically think should interrelate, but in fact do not.’’ However, the above studies did not investigate the satisfaction–performance link within the context of the budget setting. Becker and Green (1962) and Otley (1978) suggest that, in the budget setting, where managers view budgets as realistic and attainable, they are likely to be motivated to meet budgeted targets. Similarly, this study suggests that managers’ satisfaction with their budget leads them to put more effort into trying to achieve the budget targets. Therefore, the following hypothesis is proposed: H4. Manager’s budget satisfaction is positively related to budget performance. Mediating Effect of Budget Satisfaction The expectancy model of motivation provides that individuals will exert effort only if the effort has a reasonable probability of achieving a defined goal, and achieving the goal is associated with positive expectations of receiving desirable outcomes. Integrating Hypotheses 1 through 4, it can be argued that budget satisfaction is expected to mediate the direct relationship between the three fairness perceptions and budget performance. In other words, it is through budget satisfaction that fairness perceptions can influence budget performance. In short, these fairness perceptions are believed to affect budget performance through the budget satisfaction they facilitate or reinforce. It is expected that budget satisfaction, in turn, will have an impact on budget performance. Therefore, the effect of fairness perceptions on budget performance (which is the final outcome) could be indirect through budget satisfaction, an intermediate outcome. To explore this conjecture, the following hypothesis is tested: H5. Budget satisfaction mediates the direct effects of fairness perceptions on budget performance.
94
ADAM S. MAIGA
RESEARCH METHOD Sample and Procedure A questionnaire was administered to a sample of managers. The primary source of sample selection is the Industry Week series. Executive-level profit center managers at the hierarchical level immediately below the CEO were identified. In addition, to ensure that they have budgetary responsibility, the survey asked that they confirm their budgetary responsibility. A random sample of business unit managers yielded a mailing list of 593 managers from 593 business units1 from which a random sample of 290 names was selected.2 This sampling design enabled each of the listed companies an equal chance of being selected to ensure as far as possible that the sample was a representative population of manufacturing companies (Kerlinger, 1986; Lal, Dunk, & Smith, 1996). Manufacturing companies were selected because budgets play important roles in the manufacturing industry (Umpathy, 1987) and the sample enabled an assessment of the hypotheses. In addition, the choice of industry is consistent with other budget-related studies (Brownell, 1985; Brownell & McInnes, 1986; Chenhall & Brownell, 1988). A cover letter explained the purpose of the study with an exhortation for participation and cooperation. Based on the survey responses, a criterion was used to select the participants: each participant had budget responsibility in the subunit (see appendix, Part I). An abbreviated copy of the questionnaire used in the study appears in the appendix. In the first three weeks, 97 questionnaires were returned that was followed by a second mailing, which resulted in 29 new responses. Of the 126 returned questionnaires, only 92 were usable,3 giving a 31.72% response rate.4,5 Control for common method biases was accomplished through the design of the study’s procedures (procedural remedies). The design of the study’s procedures consists of (1) assuring respondents of anonymity (Podsakoff, MacKenzie, & Lee, 2003), (2) constructing the variable constructs carefully, and (3) counterbalancing the question order (Podsakoff et al., 2003). This additional approach has the effect of neutralizing some of the method biases that affect the retrieval stage by controlling the retrieval cues prompted by the question context.
Variable Measurement and Validation The variables in this study are procedural fairness, distributive fairness, interactional fairness, budget satisfaction, and budget performance. The
Fairness, Budget Satisfaction, and Budget Performance
95
data on these variables are obtained from the questionnaire (see the appendix for abbreviated questionnaire). The first fairness measure is procedural fairness. The measure of this construct includes eight procedural fairness statements on a seven-point Likert-scale ranging from (1) strongly disagree to (7) strongly agree. Six items were adapted from Magner and Johnson’s (1995) scale, which pertains to five of Levanthal’s (1980) six rules for determining the fairness of allocation procedures: (1) budgeting procedures are applied consistently across all responsibility areas, (2) budgeting procedures are applied consistently across time, (3) budgetary decisions for my area of responsibility are based on accurate information and well-informed opinions, (4) the current budgeting procedures contain provisions that allow me to appeal the budget set for my area of responsibility, (5) the current budgeting procedures conform to my own standards of ethics and morality, and (6) budgetary decision makers try hard not to favor one responsibility area over another. Additionally, this study uses two items to address Levanthal’s (1980) representativeness rule and the informational facet of procedural fairness (Greenberg, 1993), (7) the current budgeting procedures adequately represent the concerns of all responsibility areas, and (8) budgetary decision makers adequately explain how budget allocations for my responsibility area are determined. The second fairness measure is distributive fairness. Following Magner and Johnson (1995) and Greenberg (1993), distributive fairness includes managers’ responses to five items. Four items are adapted from Magner and Johnson’s (1995) distributive fairness scale. Magner and Johnson’s (1995) scale was developed for use in a budgeting environment and assesses various comparative bases (needs, expectations, and what is deserved) that managers may use when judging the fairness of distributions. An additional item addresses the interpersonal facet of distributive fairness (Greenberg, 1993). Hence, the following are used to assess distributive fairness using a sevenpoint Likert scale, with possible responses ranging from (1) strongly disagree to (7) strongly agree: (1) my responsibility area received the budget that it deserved, (2) the budget allocated to my responsibility area adequately reflects my needs, (3) my responsibility area’s budget was what I expected it to be, (4) I consider my responsibility area’s budget to be fair, and (5) my supervisor expresses concern and sensitivity when discussing budget restrictions placed on my area of responsibility. The third fairness measure is interactional fairness. The measure focuses on the supervisor’s interpersonal behavior. Specific uses include the degree to which the supervisor was sensitive to employees’ needs, considered
96
ADAM S. MAIGA
employees’ rights, and dealt with employees in an honest and dignified manner. This study uses five items from the work of Folger and Konovsky (1989) and Moorman (1991) and asks managers to indicate the extent to which they believe their supervisor did each of the following during the last budget process: (1) was honest and ethical in dealing with you, (2) showed a real interest in trying to be fair, (3) treated you with respect and dignity, (4) was sensitive to your personal needs, and (5) showed concern for your rights as an employee. These items assess interactional fairness on a seven-point Likert scale, with possible responses ranging from (1) strongly disagree to (7) strongly agree. There is debate about whether interactional fairness is a separate construct or whether it is merely the social aspect of procedural fairness (for reviews of this debate, see Bobocel & Holmvall, 2001; Cropanzano, Rupp, Mohler, & Schminke, 2001; Konovsky, 2000). Hence, an investigation is carried out on the three fairness measures to assess their validity. First, an assessment of the sampling adequacy for the fairness measures indicate that the Bartlett Test of Sphericity (w2 ¼ 935.14, Significance ¼ .000) and the Kaiser–Meyer–Olkin (KMO) Measure of Sampling Adequacy (.78) produced results in the acceptable range. Next, to examine the extent to which these measures are interrelated, this study uses factor analysis with principal component analysis and with varimax rotation to determine the grouping of the fairness measures. Three factors with eigenvalues greater than one emerged from the analysis, with the varimax rotation factor solution retaining 75.39% of the total variance in the data. Factor loadings, eigenvalues, and corresponding Cronbach alphas are provided in Table 1. The factor solutions for the defined constructs support the construct validity of the survey instrument. The Cronbach alpha coefficient is .73 for procedural fairness, .89 for distributive fairness, and .92 for interactional fairness, indicating high internal reliability for the scales (Nunnally, 1967). Convergent validity is demonstrated by each factors having multiple-question loadings in excess of .50. In addition, discriminant validity is supported, since none of the questions in the factor analysis have loadings in excess of .40 on more than one factor. Overall, these tests support the validity of the measures representing the constructs used in this study. Next, this study uses the mean score of each of the three measures for subsequent analyses. Budget Satisfaction Satisfaction with work as it relates to budget and the immediate supervisor is assessed as a feature of budget satisfaction. Three items used to measure
97
Fairness, Budget Satisfaction, and Budget Performance
Table 1.
Results of Factor Analysis of Fairness Measures.
Eigenvalues Percent variance explained (69.33) Cronbach alpha Item loading My responsibility area received the budget that it deserved The budget allocated to my responsibility area adequately reflects my needs My responsibility area’s budget was what I expected it to be I consider my responsibility area’s budget to be fair My supervisor expresses concern and sensitivity when discussing budget restrictions placed on my area of responsibility Budgeting procedures are applied consistently across all responsibility areas Budgeting procedures are applied consistently across time Budgetary decisions for my area of responsibility are based on accurate information and well-informed opinions The current budgeting procedures contain provisions that allow me to appeal the budget set for my area of responsibility The current budgeting procedures conform to my own standards of ethics and morality Budgetary decision makers try hard not to favor one responsibility area over another The current budgeting procedures adequately represent the concerns of all responsibility areas Budgetary decision makers adequately explain how budget allocations for my responsibility area are determined Was honest and ethical in dealing with you Showed a real interest in trying to be fair Treated you with respect and dignity Was sensitive to your personal needs Showed concerns for your rights as an employee
Interactional Fairness
Distributive Fairness
Procedural Fairness
5.83 29.45 0.92
3.75 19.71 0.89
2.15 11.23 0.73
0.08
0.94
0.26
0.19
0.86
0.11
0.11
0.93
0.06
0.18
0.91
0.11
0.04
0.88
0.21
0.11
0.08
0.85
0.33
0.09
0.75
0.34
0.02
0.86
0.10
0.08
0.95
0.33
0.09
0.75
0.31
0.05
0.78
0.19
0.02
0.86
0.41
0.35
0.74
0.78 0.83 0.73 0.82 0.74
0.10 0.05 0.11 0.06 0.10
0.19 0.04 0.17 0.04 0.15
98
ADAM S. MAIGA
budget satisfaction are borrowed from Smith, Kendall, and Hulin (1969) and modified for this study. Managers were asked the following: As it relates to the most recent budget process, please indicate the extent to which you were satisfied with (1) your collaboration with your supervisor and (2) the support you get from your supervisor; (3) How is your overall satisfaction? Items were rated from (1) very dissatisfied to (7) very satisfied. A factor analysis was conducted to assess its dimensionality. The Bartlett Test of Sphericity (w2 ¼ 223.35, Significance ¼ .000) and the KMO Measure of Sampling Adequacy (.74) produced results in the acceptable range. A factor analysis of the three items was subjected to a varimax rotation. The results indicate satisfactory construct validity in which the three items load above .40 level on the factor that explained 78.28% of the total variance. The Cronbach alpha coefficient is .86, which indicated high internal reliability for the scale (Nunnally, 1967). Next, this study uses the mean of this measure for subsequent analyses. Budget Performance Following Wentzel (2002), budget performance is measured with one item that specifically asked participants to self-rate their budgetary performance: Meeting budgetary targets set for my area of responsibility; this item was rated from (1) well below average to (7) well above average. This instrument was chosen because satisfaction is budget-related and is more likely to affect budgetary performance than overall job performance, which has been used in prior studies as a single global rating (see Merchant, 1981; Chenhall & Brownell, 1988). Also, as Thornton (1980) advocates, to measure performance by a self-rating method, it is necessary to guarantee respondents anonymity to avoid possible bias in the data. In fact, Thornton (1980) supports this approach for measuring performance in studies using a cognitive bias and comments: ‘‘Cognitions are an interning variable between motivational force and object performance, and should be studied.’’ Some have claimed that self-ratings of performance tend to exhibit a leniency bias compared with, say, superior ratings (Heneman, 1974, Parker, Taylor, Barrett, & Martens, 1959; Prien & Liske, 1962), others, however, disagree with the claim (Nealey & Owen, 1970). In this study, a common-method bias, inflating the fairness–budget performance correlation, is clearly a possibility, although such a bias is unlikely in the case of satisfaction because of the complex and indirect method used in its measurement. Hence, the results may, if anything, be conservative in their assessment of the amount of relationship between fairness and budget performance attributable to the linkage through satisfaction.
99
Fairness, Budget Satisfaction, and Budget Performance
RESULTS Descriptive Statistics Data for the study were obtained from 92 companies. Table 2 provides the profile of the responding companies that constitute a broad spectrum of manufacturers as defined by the 2-digit SIC codes. The classification by the primary 2-digit SIC code places the respondents in food and kindred products (11 companies), electronic and other electric equipment (27), instruments and related products (24), chemical and allied products (17), and apparel and other fabricated textile products (13). Additional information on respondents’ characteristics is provided in Table 2. The respondents to the question regarding number of years with the business unit had a mean of 5.59 years (SD ¼ 3.72) in their current position. To the number-of-years-inmanagement question, respondents indicated a mean of 13.79 years (SD ¼ 4.81). The results also show that the average number of employees is 267 (SD ¼ 139). Table 3 provides correlations for all of the variables included in the testing of hypotheses. Following James, Mulaik, and Brett (1983), both correlational Table 2. SIC Industry Code
20 36 38 28 23
Respondents’ Characteristics.
Organization Type
Number of Respondents Used in the Study
Food and kindred products Electronic and other electric equipment Instruments and related products Chemical and allied products Apparel and other fabricated textile products Total
11 27
Mean Number of years with the business unit Number of years in management Size (number of employees)
24 17 13 92 SD
5.59
3.72
13.79
4.81
267
139
100
ADAM S. MAIGA
Table 3.
Descriptive Statistics and Correlations. 1
1. Procedural fairness 2. Distributive fairness 3. Interactional fairness 4. Budget satisfaction 5. Budget performance
1.00 0.19 0.07 0.46 0.00 0.44 0.00 0.31 0.00
2
3
4
5
1.00 0.08 0.46 0.22 0.04 0.06 0.60
1.00 0.42 0.00 0.23 0.03
1.00 0.35 0.00
1.00
Correlation is significant at the 0.01 level (two-tailed). Correlation is significant at the 0.05 level (two-tailed).
and path analysis are used to investigate the hypotheses. This study uses path analysis to explore more fully the relationships among the three fairness measures, budget satisfaction, and budget performance.6 Hypotheses 1 through 4: Hypotheses 1 through 3 specify the relationships between the fairness measures and budget satisfaction. Procedural fairness, distributive fairness, and interactional fairness are each predicted to be positively related to budget satisfaction. In support of hypotheses 1 through 3, the correlations (see Table 3) indicate that all three independent variables are significantly related to budget satisfaction. The path coefficients between the three independent variables and budget satisfaction are consistent with the zero-order correlations (see Table 3), and the overall equation is significant (F ¼ 13.85, po.001). Using a one-tailed test of significance, given that all path coefficients are in the expected direction, there are significant paths (see Fig. 2) for procedural fairness (b ¼ .39 po.001), distributive fairness (b ¼ .28, po.005), and interactional fairness (b ¼ .21, po.05). Thus, these findings provide general support for hypotheses 1 through 3. Hypothesis 4 proposes a positive relationship between budget satisfaction and budget performance. Correlational analysis (see Table 3) supports this hypothesis. As expected, the zero-order correlation shows that budget satisfaction is positively related to budget performance (r ¼ .35, po.001). Path coefficients support these correlational findings for the outcome variable (i.e., budget performance), even when the three independent variables (i.e., procedural fairness, distributive fairness, and interactional fairness) are controlled for. Budget satisfaction is positively related to budget performance (b ¼ .24, po.001) (see Fig. 2). These results provide support for hypothesis 4.
101
Fairness, Budget Satisfaction, and Budget Performance
Procedural Fairness
Distributive Fairness
0.10 0.40*** 0.24*
Budget Satisfaction
0.28**
Budget Performance
0.21* 0.07 0.05
Interactional Fairness
*p < .05 **p < .005 ***p < .001
Fig. 2.
Path Diagram of Fairness, Budget Satisfaction, and Budget Performance.
These results, coupled with the support for hypothesis 1 through 3, suggest that the three fairness measures (procedural fairness, distributive fairness, and interactional fairness) are associated with budget satisfaction and that budget satisfaction is related to budget performance. However, further analysis is needed to determine whether budget satisfaction mediates the relationship between the three fairness measures and budget performance. This is to show that any direct effect of the three fairness measures on budget performance is reduced to zero when budget satisfaction is entered into the regression equation.
Mediating Effect This study uses path analysis to assess hypothesis 5, the mediating hypothesis. To ascertain the extent of mediation, this study tests the relationships between (a) the three fairness measures and budget performance, (b) budget satisfaction and budget performance, and (c) the three fairness measures and budget satisfaction. Although no direct effect between the three fairness measures and budget performance is explicitly hypothesized, these effects were estimated in the assessment of the mediating effect. Since no direct effects or reduced direct effects were found, then results supported the mediation hypothesis. Testing the mediation effect involved three steps (Baron & Kenny, 1986). The first step entailed computing the correlation between the three fairness
102
ADAM S. MAIGA
measures and budget performance to ascertain the total association between the independent and dependent variables (see Table 3). In the second step, two sets of ordinary least squares regressions were conducted. In the first set, budget satisfaction was regressed against the three fairness measures. The resulting standardized beta values represent the path coefficients of the paths from the three fairness measures in relation to budget satisfaction. This regression analysis was used to test hypotheses 1 through 3 (see Fig. 2). In the second set of regressions, budget performance was regressed against the three fairness measures and budget satisfaction. The standardized beta values represent path coefficients showing the direct paths from the independent variables to budget performance and from budget satisfaction to budget performance. These regression analyses were used to test hypothesis 4, that budget satisfaction is related to budget performance. When budget satisfaction was not controlled for, all the three fairness measures were positively associated with budget performance. However, when budget satisfaction was controlled for, the significant relationships between the three fairness measures and budget performance became insignificant. This suggests support for the hypothesis that budget satisfaction mediates the relationship between the three fairness measures and budget performance. The third step in the analysis involved decomposing the correlations between the three fairness measures and budget performance (Alwin & Hauser, 1975; James et al., 1983). The association between the three fairness measures and budget performance was examined using their zero-order correlations and standardized regression coefficients. The direct effect of the three fairness measures is the part of the total effect that is not transmitted by the mediating variable, budget satisfaction. The indirect effect of the three fairness measures on budget performance is the part of the total effect that is mediated by budget satisfaction. The spurious effect of each of the three fairness measures is due to its unanalyzed correlations with all remaining independent variables (Prescott, Kohli, & Venkatraman, 1986). These analyses allowed further examination of hypothesis 5, the extent to which budget satisfaction mediated the relationship between the three fairness measures and budget performance. Once the direct and indirect effects were obtained, the spurious effects could be calculated by subtracting the causal effects from the correlation coefficients. This analysis allowed for identifying the specific nature of the relationships between fairness, budget satisfaction, and budget performance.
103
Fairness, Budget Satisfaction, and Budget Performance
Table 4.
Decomposition of the Association Between Fairness and Budget Performance.
Procedural fairness– budget performance Distributive fairness– budget performance Interactional fairness– budget performance
E ¼ A–D
Spurious Effect A
Direct Effect B
Indirect Effect C
Total Effect D ¼ B+C
.27
.03
.17
.20
.07
.20
.10
.12
.22
.02
.36
.16
.09
.25
.11
po.05. po.01.
Table 4 breaks down the covariance between the three independent variables and the budget performance variable into direct, indirect, total, and spurious effects. Column ‘‘D ¼ B+C’’ indicates which of the three fairness measures has the strongest influence on the outcome variable. These results indicate that interactional fairness is most important in improving budget performance (total effect is .25). However, the moderate spurious calculations in Table 4, column ‘‘E ¼ A D’’ indicates that there is still some unexplained variance in the relationships between interactional fairness and the budget performance variable and between procedural fairness and budget performance. However, the purpose of this study is not to reproduce the correlation matrix but to understand the comparative contribution of direct and indirect effects. These results do provide better understanding of the magnitude of the relationships among the fairness measures, budget satisfaction, and budget performance.
SUMMARY AND DISCUSSION Using a path analytical model, the major aim of this study is to investigate the influence of fairness on budget satisfaction and the influence of budget satisfaction on budget performance. Overall, the results of this study indicate support for the theoretical framework. Fairness measures have significant relationships with budget satisfaction, and a higher level of budget satisfaction is associated with budget performance. Also, budget satisfaction
104
ADAM S. MAIGA
has a full mediation on the relationship between fairness and budget performance. This means the effect of fairness on budget performance is indirect through budget satisfaction. These findings are both intuitively and practically significant because they demonstrate the process by which fairness perceptions translate into improved budgetary performance. Thus, this study is important in practice and it contributes significantly to the literature by improving our understanding of the complex budgetary process by the recognition of budget satisfaction as a mediating variable between fairness perceptions and budget performance. The study also helps to reconcile the results reported by previous research in this area. More specifically, the results of this study suggest that fairness during budgeting appears to foster a sense of budget satisfaction, which, in turn, significantly enhances budget performance. Enhanced managers’ budget satisfaction during the budget process therefore appears to be key to maintaining budget performance. Hence, from the perspective of practice on the other hand, it is important for managers to know that fairness during the budget process can have a significant positive impact on budget satisfaction which, in turn, leads to performance. Also, in light of the estimated satisfaction–performance relationship, it appears premature to dismiss the relationship. Thus, the hypothesis that assumes a significant relationship is supported. This is an important finding as prior studies linking satisfaction to performance have been mixed. These findings are of particular interest to manufacturing units because of the critical importance to managers of achieving high budget performance. Budget performance improves with budget satisfaction, as noted above. This study suggests that satisfaction will likely be higher when higher level of perception is achieved. This model is usable for managers as a means of guiding intervention to increase budget performance. The results of this study should be assessed in light of the following limitations. First, the research framework is not exhaustive; there may be other factors, both internal and external (e.g., budget difficulty, gender, level of management control, environmental uncertainty), not included in the framework that can partially or wholly explain the results. Second, the questionnaire data are prone to common-method variance or response biases. While self-rated performance measures have been widely used in budgeting studies (e.g., Brownell & McInnes, 1986; Govindarajan, 1986; Brownell, 1982), care should be exercised in the interpretation of the results since such ratings are subjective. Third, avenues for future research still remain. For instance, this study examined the impact of fairness perceptions and budget satisfaction on budget performance in a single period. Longitudinal research is needed to
Fairness, Budget Satisfaction, and Budget Performance
105
determine if these relationships remain intact over multiple periods. Fourth, further studies may incorporate organization-prescribed rules, policies, and standards in the model to investigate their effectiveness in the pursuit of improved budget performance. Despite the limitations, the results of this study have several implications for managers and researchers. The evidence strongly suggests that the path analytic model offers a useful way for supervisors to approach managers’ budget performance, and that managers should implement practical interventions to promote fairness in budgeting. For example, by inviting subordinates to participate actively in discussions about organizational issues and involving them in the budget decision-making process may foster a sense of fairness (Ueno & Sekaran, 1992). In particular, fairness must be incorporated into development of budget satisfaction as it relates to budgetary process and the justification of attaining higher budget performance. The results of this study should enhance practitioners’ confidence in their design of their budgets.
NOTES 1. The term business unit is used to refer to a self-contained subunit (e.g., division) of a larger corporation. 2. Research budget constraints did not allow larger sample size selection. 3. The unusable returned questionnaires were either incomplete or did not meet the two selection criteria. 4. Because of contravening company policy, some preferred not to participate. 5. The use of discriminant analysis was to compare the two groups — early and late respondents (Fowler, 1993). Results revealed that the two groups did not differ significantly in either the level of the variables or in the relationship between the variables at the .05 level. This suggests that non-response bias is not likely a problem. 6. Because of the small sample size, this study uses ordinary least squares regression analysis to test the hypotheses. The Wald chi square for the theoretical model (Fig. 1) is significant (w2 ¼ 93.86, df ¼ 15, po.05). However, before presenting the study results, the theoretical model is compared to an alternative model that takes into account the correlation between procedural and interactional fairness. The chi square difference between the theoretical model and the alternate model (1.27) is not significant. Hence, the statistical results are analyzed based on the theoretical model.
REFERENCES Adams, J. S. (1963). Toward an understanding of inequity. Journal of Abnormal and Social Psychology, 65(5), 422–436.
106
ADAM S. MAIGA
Adams, J. S. (1965). Inequity in social exchange. In: L. Berkowitz (Ed.), Advances in experimental social psychology (pp. 267–299). New York: Academic Press. Alexander, S., & Ruderman, M. (1987). The role of procedural and distributive justice in organizational behaviors. Social Justice Research, 1, 177–198. Alwin, D. F., & Hauser, R. M. (1975). The decomposition effects in path analysis. American Sociological Review, 40, 37–47. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Becker, S., & Green, D. (1962). Budgeting and employee behavior. Journal of Business, 35, 392–402. Bies, R. J., & Moag, J. S. (1986). Interactional justice: Communication criteria of fairness. Research on Negotiation in Organizations, 1, 43–55. Blau, P. (1964). Exchange and power in social life. New York, NY: Wiley. Bobocel, D. R., & Holmvall, C. M. (2001). Are interactional and procedural justice different? Framing the debate. In: S. Gilliland, D. Steiner & D. Skarlicki (Eds), Research on social issues in management (pp. 85–110). Greenwich, CT: Information Age Publishing. Brockner, J., Konovsky, M., Cooper-Schneider, R., Folger, R., Martin, C., & Bies, R. J. (1994). Interactive effects of procedural justice and outcome negativity on victims and survivors of job loss. Academy of Management Journal, 37(2), 397–409. Brockner, J., & Siegel, P. (1995). Understanding the interaction between procedural and distributive justice: The role of trust. In: R. M. Kramer & T. R. Tyler (Eds), Trust in organizations: Frontiers of theory and research (pp. 391–413). Thousand Oaks, CA: Sage. Brockner, J., & Wiesenfeld, B. M. (1996). An integrative framework for explaining reactions to decisions: Interactive effects of outcomes and procedures. Psychological Bulletin, 120(2), 189–208. Brownell, P. (1982). The role of accounting data in performance evaluation, budgetary participation, and organizational effectiveness. Journal of Accounting Research, 20(1), 12–27. Brownell, P. (1985). Budgetary systems and the control of functionally differentiated organizational activities. Journal of Accounting Research, Autumn, 502–512. Brownell, P., & McInnes, M. (1986). Budgetary participation, motivation, and managerial performance. The Accounting Review, 16(4), 587–600. Brown, S. P., & Peterson, R. A. (1993). Antecedents and consequences of salesperson job satisfaction: Meta-analysis and assessment of causal effects. Journal of Marketing Research, 30(1), 63–77. 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. Churchill, G., Jr., Ford, N., & Walker, O., Jr. (1979). Personal characteristics of salespeople and the attractiveness of alternative rewards. Journal of Business Research, 7(1), 25–50. Colquitt, J. A., Conlon, D. E., Wesson, M. J., Porter, C., & Ng, K. Y. (2001). Justice at the millennium: A meta-analytic review of 25 years of organizational justice research. Journal of Applied Psychology, 86, 425–445. Cropanzano, R., & Greenberg, J. (1997). Progress in organizational justice: Tunneling through the maze. In: C. Cooper & I. T. Robertson (Eds), International review of industrial and organizational psychology (pp. 317–372). New York, NY: Wiley.
Fairness, Budget Satisfaction, and Budget Performance
107
Cropanzano, R., Rupp, D. E., Mohler, C. J., & Schminke, M. (2001). Three roads to organizational justice. In: J. Ferris (Ed.), Research in personnel and human resources management (Vol. 20, pp. 1–113). Greenwich, CT: JAI Press. Early, P. C., & Lind, E. Q. (1987). Procedural justice and participation in task selection: The role of control in mediating justice judgments. Journal of Personality and Social Psychology, 52, 1148–1160. Folger, R., & Konovsky, M. A. (1989). Effects of procedural and distributive justice on reactions to pay raise decisions. Academy of Management Journal, 32, 115–130. Fowler, F. J., Jr. (1993). Survey research methods. Newbury Park, CA: Sage. Fox, A. (1974). Beyond contract: Work, power and trust relations. London, UK: Faber and Faber. Govindarajan, V. (1986). Impact of participation in the budgetary process on managerial attitude and performance: Universalistic and contingency perspectives. Decision Sciences, 17(4), 496–513. Greenberg, J. (1993). The social side of fairness: Interpersonal and informational classes of organizational justice. In: R. Cropanzano (Ed.), Justice in the workplace: Approaching fairness in human resource management (pp. 79–103). Hillsdale, NJ: Lawrence Erlbaum Associates. Harter, J. K., Schmidt, F. L., & Creglow, A. A. (1998). A meta-analysis of the relationship of core GWA employee perceptions and business outcomes. Gallup Technical Report, Lincoln, NE. Heneman, H. G., III (1974). Comparisons of self- and superior ratings of managerial performance. Journal of Applied Psychology, 59, 638–642. Homans, G. C. (1961). Social behavior: Its elementary forms. New York, NY: Hartcourt, Brace, and World. Iaffaldano, M. T., & Muchinsky, P. M. (1985). Job satisfaction and job performance: A meta analysis. Psychological Bulletin, 97, 251–273. James, L. R., Mulaik, S. A., & Brett, J. M. (1983). Causal analysis: Assumptions, models and data. Beverly Hills, CA: Sage. Judge, A., Bono, J. E., Thoresen, C. J., & Patton, G. K. (2001). The job satisfaction–job performance relationship: A qualitative and quantitative review. Psychological Bulletin, 127(3), 376–407. Kachelmeier, S. J., & Towry, K. L. (2002). Negotiated transfer pricing: Is fairness easier said than done? The Accounting Review, 77, 571–593. Kerlinger, F. N. (1986). Foundations of behavioral research (3rd ed.). New York, NY: Holt, Rinehart and Winston, Inc. Klieman, R. S., Quinn, J. A., & Harris, K. L. (2000). The influence of employee–supervisor interactions upon job breadth. Journal of Managerial Psychology, 15(6), 587–599. Konovsky, M. A. (2000). Understanding procedural justice and its impact on business organizations. Journal of Management, 26(3), 489–511. Konovsky, M. A., & Pugh, S. D. (1994). Citizenship behavior and social exchange. Academy of Management Journal, 37, 656–669. Kumar, N., Scheer, L. K., & Steenkamp, E. M. (1995). The effects of perceived interdependence on dealer attitudes. Journal of Marketing Research, 32(3), 348–356. Lal, M., Dunk, A. S., & Smith, G. D. (1996). The propensity of managers to create budgetary slack: A cross-national re-examination using random sampling. The International Journal of Accounting, 31(4), 483–496.
108
ADAM S. MAIGA
Lau, C. M., & Lim, E. W. (2002). The intervening effects of participation on the relationship between procedural justice and managerial performance. British Accounting Review, 34, 55–78. Levanthal, G. S. (1980). What should be done with equity theory? In: K. J. Gergen, M. S. Greenberg & R. H. Willis (Eds), Social exchange: Advances in theory and research (pp. 27–55). NewYork, NY: Plenum Press. Libby, T. (1999). The influence of voice and explanation on performance in a participative budgeting setting. Accounting, Organizations and Society, 24(2), 125–137. Libby, T. (2001). Referent cognitions and budgetary fairness: a research note. Journal of Management Accounting Research, 13, 91–105. Lind, E. A., Kanfer, R., & Earley, P. C. (1990). Voice, control and procedural justice: Instrumental and non-instrumental concerns in fairness judgments. Journal of Personality and Social Psychology, 59, 952–959. Lind, E. A., & Tyler, T. R. (1988). The social psychology of procedural justice. New York, NY: Plenum Press. Lindquist, T. M. (1995). Fairness as an antecedent to participative budgeting: Examining the effects of distributive justice, procedural justice, and referent cognitions on satisfaction and performance. Journal of Management Accounting Research, 7, 122–147. Livingstone, L. P., Roberts, J. A, & Chonko, L. (1995). Perceptions of internal and external equity as predictors of outside salespeople’s satisfaction. Journal of Personal Selling and Sales Management, 15(2), 33–46. Lofquist, L., & Dawis, R. (1969). What is job satisfaction? Organizational Behavior and Human Performance, 4, 309–336. Magner, N., & Johnson, G. G. (1995). Municipal officials’ reactions to justice in budgetary resource allocation. Public Administrative Quarterly, 18(4), 439–456. Masterson, S. S., Lewis, K., Goldman, B. M., & Taylor, M. S. (2000). Integrating justice and social exchange: The differing effects of fair procedures and treatment on work relationships. Academy of Management Journal, 43, 738–748. McFarlin, D. B., & Sweeney, P. D. (1992). Distributive and procedural justice as predictors of satisfaction with personal and organizational outcomes. Academy of Management Journal, 35, 626–637. Merchant, K. A. (1981). The design of the corporate budgeting system: Influence on managerial behavior and performance. The Accounting Review, 56, 813–829. Moorman, R. H. (1991). Relationship between organizational justice and organizational citizenship behaviors: Do fairness perceptions influence employee citizenship? Journal of Applied Psychology, 76, 845–855. Moorman, R. H., Blakely, G. L., & Niehoff, B. P. (1998). Does perceived organizational support mediate the relationship between procedural justice and organizational citizenship behavior? Academy of Management Journal, 41(3), 351–357. Nealey, S. M., & Owen, T. W. (1970). A multitrait–multimethod analysis of predictors & criteria of nursing performance. Organizational Behavior and Human Performance, 5, 384–395. Netemeyer, R. G., Boles, J. S., McKee, D. O., & McMurrian, R. (1997). An investigation into the antecedents of organizational citizenship behaviors in a personal selling context. Journal of Marketing, 61(3), 85–98. Nunnally, J. C. (1967). Psychometric theory. New York: McGraw-Hill.
Fairness, Budget Satisfaction, and Budget Performance
109
Organ, D. W. (1977). A reappraisal and reinterpretation of the satisfaction–causes–performance hypothesis. Academy of Management Review, 2, 46–53. Organ, D. W. (1988). Organizational citizenship behavior: The good soldier syndrome. Lexington, MA: Lexington Books. Ostroff, C. (1992). The relationship between satisfaction, attitudes and performance: An organizational level analysis. Journal of Applied Psychology, 77, 963–974. Ostroff, C., & Schmitt, N. (1993). Configurations of organizational effectiveness and efficiency. Academy of Management Journal, 36, 1345–1361. Otley, D. T. (1978). Budget use and managerial performance. Journal of Accounting Research, 16(1), 122–131. Parker, J., Taylor, E., Barrett, R., & Martens, L. (1959). Rating scale content: III. Relationship between supervisory and self-ratings. Personnel Psychology, 12, 49–63. Petty, M. M., McGee, G. W., & Cavender, J. W. (1984). A meta-analysis of the relationships between individual job satisfaction and individual performance. Academy of Management Review, 9(4), 712–721. Podsakoff, P. M., MacKenzie, S. B., & Lee, J. Y. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of Applied Psychology, 88(5), 879–903. Prescott, J. E., Kohli, A. K., & Venkatraman, N. (1986). The market share–profitability relationship: An empirical assessment of major assertion and contradictions. Strategic Management Journal, 7, 377–394. Prien, E. P., & Liske, R. E. (1962). Assessment of higher-level personnel: Self-rating criteria: A comparative analysis of supervision ratings and incumbent self-rating of job performance. Personnel Psychology, 32(3), 187–194. Renn, R. W. (1998). Participation’s effect on task performance: Mediating roles of goal acceptance and procedural justice. Journal of Business Research, 41(2), 115–125. Sashkin, M., & Williams, R. L. (1990). Does fairness make a difference? Organizational Dynamics, 19(2), 56–71. Skarlicki, D. P., & Folger, R. (1977). Distributive and procedural fairness: Combined impact of ‘‘voice’’ and improvement on experienced inequity. Journal of Personality and Human Resources Management, 35, 108–119. Smith, P. C., Kendall, L. M., & Hulin, C. L. (1969). The measurement of satisfaction in work and retirement. Chicago, IL: Rand McNally. Solly, D. C., & Hohenshil, T. H. (1986). Job satisfaction of school psychologist in a primary rural state. School Psychology Review, 15(1), 119–126. Spector, P. (1997). Job satisfaction. Thousand Oaks, CA: Sage. Thibaut, J., & Walker, L. (1975). Procedural justice: A psychological analysis. Hillsdale, NJ: Erlbaum. Thornton, G. C., III (1980). Psychometric properties of self-appraisals of job performance. Personnel Psychology, 33, 263–271. Tyler, T. R., & Bies, R. J. (1990). Beyond formal procedures: the interpersonal context of procedural justice. In: J. S. Carroll (Ed.), Applied social psychology in business settings (pp. 77–98). Hillsdale, NJ: Erlbaum. Ueno, S., & Sekarn, U. (1992). The influence of culture on budget control practices in the U.S.A. and Japan: An empirical study. Journal of International Business Studies, 23(4), 659–674.
110
ADAM S. MAIGA
Umpathy, S. (1987). Current budgeting practices in U.S. industry: The state of the art. Quorum Publishers. 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.
APPENDIX Part I Do you have a budget responsibility in your division?______Yes_____No If you answer to the above question is yes, please answer the remaining parts of the questionnaire, otherwise stop and return the questionnaire.
Part II Distributive Fairness (Greenberg 1993; Magner & Johnson, 1995) (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. My responsibility area received the budget that it deserved. 2. The budget allocated to my responsibility area adequately reflects my needs. 3. My responsibility area’s budget was what I expected it to be. 4. I consider my responsibility area’s budget to be fair. 5. My supervisor expresses concern and sensitivity when discussing budget restrictions placed on my area of responsibility. Procedural Fairness (Greenberg, 1993; Levanthal, 1980; Magner & Johnson, 1995) (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. Budgeting procedures are applied consistently across all responsibility areas. 2. Budgeting procedures are applied consistently across time. 3. Budgetary decisions for my area of responsibility are based on accurate information and well-informed opinions. 4. The current budgeting procedures contain provisions that allow me to appeal the budget set for my area of responsibility. 5. The current budgeting procedures conform to my own standards of ethics and morality
Fairness, Budget Satisfaction, and Budget Performance
111
6. Budgetary decision makers try hard not to favor one responsibility area over another. 7. The current budgeting procedures adequately represent the concerns of all responsibility areas. 8. Budgetary decision makers adequately explain how budget allocations for my responsibility area are determined. Interactional Fairness (Folger & Konovsky, 1989; Moorman, 1991) (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. 2. 3. 4. 5.
Was honest and ethical in dealing with you. Showed a real interest in trying to be fair. Treated you with respect and dignity. Was sensitive to your personal needs. Showed concerns for your rights as an employee.
Budget Satisfaction (Smith et al., 1969) (1 ¼ strongly dissatisfied, 2 ¼ moderately dissatisfied, 3 ¼ mildly dissatisfied, 4 ¼ neutral, 5 ¼ mildly satisfied, 6 ¼ moderately satisfied, 7 ¼ strongly satisfied) 1. Collaboration with supervisor. 2. Support from your supervisor. 3. Overall satisfaction. Budget Performance (Wentzel, 2002) (1 ¼ well below average, 7 ¼ well above average) 1. Meeting budgetary targets set for my area of responsibility. Part III Please answer the following: 1. 2. 3. 4.
Number of years at this position? ___________ Number of years in management? __________ What is the number of employees at your company? ___________ Please provide your 2-digit SIC-code ____
This page intentionally left blank
112
UNDERSTANDING INVESTMENT EXPERTISE AND FACTORS THAT INFLUENCE THE INFORMATION PROCESSING AND PERFORMANCE OF INVESTMENT EXPERTS Kinsun Tam, James L. Bierstaker and Inshik Seol ABSTRACT To investigate the nature of investment expertise and factors affecting the information processing and performance of investment experts, this paper hypothesizes normative characteristics of investment expertise and compares such characteristics with actual characteristics documented in prior literature on the investment expert. Based on collective evidence from these sources, a model of investment expertise is proposed. Results support the existence of investment expertise in (1) the nature of knowledge, (2) problem solving and information search, and (3) performance. A variety of factors that could influence the information processing and performance of the investment expert, including personal, cognitive, and contextual elements, are also discussed in the paper and included in the proposed model of investment expertise.
Advances in Accounting Behavioral Research, Volume 9, 113–156 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09005-3
113
114
KINSUN TAM ET AL.
INTRODUCTION One interesting feature of the investing process is the information flow between investment experts and non-professional, non-expert investors. Investment experts acquire, process, and use accounting and non-accounting information to arrive at judgments over investment opportunities. Judgments by investment experts, because of their purported expertise, often become inputs into non-experts’ investment decisions. Since non-expert investors rely on investment experts’ judgments, the quality of these judgments could impact the financial well-being of the investing public. A parallel can be drawn between the activities of the investment expert and those of the auditor. Both the investment expert and the auditor review financial and other business information and make judgments on behalf of investors, but these reviews occur at starkly different frequencies. While the auditor expresses an opinion on a few client companies annually, the investment expert may make recommendations on a multitude of companies daily (Hirst & Hopkins, 1998). Naturally, the investment expert has more opportunities to provide his/her judgment to the investing public. The role of investment experts’ recommendations and the market’s response to these recommendations is well documented in previous research (Barber, Lehavy, McNichols, & Trueman, 2001; Krishnan & Booker, 2002). Despite investment experts’ high-reporting frequency and strong potential impact on the market, they have not been actively studied until recently (Birnberg & Shields, 1989; Hussein & Rosman, 1997). Accordingly, this study elicits evidence from financial Behavioral Accounting Research (BAR), other behavioral research, markets-based research, and biographies and autobiographies of legendary investors on the existence, nature, and factors affecting investment expertise. Making use of these resources to address a question crucial to investment decisions is in line with advocacy in the accounting literature for a multi-method and multidimensional research approach (Biggs, Selfridge, & Krupka, 1993; Kothari & Zimmerman, 1995). This study exploits the advantages of financial BAR referred to by Bamber (1993) and Bernard (1993), suggests future research to enhance financial BAR as advocated by Hussein and Rosman (1997), and answers the call for pooling the strengths of behavioral and markets-based research (Schipper, 1991; Bernard, 1993; Koonce & Mercer, 2005). This study makes three distinct contributions. First, it offers comprehensive insights on expertise in an economically important area. This paper draws insights from
Understanding Investment Expertise and Factors
115
both behavioral research and market-based research and covers investment expert biographies and autobiographies. Findings from these various sources are integrated under an expertise framework based on Be´dard and Chi (1993) and Bouwman and Bradley (1997) and a judgment model by Rohrbaugh and Shanteau (1999). Second, it proposes a model of investment expertise. The proposed model considers factors identified in previous research that influence investment expertise, and suggests new factors for investigation in future research. Third, it identifies future research opportunities. The following sections motivate research on investment expertise, develop a framework for the study, compare normative versus actual expertise, and identify factors that influence performance. The final section offers concluding remarks and discusses future research directions.
THE IMPORTANCE OF RESEARCH ON INVESTMENT EXPERTISE A fundamental question in expertise studies of any domain is whether domain-specific expertise really exists. While the expertise in some domains such as chess and music is widely demonstrated and recognized (Chase & Simon, 1973; Ericsson, Krampe, & Tesch-Romer, 1993), the same is unclear in some other domains (Shanteau, 1995). For instance, behavioral researchers have suggested that the existence of expertise is conditional, depending on whether a correct solution exists and how complex the task is to complete (Be´dard & Chi, 1993; Camerer & Johnson, 1991; Sundali & Atkins, 1994). Understanding the existence and nature of investment, expertise is particularly interesting since competition forces investment service providers (such as securities firms, mutual fund managers, and investment advisory companies) to differentiate their services from those of their rivals. Many providers resort to an emphasis on their investment expertise to entice customers. Therefore, examining the existence and nature of investment expertise is important to the investing public. Owing to the lack of regulation on the use of descriptions suggesting investment expertise, there is considerable latitude for intentional and unintentional misuse. An innocent investor taking advice from a self-proclaimed investment expert may assume unexpected risk if the expert fails to deliver quality services.
116
KINSUN TAM ET AL.
The existence (or non-existence) of investment expertise has crucial implications to the investment industry and the regulatory authorities. If investment expertise does not exist, then the claims by the providers of investment services are illegitimate and should be prohibited. If it does exist, on the other hand, it will be useful to determine its attributes and characteristics, which can then serve as a yardstick for the investing public to evaluate the expertise of the investment service providers. In addition, this set of attributes and characteristics may improve hiring and training processes of investment service providers and may enable regulatory authorities to more closely monitor the use of terms alleging expertise in investment services advertising. However, cognitive and motivational biases may exist that impair the performance of the expert (Daniel, Hirshleifer, & Teoh, 2002; McEwen & Welsh, 2001). It is important to identify these biases and consider ways to mitigate them, in order to maximize performance. For example, recent legislation aimed at enhancing the objectivity of investment service providers represents fertile ground for future research. Furthermore, if investment expertise does exist, knowing how experts’ investment performance is related to the elements of decision-making is important. For example, Rohrbaugh and Shanteau (1999) suggest that the decision-making process is the result of the integration of personal, cognitive, and contextual elements. Therefore, understanding the relationship between performance and such variables as experience, knowledge, abilities, motivation, task environment (Libby & Luft, 1993), training (Bonner, Davis, & Jackson, 1992), and prior performance (Sundali & Atkins, 1994) will provide valuable insights on investment expertise.1 For example, Bonner et al. (1992) find that training can explain knowledge in tax planning. However, the role of training in financial analysts’ knowledge and investment expertise is relatively unknown. One of the few studies that examines this issue, Jacoby et al. (2001), suggests that novice performance can be improved by training them to follow information-processing strategies commonly used by experts. Therefore, understanding investment experts’ information search and usage may help develop better training programs for future financial information users (Bouwman, Frishkoff, & Frishkoff, 1987). Currently, financial advisory services providers make expertise claims in terms of a combination of factors including experience, past performance, certifications, education, and training. The use of experience as a surrogate for expertise is accepted at least partially in various domains because of the difficulty in measuring expertise. Nevertheless, as documented in Bonner and Lewis (1990) and Libby and Luft (1993), experience, rather than
Understanding Investment Expertise and Factors
117
expertise, may be a weak explanatory variable of performance in comparison with knowledge and ability. Previous empirical research on investment experts does not show a strong positive relationship between experience and performance (Brown, 2001). If there is little or no correlation between investment experience and investment performance, the claim of expertise based on experience may be misleading (Jacob, Lys, & Neale, 1999). Likewise, if prior performance and training do not explain future investment performance, the related claims of expertise are delusive. It is imperative to have such misconceptions, if any, clarified. To determine if investment expertise exists and to investigate its nature and factors affecting investment expertise, this study reviews the properties of expertise in other domains and compares them with the properties of investment expertise documented in prior literature. The following section develops the framework of this comparison.
THE FRAMEWORK OF THIS STUDY Characteristics of expertise in general have been well investigated in prior studies (Hunter, 1983; Schmidt, Hunter, & Outerbridge, 1986). Researchers have outlined the fundamental features of expertise normative to any domain. These norms then naturally form the framework that specifies what investment expertise should be. For instance, Be´dard and Chi (1993) identify, inter alia, the following invariants of expertise:2 1. Experts know more about their domain than do novices. 2. Experts not only know more, but their knowledge is better organized. 3. On the basis of their greater knowledge and better organization, experts perform better than novices do. Likewise, Bouwman and Bradley (1997) also characterize expertise in some collateral dimensions: 1. The nature of expert knowledge (corresponding to invariants #1 and #2 in Be´dard and Chi) 2. The expert’s advantage in information search (related to invariant #2 in Be´dard and Chi) 3. Direct and surrogate measures of performance (corresponding to invariant #3 in Be´dard and Chi). Moreover, Rohrbaugh and Shanteau (1999) argue that there are three elements that influence decision-making processes: personal, cognitive, and
118
KINSUN TAM ET AL.
contextual. Personal elements refer to internally developed mechanisms for handling a given situation. Cognitive elements include information-processing strategies used to evaluate and combine various sources of inputs. Contextual elements involve environmental variables that are external to the individual (Rohrbaugh and Shanteau, 1999, p. 2). The basic model of investment expertise, shown in Fig. 1, is based on a framework consistent with Be´dard and Chi (1993) and Bouwman and Bradley (1997). The extended model in Fig. 2 includes other personal, cognitive, and contextual factors (Rohrbaugh & Shanteau, 1999) that could incrementally affect performance (or knowledge or processing) at the margin. The proposed basic model shows that the investment expert’s performance is influenced by knowledge and knowledge organization. Following this model the paper examines, through reviewing prior literature, investment expertise in three main dimensions: (1) the nature of knowledge, (2) problem representation, problem solving, and information search, and (3) performance (including surrogate performance measures). For each dimension, this study first establishes what investment expertise should be (normative), and then assesses the extent that such attributes of expertise have been (actual) established or acquired by investment service providers holding
Performance • • • •
Knowledge • Industry • Financial • Task specific
Direct measures Consensus? Reliability? Self-insight?
Knowledge Organization • Problem representation? • Information processing (configural, directive)
Note: Arrows represent directions of influence. Gaps in prior literature awaiting future research are highlighted in bold.
Fig. 1.
A Basic Model of Investment Expertise Based on Be´dard and Chi (1993).
Understanding Investment Expertise and Factors
119
Performance • • • •
Direct measures Consensus? Reliability? Self-insight?
Knowledge
Knowledge Organization
• Industry • Financial • Task specific
• Problem representation? • Information processing
( configural, directive)
Factors incrementally affecting performance (or knowledge or processing) at the margin Personal • Experience (firm-specific, task specific?) • Abilities (perceptual) • Motivation (incentives) Cognitive • Biases (cognitive, motivational) • Ethical reasoning? Contextual • Training? • Task environment (information display, disclosure method? structure? decision aids?) Note: Arrows represent directions of influence. Gaps in prior literature awaiting future research are highlighted in bold.
Fig. 2.
An Extended Model of Investment Expertise.
out to be or being regarded as experts. The extended model incorporates personal, cognitive, and contextual elements as factors incrementally affecting performance (or knowledge or processing) at the margin. That is, these elements are included as supplements rather than as a main dimension. Comparison of normative and actual expertise helps identify unanswered
120
KINSUN TAM ET AL.
questions for future research. These research questions (RQs) are numbered and presented after each comparison.
NORMATIVE VERSUS ACTUAL EXPERTISE The Nature of Knowledge The Normative Nature of Knowledge Prior research suggests that an expert possesses more, and better organized, domain-specific knowledge than do novices (Posner, 1988; Be´dard & Chi, 1993; Christ, 1993). For example, Be´dard and Chi (1993) find that the expert has deep knowledge structures and multiple links between concepts and procedures. Biggs et al. (1993) report that the expert auditor’s knowledge of client-specific events is causally linked to the knowledge of financial measures. Biggs et al. (1993) and Frederick (1991) suggest that the expert has better memory organization, which is episodic in nature. In addition, an expert’s knowledge is represented as schemata, frames, templates, or scripts (Waller & Felix, 1984; Gibbins, 1988; Riesbeck & Schank, 1989). Researchers across various domains seem to converge on the same conclusion: Be´dard and Chi (1993) on expert auditors, Chase and Simon (1973) on chess experts, Boshuizen and Schmidt (1992) on medical experts. Accordingly, the investment expert is expected to possess domain-specific knowledge, organize this knowledge in memory with templates, and link specific episodic memory to other relevant knowledge areas and concepts. The Actual Nature of Knowledge Prior research has shown that investment experts possess a large amount of domain-specific knowledge (Bouwman, 1995). For example, Anderson (1988) finds that experienced analysts possess more specialized knowledge than less experienced analysts. Bouwman et al. (1987) suggest that analysts develop task-specific knowledge. Similarly, Maines, McDaniel, and Harris (1997) find that unlike novice analysts, expert analysts are aware of industry differences, which highlights the role of industry knowledge as a component of financial analysis expertise. Domain-specific knowledge can be organized to develop templates for the experts. In Bouwman et al. (1987), concurrent verbal protocol analysis strongly supports the existence of financial templates in memory that (1) describe the characteristics of a company under review and (2) contain specific expectations for the line items of the financial statements. In a later
Understanding Investment Expertise and Factors
121
study, Bouwman (1995) finds that financial analysts immediately retrieve from memory a great deal of information about the container industry when presented with the annual financial statements of a container company. Also, Clement (1999) finds a positive relationship between analysts’ forecast accuracy and firm-specific forecasting experience. These results support the existence of templates with industry knowledge and the existence of links between episodic memory and financial statement knowledge. These actual knowledge characteristics of investment experts are consistent with the expected normative characteristics of such knowledge as suggested in prior research. Common characteristics shared between investment experts and other domain experts support the claim that investment expertise does exist. Other Factors that Incrementally Affect Knowledge As proposed in Fig. 2, personal, cognitive, and contextual elements influence investment expertise. Therefore, future research on how knowledge structures for different types of industries and judgments influence investment experts’ information processing and performance is needed (Libby, Bloomfield, & Nelson, 2002, p. 786). Such research may lead to the development of more sophisticated decision aids for investment experts (Bouwman & Bradley, 1997). RQ1. How do knowledge structures of investment experts for different industries and judgments differ from novices, and how does this influence information processing and performance? Problem Representation Normative Problem Representation A major ingredient of expert performance is the expert’s ability to mentally represent problems successfully (Be´dard & Chi, 1993). A problem representation can be defined as the manner in which people organize their perceptions of the nature, demands, and constraints of the decision situation in their memories (Newell & Simon, 1972). In the context of auditing, Christ (1993) and Bedard and Biggs (1991) find that problem representations affect problem-solving processes used to develop solutions. In addition, Kaplan and Simon (1990) and Bierstaker, Bedard, and Biggs (1999) suggest that once a problem representation is developed, it is difficult to change. Therefore, if the initially developed problem representation is incomplete, it may inhibit hypothesis generation and evaluation. Empirical evidence from
122
KINSUN TAM ET AL.
expertise research in various domains agrees on the experts’ superiority in interpreting and understanding problems: Chi, Feltovich, and Glaser (1981) in physics problems, Voss, Greene, Post, and Penner (1983) in political science problems, and Feltovich (1981) in medical problem-solving. Accordingly, the investment expert is expected to exhibit a more complete problem representation. Actual Problem Representation Few studies investigate the problem representations of investment experts, and the results are inconclusive. Yates, McDaniel, and Brown (1991) find indirect evidence that the expert’s representation is richer than that of the novice. However, they fail to find a positive relationship between problem representation and performance. The inverse expertise effect can be explained as a by-product of experts’ cue utilization (Yates et al., 1991). In other words, experts use richer representations (Murphy & Wright, 1984) that make the judgment tasks more difficult and distort the accuracy of their forecasts. However, the impact of problem representation on performance can be a function of problem complexity. For example, Matsatsinis, Doumpos, and Zopounidis (1997) discuss the importance of knowledge representation in developing expert systems, especially in complicated domains (e.g., financial analysis). These findings provide very limited support for the existence of investment expertise and inconclusive results on the positive relationship between problem representation and performance. However, the aforementioned research on knowledge suggests ways that the mental representations of investment experts and novices may differ. For example, experts’ mental representations may contain financial statement knowledge, industry templates, and episodic memories that are missing from novices’ representations. Other Factors that Incrementally Affect Problem Representation Clearly, a great deal of additional research is needed to explore how the problem representations of investment experts differ from novices, and how this influences their decision processes and performance. Process-tracing techniques, such as verbal protocol analysis and/or information search techniques (e.g., search monitor), may be particularly useful in this regard and could lead to the development of more sophisticated models of investment expertise (Bouwman & Bradley, 1997). Also, investigating how some of the contextual elements (e.g., task environment) influence problem representations may provide valuable insights on investment expertise. This leads to the following RQs:
Understanding Investment Expertise and Factors
123
RQ2. How do problem representations of investment experts differ from those of novices, and how does this influence information processing and performance? RQ3. Do investment experts have more difficulties adapting initial problem representations than novices, and what cues in the task environment might enhance or inhibit these adaptations? Problem Solving and Information Search Normative Problem Solving and Information Search Be´dard and Chi (1993) establish that experts pick the right case in analogical reasoning and better differentiate relevant from irrelevant information. For example, in auditing research, expert auditors look for information that will help them understand the overall picture while novices limit their search to information immediately applicable to the task at hand (Bedard & Mock, 1992). These observations are consistent with the presumption that experts use templates to guide their information search. In fact, Biggs, Mock, and Watkins (1988) find that expert auditors have internal schemata that allow them to organize information. Brown and Solomon (1990, 1991) and Bedard and Biggs (1991) find expert auditors to be more likely to use a configural approach of reasoning in a pattern-recognition task. Similarly, Rosman, Seol, and Biggs (1999) find that auditors use configural processing in a going-concern task, and auditors who are more accurate follow flexible strategies for financial information acquisition. Moreover, in medical research, expert radiologists use schemata to guide their information search in X-ray film analysis (Lesgold et al., 1988). Consequently, the investment expert is expected to follow a similar information search process. Actual Problem Solving and Information Search Similar to their counterparts in auditing, prior experimental research on expertise shows that investment experts use a highly configural informationprocessing approach (Slovic, 1972). For example, Hershey and Walsh (2000) find that expert financial planners show more conceptually driven information-processing patterns in investment-planning tasks compared to novices’ more data-driven patterns. Other papers use personal elements (e.g., experience) or accuracy as proxies for expertise and show that more experienced and/or accurate investment service providers’ information search follows similar strategies. For example, Bouwman (1982) finds that more experienced analysts tend to follow a directive information search strategy, where
124
KINSUN TAM ET AL.
they search for specific pieces of relevant information using a mental ‘‘checklist,’’ whereas inexperienced analysts tend to follow a lengthier sequential search strategy (i.e., moving through items in the order in which they are presented). In addition, Biggs (1984) finds that financial analysts perform structured information searches in a manner similar to those of other experts. Jacoby, Jaccard, Kuss, Troutman, and Mazursky (1984, 1985, 1986, 1987) find that better-performing analysts access different types of information, have a different pattern of information search (within-factor instead of within-stock), and access more information overall than poorer performing analysts. Anderson (1988) also finds that experienced analysts spend less time searching for information and use more directive search patterns than inexperienced analysts. Furthermore, Hunton and McEwen (1997) suggest that analysts’ performance in earnings forecasting may be linked to cognitive information search strategy. In their experiment with 60 professional financial analysts, Hunton and McEwen find that more accurate analysts use a directive information search strategy, whereas less accurate analysts rely on a sequential search strategy. Investment service providers’ historical forecasting accuracy was also linked to their search strategy observed in the experiment. In their later study, McEwen and Hunton (1999) find that accurate analysts differ from less accurate analysts in the choice of accounting information to emphasize. Accurate analysts analyze income indicators over longer time horizons and use summary indicators to a greater extent than do less accurate analysts. Overall, these findings suggest that the investment expert shares the characteristics of expertise in other domains in problem solving and information search with regard to configural information processing, thereby partially supporting the claim that investment expertise exists. Other Factors that Incrementally Affect Problem Solving and Information Search Mear and Firth (1990) find systematic differences in analyst cue weighting and combinational strategies related to a variety of demographic variables including age and experience. Another important personal element to consider is motivation. Tuttle and Burton (1999) find that monetary incentives increased analysts’ response times and cue usage. Specifically, analysts are found to use more information than in previous studies. In addition, Jacoby et al. (2001) posit that an interaction between training, ability, and motivation is likely. Future research is needed to examine how personal elements such as experience, motivation, and ability may influence investment
Understanding Investment Expertise and Factors
125
experts’ information processing and interact with contextual elements like training. RQ4. How do experience and other personal elements (e.g., motivation, confidence, and ability) influence investment experts’ information processing, and interact with contextual elements (e.g., training)? Relevant information may also differ between buy- versus sell-side analysts. One of the possible reasons includes motivational differences (a personal element) between the two types of analysts. In Williams, Moyes, and Park (1996), sell-side analysts rely more on private information from management when revising Earnings Per Share (EPS) forecasts, whereas buyside analysts more often consider other investment service providers’ opinions, market reaction, annual reports, and 10-Ks. Very few other studies compare information processing between sell-side and buy-side analysts. Future research is clearly needed. RQ5. How does the information processing of sell- and buy-side investment experts differ? Furthermore, the presentation format of financial information may also influence the information processing of investment experts. Hirst and Hopkins (1998) find that analysts are more likely to acquire and use information on unrealized gains and losses on marketable securities when that information is displayed in the statement of comprehensive income as opposed to the statement of stockholders’ equity. In addition, Hopkins (1996) finds that reporting a financial instrument in the liabilities section versus the equity section of the balance sheet affects the analysts’ stock-valuation decision. Recently, Bierstaker, Weist, and Thosar (2004) find that when an unrealized derivative gain/loss was included as a separate line item in the income statement, analysts include the gain/loss significantly more often in their P/E ratios and are more likely to list the derivative as a factor affecting their investment recommendation than when the derivative gain/loss is disclosed only in a footnote. Within the loss condition, analysts are more likely to recommend selling the stock when the derivative is recognized in the financial statements rather than disclosed in the footnotes. Results of verbal protocol analysis suggest that analysts are less likely to consider information regarding derivatives when it is contained in footnotes. Results from Bierstaker et al. are consistent with those from Harper, Mister, and Strawser (1987), Amir (1993), Imhoff, Lipe, and Wright (1995), and Davis-Friday, Folami, Liu, and Mittelstaedt (1999).
126
KINSUN TAM ET AL.
In sum, there is evidence that the information processing and performance of investment experts may be influenced by the task environment (contextual elements). Further research is needed to explore the characteristics of the task environment, such as information displays, decision aids, task structure, and disclosure methods that would aid investment experts’ access to relevant information and maximize their investment expertise. RQ6. How do characteristics of the task environment (including information display, task structure, disclosure methods, and decision aids) influence investment experts’ information processing? A summary of research on the nature of investment experts’ knowledge, problem representations, problem solving, and information search is shown in Table 1. Performance An advantage of investment expertise research is that it is possible to differentiate between better and poorer performers (Jacoby et al., 2001). Both direct and indirect measures are used in prior research to determine actual performance (Bouwman & Bradley, 1997). Direct measures reflect the economic impact of the investment expert’s decision. Indirect measures are surrogates, used when economic impact cannot be directly measured. Studies adopting indirect measures of performance typically use consensus, reliability, and self-insight as surrogates (Bouwman & Bradley, 1997; Trotman & Wood, 1991). Indirect Performance Measures Normative consensus. Consensus (i.e., agreement between experts) is the most frequently used surrogate measure of expert performance. In auditing research, for example, auditor subjects are found to exhibit a high level of consensus (Trotman & Wood, 1991; Solomon & Shields, 1995). Even though there are doubts on whether consensus is a valid surrogate of judgment accuracy (Bouwman & Bradley, 1997), studies that relate consensus to expertise apparently outnumber studies that do not. This subsection reviews prior research using consensus as a performance measure, with the assumption that investment experts should display strong consensus in their decisions. Actual consensus. In a study of the information search behavior of investment experts, Biggs (1984) asks financial analysts to pick the best company out of five in the paper industry on the basis of earning power. He finds
127
Understanding Investment Expertise and Factors
Table 1. Research on the Nature of Investment Experts’ Knowledge, Problem Representations, Problem Solving and Information Search. Authors and Date
Approach
Slovic (1972)
Experimental
Bouwman (1982)
Experimental
Biggs (1984)
Protocol analysis
Bouwman et al. (1987)
Protocol analysis
Harper et al. (1987)
Experimental
Anderson (1988)
Experimental
Mear and Firth (1990)
Experimental
Yates et al. (1991)
Experimental
Amir (1993)
Capital markets
Bouwman (1995)
Protocol analysis
Imhoff et al. (1995)
Experimental
Key Findings Investment experts use a highly configural information-processing approach More experienced analysts search for specific pieces of relevant information using a mental ‘‘checklist,’’ whereas inexperienced analysts tend to follow a lengthier sequential search strategy Financial analysts perform structured information searches in a manner similar to those of other experts Analysts develop task-specific knowledge and financial templates as they gain experience Both sophisticated and unsophisticated users of financial statements are more likely to include a pension liability in the numerator of a debt to equity ratio when the pension liability is recognized in the balance sheet rather than disclosed in a footnote Experienced analysts possess more specialized knowledge, spend less time searching for information, and use more directive search patterns than inexperienced analysts A variety of demographic variables including age and experience influence financial analysts’ cue weighting and combinational strategies Expert’s representations are richer than those of novices (based on performance) Investors underestimated the effect of postretirement benefits (PRB other than pensions) liabilities on firm value when PRB information was disclosed in footnotes Results support the existence of templates with industry knowledge, and the existence of links between episodic memory and annual report knowledge Capital markets react to obligations contained in the balance sheet, but are less responsive to footnote disclosure
128
KINSUN TAM ET AL.
Table 1. (Continued ) Authors and Date
Approach
Hopkins (1996)
Experimental
Williams et al. (1996)
Experimental
Maines et al. (1997)
Experimental
Matsatsinis et al. (1997)
Modeling
Hunton and McEwen (1997)
Retinal imaging system
Hirst and Hopkins (1998)
Experimental
Davis-Friday et al. (1999)
Capital markets
McEwen and Hunton (1999)
Retinal imaging system
Clement (1999)
Archival
Key Findings Where a financial instrument is placed in the balance sheet (the liabilities section versus the equity section) affected the impact of the financial instrument on analysts’ stock valuations Sell-side analysts rely more on private information from management when revising EPS forecasts, whereas buy-side analysts more often consider other analysts opinions, market reaction, annual reports, and 10-Ks Expert analysts are aware of industry differences that novice analysts fail to observe, indicating that industry knowledge is a component of financial analysis expertise Incorporate problem representations in developing expert systems in financial analysis In an earnings forecasting task, more accurate analysts use a directive information search strategy, whereas less accurate analysts rely on a sequential search strategy. Analysts’ historical forecasting accuracy was also linked to their search strategy observed in the experiment Analysts are more likely to acquire and use information on unrealized gains and losses on marketable securities when that information is displayed in the statement of comprehensive income as opposed to the statement of stockholders’ equity The markets treat information disclosed in footnotes as less reliable than similar information recognized in the financial statements Accurate analysts analyze income indicators over longer time horizons, and use summary indicators to a greater extent than do less accurate analysts Finds positive relationship between analysts’ forecast accuracy and firmspecific forecasting experience
129
Understanding Investment Expertise and Factors
Table 1. (Continued ) Authors and Date
Approach
Tuttle and Burton (1999) Hershey and Walsh (2000)
Experimental
Bierstaker et al. (2004)
Experimental and verbal protocol analysis
Experimental
Key Findings Monetary incentives increased analysts’ response times and cue usage Expert financial planners show more conceptually driven informationprocessing patterns compared to novices’ more data-driven pattern in investment planning tasks Within the loss condition, analysts are more likely to recommend selling the stock when the derivative is recognized in the financial statements rather than disclosed in the footnotes
considerable similarity in the subjects’ decision processes and ultimate decisions. Biggs concludes that the subjects meet Einhorn’s (1974) view of expert judgment in terms of similarity of decision behavior. Other studies, however, do not provide evidence of consensus in investment experts’ decisions. Slovic (1972), for example, follows 13 brokers’ stock-evaluation processes and finds substantial differences in the use of various cue factors. Experts (i.e., the brokers) also disagree more with one another than do novices (students). In Butler and Lang’s (1991) study, analysts behave differently by showing either persistent optimism or pessimism. In an experiment analogous to Slovic, Fleissner, and Bauman (1972), Mear and Firth (1990) study how 38 financial analysts appraise investment targets, and obtain evidence of substantial judgmental disagreement among financial analysts and considerable differences in analysts’ cue utilization patterns, including analyst cue weighting and combinational strategies. However, similar to the findings of other research studies such as Bouwman et al. (1987), analysts’ information usage appears to be related to their investment experience. In sum, research findings on the existence of consensus among investment experts are mixed. Future research is needed to reconcile these conflicting findings. RQ7. How do characteristics of the task (e.g., information display, disclosure method, task structure, and decision aids) influence investment experts’ consensus?
130
KINSUN TAM ET AL.
RQ8. How do personal elements (e.g., experience, motivation, and abilities) influence investment experts’ consensus? Normative reliability. Reliability, defined as the stability of judgment over repeated trials based on identical information, is an alternative surrogate performance measure (Solomon & Shields, 1995; Bouwman & Bradley, 1997). Even though it is not frequently used, experts normally display high reliability (Ashton, 1974; Gaumnitz, Nunamaker, Surdick, & Thomas, 1982; Be´dard, 1989). Actual reliability. In a study of financial analysts, Mear and Firth (1990) explicitly report the reliability of subjects’ judgments. They find that financial analysts utilize highly consistent judgment strategies across tasks. This result is consistent with the existence of investment expertise. However, when reliability is defined as the stability of performance rather than the stability of judgment, the result may differ. For example, Sundali and Atkins (1994) find instability of performance when the investment expert performs the repeated task of stock picking. Sundali and Atkins (1994, p. 235) remark: ‘‘The fact that an expert performed well in many previous games ... does not make that expert any more likely to do well on his next attempt.’’ Therefore, more future research on investment experts’ reliability is needed. RQ9. Why do investment experts appear to use highly consistent strategies, but are unable to achieve highly consistent performance? Normative self-insight. Another indirect measure of performance is selfinsight. Self-insight refers to the ability of decision makers to subjectively express the relative emphases placed on the available cues. It can be measured as the correlation between the real cue weights implicit in the decision maker’s statistical model with his/her subjective cue weights (Bouwman & Bradley, 1997). Limited self-insight has been reported in prior research (Ashton, 1982; Wright, 1977). Neither Be´dard (1989) nor Bouwman and Bradley (1997) recognize self-insight as a necessary characteristic of expertise. Actual self-insight. Previous studies on the investment expert’s self-insight show divergent results. In Slovic et al.’s (1972) experiment, experts exhibit less self-insight than do novices, and across all the expert subjects, selfinsight decreases with experience. This finding is consistent with Bouwman and Bradley’s (1997) hypothesis that the expert’s decision-making process is more automated. Contrary to Slovic et al.’s finding, however, Mear and Firth (1987) report a relatively high degree of self-insight displayed by financial analysts on risk and return judgments. In sum, research findings on
Understanding Investment Expertise and Factors
131
the self-insight of investment experts are mixed, and offer no insight on the existence of investment expertise. Future research that could reconcile these conflicting findings would be valuable. RQ10. Under what conditions (contextual elements) are investment experts more or less likely to demonstrate self-insight? RQ11. Does self-insight change with personal elements (e.g., experience, motivation, abilities)? Direct Performance Measures Normative performance. Prior empirical research confirms experts’ superior performance in chess and music (Posner, 1988; Ericsson et al., 1993). In addition, Schraagen (1993) finds that experts outperform novices in designing experiments for comparing the taste of various brands of colas. Similarly, expert auditors, bank loan officers, and accountants outperform novices in making going-concern judgments, predicting bankruptcy, and identifying corporate tax issues (Ashton, 1985; Weber, 1978; Zimmer, 1981; Bonner et al., 1992).3 Be´dard (1991) finds that the quality of audit decisionmaking in an audit-planning context is better among experts than among novices. Abdolmohammadi and Wright (1987) and Colbert (1989) confirm the superior performance of the expert in unstructured and semi-structured auditing tasks, but not structured tasks. Marchant (1989) finds expert auditors outperform novices when a familiar situation is encountered. Accordingly, if investment expertise does exist, superior performance should be one of its characteristics, especially on unstructured (or semi-structured) tasks. In particular, investment experts should make significantly better judgments (i.e., on security prices, earnings, dividends, market values, etc.) than do novices, and thereby produce better performance from their investment decisions. Actual performance. Insights on directly measured performance come from three types of publications, namely, (1) capital markets research literature, (2) literature on laboratory experiment using investment professionals as subjects, and (3) biographies and autobiographies on individual investment experts. Capital Markets Research. One argument against the existence of investment expertise in capital markets research is the efficient market hypothesis (Fama, 1970). Given the predominant belief that capital markets are efficient in the semi-strong form (Copeland & Weston, 1988), experts acting without inside information are expected to deliver a level of performance
132
KINSUN TAM ET AL.
consistently indistinguishable from that of novices. Ball (1989) confirms this belief with 20 years of evidence from the literature. Moreover, the literature on event studies clearly suggests that capital markets are efficient with respect to publicly available information (Fama, 1991).4 This implies that even investment experts cannot maintain superior performance based on public information.5 Moreover, with minor exceptions, capital markets researchers have not been able to discover trading strategies and heuristics that consistently generate an excess profit.6 Empirical evidence from capital markets research concurs with the expectation that consistent superior performance over time is unlikely. For example, Treynor and Mazuy (1966) evaluate the performance of 47 mutual funds and conclude that no person can outguess the market. Sometimes, professional fund managers report even lackluster performance. Jensen (1968, 1969) and Brinson, Singer, and Beebower (1991) find that returns to investors in funds are on average about 1 percent/year below the market portfolio. Both theoretically and empirically, conclusions from this line of capital markets research do not support the existence of investment expertise. Results from capital markets research on investment service providers’ forecast accuracy, however, have shown mixed results. Some research (Richards, 1976; Brown & Rozeff, 1980; O’Brien, 1987; O’Brien, 1990; and Butler & Lang, 1991) does not support the existence of investment experts who could consistently provide relatively more accurate forecasts over time. Other studies, however, document that some financial analysts can persistently outperform the others. For instance, Givoly and Lakonishok (1984) review earlier studies of the analyst’s forecasts and come to the conclusion that analysts produce more accurate earnings predictions than those generated by naive models. Brown, Richardson, and Schwager (1987) compare the accuracy of the Value Line Investment Survey with that of a time-series forecasting model and find evidence of the investment expert’s forecast superiority.7 Stickel (1992) shows that Institutional Investor AllAmerican analysts’ forecasts are more accurate than non-All-American analysts’ forecasts, and that capital market participants believe differences exist in analysts’ forecast accuracy.8 Sundali and Atkins (1994) study experts whose forecasts appear on the Investment Dartboard column of the Wall Street Journal and find that the experts outperform both market averages and randomly thrown darts in picking stocks. Sinha, Brown, and Das (1997) also find differences in financial analysts’ forecast accuracy, and attribute prior research’s failure to find such differences to inadequate control for the recency effect. Using a simple model based on past forecast
Understanding Investment Expertise and Factors
133
accuracy, Brown (2001) demonstrates that the most accurate analysts can be identified. Other Factors that Incrementally Affect Performance. Many previous studies show the influence of personal elements (e.g., experience, ability) on performance. For example, Mikhail, Walther, and Willis (1997) and Clement (1999) report that financial analysts’ performance in earnings forecasts improves with experience, suggesting that expert performance in financial analysis does exist and can be achieved over time. Ding and Wermers (2004) find that experience matters for growth-oriented mutual fund managers, but not other types of funds. The stock-picking track record of the manager, however, is a stronger predictor of manager’s performance for all types of funds. However, questions remain over expertise in earnings forecasting. Ding and Wermers (2004) suggest that experience could proxy for greater access to corporate managers. Jacob et al. (1999) caution that experience could be a spurious explanatory variable driven by survival bias. Hong, Kubik, and Soloman (2000) suggest that analysts’ ability, not experience, explains variations in earnings forecast accuracy, but this is inconsistent with expertise research in auditing that attributes superior performance to both experience and ability (Bierstaker & Wright, 2001). Future research examining the role of personal elements (e.g., experience, ability, and motivation) as well as other contextual elements (e.g., features of the task environment) is clearly needed. RQ12. Why do investment experts outperform novices in earnings forecasts but not in delivering investment profits? RQ13. Are certain personal elements (e.g., experience, abilities, and motivation) more important to investment expertise than others? RQ14. How do personal elements (e.g., experience, abilities, and motivation) influence investment experts’ performance? RQ15. How do characteristics of the task (e.g., information display, disclosure method, task structure, and decision aids) influence investment experts’ performance? Other interesting personal and cognitive elements need to be considered. There is evidence that investment service providers are systematically biased in their forecasts and recommendations perhaps due to agency problems and motivational incentives (Daniel et al., 2002). Stock recommendations are typically buys over sells by a seven to one ratio (Womack, 1996), and
134
KINSUN TAM ET AL.
forecasts are generally optimistic at the 12-month or longer horizons, but pessimistic at 3-month or less (Brown, 2001). Such biases can be related to personal or cognitive elements. The reasons behind such biases are not clear. Additional research to explore the effect of both motivational and cognitive biases on performance (McEwen & Welsh, 2001) and means of mitigating those biases through regulatory changes (such as removing the presence of certain incentives schemes and investment banking relationships) is timely and needed. Furthermore, future research could address how cognitive elements such as ethical reasoning (Louwers, Ponemon, & Radtke, 1997) and pre-decisional distortion (Wilks, 2002) interact with other contextual elements (e.g., incentives) and how they influence performance. RQ16. How do cognitive elements (e.g., motivational biases) impair the performance of investment experts, and how could those biases be mitigated? RQ17. How does investment experts’ ethical reasoning (a cognitive element) influence their information processing and performance in the presence of incentives? RQ18. Do investment experts experience pre-decisional distortion (a cognitive element) during information processing? Laboratory Experiment. An alternative approach to capital markets research, which typically employs archival data, is to examine the investment expert’s performance in an experiment. Heitner (1991) asks analysts to rank a subset of oil and gas companies in the order of expected returns over a 6month period and compares the actual returns of top rank stocks with bottom rank stocks. He finds that analysts could identify the winner stocks from the losers. Other Factors that Incrementally Affect Performance. Heitner’s (1991) result suggests that contrary to the belief in market efficiency in capital markets research, the investment expert does produce superior performance.9 However, future research is needed to isolate personal, cognitive, and contextual elements that differentiate investment expertise. Ghosh and Whitecotton (1997), for example, find that perceptual ability and tolerance for ambiguity are related to financial analysts’ earnings forecasting accuracy. Another personal element that appears to have been largely unexplored in laboratory experiments despite ample support in the behavioral accounting literature (Bouwman & Bradley, 1997) is the role of task-specific experience in investment expertise.
Understanding Investment Expertise and Factors
135
The influence of contextual variables on performance is another area deserving further investigation. Hunton and McEwen (1997) suggest motivational incentives, such as whether the investment service provider’s brokerage follows the firm and whether an underwriting relationship exists, increase the propensity of investment service providers to make more optimistic and less accurate earnings predictions. Eames, Glover, and Kennedy (2002) suggest that forecast optimism is related to an unconscious desire to justify favorable stock recommendations. Recent research by Tan, Libby, and Hunton (2002), for example, suggests that analysts’ forecasts may be affected by firms’ earnings pre-announcement strategies. Irvine (2004) suggests that analysts use positive stock recommendations to a greater extent than biased forecasts to generate higher trading commissions. Investment expert’s performance may also interact with decision aids and incentives. For example, Whitecotton (1996) finds that decision aid users with more financial forecasting experience have higher levels of earnings forecast accuracy. Ashton (1990) suggests that monetary incentives caused analysts’ performance to decrease in a bond-ratings task because they attempt to ‘‘beat the aid’’ and are unable to do so. So, additional research should be conducted that examines how the performance of the investment expert may be enhanced or inhibited by decision aids (a contextual element). Furthermore, although prior research on auditing expertise has shown that experts outperform novices in unstructured or semi-structured tasks but not in structured tasks (Abdolmohammadi & Wright, 1987; Colbert, 1989), there is little research on how task structure influences investment experts’ performance. It would be useful to develop a taxonomy of the investment expert’s tasks such as earnings forecast, securities price forecast, stock recommendations, report writing, dividends forecasts, and market value assessments. Such a taxonomy could include the structure inherent in each task, decision aids used to perform the task, and the rank and type of investment service provider who would typically perform the task (similar to Abdolmohammadi, 1999; Abdolmohammadi & Usoff, 2001 for audit tasks). Investment service providers’ performance on several of these tasks has not been examined in prior research, and future research in this area is needed. RQ19. What types of tasks do investment experts perform, how structured are these tasks, and what decision aids (contextual elements) are available for performing them?
136
KINSUN TAM ET AL.
RQ20. What is the role of task-specific experience (a personal element) in investment expertise? RQ21. What types of abilities (personal elements) are important for performing the tasks of the investment expert? RQ22. What aspects of the task environment (contextual elements) potentially influence the motivation (a personal element) and potentially bias the information processing (cognitive element) and performance of the investment expert? Biographies and Autobiographies. Biographical and autobiographical publications on individual ‘‘legendary’’ investment managers and professional investors serve as a third source of insights on the performance of investment experts. These publications typically describe experts generating consistently superior investment performance on multi-billion dollar portfolios. The three most salient instances involve Peter Lynch, George Soros, and Warren Buffett.10 Lynch is the ex-manager of Fidelity Magellan Fund, the largest of its kind in the world. In the 10 years when he was the fund manager, the fund grew at an average rate of 34 percent per year, substantially outperforming the market. Soros maintains an even stronger record. He launched the Quantum Fund in 1969 (and a few more funds subsequently) and grew the total assets of all the funds to $10 billion in 1995. Soros’ Quantum Fund returned on average 35 percent each year between 1969 and 1994. Buffett manages Berkshire Hathaway, a company with a $35-billion (as of December 31, 2003) diversified equity investment portfolio. Berkshire Hathaway’s book value per share grew at 22 percent per year from 1964 to 2003, outperforming the market by a sizable margin.11 In general, these publications claim that superior performance of the experts is well acknowledged within the investment profession, supporting the claim for the existence of expertise. RQ23. What are the characteristics (personal elements) of well-known investment experts that would lend insight into the nature of investment expertise? Summary on Performance and Surrogate Measures of Performance. The findings from research on the reliability of the expert’s judgment, laboratory studies, and biographical and autobiographical publications are consistent with the existence of investment expertise. These findings contradict the
Understanding Investment Expertise and Factors
137
general result from some previous capital markets research on market efficiency. However, this contradiction is reconcilable. In particular, capital markets research typically entails taking averages of many individuals, but the mediocre performance attained by the ‘‘average’’ individual does not necessarily preclude the possibility of a small percentage of experts earning abnormal returns. Evidence from capital markets research against the existence of investment expertise should be considered together with this caveat. Other evidence from performance and performance surrogates suggests that investment expertise does exist but could be impaired or enhanced by a variety of factors. Table 2 summarizes the research on performance. Summarizing the Comparison of Normative Versus Actual Expertise. There is considerable evidence that the investment expert exhibits attributes comparable to that of experts in other domains. Taken individually, some streams of evidence may lack statistical power owing to a limited number of observations. When considered collectively, however, substantial evidence points to the existence of investment expertise. Table 3 summarizes these results. The proposed basic and extended models of investment expertise show factors and their relationships that are previously discussed in the paper, and highlight opportunities for future research.
CONCLUSIONS To investigate the nature of investment expertise and factors affecting the information processing and performance of investment experts as well as future implications, this study reviews the characteristics of the investment expert’s nature of knowledge, problem representations, problem-solving and information search heuristics, and performance documented in prior research. This study also compares these characteristics to those of experts in other domains. Results suggest that the investment experts, consistent with experts in other domains, have more and better-organized knowledge, may have more complete problem representations, use a configural approach in problem-solving, guide information searches with templates, and outperform novices. Therefore, when collectively considered, the evidence revealed by this study provides substantial support on the existence of investment expertise.
138
KINSUN TAM ET AL.
Table 2. Authors and Date
Research on Investment Experts’ Performance. Approach
Treynor and Mazuy (1966)
Capital markets
Jensen (1968, 1969)
Capital markets
Slovic (1972)
Experimental
Slovic et al. (1972)
Experimental
Biggs (1984)
Experimental
Givoly and Lakonishok (1984) Jacoby et al. (1984)
Literature review
Jacoby et al. (1985)
Experimental
Jacoby et al. (1986)
Experimental
Jacoby et al. (1987)
Experimental
Bouwman et al. (1987)
Verbal protocol analysis
Brown et al. (1987)
Capital markets
Experimental
Key Findings Evaluated the performance of 47 mutual funds and conclude that no person can outguess the market Returns to investors in funds are on average about 1 percent per year below the market portfolio Experts disagree more with one another than do novices in the stock-evaluation processes Experts exhibit less self-insight than do novices, and across all the expert subjects, self-insight decreases with experience Financial analysts show considerable similarities of decision behavior Analysts produce more accurate earnings predictions than those generated by naı¨ ve models Better-performing analysts are more likely to ignore irrelevant outcome feedback than poorer performing analysts Better-performing analysts acquire different types of information than poorer performing analysts Better-performing analysts acquire more information, different types of information, and in a different sequence than poorer performing analysts Better-performing analysts tend to use within-factor information search, whereas poorer performing analysts more often engage in within-stock information search Investment experts have financial templates in memory that describe what a company looks like, and contain specific expectations for the line items of the financial statements Compares the accuracy of the Value Line Investment Survey with that of a timeseries forecasting model, and find evidence of the investment analyst’s forecast superiority
139
Understanding Investment Expertise and Factors
Table 2. (Continued ) Authors and Date
Approach
Mear and Firth (1987)
Experimental
Ball (1989)
Literature review
Ashton (1990)
Experimental
Mear and Firth (1990)
Experimental
Brinson et al. (1991)
Capital markets
Butler and Lang (1991) Fama (1991)
Experimental
Heitner (1991)
Experimental
Stickel (1992)
Capital markets
Lynch and Rothchild (1994, 2000) Sundali and Atkins (1994)
Autobiographies
Hagstrom (1995)
Biography
Whitecotton (1996)
Experimental
Womack (1996)
Capital markets
Event study
Capital markets
Key Findings A relatively high degree of self-insight was displayed by financial analysts on risk and return judgments 20 years of evidence suggests market efficiency Performance decreased in a bond-ratings task when analysts are given monetary incentives Financial analysts show substantial judgmental disagreement among themselves. Financial analysts utilize highly consistent judgment strategies across tasks Returns to investors in funds are on average about 1 percent/year below the market portfolio Analysts behave differently by showing either persistent optimism or pessimism Capital markets are efficient with respect to publicly available information Analysts could identify the winner oil and gas stocks from the losers Institutional All-American analysts’ forecasts are more accurate than nonAll-American analysts’ forecasts, and the capital market participants believe differences exist in analysts’ forecast accuracy Describe the outstanding performance of Fidelity Magellan during Lynch’s tenure Experts outperform both market averages and randomly thrown darts in picking stocks Describes and analyzes Warren Buffett’s success and strategies Decision aid users with more financial forecasting experience have higher levels of earnings forecast accuracy Stock recommendations are typically buys over sells by a seven to one ratio
140
KINSUN TAM ET AL.
Table 2. (Continued ) Authors and Date Ghosh and Whitecotton (1997) Hunton and McEwen (1997)
Approach Experimental
Experimental
Key Findings Perceptual ability and tolerance for ambiguity are related to financial analysts’ earnings forecasting accuracy Motivational incentives such as the analyst’s brokerage follows the firm, or an underwriting relationship, increase the propensity of analysts to make more optimistic and less accurate earnings predictions
Mikhail et al. (1997) Capital markets
Sinha et al. (1997)
Capital markets
Slater (1997)
Biography
Hong et al. (2000)
Capital markets
Clement (1999)
Capital markets
Brown (2001)
Model
McEwen and Welsh (2001)
Literature review
Jacoby et al. (2001)
Experimental
Eames et al. (2002)
Experimental
Financial analysts’ performance in earnings forecasts improves with experience Find differences in financial analysts’ forecast accuracy, and attribute prior research’s failure to find such differences to inadequate control for the recency effect Gives an account of Soros’ investment success Analysts’ ability, not experience, explains variations in earnings forecast accuracy Financial analysts’ performance in earnings forecasts improves with experience Using a simple model based on past forecast accuracy, the most accurate analysts can be identified. Forecasts are generally optimistic at the 12-month or longer horizons, but pessimistic at 3month or less Cognitive and motivational biases may exist that impair the performance of the investment expert Training with the information-accessing strategies of better-performing securities analysts enhanced novices’ performance Forecast optimism is related to an unconscious desire to justify favorable stock recommendations. Other factors in the analyst’s task environment may influence their performance
141
Understanding Investment Expertise and Factors
Table 2. (Continued ) Authors and Date
Approach
Tan et al. (2002)
Experimental
Wilks (2002)
Experimental
Ding and Wermers (2004)
Capital markets
Irvine (2004)
Capital markets
Key Findings Analysts’ forecast may be affected by firms’ earnings pre-announcement strategies Auditors’ going–concern judgments are influenced by pre-decisional distortion during information processing Experience matters for growth-oriented mutual fund managers, but not other types of funds. The stock-picking track record of the manager was a stronger predictor of manager’s performance for all types of funds Analysts use positive stock recommendations to a greater extent than biased forecasts to generate higher trading commissions
FUTURE RESEARCH DIRECTIONS There have been few studies on the investment expert in areas such as problem representation, information search processes, ethical reasoning, regulation, task structure and general task taxonomy, task-specific experience (including sell- versus buy-side analysts), training, and indirect measures of performance including consensus and reliability. In addition, findings on investment experts’ self-insight are mixed. Future research exploring these areas may help correct this deficiency and provide critical evidence on investment expertise. A summary of the RQs posed in this study is shown in the appendix. The existence of investment expertise suggests that there may be grounds for the providers of investment services to claim to be experts. Whether these claims are legitimate depends on the quality of future performance. Since future performance is not ex ante observable, surrogate performance measures are needed (e.g., past performance). However, previous studies show mixed results. For example, Sundali and Atkins (1994) demonstrate that past performance may not be an appropriate surrogate. On the other hand, Brown (2001) argues that past performance is as important as other characteristics (e.g., experience). Future research is needed to reconcile these conflicting findings.
142
Table 3.
Dimension
Decision Elements
Personal
Summary of Normative and Actual Investment Expertise.
Knowledge
Experience Abilities
Experts have more and better organized knowledge than novices Experts have more task-specific experience than novices Experts will have a higher level of ability than novices
Motivation
Experts’ level of motivation will affect their performance
Biases
Experts’ performance will be influenced by cognitive and motivational biases Experts have more complete problem representations than novices
Problem representation
Actual
Experts have more and better organized knowledge than novices Future research needed Perceptual ability and tolerance for ambiguity are related to financial analysts’ earnings forecasting accuracy Incentives such as the analyst’s brokerage follows the firm, or an underwriting relationship, increase the propensity of analysts to make more optimistic and less accurate earnings predictions Cognitive and motivational biases may exist that impair the performance of the investment expert Experts’ mental representations may contain financial statement knowledge, industry templates, and episodic memories that are missing from novices’ representations
KINSUN TAM ET AL.
Cognitive
Normative
Ethical reasoning Contextual
Task environment
Training
Performance
Consensus Evidence on consensus of investment experts is mixed Reliability
Performance
Experts are guided by templates and schemata and use a configural approach to information acquisition. Novices focus on surface features of the problem Experts’ ethical reasoning will affect their performance Features of the task environment will influence expert information processing and performance Training will enhance experts’ knowledge and performance
Evidence of experts configural information search exists, but more evidence is needed on how the task environment influences information processing
Future research is needed Information presentation format influences expert information processing and performance Training appears to enhance novices’ performance. Future research on experts is needed
Experts demonstrate a high level of consensus
Experts’ judgments exhibit high reliability
Self-insight
Experts have little self-insight
Task performance
Experts outperform novices, especially on familiar, complex, and unstructured tasks
Understanding Investment Expertise and Factors
Problem solving and information processing
Investment experts demonstrate high reliability in their judgment strategies, but not performance Self-insight may decrease with experience, but findings are mixed Evidence generally points to the superior performance of investment experts, however, biases, incentives, and other factors may also influence performance
143
144
KINSUN TAM ET AL.
In addition, although analysts appear to differ in expertise (Stickel, 1992; Sinha et al., 1997), the reasons for these differences remain unclear. As previously mentioned, Mikhail et al. (1997) find that analysts’ forecast errors decrease as they gain experience, which is consistent with the ‘‘learning by doing’’ model. Clement (1999) also finds that more experienced analysts have smaller forecast errors. Nevertheless, Jacob et al. (1999) suggest that more experienced analysts perform better because of survival bias (i.e., weaker performing analysts are forced out of the profession) and conclude that analyst performance is a function of ability rather than experience. Future research is needed to examine the type of experience and abilities needed to maximize the performance of the investment expert. A limitation of these studies is that researchers have had to infer the effects of experience and ability on investment expertise using aggregate financial data. An advantage of experimental research is that it can isolate the effects of variables on a specific group of financial statement users (McDaniel & Hand, 1996). Future behavioral research is needed to isolate the effects of personal elements (e.g., experience, ability), and contextual elements (e.g., task structure and decision aids) on investment expert performance to help determine if training will improve investment expertise (Koonce & Mercer, 2005). The roles of deliberate practice (Ericsson et al., 1993) and motivation (Tuttle & Burton, 1999) should also be explored. In addition, aspects of the task environment that may enhance or inhibit the performance of the investment expert, including incentives and information disclosure should be investigated. One promising area of future expertise research is to model investment expertise so that it could be used, for example, to train novices. This is particularly important since the number of new investors has grown dramatically. The number of households owning stocks doubled in the 1990s (Foust, 1997), for example; and many of these new investors did not have any previous investment experience (Himelstein, 1997). Currently, relatively little research exists on investment expertise training, despite the fact that small improvements in performance could lead to dramatic increases in profits (Jacoby et al., 2001). Two popular areas of modeling, computational modeling (Biggs et al., 1993; Garfinkel, 1995) and artificial neural networks (Tam & Kiang, 1992; Yoon, Guimaraes, & Swales, 1994; Chiang, Urban, & Baldridge, 1996), may yield new insights on how investment experts acquire, combine, and evaluate information to achieve superior performance. Additional research on the relative merits of using prior performance, knowledge, ability, training, motivation, ethical reasoning, and experience
Understanding Investment Expertise and Factors
145
as performance predictors may also offer valuable insights to the Securities and Exchange Commission (SEC) over the monitoring of financial services providers claiming to be experts.12 Investment services providers seeking to improve their hiring decisions may benefit from these studies. Furthermore, research on the effectiveness of recent legislation aimed at enhancing analyst objectivity by removing certain business relationships and incentive schemes is important and needed. Another limitation of capital markets studies is that they provide little information about investment experts’ information processing (Wahlen et al., 2000). Future behavioral research needs to explore factors affecting investment experts’ information processing and linkages between information processing and investment expertise (Hunton & McEwen, 1997). As an example of research exploring these factors, Bierstaker et al. (2004) find that information presentation format influences the manner in which analysts process information regarding derivatives. More future research along similar lines may help educate users of financial information and accounting disclosures as well as clarify uncertainties over the investment expert’s performance. For example, what task structure, decision aids, and incentives enable the investment expert to perform at the highest level? What biases may impair the performance of the investment expert (Daniel et al., 2001; McEwen & Welsh, 2001)? How could these biases be mitigated to maximize expert performance? Since human analysts are subject to biases, could a computer expertise model outperform the analysts? These are important questions that only future research can answer.
NOTES 1. Sundali and Atkins (1994) find that prior performance cannot explain future performance in investment analysis tasks. 2. There are situations where the expert does not excel, namely, when a correct solution procedure does not exist, when the situation is not sufficiently understood, and when there is a mismatch between expert and the task. 3. However, experts are sometimes outperformed by simple linear models (Bouwman & Bradley, 1997). 4. The other two branches of efficiency studies, namely, the test of returns predictability and the test for insider information, suffer from the joint-hypothesis problem and do not produce conclusive evidence (Fama, 1991). 5. However, recent research also suggests that markets may be less efficient than previously thought (Berg, Dickhaut, & McCabe, 1995 Berg et al 1995; Lee, 2001; Bloomfield & Wilks, 2000). 6. There are occasional reports of trading strategies that consistently yield an excess profit (Bartov & Bodnar, 1994; Harris & Ohlson, 1990). However, according
146
KINSUN TAM ET AL.
to Ball’s (1989) argument, these results are inconclusive owing to the joint-hypothesis problem or the use of outlying data. 7. Value Line Investment Survey represents the forecast of one or two analysts. 8. The Institutional Investor selects its All-American Research Team of superior earnings forecasters based on a survey of money managers. 9. One experiment-based research arrives at a different conclusion. Yates et al. (1991) ask undergraduate and graduate students in a finance course to make probabilistic forecasts of quarterly changes in the stock prices and earnings of publicly trades companies and find that the overall accuracy of both price and earnings forecasts is very modest. Instead of regarding Yates et al.’s (1991) finding as a disproof of the expert analyst’s superior performance, it may be more appropriate to count it as supportive evidence that novices (in this case, students at two levels) are unable to produce superior performance. 10. The performance and strategies of Peter Lynch are described in the books One up on Wall Street (Lynch & Rothchild, 1994) and Beating the Street (Lynch & Rothchild, 2000). Slater (1997) gives an account of Soros’ investment success in Soros: The Life, Times, & Trading Secrets of the World’s Greatest Investor. In addition, the August 23, 1993 issue of The Business Week claims that ‘‘no other investor has produced better results [than Soros].’’ Hagstrom (1995) describes and analyzes Warren Buffett’s success and strategies in The Warren Buffett Way: Investment Strategies of the World’s Greatest Investor. Although these publications are not subject to rigorous scrutiny as are accounting research papers, their accuracy is still verifiable to various extents. Lynch’s Magellan Fund is a mutual fund regulated by the Investment Company Act of 1940. The Act provides for registration, full disclosure, and regulation of investment companies to prevent fraudulent abuses. Soros’ Quantum Fund was registered in Netherlands Antilles, and therefore subject to foreign instead of US regulations. Disclosures about the Quantum Fund are not readily available. However, Soros has earned the title of ‘‘the World’s Greatest Money Manager’’ from the Institutional Investor magazine in 1981 and a similar accolade from the Business Week in 1993. If we adopt Shanteau’s (1988) operational definition of experts as ‘‘those to be so considered by colleagues,’’ then there are grounds for recognizing Soros’ performance. Lastly, Buffett’s Berkshire Hathaway is a listed company under the governance of the SEC. Public records of Berkshire Hathaway can verify the accuracy of the descriptions about Buffett. 11. The performance in terms of returns accomplished by these three experts is not adjusted for risk. For fair comparison, risk-adjusted returns should be used instead. 12. The performance in terms of returns accomplished by these three experts is not adjusted for risk. For fair comparison, risk-adjusted returns should be used instead.
ACKNOWLEDGMENTS The authors are very grateful to the editor and two anonymous reviewers for many thoughtful comments. The authors thank Stan Biggs for reviewing an earlier version of this paper. The authors are thankful to participants at the
Understanding Investment Expertise and Factors
147
ABO Research Forum at the 2003 American Accounting Association Annual Meeting in Honolulu, especially Ping Lin, for helpful comments.
REFERENCES Abdolmohammadi, M. (1999). A comprehensive taxonomy of audit task structure, professional rank and decision aids for behavioral research. Behavioral Research in Accounting, 11, 51–92. Abdolmohammadi, M., & Usoff, C. (2001). The assessment of task structure, knowledge base, and decision aids for a comprehensive inventory of audit tasks. Westport, CT: Quorum Books. Abdolmohammadi, M., & Wright, A. (1987). An examination of the effects of experience and task complexity on audit judgments. The Accounting Review, 62(1), 1–14. Amir, E. (1993). The market valuation of accounting information: The case of post-retirement benefits other than pensions. The Accounting Review, 68(4), 703–724. 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. Ashton, H. (1985). Does consensus imply accuracy in accounting studies of decision making? The Accounting Review, 60(2), 173–185. Ashton, R. (1974). Cue utilization and expert judgments: A comparison of independent auditors with other judges. Journal of Applied Psychology, 59(August), 437–444. Ashton, R. (1982). Human information processing research in accounting. American Accounting Association Monograph Number 17 (Chapter 1). Sarasota: American Accounting Association. Ashton, R. (1990). Pressure and performance in accounting decision settings: Paradoxical effects of incentives, feedback, and justification. Journal of Accounting Research, 28(Suppl.), 148–180. Ball, R. (1989). What do we know about stock market ‘‘efficiency’’? In: R. Guimaraes, B. Kingsman & S. Taylor (Eds), A reappraisal of the efficiency of financial markets (pp. 25–55). Berlin, Germany: Springer. Bamber, E. (1993). Opportunities in behavioral accounting research. Behavioral Research in Accounting, 5, 1–29. Barber, B., Lehavy, R., McNichols, M., & Trueman, B. (2001). Can investors profit from the prophets? Security analyst recommendations and stock returns. The Journal of Finance, 56(April), 531–563. Bartov, E., & Bodner, G. (1994). Firm valuation, earnings expectations, and the exchange-rate exposure effect. The Journal of Finance, 49(December), 1755–1785. Be´dard, J. (1989). Expertise in auditing: Myth or reality? Accounting, Organizations and Society, 14, 113–131. Be´dard, J. (1991). Expertise and its relation to audit decision quality. Contemporary Accounting Research, 8(1), 198–222. Bedard, J., & Biggs, S. (1991). Pattern recognition, hypothesis generation and auditor performance in analytical task. The Accounting Review, 66(3), 622–642. Be´dard, J., & Chi, M. T. H. (1993). Expertise in auditing. Auditing: A Journal of Practice and Theory, 12(Suppl.), 21–45.
148
KINSUN TAM ET AL.
Bedard, J., & Mock, T. (1992). Expert and novice problem-solving behavior in audit planning. Auditing: A journal of Practice and Theory, 11(Suppl.), 1–20. Berg, J., Dickhaut, J., & McCabe, K. (1995). The individual versus the aggregate. In: R. Ashton & A. Ashton (Eds), Judgment and decision-making research in accounting and auditing (pp. 102–134). New York, NY: Cambridge University Press. Bernard, V. (1993). Stock price reaction to earnings announcements: A summary of recent anomalous evidence and possible explanation. In: R. Thaler (Ed.), Advances in behavioral finance. New York, NY: Russell Sage Foundation. Bierstaker, J., Bedard, J., & Biggs, S. (1999). The role of problem representation shifts on auditor decision processes in analytical procedures. Auditing: A Journal of Practice and Theory, 18(1), 18–36. Bierstaker, J., Weist, D., & Thosar, S. (2004). Analysts’ responses to alternative methods of reporting unrealized gains and losses on derivatives. Academy of Accounting Studies Journal, 8(1), 1–28. Bierstaker, J., & Wright, S. (2001). A research note concerning practical problem-solving ability as a predictor of performance in auditing tasks. Behavioral Research in Accounting, 13, 49–62. Biggs, S. (1984). Financial analysts’ information search in the assessment of corporate earning power. Accounting, Organization and Society, 9, 313–323. Biggs, S., Mock, T., & Watkins, P. (1988). Auditor’s use of analytical review in audit program design. The Accounting Review, 63(1), 148–161. Biggs, S., Selfridge, M., & Krupka, G. R. (1993). A computational model of auditor knowledge and reasoning processes and the going-concern judgment. Auditing: A Journal of Practice and Theory, 12(Suppl.), 82–99. Birnberg, J., & Shields, J. (1989). Three decades of behavioral accounting research: A search for order. Behavioral Research in Accounting, 1, 23–74. Bloomfield, R., & Wilks, T. J. (2000). Disclosure effects in the laboratory: Liquidity, depth, and cost of capital. The Accounting Review, 75(January), 13–41. Bonner, S., Davis, J., & Jackson, B. (1992). Expertise in corporate tax planning: The issue identification stage. Journal of Accounting Research, 30(Suppl.), 1–28. Bonner, S., & Lewis, B. (1990). Determinants of auditor expertise. Journal of Accounting Research, 28(Suppl., April), 1–28. Boshuizen, H., & Schmidt, H. (1992). On the role of biomedical knowledge in clinical reasoning by experts, intermediates and novices. Cognitive Science, 16, 153–184. Bouwman, M. (1982). The use of accounting information: Expert versus novice behavior. In: G. Ungson & D. Braunstein (Eds), Decision making: An interdisciplinary inquiry (pp. 134– 167). Boston, MA: Kent Publishing Company. Bouwman, M. (1995). Knowledge elicitation and representation: An example from financial analysis. In: M. Vasarhelyi (Ed.), Artificial intelligence in accounting and auditing, (Vol. 3, pp. 21–52). Princeton, NJ: Markus Wiener Publishers. Bouwman, M., & Bradley, W. (1997). Judgment and decision-making: Part II. Expertise, consensus and accuracy. In: V. Arnold & S. G. Sutton (Eds), Behavioral accounting research: Foundation and frontiers (pp. 89–133). Sarasota, FL: American Accounting Association. Bouwman, M., Frishkoff, P. A., & Frishkoff, P. (1987). How do financial analysts make decisions? A process model of the investment screening decision. Accounting, Organization, and Society, 12(1), 1–29.
Understanding Investment Expertise and Factors
149
Brinson, G. P., Singer, B. D., & Beebower, G. L. (1991). Determinants of portfolio performance II: An update. Financial Analysts Journal, 47(3), 40–48. Brown, C., & Solomon, I. (1990). Auditor configural information processing in control risk assessment. Auditing: A Journal of Practice and Theory, 9, 17–38. Brown, C., & Solomon, I. (1991). Configural information processing in auditing: The role of domain-specific knowledge. The Accounting Review, 66, 100–119. Brown, L. (2001). How important is past analyst forecast accuracy? Financial Analysts Journal, 57(6), 44–49. Brown, L., & Rozeff, M. S. (1980). Analysts can forecast accurately!. Journal of Portfolio Management, 6, 31–34. Brown, L. D., Richardson, G. D., & Schwager, S. J. (1987). An information interpretation of financial analyst superiority in forecasting earnings. Journal of Accounting Research, 25(1), 49–67. Butler, K., & Lang, L. (1991). The forecast accuracy of individual analysts: Evidence of systematic optimism and pessimism. Journal of Accounting Research, 29(1), 150–156. Camerer, C. F., & Johnson, E. J. (1991). The process-performance paradox in expert judgment: How can experts know so much and predict so badly? In: K. A. Ericsson & J. Smith (Eds), Towards a general theory of expertise: Prospects and limits (pp. 195–217). New York: Cambridge Press. Chase, G., & Simon, A. (1973). The mind’s eye in chess. In: G. Chase (Ed.), Visual information processing. London, England: Academic Press. Chi, M., Feltovich, P., & Glaser, R. (1981). Categorization and representation of physics problems by experts and novices. Cognitive Science, 5(2), 121–151. Chiang, E., Urban, T., & Baldridge, G. (1996). A neural network approach to mutual fund net value forecasting. Omega International Journal of Management Science, 24, 205–215. Christ, M. (1993). Evidence on the nature of audit planning problem representations: An examination of auditor free recalls. The Accounting Review, 68(April), 304–321. Clement, M. (1999). Analyst forecast accuracy: Do ability, resources, and portfolio complexity matter? Journal of Accounting and Economics, 27, 285–303. Colbert, J. L. (1989). The effect of experience on auditor’s judgments. Journal of Accounting Literature, 8, 137–149. Copeland, E., & Weston, J. (1988). Financial theory and corporate policy (3rd ed.). Reading, MA: Addison-Wesley. Daniel, K., Hirshleifer, D., & Teoh, S. (2002). Investor psychology in capital markets: Evidence and policy implications. Journal of Monetary Economics, 49(1), 139–209. Davis-Friday, P., Folami, L., Liu, C., & Mittelstaedt, H. (1999). The value relevance of financial statement recognition vs. disclosure: Evidence from SFAS No. 106. The Accounting Review, 74(October), 403–423. Ding, B., & Wermers, R. (2004). Mutual fund stars: The performance and behavior of U.S. fund managers. http://ssrn.com/abstract=521483. Eames, M., Glover, S., & Kennedy, J. (2002). The association between trading recommendations and broker-analysts earnings forecasts. Journal of Accounting Research, 40(1), 85– 104. Einhorn, H. J. (1974). Expert judgment: Some necessary conditions and an example. Journal of Applied Psychology, 59, 562–571. Ericsson, K., Krampe, R., & Tesch-Romer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100, 363–406.
150
KINSUN TAM ET AL.
Fama, E. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 25, 383–417. Fama, E. (1991). Efficient capital markets: II. Journal of Finance, 46(5), 1575–1617. Feltovich, P. (1981). Knowledge based components of expertise. Unpublished doctoral dissertation. University of Minnesota. Foust, D. (1997). Greenspan is right to stake out Wall Street. Business Week, March 17, p. 31. Frederick, D. (1991). Auditors’ representation and retrieval of internal control knowledge. The Accounting Review, 66, 240–258. Garfinkel, S. (1995). Is your computer a smarter investor than Warren Buffett or Paul Tudor Jones? Not yet. Worth, December/January. www.simson.net/clips/1995/95.Worth.Buffett.txt Gaumnitz, B., Nunamaker, R., Surdick, J., & Thomas, F. (1982). Auditor consensus in internal control evaluation and audit program planning. Journal of Accounting Research, 20(Autumn), 745–755. Ghosh, D., & Whitecotton, S. M. (1997). Some determinants of analysts’ forecast accuracy. Behavioral Resarch in Accounting, 9(Suppl.), 50–68. Gibbins, M. (1988). Knowledge structures and experienced auditor judgment. In: A. Bailey (Ed.), Auditor productivity in the year 2000: 1987 proceeding of the eleventh Arthur Young Professors’ roundtable (pp. 51–73). Reston, VA: Council of the Arthur Young Professors. Givoly, D., & Lakonishok, J. (1984). Properties of analyst’s forecast of earnings: A review and analysis of the research. Journal of Accounting Literature, 3, 117–148. Hagstrom, R. (1995). The Warren Buffett way: Investment strategies of the world’s greatest investor. New York: Wiley. Harper, R. M., Mister, W. G., & Strawser, J. R. (1987). The impact of new pension disclosure rules on perception of debt. Journal of Accounting Research, 25, 327–330. Harris, T., & Ohlson, J. (1990). Accounting disclosures and the market’s valuation of oil and gas properties: Evaluation of market efficiency and functional fixation. The Accounting Review, 65(October), 764–780. Heitner, P. (1991). Isn’t it time to measure analysts’ track records? Financial Analysts Journal, 47(3), 5–6. Hershey, D. A., & Walsh, D. A. (2000). Knowledge versus experience in financial problem solving performance. Current Psychology, 19, 261–291. Himelstein, L. (1997). Schwab is ‘‘Fighting on Three Fronts’’. Business Week, March 10, pp. 94–95. Hirst, E. D., & Hopkins, P. E. (1998). Comprehensive income reporting and analysts’ valuation judgments. Journal of Accounting Research, 36(Suppl.), 47–75. Hong, H., Kubik, J., & Soloman, A. (2000). Security analysts’ career concerns and herding of earnings forecasts. Rand Journal of Economics, 31(1), 121–144. Hopkins, P. E. (1996). The effect of financial statement classification of hybrid financial instruments on financial analysts’ stock price judgments. Journal of Accounting Research, 34(Suppl.), 33–50. Hunter, J. (1983). A causal analysis of cognitive ability, job knowledge, job performance, and supervisory ratings. In: F. Landy, S. Zeddeck & J. Cleveland (Eds), Performance measurement and theory (pp. 257–266). Hillsdale, NJ: Erlbaum. 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(4), 497–516.
Understanding Investment Expertise and Factors
151
Hussein, M., & Rosman, A. (1997). The ex ante role of behavioral research in setting financial accounting policy. In: V. Arnold & S. G. Sutton (Eds), Behavioral accounting research: Foundations and frontiers (pp. 222–246). Sarasota, FL: American Accounting Association. Imhoff, E. A., Lipe, R. C., & Wright, D. W. (1995). Is footnote disclosures an adequate alternative to financial statement recognition? Journal of Financial Statement Analysi, 1 (Fall), 70–81. Irvine, P. (2004). Analysts’ forecasts and brokerage-firm trading. The Accounting Review, 79(1), 125–149. Jacob, J., Lys, T., & Neale, M. (1999). Expertise in forecasting performance of security analysts. Journal of Accounting and Economics, 28, 51–82. Jacoby, J., Jaccard, J. J., Kuss, A., Troutman, T., & Mazursky, D. (1987). New directions in behavioral process research: Implications for social psychology. Journal of Experimental Social Psychology, 24, 146–174. Jacoby, J., Kuss, A., Mazursky, D., & Troutman, T. (1985). Effectiveness of security analyst information accessing strategies: A computer interactive assessment. Computers in Human Behavior, 1, 95–113. Jacoby, J., Mazursky, D., Troutman, T., & Kuss, A. (1984). When feedback is ignored: The disutility of outcome feedback. Journal of Applied Psychology, 69, 531–545. Jacoby, J., Morrin, M., Johar, G., Gurham, Z., Kuss, A., & Mazursky, D. (2001). Training novice investors to become more expert: The role of information accessing strategy. Journal of Psychology and Financial Markets, 2(2), 69–79. Jacoby, J., Troutman, T., Kuss, A., & Mazursky, D. (1986). Experience and expertise in complex decision making. In: R. Lutz (Ed.), Advances in consumer research, (Vol. 13, pp. 469–472). MD: Association for Consumer Research. Jensen, M. C. (1968). The performance of mutual funds in the period 1945–1964. Journal of Finance, 23(2), 389–416. Jensen, M. C. (1969). Risk, the pricing of capital assets, and the evaluation of investment portfolios. Journal of Business, 62, 167–247. Kaplan, C., & Simon, H. (1990). In search of insight. Cognitive Psychology, 22, 374–419. Koonce, L., & Mercer, M. (2005). Using psychology theories in archival financial accounting research. Journal of Accounting Literature, 24, 221–242. Kothari, S., & Zimmerman, J. (1995). Price and return models. Journal of Accounting and Economics, 20(September), 155–192. Krishnan, R., & Booker, D. M. (2002). Investors’ use of analysts’ recommendations. Behavioral Research in Accounting, 14, 129–156. Lee, C. M. C. (2001). Market efficiency and accounting research: A discussion of: ‘‘Capital Market Research in Accounting’’ by S. P. Kothari. Journal of Accounting and Economics, 31(September), 233–253. Lesgold, A., Rubinson, H., Feltovitch, P., Glaser, R., Klopfer, D., & Wang, Y. (1988). Expertise in a complex skill: Diagnosing X-ray pictures. In: M. Chi, R. Glaser & M. J. Farr (Eds), The nature of expertise (pp. 453–494). Hillsdale, NJ: Erlbaum. Libby, R., Bloomfield, R., & Nelson, M. (2002). Experimental research in financial accounting. Accounting Organizations and Society, 27(8), 775–810. Libby, R., & Luft, J. (1993). Determinants of judgment performance in accounting settings: Ability, knowledge, motivation, and environment. Accounting, Organizations and Society, 18(5), 425–450.
152
KINSUN TAM ET AL.
Louwers, T., Ponemon, L., & Radtke, R. (1997). Examining accountants’ ethical behavior: A review and implications for future research. In: V. Arnold & S. G. Sutton (Eds), Behavioral accounting research: Foundation and frontiers (pp. 188–221). Sarasota, FL: American Accounting Association. Lynch, P., & Rothchild, J. (1994). Beating the street: The best-selling author of (one up on Wall Street) shows you how to pick winning stocks and mutual funds. New York, NY: Simon & Schuster. Lynch, P., & Rothchild, J. (2000). One up on Wall Street: How to use what you already know to make money in the market. New York, NY: Simon & Schuster. Maines, L. A., McDaniel, L. S., & Harris, M. S. (1997). Implications of proposed segmented reporting standards for financial analysts investment judgements. Journal of Accounting Research, 35(Suppl.), 1–24. Marchant, G. (1989). Analogical reasoning and hypothesis generation in auditing. The Accounting Review, 64(3), 500–513. Matsatsinis, N. F., Doumpos, M., & Zopounidis, C. (1997). Knowledge acquisition and representation for expert systems in the field of financial analysis. Expert Systems with Applications, 12, 247–262. McDaniel, L. S., & Hand, J. R. M. (1996). The value of experimental methods for practicerelevant accounting research. Contemporary Accounting Research, 13(Spring), 339–352. McEwen, R., & Hunton, J. (1999). Is analyst forecast accuracy associated with accounting information use? Accounting Horizons, 13(1), 1–16. McEwen, R., & Welsh, M. J. (2001). The effect of bias on decision usefulness: A review of behavioral financial accounting research. Advances in Accounting Behavioral Research, 4, 3–24. Mear, R., & Firth, M. (1987). Cue usage and self-insight of financial analysts. The Accounting Review, 62(1), 176–182. Mear, R., & Firth, M. (1990). A parsimonious description of individual differences in financial analyst judgment. Journal of Accounting, Auditing and Finance, 5(4), 501–520. Mikhail, M., Walther, B., & Willis, R. (1997). Do security analysts improve their performance with experience? Journal of Accounting Research, 35(Suppl.), 131–157. Murphy, G., & Wright, J. (1984). Changes in conceptual structure with expertise: Differences between real-world experts and novices. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10, 144–155. Newell, A., & Simon, H. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-Hall. O’Brien, P. (1987). Individual forecasting ability. Managerial Finance, 13, 3–9. O’Brien, P. (1990). Forecast accuracy of individual analysts in nine industries. Journal of Accounting Research, 28(Autumn), 286–304. Posner, M. (1988). Introduction: What is it to be an expert? In: M. Chi, R. Glaser & M. Farr (Eds), The nature of expertise (pp. xxix–xxxvi). Hillsdale, NJ: Erlbaum. Richards, R. M. (1976). Analysts’ performance and the accuracy of corporate earnings forecasts. The Journal of Business, 49, 350–357. Riesbeck, C., & Schank, R. (1989). Inside case-based reasoning. Erlbaum: Hillsdale, NJ. Rohrbaugh, C. C., & Shanteau, J. (1999). Context, process, and experience: Research on applied judgment and decision making. In: F. Durso (Ed.), Handbook of applied cognition (pp. 115–139). NY: Wiley. Rosman, A. J., Seol, I., & Biggs, S. (1999). The effect of stage of development and financial health on auditor decision behavior in the going-concern task. Auditing: A Journal of Practice and Theory, 18, 37–54.
Understanding Investment Expertise and Factors
153
Schipper, K. (1991). Commentary on analysts’ forecast. Accounting Horizon, 5(December), 105–121. Schmidt, F., Hunter, J., & Outerbridge, A. (1986). Impact of job experience and ability on job knowledge, work sample performance, and supervisory ratings of job performance. Journal of Applied Psychology, 71(3), 432–439. Schraagen, J. (1993). How experts solve a novel problem in experimental design. Cognitive Science, 17, 285–309. Shanteau, J. (1988). Psychological characteristics and strategies of expert decision makers. Acta Psychologica, 68, 203–215. Shanteau, J. (1995). Expert judgment and financial decision making. In: B. Green (Ed.), Risk behavior and risk management (pp. 15–31). Stockholm: Risk Research Group at the Stockholm University. Sinha, P., Brown, L., & Das, S. (1997). A re-examination of financial analysts’ differential earnings forecast accuracy. Contemporary Accounting Research, 14, 1–42. Slater, R. (1997). Soros: The life, times, and trading secrets of the worlds greatest investor. Burr Ridge, II, USA: Irwin Professional. Slovic, P. (1972). Psychological study of human judgment: Implications for investment decision making. Journal of Finance, 27, 779–799. Slovic, P., Fleissner, D., & Bauman, W. S. (1972). Analyzing the use of information in investment decision making: A methodological proposal. Journal of Business, 45, 283–301. Solomon, I., & Shields, M. (1995). Judgment and decision research in auditing. In: R. Ashton & A. Ashton (Eds), Judgment and decision making research in accounting and auditing (pp. 137–175). New York: Cambridge University Press. Stickel, S. (1992). Reputation and performance among security analysts. Journal of Finance, 47, 1811–1836. Sundali, J., & Atkins, A. (1994). Expertise in investment analysis: Fact or fiction. Organizational Behavior and Human Decision Processes, 59, 233–241. Tam, K., & Kiang, M. (1992). Managerial application of neural networks: The case of bank failure predictions. Management Science, 38(7), 926–947. Tan, H. T., Libby, R., & Hunton, J. E. (2002). Analysts’ reactions to earnings pre-announcement strategies. Journal of Accounting Research, 40(1), 223–246. Treynor, J. L., & Mazuy, K. (1966). Can mutual funds outguess the market? Harvard Business Review, 44, 131–136. Trotman, K., & Wood, R. (1991). A meta-analysis of studies of internal control judgments. Journal of Accounting Research, 29(1), 180–192. Tuttle, B., & Burton, F. G. (1999). The effects of a modest incentive on information overload in an investment analysis task. Accounting, Organizations and Society, 24(8), 673–687. Voss, J., Greene, T., Post, T., & Penner, B. (1983). Problem-solving skills in the social sciences. In: G. H. Bower (Ed.), The Psychology of learning and motivation: Advances in research and theory, 17 (pp. 165–213). New York, NY: Academic Press. Wahlen, J. M., Boatsman, J. R., Herz, R. H., Jonas, G. J., Palepu, K. G., Ryan, S. G., Schipper, K., Schrand, C. M., & Skinner, D. J. (2000). Response to the FASB preliminary views: Reporting financial instruments and certain related assets and liabilities at fair value. Accounting Horizons, 14(4), 501–508. Waller, W. S., & Felix Jr., W. L. (1984). Cognition and the auditor’s opinion formulation process: A schematic of interactions between memory and current audit evidence. In: S. Moriarty & E. Joyce (Eds), Decision making and accounting, current research (pp. 17– 48). Norman, OK: University of Oklahoma Press.
154
KINSUN TAM ET AL.
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. Whitecotton, S. (1996). The effects of experience and confidence on decision aid reliance: A causal model. Behavioral Research in Accounting, 8, 194–216. Wilks, J. (2002). Predicisional distortion of evidence as a consequence of real-time audit review. The Accounting Review, 77(1), 51–71. Williams, P., Moyes, G., & Park, K. (1996). Factors affecting earnings forecast revisions for the buy-side and sell-side analyst. Accounting Horizons, 10(3), 112–121. Womack, K. (1996). Do brokerage analysts’ recommendations have investment value? Journal of Finance, 51, 137–168. Wright, W. F. (1977). Self-insight into the cognitive processing of financial information. Accounting, Organization and Society, 2, 323–331. Yates, J. F., McDaniel, L. S., & Brown, E. S. (1991). Probabilistic forecast of stock price and earnings: The hazards of nascent expertise. Organizational Behavior and Human Decision Processes, 56, 60–79. Yoon, Y., Guimaraes, T., & Swales, G. (1994). Integrating artificial neural networks with rulebased expert systems. Decision Support Systems, 11, 497–507. Zimmer, I. (1981). A comparison of the prediction accuracy of loan officers and their linearadditive models. Organizational Behavior and Human Performance, 17, 69–74.
APPENDIX: FUTURE RESEARCH QUESTIONS Knowledge RQ1. How do knowledge structures of investment experts for different industries and judgments differ from novices, and how does this influence information processing and performance? Problem Representation RQ2. How do problem representations of investment experts differ from those of novices, and how does this influence information processing and performance? RQ3. Do investment experts have more difficulties adapting initial problem representations than novices, and what cues in the task environment might enhance or inhibit these adaptations? Information Processing RQ4. How do experience and other personal elements (e.g., motivation, confidence, and ability) influence investment experts’ information processing and interact with contextual elements (e.g., training)?
Understanding Investment Expertise and Factors
155
RQ5. How does the information processing of sell- and buy-side investment experts differ? RQ6. How do characteristics of the task environment (including information display, task structure, disclosure methods, and decision aids) influence investment experts’ information processing? Consensus RQ7. How do characteristics of the task (e.g., information display, disclosure method, task structure, and decision aids) influence investment experts’ consensus? RQ8. How do personal elements (e.g., experience, motivation, and abilities) influence investment experts’ consensus? Reliability RQ9. Why do investment experts appear to use highly consistent strategies, but are unable to achieve highly consistent performance? Self-insight RQ10. Under what conditions (contextual elements) are investment experts more or less likely to demonstrate self-insight? RQ11. Does self-insight change with personal elements (e.g., experience, motivation, abilities)? Performance RQ12. Why do investment experts outperform novices in earnings forecasts but not in delivering investment profits? RQ13. Are certain personal elements (e.g., experience, abilities, and motivation) more important to investment expertise than others? RQ14. How do personal elements (e.g., experience, abilities, and motivation) influence investment experts’ performance? RQ15. How do characteristics of the task (e.g., information display, disclosure method, task structure, and decision aids) influence investment experts’ performance?
156
KINSUN TAM ET AL.
RQ16. How do cognitive elements (e.g., motivational biases) impair the performance of investment experts, and how could those biases be mitigated? RQ17. How does investment experts’ ethical reasoning (a cognitive element) influence their information processing and performance in the presence of incentives? RQ18. Do investment experts experience pre-decisional distortion (a cognitive element) during information processing? RQ19. What types of tasks do investment experts perform, how structured are these tasks, and what decision aids (contextual elements) are available for performing them? RQ20. What is the role of task-specific experience (a personal element) in investment expertise? RQ21. What types of abilities (personal elements) are important for performing the tasks of the investment expert? RQ22. What aspects of the task environment (contextual elements) potentially influence the motivation (a personal element) and potentially bias the information processing (cognitive element) and performance of the investment expert? RQ23. What are the characteristics (personal elements) of well-known investment experts that would lend insight into the nature of investment expertise?
THE INFLUENCE OF OUTCOME KNOWLEDGE ON JUDGES AND JURORS’ EVALUATIONS OF AUDITOR DECISIONS: A REVIEW AND SYNTHESIS OF PRIOR RESEARCH D. Jordan Lowe and Philip M. J. Reckers ABSTRACT During the last several years, a stream of research has evolved that investigates the influence of outcome information on evaluation judgments in an auditor legal liability context. These studies have included judges and jurors and have utilized different cases and scenarios. Our objective in this paper is to review and discuss insights from this stream of research. This research consists of three phases. Phase 1 focuses on the robust manifestation of outcome effects in an audit legal liability context, Phase 2 examines the effectiveness of selected mitigation strategies in moderating outcome effects, and Phase 3 begins the process of developing a preliminary theoretical framework. We also discuss future research that could be done to better understand outcome effects and to test operational responses and proposed remedies.
Advances in Accounting Behavioral Research, Volume 9, 157–178 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09006-5
157
158
D. JORDAN LOWE AND PHILIP M.J. RECKERS
INTRODUCTION The large and still growing number of lawsuits filed against auditors and public accounting firms has imposed a tremendous cost on the profession. Legal exposure has caused the accounting profession to reevaluate its role and product mix. It has also prompted a search for viable legal defense options, including legislated judicial reform. In 1992, major lawsuits worldwide against the large public accounting firms totaled more that $30 billion in requested damages (Arthur Andersen et al., 1992). The legal woes of the profession have only intensified in the decade since then. Progressively, academic researchers have expressed increased interest and called for more systematic research of the U.S. legal system (Kinney, 1993, 1994; Lochner, 1993; Palmrose, 1991). The accounting profession argued that an expectations gap was the root cause of much of its legal woes. The premise was that an expectations gap arose out of diverging perceptions by the accounting profession and third parties regarding the profession’s role, responsibilities, and related performance (Balachandran, 1993; Jennings, Kneer, & Reckers, 1993). However, as the research community began to carefully examine this issue, they found that the different time perspectives from which auditors and judges/jurors viewed auditor performance was the more dominant underlying cause of any user/preparer gap. In the current U.S. legal system, evaluative judgments are made after the fact, whereas the decisions being evaluated were made at an earlier point in time. That is, when an auditor is charged with negligent conduct, judges and jurors must decide ex-post (with hindsight) whether the auditor exercised ‘‘due professional care’’ (in foresight). Judges and jurors are expected to provide evaluation judgments based upon the defendant auditors’ behavior and/or decision process prior to the occurrence of the negative outcome (Devitt, Blackmar, & Wolff, 1987; Sand, Siffert, Loughlin, Reiss, & Batterman, 1997). However, judges and jurors have outcome knowledge that may impede their ability to mentally recreate the situation the auditor faced in foresight or reestablish the ex-ante predictability of subsequent events (Schkade & Kilbourne, 1991). Ex-post, it may be clear which audit procedures if any ‘‘could have been’’ performed, and how evidence ‘‘should have been’’ evaluated to uncover problematic client conditions (Buchman, 1985; Kinney, 1993). Thus, the auditors may be held to an unrealistic standard in that they are expected to anticipate all negative outcomes and take arguably overly conservative a priori steps to avoid them. Outcome knowledge can exert two different but related effects (Hawkins & Hastie, 1990). First, there may be a tendency for individuals with outcome
The Influence of Outcome Knowledge
159
knowledge to overestimate their ability to have predicted a given outcome ex-ante (hindsight effects). Second, evaluators’ judgments of others’ decisions may be biased by knowledge of outcomes, such that negative (positive) outcomes result in more unfavorable (favorable) evaluations (outcome effects). Outcome effects were given primary attention as they relate most directly to matters of legal liability. The purpose of this article is to review and discuss insights to be gleaned from several studies that investigated the influence of outcome information on judges and jurors’ evaluations of auditors. While some other research has examined outcome effects in other accounting-related scenarios,1 the focus in this article will be upon a stream of research, which is specific to the public accounting profession in a legal context. This research stream is divided into three phases. Phase 1 focuses on the robust manifestation of outcome effects in an audit legal liability context, Phase 2 examines the effectiveness of selected mitigation strategies in moderating outcome effects, and Phase 3 develops a preliminary theoretical framework that would explain how outcome effects operate, and which hopefully facilitates future thought and research.2 The final section of the paper suggests opportunities for future research. An overview of these research phases and corresponding articles can be seen in Exhibit 1 and Exhibit 2.
OUTCOME EFFECTS IN AN AUDIT LEGAL LIABILITY CONTEXT Pioneering research in this area sought to determine whether outcome effects found elsewhere also manifested in an audit legal liability context. Specifically, this research sought to determine whether professionally trained and experienced judges would be susceptible to outcome effects when evaluating allegations of auditor negligence.3 Anderson et al. (1993a, 1993b) conducted two research studies with general jurisdiction judges at the National Judicial College and with audit seniors from a single Big Four public accounting firm. Judges were given the task of evaluating auditors’ performance in a scenario in which auditors failed to discover material asset impairments owing to obsolescence. Results indicated that judges’ evaluations of auditor performance were dependent upon outcome information. That is, judges provided higher evaluations of auditor performance (the audit steps taken) in the presence of favorable outcome information and lower evaluations in the presence of unfavorable outcome information. This finding supports the proposition that outcome knowledge can potentially
160
D. JORDAN LOWE AND PHILIP M.J. RECKERS
Exhibit 1. Overview of Research. EXHIBIT 1 Overview of Research
Outcome Effects in an Audit Legal Liability Context Anderson, Lowe, & Reckers (1993) Anderson, Jennings, & Reckers (1993)
Mitigation Strategies
Hindsight Strategies
Foresight Strategies
Lowe & Reckers (1994) Anderson et al (1997) Anderson & Reckers (1998) Lee, Lowe, & Reckers (1998)
Lowe & Reckers (2000)
Developing a Preliminary Theoretical Framework of Outcome Effects Jennings, Lowe, & Reckers (1998) Lowe & Reckers (2002)
restrict judges’ ability to objectively evaluate auditor performance, retrospectively. This finding was important because professional liability lawsuits may be tried before either judges or juries alone, at the defendant’s discretion; and there was an implicit assumption that judges’ decisions and awards were less volatile and more merit based. Auditor participants in this research were found to exhibit similar outcome effects, suggesting that increased domain knowledge does not necessarily reduce the magnitude of these effects.
161
The Influence of Outcome Knowledge
Exhibit 2. Summary of Research. Study
Subject Group
Major Results
Outcome effects in an audit legal liability context Anderson, Lowe, and Judges, auditors Outcome knowledge influenced judges’ and Reckers (1993b) auditors evaluations of auditor decisions. Judges provided significantly lower evaluations of auditors’ performance than did auditors, reflective of an expectation gap. Results were also consistent with a cognitive interpretation, such that higher relevance was given to negative (positive) cues when the outcome was also negative (positive) and therefore congruent Anderson, Jennings, Judges, auditors Outcome knowledge influenced judges and and Reckers auditors’ evaluations of auditor (1993a) decisions. Judicial attitudes about the role of the auditors were statistically correlated with their evaluations. The evaluator’s recognition of the presence of outcome knowledge or in its propriety of use did not alter the influence of outcome information Mitigation strategies Hindsight mitigation strategies Lowe and Reckers Jurors (1994)
Anderson, Jennings, Lowe, and Reckers (1997)
Judges
Anderson and Reckers (1998)
Jurors
Outcome knowledge influenced jurors’ evaluations of the auditor’s judgments. A mitigation strategy in which subjects, which were given alternative outcomes to consider was found to be effective in mitigating the influence of outcome knowledge Outcome knowledge influenced judges’ evaluations of auditors’ performance. The alternative outcomes mitigation strategy that was successful in mitigating outcome effects in Lowe and Reckers (1994) was not found to be effective for judges. However, the mitigation strategy that redirected attention away from the plaintiff (to other stakeholders) was effective in mitigating outcome effects The alternative outcomes mitigation strategy was found to be effective in mitigating outcome effects for educated jurors but not for less educated jurors
162
D. JORDAN LOWE AND PHILIP M.J. RECKERS
Exhibit 2. (Continued ) Study Lee, Lowe, and Reckers (1998)
Subject Group MBA students
Foresight mitigation strategies Lowe and Reckers Auditors (2000)
Major Results Higher ex-ante expectations of audit practice lead to a greater ‘‘surprise’’ and subsequently to greater outcome effects. Subjects with higher moral development are more susceptible to outcome effects A foresight mitigation strategy that encourages auditors to consider a potential damaging legal scenario that could develop from ex-ante information was found to be effective in mitigating the influence of outcome knowledge
Developing a preliminary theoretical framework of outcome effects Jennings, Lowe, and Judges Judges’ evaluation of auditor decisions Reckers (1998) were directly related to the degree of outcome foreseeability (causality), such that hindsight effects were found under conditions of a foreseeable outcome, were reduced for a partially foreseeable outcome, and were found to be nonexistent for an unforeseeable outcome Lowe and Reckers Auditors, jurors, The expectation gap may be due to two (2002) judges components – (1) the difference between ex-ante judgments of auditors and nonauditors who do not have outcome knowledge and (2) the difference between nonauditor judgments that are made with and without negative outcome knowledge. Findings indicate that if the judgment directly involves performance evaluation, the components are likely to compound each other
The findings of these studies suggested a cognitive explanation of the phenomenon. This appeared to be the case because study participants assigned higher relevance scores to negative information items (i.e., items predictive of negative outcomes that were embedded in the case scenario) when the outcome was unfavorable and assigned higher relevance scores to positive information items when the outcome was favorable. From a cognitive perspective, the findings are consistent with the participants processing
The Influence of Outcome Knowledge
163
information in a temporally backward mode, from the given outcome, to the antecedent conditions. That is, individuals focused their attention on the given outcome and tried to explain the occurrence by creating causal links to predecessor events and actions. Once this causal framework was developed; individuals experienced difficulty seeing how alternative outcomes could have occurred or responsibly could have been expected by the auditor (Schkade & Kilbourne, 1991). Under this explanation, informational data causally linked to the outcome would appear to be more relevant and important in determining the event’s outcome (e.g., Baron & Hershey, 1988; Fischhoff, 1975). This early research also indicated that individuals were seemingly unaware of this cognitive process and the effect that outcome knowledge had on their perceptions and judgments. Individuals’ self-assessments of their use of outcome knowledge were largely inaccurate. In fact, those who expressed the belief that the use of outcome information was inappropriate were found to use this information just as much as those who thought the use was appropriate.
MITIGATION STRATEGIES The results from Anderson et al. (1993a, 1993b) suggested that the public accounting profession might be significantly and negatively affected by outcome effects in the U.S. civil liability system. Therefore, the next logical step was to determine whether non-normative outcome effects could be mitigated or at least moderated. That is, researchers investigated several mitigation strategies with the intent of overcoming outcome effects, and thereby achieving a more objective evaluation of auditor performance. This was considered to be a complex and difficult task as outcome effects have been found to be robust and resistant to prior mitigation efforts elsewhere in other contexts (see Fischhoff, 1982 and Hawkins & Hastie, 1990 for a review). Hindsight Mitigation Strategies To devise an effective mitigation strategy requires a fundamental understanding of outcome effects. Arguably, this requisite understanding was lacking in many early efforts. Preliminary research posited that the phenomenon was caused by individuals’ tendency to focus on a given outcome and interpret antecedent behavior or events in a backward processing mode.4 If so, a mitigation strategy focusing on breaking down the causal
164
D. JORDAN LOWE AND PHILIP M.J. RECKERS
links between outcome knowledge and antecedent behavior and events might be effective in mitigating outcome effects. Further, the perceived linkage between the given outcome and prior behaviors and events could potentially be weakened if individuals were encouraged to consider other viable alternative outcomes. That is, the association of ‘‘alternative outcomes’’ with the same set of antecedent events and judgments might reduce the perceived inevitability of the actual outcome (Arkes, 1991; Janoff-Bulman, Timko, & Carli, 1985). The hoped-for effect of having individuals consider alternative outcomes was to shift participants from a hindsight to a foresight perspective, such that decision makers and evaluators’ perspectives are equated. Lowe and Reckers (1994) conducted an initial mitigation study using a scenario embedded with a severe negative outcome – audit client bankruptcy. Using such a severe negative outcome created a strong test for mitigation efforts, but was representative of the context of many lawsuits against audit firms. A three-step mitigation strategy was devised as a means of overcoming outcome effects. The strategy dictated that participants were (1) given two alternative outcomes (bankruptcy/solvency) to consider, (2) asked to assess the probability that the alternative outcomes could have occurred (see Arkes, Faust, Guilmette, & Hart, 1988), and (3) then asked to generate their own alternative outcome, etc., all before rendering judgments. The second and third steps of the mitigation strategy were specifically included to encourage participants to elaborate on (instantiate) the alternative outcomes (Arkes, 1989, 1991). Prospective jurors were selected to serve as participants in this study.5 The experiment was conducted at the county courthouse of a large metropolitan city. Jurors were compensated for completing this task. Different subsets of jurors were provided with (a) no outcome (control group), (b) a negative outcome, or (c) a negative outcome with the alternative outcomes mitigation strategy. In spite of receiving instructions to base their responses on information available before learning of an outcome, jurors (under non-mitigation experimental treatments) tended to bias auditor evaluation judgments in the direction of the negative (bankruptcy) outcome. Outcome knowledge of the audit client’s bankruptcy consistently resulted in lower evaluations of the auditor’s performance. Consistent with previous research, these jurors also tended to assign higher relevance scores to information items, suggestive of negative outcomes when the eventual outcome was indeed unfavorable (compared to the control group). Thus, congruence could be found between manipulated outcome and assigned relevance scores for information items embedded in the case scenario.
The Influence of Outcome Knowledge
165
Of most importance, the alternative outcomes mitigation strategy was found to be effective in significantly reducing outcome effects. Among jurors given a negative outcome, those jurors with a mitigation strategy provided significantly higher evaluations of the auditor’s performance. In addition, jurors’ ratings of the relevance of the embedded negative information items were significantly reduced by the mitigation strategy. These results provided support for the potential effectiveness of an experimenter-devised strategy in mitigating outcome effects within the context of the legal liability system. These results were especially encouraging, given that the mitigation strategy had to overcome a particularly strong outcome manipulation consisting of the audit client’s bankruptcy and significant stockholder losses. Given the success of the alternative outcomes mitigation strategy with jurors, a logical next step was to determine whether mitigation strategies would likewise be effective with judges. Anderson et al. (1997) designed such a study. Their study aimed to test the effectiveness of the alternative outcomes mitigation strategy (Lowe & Reckers, 1994) as well as a second mitigation strategy, which was specifically designed for judges. In the second strategy, rather than appealing to judges to consider alternative outcomes, Anderson et al. (1997) reasoned that it may be equally effective to appeal them to consider alternative stakeholders. In the courtroom, the focus of attention may be exclusively on individuals alleging damages. The alternative stakeholders’ mitigation strategy redirects judges’ attention from one party claiming damages to other parties, who might have been damaged if the auditor behaved otherwise. Redefining (clarifying) the auditors’ responsibility to society as one that includes a responsibility to a variety of stakeholders, and by doing so, redirecting the attention of the court to other stakeholders (e.g., preexisting stockholders, creditors, and employees), may assist the court in better understanding the auditors’ decision-making process.6 That is, the ex-post evaluator might come to better understand the exante situation faced by the auditor. To test these two mitigation strategies, Anderson et al. (1997) conducted an experiment with general jurisdiction judges at the National Judicial College. A case instrument that was previously used in Anderson et al. (1993a, 1993b) was adapted to include a no outcome (control) group as well as the two mitigation strategies. The findings indicated that judges provided significantly lower evaluations in the presence of unfavorable outcome information, compared to those judges receiving no outcome information. With respect to the two mitigation efforts examined, the mitigation strategy in which attention was redirected away from the plaintiff (to other stakeholders) was effective in mitigating outcome effects. Conversely, the mitigation strategy that
166
D. JORDAN LOWE AND PHILIP M.J. RECKERS
introduced alternative outcomes (and found to be successful with jurors in Lowe & Reckers, 1994) was not effective in mitigating outcome effects. The relative effectiveness of the two mitigation strategies arguably was dependent on the participant group (e.g., judges versus jurors). Prior research (see Howe, 1991; Kalven & Zeisel, 1966; Pennington & Hastie, 1990) had indeed shown that judges and jurors integrate and assess information in different ways. The manner in which individuals process alternative outcome information and assess their likelihood might influence the extent that the mitigation process occurs (Hoch, 1985). Judges are accustomed to assessing outcomes ex-post, and prescriptively are to attend only the facts and interpretations of the law. Thus, they may be less apt to consider or give substantial weight to ex-ante, hypothetical alternative outcomes. This reasoning was advanced to explain why the alternative outcomes mitigation strategy was effective in mitigating outcome effects with judicially inexperienced jurors (see Lowe & Reckers, 1994), but was ineffective in mitigation efforts with judges. A secondary objective of Anderson et al. (1997) was to determine whether differential outcome effects would be elicited by favorable and unfavorable outcomes. Prior research had not included positive and negative outcomes with a ‘‘no’’ outcome (control) group to measure the extent of these effects. The results indicated that outcome effects were not symmetric, as the positive outcome did not elicit significant outcome effects. An explanation for this finding may be due to the positive outcome being perceived as being less salient than a negative outcome. This explanation is consistent with prior studies showing that negative events are given greater weight than positive events (e.g., Anderson & Maletta, 1994; Ashton & Ashton, 1988; Mizerski, 1982). In addition, a negative outcome would likely be considered as an event occurrence (inventory obsolescence) while the continued usefulness of a product would be considered as a nonevent. Because individuals appear to have cognitive difficulty in processing nonevents (Christensen-Szalanski & Fobian-Willham, 1991; Fischhoff, 1977; Rachlinski, 1998), a positive outcome (or nonevent) might have only a moderate influence upon an individual’s judgment process.7 This preliminary research provided some important insights into the mitigation of outcome effects. Similar to these studies, other researchers have also had mixed success with mitigation strategies across other contexts (see Hawkins & Hastie, 1990). Given the disparity of results across several contexts, some have suggested that individual differences may influence the effectiveness of mitigation strategies. Anderson and Reckers (1998) chose to examine education level, as this is an important factor in the juror selection
The Influence of Outcome Knowledge
167
process (voir dire). The authors proposed that individuals’ level of education is associated with their reasoning ability. Educated individuals may be more able to consider and generate alternative outcomes, and thus may be more receptive to the alternative outcomes mitigation strategy. The objective of this study was to examine the influence (and interaction) of jurors’ education level on the mitigation of outcome effects. Anderson and Reckers (1998) performed this study with prospective jurors at the county courthouse of a large metropolitan city. Jurors were partitioned into two groups – a high and a low education group. Results revealed a significant interaction between outcome knowledge (no outcome, negative outcome, and negative outcome with mitigation strategy) and education level. More specifically, the negative outcome with mitigation strategy was found to be effective in mitigating outcome effects for more educated jurors but not for less educated jurors. Thus, one needs to be aware of jurors’ overall education level to determine whether a mitigation strategy may be effective in a court of law. Lee et al. (1998) investigated two other individual differences that they hypothesized would influence participants’ ability to respond to mitigation efforts. The first individual difference examined was the expectations of the audit function. The authors reasoned that the degree of surprise or unexpected nature of an event enhances the influence of outcome knowledge (Hoch & Loewenstein, 1989; Reimers & Butler, 1992). Surprise causes individuals to engage in enhanced sense-making efforts to resolve any apparent differences between the reported outcome and antecedent evidence so that a coherent story is maintained (Janoff-Bulman et al., 1985; Mazursky & Ofir, 1990). In the context of audit legal liability, differential foresight expectations of auditor conduct/judgment should lead to correspondingly different degrees of surprise when presented with an unfavorable outcome. Jurors that have high expectations of the audit function should be very surprised to find a company having severe financial difficulties shortly after receiving an unqualified opinion, thus exhibiting outcome effects in their evaluation of the auditor’s performance. Correspondingly, these same jurors should have their surprise significantly reduced by considering alternative (benign) outcomes through the mitigation strategy. A second individual difference that could affect the magnitude of outcome effects as well as subsequent mitigation efforts is an individual’s level of moral development. Moral development could be expected to affect the perceived significance of others’ losses to self. Kadous (2001) suggests that jurors will find it disturbing to believe that negative outcomes can arise and that no one is responsible for preventing them. This reasoning seems
168
D. JORDAN LOWE AND PHILIP M.J. RECKERS
particularly likely to apply to individuals at higher levels of moral development. Such individuals follow ‘‘underlying principles such as equity and fair play’’ while individuals at lower levels of moral development focus only on themselves (Kaplan, Newberry, & Reckers, 1997). When an individual with a high level of moral development perceives that others (i.e., investors) have experienced losses, he or she is more likely to be sensitive to their predicament and seek reasons and/or remedies for this occurrence (Ponemon, 1992, 1993). A plausible explanation that could be conceived in hindsight is the poor performance by the auditors. Encouraging individuals with a high degree of moral development to consider alternative outcomes should lead them away from blaming auditors and toward other circumstances that existed at the time of the audit. Lee et al. (1998) conducted an experiment with Masters of Business Administration students from two universities serving in their role as jurors. Similar to other studies, these participants were exposed to an outcome manipulation and (where appropriate) the alternative outcomes mitigation strategy. In addition, students completed an attitudes questionnaire related to the audit function as well as the Defining Issues Test (DIT), as developed by Rest (1979). Findings indicated that individuals with high expectations of the audit function and/or a high level of moral development were (1) more likely to succumb to outcome effects but were also (2) more inclined to yield to selected efforts to mitigate these effects. These results (as well as those from Anderson & Reckers, 1998) have implications for the process for voir dire in courtroom scenarios in which outcome effects may be evident in juror evaluations of auditor decisions. As only certain individuals were affected by the consideration of alternative outcomes as a mitigation strategy, attorneys may want to consider exercising peremptory challenges to reject individuals with certain characteristics. Foresight Mitigation Strategies Prior research has focused almost exclusively on hindsight mitigation strategies whereby the hindsight perspective is altered so as to coincide with the foresight perspective (Reimers & Butler, 1992). Given the mixed results in research utilizing hindsight strategies (see Fischhoff, 1982 and Hawkins & Hastie, 1990), some have suggested that foresight mitigation strategies should also be considered as a means to mitigate the effects of outcome knowledge (Arkes et al., 1988; Creyer & Ross, 1993; Reimers & Butler, 1992). Foresight mitigation strategies attempt to modify the foresight perspective to be consistent with a hindsight perspective.
The Influence of Outcome Knowledge
169
Lowe and Reckers (2000) reasoned that foresight mitigation strategies could be particularly useful to auditors who are made aware of (1) the extreme negative outcomes that may befall the public accounting firm as a result of their incorrect judgments, (2) the risk that their performance will be judged in hindsight by external parties, and (3) the knowledge that influence the hindsight evaluation of external, ex-post evaluators may be difficult (Boatsman, Moeckel, & Pei, 1997). Foresight mitigation strategies could be implemented as decision aids in the auditors’ planning and judgment process (Brown & Solomon, 1993; Reimers & Butler, 1992). The decision aid could prompt auditors to explicitly consider the most damaging legal scenario that could develop from the ex-ante facts and information. If the potential legal consequences of an auditor’s judgment are considered in foresight, the auditor would be more likely to perform the audit with a hindsight perspective. That is, the auditor could consider a potential outcome ex-ante so that the foresight judgment approximates the hindsight judgment of a potential evaluator.8 Lowe and Reckers (2000) conducted an experiment with audit seniors from a single Big Four firm. Auditors completed the experimental materials during firm training sessions with one of the researchers in attendance. A case narrative was provided, which concluded with auditors having to assess possible inventory obsolescence. Participant subsets were given (a) no outcome, (b) a negative (hindsight) outcome, and (c) no outcome with a foresight strategy embedded into a decision aid. The foresight decision aid prompted auditors to consider the possible legal consequences of a nonreporting decision choice (i.e., not to report issues related to inventory). That is, the authors wanted to encourage participants to consider the potential legal exposure of the firm in a court of law. Results indicated that the foresight decision aid was effective in mitigating the influence of outcome knowledge. This is the first study we are aware of that has utilized a foresight strategy to completely mitigate these effects. Auditors who were encouraged to describe a negative outcome in foresight provided responses that were almost identical to auditors who were given actual outcome knowledge. These results suggest that auditors’ ex-ante decision processes can be altered, so their judgments approximate those of ex-post evaluators. This has implications for audit legal liability as it suggests that rather than focusing exclusively on influencing judges and jurors’ expectations in hindsight, it may be expedient to also modify auditor’s expectations in foresight. However, this study did not address the additional audit costs of doing so, relative to the additional legal costs of not doing so.
170
D. JORDAN LOWE AND PHILIP M.J. RECKERS
DEVELOPING A PRELIMINARY THEORETICAL FRAMEWORK Some researchers have begun to synthesize previous research and develop a preliminary theoretical framework of outcome effects. The objective of moving forward and building a framework is to (1) advance a better understanding of the phenomenon, (2) reconcile existing research, (3) provide guidance for future research, and eventually (4) offer insights to resolving audit legal liability issues. While a cohesive framework has not as yet been established, two recent studies (Jennings et al. 1998; Lowe & Reckers, 2002) provide initial impetus in this process. Jennings et al. (1998) reasoned that the extent to which individuals succumb to the influence of outcome knowledge may relate to the perceived feasibility (or ease) of reconstructing a causal relationship between the outcome and antecedent conditions. That is, outcome effects arguably rely upon the ability of the evaluator, ex-post, to reconstruct a cogent causal series of related events leading up to the outcome, and the tendency to overestimate the ex-ante likelihood of the reconstructed scenario. A foreseeable outcome should be conducive for assimilating the causal links between the outcome and antecedent factors. Conversely, less foreseeable outcomes that create some surprise in individuals arguably should have the effect of inhibiting backward reconstruction and causal relations (Hawkins & Hastie, 1990; Wasserman, Lempert, & Hastie, 1991). An unforeseeable outcome should have the effect of counteracting the hindsight tendency to integrate the outcome into individuals’ knowledge structures. When outcome feedback is truly unexpected, individuals may be more apt to acknowledge their surprise that, in turn, should have the effect of moderating outcome effects. In sum, outcome effects would be expected to be greatest when an outcome can be fit into a causal reconstruction leading back to antecedent conditions. However, an unforeseeable outcome is expected to result in little if any outcome effects. Jennings et al. (1998) assessed whether the perceived causality between an outcome and antecedent conditions affects the magnitude of judges’ outcome effects. That is, do negative client outcomes always lead to elevated evaluations of auditor negligence or legal responsibility; or, must surrounding conditions easily lend themselves to causal reconstruction for this to happen? This is an important point as the accounting profession is concerned that judicial decisions are often insensitive to the causal nature and merits of the case (Cloyd, Frederickson, & Hill, 1996; Palmrose, 1997). The
The Influence of Outcome Knowledge
171
public accounting profession asserts that the present system ‘‘makes it both easy and financially rewarding to file claims regardless of the merits of the case’’ (Arthur Andersen et al., 1992, p. 1). Although causality is a necessary condition for recovery, auditor groups suggest that lawsuits are filed against auditors when proximate cause is lacking (American Institute of Certified Public Accountant (AICPA), 1993; Cloyd et al., 1996). Jennings et al. (1998) conducted an experiment to examine the relationship between causality and outcome effects. Their study was performed with general jurisdiction judges from the National Judicial College. Judges were given an inventory obsolescence case that manipulated varying degrees of outcome foreseeability – (1) no outcome, (2) a negative outcome from foreseeable causes, (3) a negative outcome from partially foreseeable causes, and (4) a negative outcome from unforeseeable causes. As predicted, judges provided progressively lower auditor evaluations in relation to increasingly foreseeable outcomes. Outcome effects were evidenced for the foreseeable outcome and to a lesser extent for the partial foreseeable outcome group. However, when judges were provided with a negative outcome from unforeseeable causes, outcome effects were not observed. This knowledge provides some assurance to the public accounting profession that outcome foreseeability (causality) is a relevant factor in judges’ evaluations. If judges had been shown to be highly insensitive to the causal nature of events, then that would suggest adherence to a strict liability philosophy. Further, Jennings et al. (1998) show that recognition of the relative unforeseeability of an outcome tends to moderate outcome effects. Defense counsel could be encouraged to focus on the external auditors’ foresight perspective by emphasizing the relatively unforeseeable nature of the eventual outcome. Further research related to the influences of causality and foreseeability would be beneficial to the academic and professional communities. Lowe and Reckers (2002) provide another step toward the development of a theoretical framework by examining the relationship of outcome information and the expectation gap. The expectation gap has been purported to be due to one of the two components. The first component represents the difference between ex-ante judgments of auditors versus financial statement users (i.e., investors, creditors, employees, legislators, etc.) who do not have outcome knowledge. The second component connotes the difference between users’ judgments that are made with and without negative outcome knowledge. Prior research has attempted to show that the expectation gap may be due to the first component (e.g., Anderson et al., 1993a, 1993b) or to the second component (e.g., Lowe & Reckers, 1994; Reimers & Butler,
172
D. JORDAN LOWE AND PHILIP M.J. RECKERS
1992). With the exception of Kinney and Nelson (1996), research has not attempted to examine both components. Regarding judgments made prior to receipt of outcome information (component one), we expect that auditors would provide higher ex-ante evaluations of fellow auditors than would non-auditors. That is, auditors may be expected to be less critical of their peers than non-auditors due to (1) group member affinity, (2) divergent attitudes and expectations between the two groups (Anderson et al., 1993a, 1993b; Arrington, Hillison, & Williams, 1983, Arrington, Bailey, & Hopwood, 1985; Lowe, 1994), and (3) value implications to their own work. With respect to judgments rendered after receipt of outcome information (component two), we would expect negative outcome knowledge to influence non-auditors, such that their ex-post evaluation judgments would be less favorable than their ex-ante judgments (Baron & Hershey, 1988; Hawkins & Hastie, 1990; Kadous, 2001). We propose that these two components would be compounded in an audit legal liability scenario in which performance evaluations take place. Lowe and Reckers (2002) tested these propositions by utilizing existing data from two prior studies (Lowe & Reckers, 1994; Anderson et al., 1997). However, to perform appropriate tests, Lowe and Reckers had to collect additional data for each study to form cells in which auditors made ex-ante evaluation judgments. Results indicated that in pre-outcome judgments, auditors provided significantly higher ex-ante evaluation judgments of fellow auditors than did their juror/judge counterparts. Regarding post-outcome judgments, jurors and judges provided significantly lower evaluations when told of a negative outcome than when they had no outcome knowledge. Finally, when the pre- and post-outcome judgments were combined, the effects were compounded such that non-auditors ex-post judgments were significantly lower than the ex-ante judgments made by auditors. This preliminary analysis suggests that if the judgment involves performance evaluation (as would be done in a court of law), the components are likely to compound each other resulting in significant outcome effects. Further research in this pivotal area would benefit both the academic and professional communities.
FUTURE RESEARCH While we believe that substantive progress has been made over the last decade, much research remains to be done both to better understand outcome effects and to test operational responses and proposed remedies. To enhance our understanding of outcome effects and their influence on the
The Influence of Outcome Knowledge
173
accounting profession, the following research questions are proposed: 1. What are the underlying causes of the differences observed across user groups (e.g., explain the differences observed between jurors and judges)? Why are some mitigation efforts more successful with some groups versus others? 2. Individual differences (e.g., education level, gender, expertise, etc.) are often relevant for the juror selection process (voir dire). What individual differences would be interesting to examine that would be relevant to outcome effects and their mitigation? 3. While prior research has found some success with the alternative outcomes and alternative stakeholders mitigation strategies, what other methods could be explored and tested? 4. Will new corporate governance regimes dictated by Sarbanes–Oxley (SOX) reforms create an environment in which perceived auditor culpability is increased in hindsight because the comparative contribution of the corporate unit is perceived as lessened? 5. What foresight strategies could be designed and embedded into decision aids that would encourage auditors to make decisions in light of potential legal outcomes? Could these decision aids be designed such that they are not overly obtrusive? 6. What near and long-term costs and benefits are there to the accounting profession of proceeding with mitigation strategies? In conclusion, the research discussed in this paper was conducted during times when the profession enjoyed a higher public reputation (i.e., preEnron). Future research will occur in a different environment. For example, Reckers, Jennings, Lowe, and Pany (2005) report significant attitude changes among judges pre- and post-Enron/SOX. Prospective research will need to both replicate and extend past research. Research is also encouraged that would examine a wider set of case scenarios that are representative of the spectrum of cases litigated today. Evidence exists that in the post-SOX era, different types of litigation is reaching the courts.
NOTES 1. In a managerial setting, researchers have examined how outcome effects influence the performance evaluation and how outcome controllability moderates these effects (Brown & Solomon, 1987, 1993; Fisher & Selling, 1993; Ghosh & Lusch, 2000; Lipe, 1993; Tan & Lipe, 1997). In a tax setting, researchers have investigated whether tax professionals are able to use outcome information in a manner consistent with the law (Helleloid, 1988) and whether practice risk influences these effects
174
D. JORDAN LOWE AND PHILIP M.J. RECKERS
(Kadous & Magro, 2001). Finally, other research has examined how outcome effects are influenced by negative effect (Kadous, 2001). 2. These phases of research represent a comprehensive research program by the authors. The three phases of research were progressive, such that each research project built upon each other. 3. Judges represent an important group given that judges (1) are involved in a significant number of court cases against auditors, (2) control important elements of the trial (i.e., the admissibility of evidence, appropriateness of witnesses, and directions to the jury), (3) preside over appeal hearings, and (4) have the power to render summary judgment when there is no valid issue of material fact (Anderson, Maletta, & Wright, 1998; Black, 1990). 4. This ‘‘association error’’ is an unintended cost of an adaptive association-based semantic memory system (Arkes, 1991). The automatic nature of these semantic associations becomes a cost when judgmentally irrelevant semantic associations influence the decision process (Dellarosa & Bourne, 1984). 5. Palmrose (1991) reports that from a comprehensive sample of auditor civil trials (tried to verdict) between 1960–1990, 81 percent were jury trials. 6. It is also important to note that the auditor is ethically and professionally bound by the professional code of conduct to consider the potentially deleterious effects of an inappropriately issued qualified opinion on other stakeholders (e.g., Balachandran, 1993; Shaver, 1985). 7. Schkade and Kilbourne (1991) postulate that some individuals, not given outcome information, may assume a positive outcome since they do not receive any contrary information. 8. Unlike other researches discussed in this article, hindsight and not outcome effects are examined. The reason for this is that foresight mitigation strategies necessarily involve likelihood judgments and not evaluation judgments.
ACKNOWLEDGMENTS The authors wish to acknowledge the helpful comments from participants at the 2004 Accounting, Behavior, and Organizations Research Conference and the participants at the 2004 Annual Meeting of the American Accounting Association. This research program has been assisted by the Ernst & Young Foundation, by the KPMG Peat Marwick Research Opportunities in Auditing Program, as well as by other University-sponsored programs.
REFERENCES American Institute of Certified Public Accountants (AICPA). (1993). Meeting the financial reporting needs of the future: A public commitment from the public accounting profession. New York, NY: American Institute of Certified Public Accountants.
The Influence of Outcome Knowledge
175
Anderson, B. H., & Maletta, M. M. (1994). Auditor attendance to negative and positive information: The effect of experience-related differences. Behavioral Research in Accounting, 6, 1–20. Anderson, B. H., Maletta, M. M., & Wright, A. (1998). Perceptions of auditor responsibility: Views of the judiciary and the profession. International Journal of Auditing, 2(3), 215–232. Anderson, J. C., Jennings, M. M., Lowe, D. J., & Reckers, P. M. J. (1997). The mitigation of hindsight bias in judges’ evaluation of auditor decisions. Auditing: A Journal of Practice and Theory, 18(2), 20–39. Anderson, J. C., Jennings, M. M., & Reckers, P. M. J. (1993a). The presence of hindsight bias in peer and judicial evaluation in public accounting litigation. Tort and Insurance Law Journal, 28(3), 462–479. Anderson, J. C., Lowe, D. J., & Reckers, P. M. J. (1993b). Evaluation of auditor decisions: Hindsight bias effects and the expectation gap. Journal of Economic Psychology, 14(4), 711–737. Anderson, J. C., & Reckers, P. M. J. (1998). Mitigating hindsight bias in jurors’ evaluation of auditor decisions: Considering alternative outcomes and the education of jurors. Advances in Accounting, 16, 221–237. Arkes, H. R. (1989). Principles in judgment/decision making research pertinent to legal proceedings. Behavioral Sciences and the Law, 7(4), 429–456. Arkes, H. R. (1991). Costs and benefits of judgment errors: Implications for debiasing. Psychological Bulletin, 110(3), 486–498. Arkes, H. R., Faust, D., Guilmette, T. J., & Hart, K. (1988). Eliminating the hindsight bias. Journal of Applied Psychology, 73(2), 305–307. Arrington, C. E., Bailey, C. D., & Hopwood, W. S. (1985). An attribution analysis of responsibility assessment for audit performance. Journal of Accounting Research, 23(1), 1–20. Arrington, C. E., Hillison, W., & Williams, P. (1983). The psychology of expectation gap: Why is there so much dispute about auditor responsibility? Accounting and Business Research, 13(52), 243–250. Arthur Andersen & Co., Coopers & Lybrand, Deloitte & Touche, Ernst & Young, KPMG Peat Marwick, & Price Waterhouse. (1992). A liability crisis in the United States: Impact on the accounting profession. Journal of Accountancy, 174(5), 19–23. Ashton, A. H., & Ashton, R. H. (1988). Sequential belief revision in auditing. The Accounting Review, 63, 623–641. Balachandran, B. V. (1993). Introduction: Auditors’ legal liability in the United States. Journal of Economics and Management Strategy, 2(3), 333–338. Baron, J., & Hershey, J. C. (1988). Outcome bias in decision evaluation. Journal of Personality and Social Psychology, 54(4), 569–579. Black, H. C. (1990). Black’s law dictionary. St Paul, MN: West Publishing Co. Boatsman, J. R., Moeckel, C. L., & Pei, K. W. (1997). The effects of decision consequences on auditors’ reliance on decision aids in audit planning. Organizational Behavior and Human Decision Processes, 71, 211–248. Brown, C. E., & Solomon, I. (1987). Effects of outcome information on evaluations of managerial decisions. The Accounting Review, 62(3), 564–577. Brown, C. E., & Solomon, I. (1993). An experimental investigation of explanations for outcome effects on appraisals of capital-budgeting decisions. Contemporary Accounting Research, 10(1), 84–117.
176
D. JORDAN LOWE AND PHILIP M.J. RECKERS
Buchman, T. A. (1985). An effect of hindsight on predicting bankruptcy with accounting information. Accounting, Organizations & Society, 10(3), 267–285. Christensen-Szalanski, J. J., & Fobian-Willham, C. S. (1991). The hindsight bias: A metaanalysis. Organizational Behavior and Human Decision Processes, 48(1), 147–168. Cloyd, C. B., Frederickson, J. R., & Hill, J. W. (1996). Motivating factors in lawsuits against independent auditors: Experimental evidence on the importance of causality. Journal of Accounting and Public Policy, 15(3), 185–218. Creyer, C., & Ross, W. T., Jr. (1993). Hindsight bias and inferences in choice: The mediating effect of cognitive effort. Organizational Behavior and Human Decision Processes, 55, 61–77. Dellarosa, D., & Bourne, L. E. (1984). Decisions and memory: Differential retrievability of consistent and contradictory evidence. Journal of Verbal Learning and Verbal Behavior, 23, 669–682. Devitt, E. J., Blackmar, C. B., & Wolff, M. A. (1987). Federal jury practice and instructions: Civil (4th Ed.). St Paul, MN: West Publishing. Fischhoff, B. (1975). Hindsight 6¼ foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288–299. Fischhoff, B. (1977). Perceived informativeness of facts. Journal of Experimental Psychology: Human Perception and Performance, 3(2), 349–358. Fischhoff, B. (1982). Debiasing. In: D. Kahneman, P. Slovic & A. Tversky (Eds), Judgment under uncertainty: Heuristics and biases (pp. 422–444). New York, NY: Cambridge University Press. Fisher, J., & Selling, T. I. (1993). The outcome effect in performance evaluation: Decision process observability and consensus. Behavioral Research in Accounting, 5, 58–77. Ghosh, D., & Lusch, R. F. (2000). Outcome effect, controllability, and performance evaluation of managers: Some field evidence from multi-outlet businesses. Accounting Organizations and Society, 25, 411–425. Hawkins, S. A., & Hastie, R. (1990). Hindsight: Biased judgments of past events after the outcomes are known. Psychological Bulletin, 107(3), 311–327. Helleloid, R. T. (1988). Hindsight judgments about taxpayers’ expectations. The Journal of the American Taxation Association, 10(1), 31–46. Hoch, S. J. (1985). Counterfactual reasoning and accuracy in predicting personal events. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11(4), 719–731. Hoch, S. J., & Loewenstein, G. F. (1989). Outcome feedback: Hindsight and information. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15(4), 605–619. Howe, E. S. (1991). Integration of mitigation, intention, and outcome damage information, by students and circuit court judges. Journal of Applied Social Psychology, 21, 875–895. Janoff-Bulman, R., Timko, C., & Carli, L. L. (1985). Cognitive biases in blaming the victim. Journal of Experimental Social Psychology, 21(2), 161–177. Jennings, M. M., Kneer, D. C., & Reckers, P. M. J. (1993). The significance of audit decision aids and pre-case jurists’ attitudes on perceptions of audit firm culpability and liability. Contemporary Accounting Research, 9(2), 489–507. Jennings, M. M., Lowe, D. J., & Reckers, P. M. J. (1998). Causality as an influence on hindsight bias: An empirical examination of judges’ evaluations of professional audit judgment. Journal of Accounting and Public Policy, 21, 143–167. Kadous, K. (2001). Improving jurors’ evaluations of auditors in negligence cases. Contemporary Accounting Research, 18(3), 425–444.
The Influence of Outcome Knowledge
177
Kadous, K., & Magro, A. M. (2001). The effects of exposure to practice risk on tax professionals’ judgments and recommendations. Contemporary Accounting Research, 18(3), 451–475. Kalven, H., Jr., & Zeisel, H. (1966). The American Jury. Boston, MA: Little, Brown. Kaplan, S. E., Newberry, K., & Reckers, P. M. J. (1997). An examination of the effect of moral reasoning on taxpayers’ judgments. The Journal of the American Taxation Association, 19(2), 38–54. Kinney, W. R., Jr. (1993). Auditors’ liability: Opportunities for research. Journal of Economics and Management Strategy, 2(3), 349–360. Kinney, W. R., Jr. (1994). Audit litigation research: Professional help is needed. Accounting Horizons, 8(2), 80–86. Kinney, W. R., Jr., & Nelson, M. W. (1996). Outcome information and the expectation gap: The case of loss contingencies. Journal of Accounting Research, 34(2), 281–299. Lee, T., Lowe, D. J., & Reckers, P. M. J. (1998). The influence of moral development and predecisional attitudes in the mitigation of hindsight bias. Advances in Accounting, 16, 239–252. Lipe, M. G. (1993). Analyzing the variance investigation decision: The effects of outcomes, mental accounting, and framing. The Accounting Review, 68, 748–764. Lochner, P. R. (1993). Accountants’ legal liability: A crisis that must be addressed. Accounting Horizons, 7(2), 92–96. Lowe, D. J. (1994). The expectation gap in the legal system: Perception differences between auditors and judges. Journal of Applied Business Research, 10(3), 39–44. Lowe, D. J., & Reckers, P. M. J. (1994). The effects of hindsight bias on jurors’ evaluations of auditor decisions. Decision Sciences, 25(3), 401–426. Lowe, D. J., & Reckers, P. M. J. (2000). The use of foresight decision aids in auditor judgments. Behavioral Research in Accounting, 12, 97–118. Lowe, D. J., & Reckers, P. M. J. (2002). A preliminary framework in examining the influence of outcome information on evaluations of auditor decisions. Advances in Accounting, 19, 177–187. Mazursky, D., & Ofir, C. (1990). I could never have expected it to happen: The reversal of the hindsight bias. Organizational Behavior and Human Decision Processes, 46(1), 20–33. Mizerski, R. W. (1982). An attribution explanation of the disproportionate influence of unfavorable information. Journal of Consumer Research, 9, 301–310. Palmrose, Z. (1991). Trials of legal disputes involving independent auditors: Some empirical evidence. Journal of Accounting Research, 29(Supplement), 149–184. Palmrose, Z. (1997). Audit litigation research: Do the merits matter? An assessment and directions for future research. Journal of Accounting and Public Policy, 16, 355–378. Pennington, N., & Hastie, R. (1990). Practical implications of psychological research on juror and jury decision making. Personality and Social Psychology Bulletin, 16(1), 90–105. Ponemon, L. A. (1992). Ethical reasoning and selection-socialization in accounting. Accounting, Organizations and Society, 17(3/4), 239–258. Ponemon, L. A. (1993). The influence of ethical reasoning on auditors’ perceptions of management’s integrity and competence. Advances in Accounting, 11, 1–29. Rachlinski, J. J. (1998). A positive psychological theory of judging in hindsight. The University of Chicago Law Review, 65(2), 571–625. Reckers, P. M. J, Jennings, M., Lowe, D. J., & Pany, K. (2005). Judges’ attitudes toward the public accounting profession. Working Paper, Arizona State University, Arizona.
178
D. JORDAN LOWE AND PHILIP M.J. RECKERS
Reimers, J. L., & Butler, S. A. (1992). The effect of outcome knowledge on auditors’ judgmental evaluations. Accounting, Organizations, and Society, 17(2), 185–194. Rest, J. R. (1979). Development in judging moral issues. Minneapolis, MN: University of Minnesota Press. Sand, L. B., Siffert, J. S., Loughlin, W. P., Reiss, S. A., & Batterman, N. (1997). Modern federal jury instructions, Vol. 4. San Francisco, CA: Matthew Bender & Co. Schkade, D. A., & Kilbourne, L. M. (1991). Expectation-outcome consistency and hindsight bias. Organizational Behavior and Human Decision Processes, 49(1), 105–123. Shaver, K. G. (1985). The analysis of blame: Cause, responsibility, and blameworthiness. New York, NY: Springer. Tan, H., & Lipe, M. G. (1997). Outcome effects: The impact of decision process and outcome controllability. Journal of Behavioral Decision Making, 10, 315–325. Wasserman, D., Lempert, R. O., & Hastie, R. (1991). Hindsight and causality. Personality and Social Psychology Bulletin, 17(1), 30–35.
WHY YOU SHOULD CONSIDER SEM: A GUIDE TO GETTING STARTED Cindy Blanthorne, L. Allison Jones-Farmer and Elizabeth Dreike Almer ABSTRACT Structural Equation Modeling (SEM) offers researchers additional flexibility and enhanced research conclusions, yet it is still underutilized in accounting. This may be in part because many researchers are not sufficiently familiar with SEM. SEM can be difficult to apply, especially if the research study was not appropriately planned to accommodate the necessary assumptions and data requirements. This article helps researchers overcome some barriers to using SEM by providing a simple guide to effectively planning a study suitable for an SEM analysis while also suggesting references and additional reading on the topic. To further encourage the use of SEM, the practical benefits of using SEM over the traditional regression approach for some research situations are also explained. Finally, a comparison of a regression and an SEM analysis of the same data testing the same theoretical model is included in the Appendices A and B in order to compare the differences in the research conclusions obtained by the two methods of analysis.
Advances in Accounting Behavioral Research, Volume 9, 179–207 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09007-7
179
180
CINDY BLANTHORNE ET AL.
INTRODUCTION Structural Equation Modeling (SEM)1 was first promoted for use in accounting by Gregson (1992), who in his Accounting Horizons article, suggested that the method was underutilized by accounting researchers. Now more than 10 years after Gregson’s initial commentary, SEM is still not extensively used in behavioral accounting research. For example, since its inception in 1998 through 2004, Advances in Accounting Behavioral Research published a total of 78 articles, 56 of which contained data. Of the articles with data, only 5 studies (8.9%) used SEM. More significantly, 13 studies (23.2%) potentially could have been approached using SEM as they used mediation models, simple averages of questionnaire items or factor scores coupled with traditional analysis methods such as regression. Similarly, between 1994 and 2004, only 15.1% of articles containing data in Behavioral Research in Accounting utilized SEM. Another 21.0% of the studies with data could have potentially been approached using SEM. This limited use of SEM may be surprising given that SEM offers behavioral accounting researchers several advantages over traditional analyses. Briefly, SEM allows researchers to control for measurement error when using latent constructs such as attitude or job satisfaction (Bollen, 1989, pp. 16–17), to investigate modeled path coefficients simultaneously, to test for the overall consistency between the data and the hypothesized model, and to test for mediating relationships between variables in a more straightforward manner than traditional methods (Baron & Kenny, 1986, p. 1177). Before faulting accounting researchers for not employing SEM, some of the reasons for its limited use should be acknowledged. First, SEM was only available in the curriculum for the most recent PhD accounting graduates, and SEM can be difficult to learn on one’s own without the assistance of a structured course. Consequently, many researchers are unaware of the benefits of SEM over a traditional analysis when analyzing behavioral data. With so many varied sources on the topic, accounting researchers may also have difficulty finding readable textbooks and articles providing practical guidance on conducting SEM research in a constrained setting. Further, fitting structural models to data is often problematic, especially if careful planning is not used to design the experiment before the data are gathered. Guidance for the many potentially problematic junctures in an SEM analysis can be found only by accessing multiple sources. Finally, even if all the appropriate references are found, certain statistical assumptions and sample size requirements must be met that can be difficult to achieve in practice. In
Why You should Consider SEM
181
this article, we hope to help researchers overcome some of these barriers that account for the limited use of SEM in behavioral accounting. The primary goal of this article is to provide a simple guide to effectively planning a study suitable for an SEM analysis, while also suggesting references and further reading on the topic of SEM. This article gleans points relevant to behavioral accounting researchers from the vast body of SEM literature. Throughout, the goal is to summarize the current ‘‘best practices’’ and ‘‘rules of thumb’’ while also providing the reader with guidance for obtaining more information on the topics discussed. As such, this article is intended as a starting point for SEM novices who, after reading this article, should still consult the primary sources referenced here to enhance their understanding of methods presented. This discussion should also be useful to PhD students learning SEM and to more experienced SEM users who may find themselves seeking answers to common problems encountered in SEM. Finally, this information may benefit reviewers as they consider the appropriateness and feasibility of SEM use for papers submitted. As a context to present ‘‘best practices’’ and ‘‘rules of thumb’’ for conducting an SEM study, the article follows the analysis of a behavioral tax compliance data set. This research example utilizes the Theory of Reasoned Action (TRA) (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980) as its theoretical foundation, and allows for testing a mediating relationship between predictors and an outcome. The example illustrates various problematic junctures that are typically encountered in an SEM analysis. Additionally, this article provides helpful suggestions for planning an effective SEM study and details why it is beneficial to use SEM rather than a traditional method. Finally, for readers interested in comparing SEM to a traditional regression analysis, the regression analysis and explicit comparison to SEM is included in the Appendices A and B.
RESEARCH EXAMPLE Because a sound theoretical foundation is necessary to conduct a valid SEM analysis, it is important to ensure that an appropriately validated theory is applicable to the research questions of interest, research instrument and data collected. The behavioral tax compliance study in our research example utilizes the TRA to examine underreporting tendencies of taxpayers. As such, a brief background on TRA is presented first, followed by details of the analysis. Throughout the analysis, guidance for a smoother application of SEM is highlighted.
182
CINDY BLANTHORNE ET AL. Attitude Intention
Behavior
Subjective Norm
Fig. 1.
The Theory of Reasoned Action.
The Theory of Reasoned Action Briefly, TRA (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980) posits that behavior is immediately preceded by intention and that intention is based upon the attitude and subjective norm of an individual (see Fig. 1). If people generally do what they intend to do, then the determinants of intention are of primary interest. TRA was developed to assess the determinants of intention. Attitude is based on a person’s belief that a behavior will lead to a given outcome. Subjective norm is the perception of approval by relevant referents concerning the behavior in question. Intention mediates the relationships between attitude and behavior as well as between subjective norm and behavior. Stated another way, attitude and subjective norm have direct effects on intention but affect behavior only indirectly (through intention).2 Because the TRA posits a mediating relationship between predictors and an outcome, SEM is a potentially appropriate method of analysis. Sample Characteristics The data analyzed in the current study is a subset of measures examined in Blanthorne (2000). The purpose of Blanthorne (2000) was to compare income underreporting tendencies of taxpayers who have the real-life opportunity to underreport income versus those who do not. Individuals having opportunity are of particular interest to the IRS because they are the least compliant group of taxpayers. Thus, approximately one-half of the 345 subjects were swap meet vendors and other taxpayers who had the opportunity to underreport income on their tax returns. The sample characteristics are given in Table 1. The predictions made in Blanthorne (2000) were that the influence of attitude, subjective norms, and ethics hinged on the perception of control that taxpayers believed that they have over underreporting income. Specifically, taxpayers who have opportunity would perceive such control to underreport income, which, in turn, would enable them to attend to the remaining variables. Blanthorne found that ethics was the dominant
Why You should Consider SEM
Table 1.
183
Sample Data Characteristics.
Response rate Mean age Percent male Percent of income from self-employment Percent completed at least bachelors’ degree Percent owned own business Percent self or spouse prepared tax return
78% 39.7 59% 53% 61% 56% 51%
predictor and that all other predictors of underreporting income were less significant. In the current study, a subset of measures captured in Blanthorne (2000) is utilized to specifically highlight the issues related to the use of SEM. Questionnaire Items of Interest Before demonstrating the resulting structural model for our research example, it is necessary to understand the questionnaire items of interest (see Appendix A for specific questions). Most of the questionnaire items were rated on a seven-point Likert scale. The measures for attitude and subjective norm were based on the five most commonly cited items (for each variable) elicited from pre-tests.3 For instance, in regard to subjective norms, pre-test respondents were asked to list the people who influenced their tax reporting decisions. All five of the original items from the pre-test were retained to measure subjective norm. Four of the original five items were retained to measure attitude.4 Two items were used to measure respondent’s intention to underreport income on their tax return. This measure of intention is similar to that used in most prior applications of the theory (see Madden, Ellen, & Ajzen, 1992 for a review). Originally four items were selected to measure the respondents’ underreporting behavior. Respondents were presented with two hypothetical situations and asked how much of the available income they would report on a tax return. Respondents were also asked two additional questions about their past tax reporting behavior using a seven-point Likert scale. Initial confirmatory factor analyses (CFAs) showed that the four items measured two different constructs, with the hypothetical items measuring one construct and the past behavior items measuring another. As with most theories, applicability of the TRA is subject to boundary conditions, one of which is the level of specificity between the measure of intention and the measure of behavior (Fishbein & Ajzen, 1975; Ajzen & Fishbein, 1980).
184
CINDY BLANTHORNE ET AL.
Table 2. Construct Attitude Subjective norm Intention Behavior
Construct Reliability.
Coefficient Alpha
Composite Reliability
0.78 0.84 0.75 0.72
0.77 0.88 0.78 0.81
Because the measure of the respondent’s past behavior logically relates more closely to their intention to underreport than their responses to a hypothetical situation, we retained the measures of past behavior in our model. Cronbach’s Alphas and composite construct reliabilities were computed to assess the internal consistency (reliability) of the constructs attitude, subjective norm, intention, and behavior. These values are given in Table 2. The alpha values, range from 0.72 to 0.84. The composite construct reliabilities were computed using the SEM results according to Hair, Anderson, Tatham, and Black (1998, p. 612). Lampe, Conover, Bline, and Sutton (1999) discuss the uses and definition of Cronbach’s Alpha for assessing construct reliability. The use of the SEM-based composite reliability is discussed in Hair et al. (1998, p. 612).
SEM ANALYSIS Measurement Model Consistent with the recommendations of Anderson and Gerbing (1988), the first step to completing an SEM analysis is the fitting of the measurement or factor analytic portion of the model. The measurement model for this application of TRA is presented in Fig. 2a. Variables that cannot be directly observed are referred to as latent constructs. The oval shaped symbols represent the four latent constructs (attitude, subjective norm, intention, and behavior). Information about latent constructs is obtained indirectly by observing several indicators or variables affected by the latent construct (Long, 1983, p. 11). These observed variables or indicators are often responses to questionnaire items. The rectangular shaped symbols in Figs. 2a and 2b represent observed variables (att1, att2, etc.). The circular symbols with arrows pointing to each of the observed variables are error or disturbance terms similar to residuals in a regression equation. Notice that
185
Why You should Consider SEM (a) 1
att1
1
att2 Attitude
1
att3
1
1
att4 1
1
int1
1
int2
1
1
b1
b2
1
1
Intention
Behavior
sn1
1
sn2
1
sn3
1
Subjective Norm
1
sn4
1
sn5
(b) 1
att1
1
att2 Attitude
1
att3
1
1
att4 1
1
int1
int2
1
1
b1
b2
1
1
Intention
Behavi or
1
1 db
di 1 1 1 1 1
Fig. 2.
sn1 sn2 sn3
1
Subjective Norm
sn4 sn5
(a) Measurement Model. (b) Full SEM Model.
186
CINDY BLANTHORNE ET AL.
the arrows between the latent constructs and the related observed variables point from the latent constructs to the observed variables because the observed variables are believed to be effect indicators (as opposed to causal indicators). That is, underlying attitude is believed to affect responses to certain questions regarding attitude rather than the responses shaping attitudes. The curved arrows between the latent constructs represent simple correlations between the two associated constructs. The fitting of the measurement portion of a model in SEM is analogous to using CFA. Taking this important step before adding the structural or directional relationships between latent constructs has several important benefits. Most importantly, it allows the researcher to separate inconsistencies with the measurement model from inconsistencies with the structural portion of the model. Inconsistencies in the measurement model could manifest with indicators that do not load properly on a latent construct, poor construct reliability, lack of discriminant validity, or poor overall fit. Structural Model The structural portion of an SEM model refers to the direct and indirect effects among latent constructs and/or observed variables. Once the measurement model has been fit, the structural components can be added to the measurement model in order to develop the full model for analysis. The full model for this application of TRA is pictured in Fig. 2b using standard symbols for each of the model elements (Kline, 2005, pp. 66–74). The straight arrows in Fig. 2b replace some of the curved arrows, and indicate direct effects between latent constructs. The direction of the arrows represents the theoretical direction of the proposed effects. For example, attitude is believed to directly affect intention, but intention is not believed to affect attitude. Thus, attitude about a behavior is expected to precede intention, and a straight arrow points from attitude to intention. Similarly intention precedes the eventual behavior as indicated by a straight arrow pointing from intention to behavior. AMOS 5.05 was used to obtain the parameter estimates for the models pictured in Figs. 2a and 2b. The parameter estimates for the final model given in Fig. 2b as well as model fit indices are given in Table 3.6 Model Fit An important step in an SEM analysis is assessing model fit. While there are many possible goodness of fit indices (GFIs), we discuss some of the more
187
Why You should Consider SEM
Table 3.
SEM Results.
Goodness of Fit Indices w2 190.463 (df ¼ 57, p ¼ 0.003) w2/df ¼ 1.587 GFI 0.930 IFI 0.934 CFI 0.931 RMSEA 0.041 (90% CI [0.024,0.057]) Factor Loadings
Squared Multiple Correlations
Factor
Indicator
Attitude
att1 att2 att3 att4 sn1 sn2 sn3 sn4 sn5 int1 int2 b1 b2
Subjective norm
Intention Behavior
Standardized
Unstandardized
p-value
0.684 0.693 0.568 0.745 0.701 0.896 0.771 0.756 0.708 0.701 0.884 0.714 0.926
0.966 0.81 0.632 1.00 1.118 1.382 1.272 1.00 0.997 1.00 1.405 1.00 1.617
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Intention Behavior int1 int2 b1 b2 sn1 sn2 sn3 sn4 sn5 att1 att2 att3 att4
0.109 0.766 0.492 0.781 0.510 0.858 0.491 0.802 0.594 0.572 0.501 0.468 0.481 0.322 0.555
Regression Weights
Intention ’ Attitude Intention ’ Subjective norm Behavior ’ Intention
Coefficient
Test statistic
p-value
0.191 0.266 0.648
3.394 4.031 8.969
0.001 0.000 0.000
commonly accepted measures and highlight some rules of thumb for determining model fit. The GFIs for the full model in our example indicate that the data are consistent with the proposed model. Although the w2 test of fit is statistically significant (suggesting inadequate fit), it is well known that this test is overly restrictive when large samples such as this one are used (Kline, 2005). Kline suggests that a more useful measure of fit is to divide the w2 statistic by its degrees of freedom. Any ratio below three is indicative
188
CINDY BLANTHORNE ET AL.
of a well-fitting model (Kline, 2005); thus, the calculated value of 1.587 supports that the model fit is adequate. Because no one measure of fit sufficiently summarizes all of the statistical properties of a model, it is common to examine additional measures of model fit (Kline, 2005). Another measure of model fit, the GFI is 0.930, indicating that about 93% of the covariance matrix (of the observed variables) is accounted for by the proposed model (Kline, 2005). It is desirable to obtain a GFI value close to one, although there is no commonly accepted threshold for this measure. Two other common measures of fit are the Incremental Fit Index (IFI) and Comparative Fit Index (CFI), and it is generally accepted that IFI and CFI values above 0.95 indicate a well-fitting model and values below 0.90 indicate that substantial improvements can be made to the model (Bollen, 1989, p. 274). The IFI and CFI for this model are respectively 0.934 and 0.931. These values are slightly lower than the recommended 0.95 cutoff for a well-fitting model, thus they indicate that the model fit is only moderately good. The fit index titled Root Mean Square Error Approximation (RMSEA) for this model is 0.041 with a 90% confidence interval of (0.024, 0.057). The general heuristic for a well-fitting model is to obtain an RMSEA value below 0.08 (Hair et al., 1998, p. 656).7 Taken together, the above fit measures indicate that the model fit is good; however, the individual components of the model (i.e., strength and direction of relationships among variables and validity of factor solutions) are of primary interest. Table 3, shows that all of the indicators to the latent constructs (i.e., factor loadings) are statistically significant, which indicates a reasonable factor solution.8 Additionally, the squared multiple correlation coefficients, which give the proportion of the variability in the item indicators that is due to the respective latent construct, each range from 0.322 to 0.858. A high squared multiple correlation coefficient for an indicator implies that it is strongly related to the latent construct identified by the other specified indicators. Further, the squared multiple correlation coefficient for the latent construct behavior indicates that about 77% of the variability in behavior is accounted for by the model. The regression weights indicate that attitude and subjective norm are significantly related to intention (p ¼ 0.001 and p ¼ 0.000, respectively), and that intention is significantly related to behavior (p ¼ 0.000). Test of Mediation To conduct a formal test for the mediating effect of intention on the attitude–behavior and the subjective norm–behavior relationships, the orig-
189
Why You should Consider SEM
Attitude
Intention 1
Behavior 1 db
di
Subjective Norm
Fig. 3.
Modified SEM Model.
inal model in Figs. 2a and 2b will be compared to an alternate model shown in Fig. 3.9 Here, the original SEM model was modified by including a direct arrow between subjective norm and behavior as well as a direct arrow between attitude and behavior. If model improvement is achieved by including these direct arrows between the predictors (attitude and subjective norm) and the outcome (behavior), partial mediation of one or both of the relationships is supported. However, if there is no improvement in the fit of the model when including these direct paths, complete mediation of the relationships through intention is supported (Holmbeck, 1997). A formal test for comparing the model in Figs. 2a and 2b to the one in Fig. 3 gives a w2 test statistic of 0.02 (p ¼ 0.990). This implies that the two additional paths in Fig. 3 do not differ significantly from zero. Thus, the SEM model in Fig. 2b is retained, and the completely mediated attitude–intention–behavior as well as the subjective norm–intention–behavior relationships are supported. The practical interpretation of this finding is that attitude and subjective norms indirectly affect underreporting behavior through the individual’s intention to underreport taxable income.
190
CINDY BLANTHORNE ET AL.
PLANNING AN EFFECTIVE SEM STUDY The preceding section provides guidance for performing an SEM analysis. But before a researcher would get to the analysis stage, there are several additional SEM issues to consider in the planning process. Develop one or more Initial Models The first step in bringing a research idea to fruition is to carefully develop a hypothesized model. While this is an important step that should precede data gathering when using any statistical method, it is a critical step when designing an SEM study. While regression or ANOVA can be used in an exploratory setting, SEM requires a strong theoretical foundation for the model. SEM analyses allow researchers to draw more solid conclusions from their research provided the models are theory based, the data are appropriate for the model, and the statistical assumptions are reasonable. Failing to specify the model before the data are gathered can lead to disappointing results as the researcher may find there is too little data, too few items to measure a construct, or an important theoretical variable was unintentionally omitted in the data gathering. Before gathering data for an SEM analysis, it is critical to insure that the model is identified. Model identification deals with the ability to uniquely estimate the model parameters for a given set of observations. If a model is underidentified (contains too few observed variables to estimate the parameters), computer programs will not fit the model, and the researcher must either (1) gather more variables on the existing sample or (2) make modifications to the model. The first option is likely impossible in a situation where subject availability is constrained such as occurs in behavioral accounting research, and the second may lead to weaker research conclusions; thus, it is highly recommended that researchers determine if a hypothesized model is identified before gathering the data. Kline (2005) and Bollen (1989), among others give methods for determining the identification status of SEM models. Carefully select the Indicators to the Latent Constructs A contributor to model identification is the presence of a sufficient number of observed variables or indicators to measure a latent construct. In most cases, a researcher needs at least three indicators per latent construct in order for the measurement portion of the model to be identified. Kenny’s (1979, p. 143)
Why You should Consider SEM
191
heuristic about the number of indicators is ‘‘Two might be fine, three is better, four is best, and anything more is gravy.’’ The goal is to end up with at least three indicators per latent construct, so we recommend designing your questionnaires with more items than you will need. Some indicators may need to be discarded due to lack of internal consistency or other problems. In addition to having enough indicators, it is important to choose the right indicators to the latent constructs. The indicators that should be initially included in the instrument depend on the a priori evidence (i.e., theory and/ or prior literature) of the strength of the relationship between each indicator and its related latent construct. Obtain a Large Sample Statisticians always recommend getting more data, but behavioral accounting researchers can be seriously constrained from doing this because professional accountants often serve as subjects. Unfortunately this does not change the fact that a large sample is needed for an SEM analysis to provide reliable results. Research has shown that results of SEM analysis are not reliable when based on small samples. For example, Anderson and Gerbing (1988) indicated that 100–150 subjects is the minimum satisfactory sample size when using SEM, and Boomsma (1983) recommended at least 400 observations. Most textbooks on the subject suggest between 250 and 500 observations are necessary (e.g., Schumacker & Lomax, 1996; Kline, 2005; Loehlin, 2004). This leaves behavioral accounting researchers in a difficult position of needing access and cooperation of more participants (often professional accountants) than is often practical to obtain. In fact, many behavioral researchers who did not plan for SEM may not have gathered enough data. For instance, Almer and Kaplan (2002, p. 14) commented that they ‘‘rejected structural equation modeling because y [it] requires the use of a much larger sample size than [they] have.’’ In fact, 5 of the 13 potential SEM studies published in Advances in Accounting Behavioral Research have less than 100 subjects and only 4 of the studies have greater than 250 subjects. Although Behavioral Research in Accounting has more potential SEM studies, 25 in the past 10 years, only 15 of the studies have greater than 100 subjects and only 2 have greater than 250 subjects. Access to larger sample sizes is often only possible through the use of student subjects, which itself can be fraught with problems when students are intended surrogates for professional accountants (Walters-York & Curatola, 1998). In short, SEM sample size requirements may force researchers to make tradeoffs between appropriateness of subject-task matching and use of optimal statistical techniques.
192
CINDY BLANTHORNE ET AL.
Cleanse the Data As with any statistical methodology, SEM comes with a list of assumptions. The bad news is that the assumptions can often be difficult to strictly meet in practice. The good news is that there are often remedies, some of which are outlined below. One of the primary assumptions necessary for SEM is that the observed variables follow a multivariate normal distribution. This differs from the normality assumption in ANOVA or regression in that with the traditional methods, only the model errors are required to follow a univariate normal distribution. The normality assumption in SEM is much more stringent and applies to all observed variables.10 If the data do not follow a multivariate normal distribution, the standard errors and hypothesis tests may be unreliable. Tabachnick and Fidell (2001, Chapter 4) discuss methods for assessing multivariate normality and offer several options for transforming non-normal data. In cases when transformations of the data do not result in approximate normality, alternate estimation methods within SEM may be used. Several software packages allow the use of Asymptotically Distribution Free (ADF) estimation methods that do not require the assumption of multivariate normality. Additionally, AMOS 4.0 allows the researcher to use a bootstrap approach to hypothesis testing within an SEM model. The bootstrap is a computationally intensive method of estimating standard errors that does not require assumptions regarding the distribution of the data. West, Finch, and Curran (1995) discuss the effects of nonnormality on SEM models and offer several remedies to this problem. In addition to problems of nonnormality, outliers and influential cases can also affect the SEM analysis. Because SEM models the covariance structure among the observed variables, any outlier or influential point that affects the covariance between variables can affect the estimated model parameters. Generally the best approach is to identify and deal with outliers before beginning your SEM analysis. Whether to omit outliers or to retain them is a decision that depends on the circumstances surrounding the origin of the case in question, the sample size, and the importance of this case to the research conclusions. Detailed methods for identifying and dealing with outliers and influential points are discussed in detail in Tabachnick and Fidell (2001, Chapter 4). A commonly used moniker for SEM models is Linear Structural Relationships (LISREL). True to its name, SEM is a method for studying linear relationships among variables, so one must insure that the relationships among the observed variables are, indeed, linear. If one or more of the bivariate relationships between observed variables is not linear (as indicated
Why You should Consider SEM
193
by a simple scatter plot), linearizing transformations can be used before fitting the SEM models. Unfortunately, transformations can also make interpretations difficult. Linearizing transformations are also discussed in Tabachnick and Fidell (2001, Chapter 4). Clearly present the Results Although not part of the planning process, presentation of the results is an important element of an effective analysis and should be briefly discussed. A clear presentation of the analysis in a research article is crucial to the usefulness of the analysis for practical understanding as well as furthering research in an area. A key component to assessing an SEM model is the use of GFIs. Researchers should present several of these indices including but not limited to the w2 test of fit, the w2 divided by its degrees of freedom, IFI, NFI, and RMSEA. In addition to measures of overall fit, measures of components of fit including standardized factor loadings, structural weights, and squared multiple correlation coefficients should be included. Some measures of construct reliability should also be included. Our preference is the composite reliability that can be computed directly from the SEM output (see, e.g., Hair et al., 1998, p. 12), but many choose to present Cronbach’s Alpha coefficient. A detailed discussion of the benefits and drawbacks of the use of Chronbach’s Alpha coefficient is given in Lampe et al. (1999). It is often important to make information available to the reader so that an analysis can be reproduced. Thus, it is important to include the means and standard deviations of the observed variables, as well as the correlations between the observed variables. From this information, a reader can reproduce the covariance matrix of the observed data and easily reproduce an SEM analysis. If space constraints limit the ability of authors to present these detailed descriptive statistics, the information could be made available on personal websites or by request of the authors. Additional information on the effective reporting of SEM results is available in Shook, Ketchen, Hult, and Kacmar (2004).
WHY USE SEM? So given all the constraints on planning and conducting an SEM project, a reasonable question to ask is, ‘‘Why use SEM? Isn’t it possible to simply use regression to test for similar relationships?’’ In addition to accounting for random measurement error, SEM offers researchers added flexibility over a
194
CINDY BLANTHORNE ET AL.
regression approach and gives opportunities for an enhanced analysis. As is evident when the data from our research example is analyzed using a regression approach (see Appendix B), conclusions may differ from those found under SEM for a variety of reasons. The following section highlights some of the enhancements of the benefits of SEM over traditional methods. SEM Accounts for Random Measurement Error Traditional regression models are equivalent to SEM models if the latent constructs are assumed to be directly observed or measured without error. Thus, if the latent constructs are measured with negligible error, a researcher can expect to obtain similar results when using comparable traditional and SEM models. Many behavioral researchers work with latent constructs, and these constructs are often measured with substantial and unavoidable error. Ignoring measurement error and using regression rather than SEM in situations when reliability is not high can undermine attempts to estimate the effect of one variable on another and may cause a researcher to miss significant relationships or to draw erroneous conclusions. Bollen (1989, Chapter 5) summarizes the effects of ignoring measurement error on traditional regression models. Using our research example, we can illustrate the reason that the regression and SEM analysis of the same data using comparable models provide different research conclusions. Regression and SEM analyses will give equivalent results for comparable models provided the latent constructs are measured without error. To illustrate the effect of measurement error on the subjective norm–intention–behavior relationship, an SEM model can be formed using the averaged items as proxies for the latent constructs. Using this approach, the reliability of the proxies can be varied to show the effect of measurement error on the conclusions.11 Table 4 gives the coefficients and p-values when varying the reliability of intention from 1.0 to 0.75. As the reliability of intention decreases, the coefficient between subjective norm and intention remains the same, but the other two coefficients change markedly. Specifically, the relationship between intention and behavior becomes stronger as indicated by larger significant coefficients. Conversely, the relationship between subjective norm and behavior becomes weaker as indicated by smaller coefficients that become insignificant as the reliability decreases. In fact, with what is considered to be a high reliability of 0.85, the relationship between subjective norm and behavior is no longer statistically significant (at the 5% level). The measurement error in intention that was not accounted for in the regression analysis affected the results leading to conclusions that were inconsistent with TRA. This error in intention was
Why You should Consider SEM
Table 4.
Comparison of Reliability. Reliability
Intention ’ Subjective norm Behavior ’ Intention Behavior ’ Subjective norm
1.00
0.95
0.90
0.85
0.80
0.75
0.373 (0.000) 0.569 (0.000) 0.117 (0.009)
0.373 (0.000) 0.602 (0.000) 0.105 (0.022)
0.373 (0.000) 0.638 (0.000) 0.092 (0.046)
0.373 (0.000) 0.678 (0.000) 0.077 (0.094)
0.373 (0.000) 0.724 (0.000) 0.060 (0.192)
0.373 (0.000) 0.777 (0.000) 0.040 (0.424)
Note: p-values shown in parentheses.
195
196
CINDY BLANTHORNE ET AL.
accounted for in the SEM and the results confirmed consistency between the TRA model and the data. SEM can Control for some Types of Non-Random Error SEM not only allows a researcher to control for the random measurement error that is inherent to measuring latent constructs, but it also allows a researcher to control for some types of nonrandom measurement error. For example, suppose the latent construct, attitude, will be measured on two separate occasions using the same sample of individuals and the same instrument. The goal of the analysis is to show changes in attitude over time. A certain portion of the variability in each item measuring attitude will be due to the true score, or the latent construct of interest. The remaining variability in the item that is not due to the latent construct is measurement error. It may be unreasonable to assume that the measurement error in a given item observed at the first time period is uncorrelated with the measurement error of the same item observed at the second time period. A traditional regression model allows no way to control for this type of error correlation. In an SEM model, a researcher can specify that the two measurement errors correlate over time, and the model fit may be greatly improved. This allows researchers the flexibility to analyze their data using a variety of more complicated theoretical models, and to adapt these models to their research methodologies. SEM offers researchers many options to modify and test theory, and the ability to correlate error terms is an important component to this flexibility. However, a strong warning is offered against adding correlation terms among errors simply to improve model fit. Error terms should only be correlated when the theory or research methodology suggests that the correlations are appropriate. As with any statistical model, the more constraints placed on a model in terms of additional parameters, the better the model will fit. For example, adding more predictors to a regression model will always increase R2, even if the predictors are not meaningful. In the same way, adding more correlation terms to an SEM model will often improve the goodness of fit measures. Cortina (2002) outlines conditions under which it is appropriate to allow error terms to correlate in an SEM analysis. SEM allows for Straightforward Evaluation of Convergent and Discriminant Validity The concepts of convergent and discriminant validity are central to the evaluation of the CFA or measurement portion of an SEM. Briefly,
Why You should Consider SEM
197
convergent validity measures the degree to which the indicators of a latent construct measure the same construct. The indicators are presumed to measure the same construct if the intercorrelations among the indicators are moderately high. The squared multiple correlations given in Table 3 for the questionnaire items give a measure of how each item relates to its respective latent construct. A high squared multiple correlation coefficient for an indicator implies that it is strongly related to the latent construct identified by the other specified indicators. For example, the squared multiple correlation coefficient for sn2 is 0.802. This implies that about 80% of the variability in this indicator is due to the latent construct subjective norm. The remaining 20% is due to measurement error. Discriminant validity measures the degree to which two or more latent constructs measure different constructs. For example, the estimated correlation between attitude and subjective norm in our example is 0.231. While the two latent constructs are related, the correlation is not so high that one would believe they are redundant factors. A correlation coefficient of 0.85 or higher is often the heuristic used to indicate a lack of discriminant validity (Kline, 2005, p. 60).12 SEM gives a Global View One of the primary benefits of SEM over traditional models is that SEM takes an ‘‘all-at-once’’ approach to answering several research questions. If regression is used to analyze our data set, six models are used to piece together the results necessary to answer the basic research question (see Appendix B). Further, the regression analysis is often preceded by an exploratory factor analysis of the data in order to define the factors to be included in the regression model. When using an SEM approach, analysis of the measurement portion of the model allows testing for convergent and discriminant validity. When adding the structural relationships to the model, an SEM analysis provides overall measures of goodness of fit of both the regression and factor analytic portions of the model. The overall goodness of fit measures in SEM give the researcher tools to summarize the consistency between the data and the proposed theoretical model. There are no similar measures available when using most traditional methods. SEM emphasizes Theory Testing SEM is not an exploratory technique and should not be used for exploring your data to determine what relationships might exist. SEM is used ideally
198
CINDY BLANTHORNE ET AL.
to test the applicability of a priori theories to a research scenario; however, in practice SEM models are not strictly confirmatory either. A strictly confirmatory model would occur when a researcher specified a model, gathered the data, tested the model, and if the model did not fit, abandoned the idea for another project. Few researchers have the luxury of simply giving up when a model does not fit. The most common use of SEM occurs when the original model does not fit the data, and the researcher carefully modifies the model and retests the same data (Jo¨reskog, 1993). The a priori nature of SEM allows the analyst to draw stronger theoretical conclusions than when using exploratory analysis methods.
CONCLUDING REMARKS As the sophistication of behavioral research in accounting increases, so does the need for more sophisticated statistical methods. Because SEM is a more sophisticated method that can, if properly applied, yield more solid inferences in the presence of measurement error, it is becoming increasingly expected in behavioral accounting research. To this end, our paper makes strides in helping accounting researchers meet the growing expectation for SEM. This paper provides a research guide for effectively planning a study suitable for an SEM analysis. The guide discusses key elements that should be present, and provides advice for avoiding common pitfalls when designing the study. The guide also provides multiple tips to facilitate smoother application of the method and suggests understandable references for further reading on the topic of SEM. In short, by helping researchers unfamiliar with SEM better understand why it is useful and helping researchers familiar with SEM use it more efficiently and effectively, we hope to encourage the intentional use of SEM within behavioral accounting research.
NOTES 1. SEM models are referred to by many names including Linear Structural Relationships (LISREL) models, simultaneous equation models, covariance structure models, and latent variable models. SEM models that contain both factor analytic portions (CFA) and structural portions such as regression analysis are often referred to as hybrid SEM models (Kline, 2005). 2. See Loehlin (2004, p. 8) or Schumacker and Lomax (1996, p. 90) for a discussion of direct and indirect effects.
Why You should Consider SEM
199
3. Pre-tests are commonly used in conjunction with the TRA and recommended by Madden et al. (1992). 4. One item, bragging rights, was dropped from the analysis due to a lack of convergent validity. See discussion of convergent validity in the section ‘‘Why Use SEM.’’ 5. AMOS 5.0 (Arbuckle, 2003) is a computer program that is commonly used to fit structural equation models. Kline (2005, Chapter 4) discusses common SEM software packages. 6. Asymptotically Distribution Free Estimation (ADF) was used because of violations of the multivariate normality assumption. The ADF estimation method is recommended as a remedy for violations of the normality assumption in West et al. (1995). The multivariate normality assumption for SEM is discussed in a later section. 7. For more information regarding overall goodness of fit indices in SEM, see Bollen (1989, pp. 256–289) and Hair et al. (1998, pp. 653–661). 8. The coefficients for the indicators att4, sn4, int1, and b1 were each set equal to one in order to provide scales for the latent constructs. Tests of significance cannot be performed on scaling indicators. 9. The indicators to the latent constructs have been removed from Fig. 3 to simplify the picture. The indicators used are the same as those used in the model pictured in Fig. 2. 10. One exception to this would be categorical variables such as gender or group membership, which can be included using a multigroup approach. 11. The reliability of the proxy variable can be set by forcing the error variance equal to (1 a)s2, where a is the desired reliability, and s2 is the variance of the variable. 12. See Kline (2005) and Byrne (2001, Chapter 3) for formal methods to test for discriminant validity in SEM. 13. Using simple averages rather than factor scores is commonly accepted when using a validated instrument (Kline, 2005, p. 207). 14. The difference between the total effect of the predictor on the outcome (0.330) in model 4 and the direct effect of the predictor on the outcome (0.117) in model 6 is equal to the product of the coefficient between the predictor and mediator (0.373) and the coefficient between the mediator and the outcome (0.569). Testing whether this product is significantly different than zero gives a formal test for partial mediation. The standard error of this product (0.3730.569) is equal to the square root of mo2spm2+pm2smo2+spm2smo2, where mo and pm are the regression coefficients for the M O and P M relationships (0.569 and 0.373, respectively), and smo and spm are the respective standard errors of these coefficients. For this example, the standard error is calculated as O(0.5692)(0.0782)+(0.3732)(0.0312)+(0.0782)(0.0312) ¼ O0.002 ¼ 0.046. Dividing the product (0.3730.569 ¼ 0.212) by the standard error gives a z-score (z ¼ 0.212/0.046 ¼ 4.62), which serves as a test statistic for partial mediation. Using a standard normal distribution, the p-value for this test statistic is p ¼ 0.000. Because this product is significantly different than zero, there is evidence to support that the relationship between subjective norm and the behavior of underreporting income on tax returns is only partially mediated by the intention toward this behavior. Preacher and Leonardelli (2003) provide an online calculator for the test of partial mediation.
200
CINDY BLANTHORNE ET AL.
REFERENCES Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood-Cliffs, NJ: Prentice-Hall. Almer, E. D., & Kaplan, S. E. (2002). The effects of flexible work arrangements on stressors, burnout, and behavioral job outcomes in public accounting. Behavioral Research in Accounting, 14, 1–34. Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411–423. Arbuckle, J. L. (2003). Amos 5.0 [computer software]. Chicago, IL: SmallWaters. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Blanthorne, C. (2000). The role of opportunity and beliefs on tax evasion: A structural equation analysis. Unpublished doctoral dissertation, Arizona State University. Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley-Interscience. Boomsma, A. (1983). On the robustness of LISREL (maximum likelihood estimation) against small sample size and nonnormality. Amsterdam: Sociometric Research Foundation. Byrne, B. M. (2001). Structural equation modeling with AMOS: Basic concepts, applications and programming. Mahwah, NJ: Lawrence Erlbaum Associates, Inc. Cortina, J. M. (2002). Big things have small beginnings: An assortment of ‘‘minor’’ methodological misunderstandings. Journal of Management, 28, 339–362. Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley. Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects in counseling psychology research. Journal of Counseling Psychology, 51, 115–134. Goodman, L. A. (1960). On the exact variance of products. Journal of the American Statistical Association, 55, 708–713. Gregson, T. (1992). The advantages of LISREL for accounting researchers. Accounting Horizons, 6, 42–48. Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis (5th ed.). Upper Saddle River, NJ: Prentice-Hall. Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: Examples from the child-clinical and pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65, 599–610. Joreskog, K. G. (1993). Testing structural equation models. In: K. A. Bollen & J. S. Lang (Eds), Testing structural equation models (pp. 294–316). Newbury Park, CA: Sage Publications. Kenny, D. A. (1979). Correlation and causality. New York: Wiley. Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). New York: The Guildford Press. Lampe, J. C., Conover, W. J., Bline, D. M., & Sutton, S. G. (1999). Uses and misuses of Cronbach’s alpha: Implications for behavioral researchers. Advances in Accounting Behavioral Research, 2, 283–307. Loehlin, J. C. (2004). Latent variable models – an introduction to factor, path and structural analysis (4th ed.). Hillsdale, NJ: Lawrence Erlbaum Associates. Long, J. S. (1983). Covariance structure models – an introduction to LISREL. Quantitative Applications in the Social Sciences, series no. 07-034. Newberry Park, CA: Sage.
201
Why You should Consider SEM
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7, 83–104. Madden, T. J., Ellen, P. S., & Ajzen, I. (1992). A comparison of the theory of planned behavior and the theory of reasoned action. Personality and Social Psychology Bulletin, 18, 3–9. Preacher, K., & Leonardelli, G. (2003). Calculation for the Sobel test: An interactive calculation tool for mediation tests. Retrieved June 22, 2004 from http://www.unc.edu/preacher/ sobel/sobel.htm. Schumacker, R. E., & Lomax, R. G. (1996). A beginner’s guide to structural equation modeling. Mahwah, NJ: Lawrence Erlbaum Associates. Shook, C. L., Ketchen, D. J., Hult, G. T. M., & Kacmar, K. M. (2004). An assessment of the use of structural equation modeling in strategic management research. Strategic Management Journal, 25, 397–404. Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Needham Heights, MA: Allyn and Bacon. Walters-York, L. M., & Curatola, A. P. (1998). Recent evidence on the use of students as surrogate subjects. Advances in Accounting Behavioral Research, 1, 123–143. West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with nonnormal variables: Problems and remedies. In: R. H. Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 56–75). Thousand Oaks, CA: Sage.
APPENDIX A. QUESTIONNAIRE ITEMS Measures of Attitude ‘‘______________ has the following type of effect on whether or not I would underreport income on my income tax return:’’
att1: paying less tax att2: the effort required to prepare my tax return att3: feeling that I have ‘‘beat the system’’ att4: having more money left in my pocket att5: being able to brag about underreporting
Discourages Underreporting 1 2
3
No Effect 4
5
Encourages Underreporting 6 7
1
2
3
4
5
6
7
1
2
3
4
5
6
7
1
2
3
4
5
6
7
1
2
3
4
5
6
7
202
CINDY BLANTHORNE ET AL.
Measures of Subjective Norm ‘‘My _______________ has the following type of effect on whether or not I would underreport income on my income tax return:’’ Note: If a certain person (or group) is not applicable to you, please select ‘‘no effect.’’
sn1: spouse/significant other sn2: family sn3: tax return preparer sn4: friends sn5: business contacts/ peers
Discourages Underreporting 1 2 1 1 1 1
2 2 2 2
3
No Effect 4
5
3 3 3 3
4 4 4 4
5 5 5 5
Encourages Underreporting 6 7 6 6 6 6
7 7 7 7
Measure of Intention Int1: I was determined to underreport income on the last income tax return I filed. Definitely Disagree 1
2
3
4
5
6
Definitely Agree 7
Int2: I intended to underreport income on the last income tax return I filed. Definitely Disagree 1
2
3
4
5
6
Definitely Agree 7
Measures of Behavior b1: On your 1998 tax return, how much of the income that you believed was taxable did you include on your tax return Included All 1
2
3
4
5
6
Included None 7
b2: In past years, approximately how often do you think you underreported your income on your tax return? Never Always 1 2 3 4 5 6 7
Why You should Consider SEM
203
b3: Karen Chip operates a consulting business out of her home. Although she has adequate office space in her home, she is generally required to travel to client’s offices to advise them about software decisions. Karen has been in business for 5 years. Karen describes the first 3 years as rather ‘‘lean’’ and fees generated during those years were just enough to cover expenses and pay her a small salary. Luckily, Karen’s consulting fees have continued to rise while her expenses have remained fairly stable. During 1998, Karen received a few small cash payments from clients. The total received was $5,000. The fees were not reported to the IRS. If you were Karen, how much (if any) of the $5,000 cash fees would you include on your 1998 income tax return? ENTER AMOUNT HERE: $__________ b4: Bob lives near a large state university. During 1998 Bob rented out a bedroom to a student. This is Bob’s first and only rental agreement. The agreement included the use of one bedroom, a private bathroom and kitchen privileges. The student paid Bob a total of 10 rental payments during 1998. Each rental payment was in the amount of $250 and was delivered to Bob in cash on the first of the month. Thus, Bob received a total of $2,500 for the use of his home. Rental income is taxable. If you were Bob, how much (if any) of the $2,500 rental amount would you include on your 1998 income tax return? ENTER AMOUNT HERE: $ __________
APPENDIX B. ALTERNATIVE TEST OF MEDIATION Traditional Analysis using Regression Using the multistep regression approach outlined in Baron and Kenny (1986) and Frazier, Tix, and Barron (2004), regression analysis is used to determine if relationships between attitude and behavior as well as subjective norm and behavior are mediated by intention. If, for example, intention is found to be a mediator between attitude and behavior, this implies that intention toward the behavior explains the relationship between attitude toward the behavior and the eventual behavior. The procedure for testing mediation effects in regression is a four-step method that includes: (1) testing for significant relationships between the predictors, attitudes, and subjective norms and outcome behavior variables
204
CINDY BLANTHORNE ET AL.
(P O), (2) testing for significant relationships between the predictors and mediator, intentions (P–M), (3) testing for significant relationships between the mediator and outcome variable (M O), and (4) showing that the strength of the relationship between the predictor and the outcome is significantly reduced when the mediator is added to the model. To test the model given in Fig. 1, the average of the respective questionnaire items representing the constructs of attitude, subjective norm, intention, and behavior are used.13 Three regression models will be used to test whether intention is a mediator for the relationship between attitude and behavior. First, behavior is regressed on attitude to establish the P O relationship. Next, intention is regressed on attitude to establish the P M relationship. Finally, behavior is regressed on both attitude and intention to establish the M O relationship. The results of these regression analyses appear in Table A1. The three regression models complete the first three steps to study the attitude–intention–behavior relationship. In the first model, attitude is a significant predictor of behavior (p ¼ 0.000) that establishes the P O relationship. The second model indicates attitude is a significant predictor of the intention (p ¼ 0.000) establishing the P M relationship. Finally, the Table A1.
Regression – Test of Intention as a Mediator between Attitude and Behavior. Coefficient
Standard error
t-statistic
p-value
Model 1: Response Variable Behavior (P–O) Constant 0.922 0.242 Attitude 0.234 0.055 R2: 0.051
3.815 4.276
0.000 0.000
Model 2: Response Variable Intention (P–M) Constant 0.817 0.298 Attitude 0.319 0.068 R2: 0.061
2.743 4.732
0.006 0.000
Model 3: Response Variable Behavior (M–O) Constant 0.449 0.171 Attitude 0.049 0.040 Intention 0.580 0.031 R2: 0.535
2.618 1.239 18.870
0.009 0.216 0.000
205
Why You should Consider SEM
third model shows that when attitude is included in the model to predict behavior, intention is a significant predictor of behavior (p ¼ 0.000) that establishes the M O relationship. Inspection of the first and third models allows us to complete the fourth step of the test for mediation, to show that the relationship between the predictors and the outcome is significantly reduced when intention is added to the model. The coefficient for attitude is 0.234 (p ¼ 0.000) in the first model that does not include intention and drops to 0.049 (p ¼ 0.216) in the third model that includes intention. Attitude was originally shown to predict behavior, but once intention was added to the model, the coefficient for attitude was no longer significantly different than zero. This implies that intention is a complete mediator for the relationship between attitude and behavior. In other words, intention toward underreporting income on a tax return (behavior) completely explains the relationship between attitude toward this behavior and the actual behavior. Three additional regression models will be used to establish intention as a mediator for the relationship between subjective norm and behavior. Behavior is regressed on subjective norm to establish the P O relationship. Next, intention is regressed on subjective norm to establish the P M relationship. Finally, behavior is regressed on both subjective norm and intention to establish the M O relationship. The results of these regression analyses are given in Table A2. Table A2.
Regression – Test of Intention as a Mediator between Subjective Norm and Behavior. Coefficient
Standard error
t-statistic
p-value
Model 4: Response Variable Behavior (P–O) Constant 0.738 0.233 Subjective norm 0.330 0.063 R2: 0.075
3.171 5.273
0.002 0.000
Model 5: Response Variable Intention (P–M) Constant 0.841 0.290 Subjective norm 0.373 0.078 R2: 0.062
2.894 4.780
0.004 0.000
Model 6: Response Variable Behavior (M–O) Constant 0.260 0.166 Subjective norm 0.117 0.046 Intention 0.569 0.031 R2: 0.542
1.564 2.579 18.663
0.119 0.010 0.000
206
CINDY BLANTHORNE ET AL.
The subjective norm–intention–behavior relationship can be explained using the output in Table A2. In the fourth model, subjective norm significantly predicts behavior (p ¼ 0.000) that confirms the P O relationship. The fifth model indicates that subjective norm is a significant predictor of the intention (p ¼ 0.000) confirming the P M relationship. Finally, the sixth model shows that when subjective norm is included in the model to predict behavior, intention is a significant predictor of behavior (p ¼ 0.000) that confirms the M O relationship. Notably, in the sixth model, subjective norm is still a significant predictor of behavior (p ¼ 0.010) when intention is included. This implies that intention may only partially mediate the relationship between subjective norm and behavior. In other words, intention only accounts for some of the relationship between the predictor and the outcome. To establish partial mediation of this relationship, it is not sufficient to simply show that the coefficient for the predictor (subjective norm) is statistically different than zero when the mediator (intention) is included in the model to predict the response (behavior). MacKinnon, Lockwood, Hoffman, West, and Sheets (2002) compared several methods for formally testing partial mediation. Here, the method developed by Goodman (1960) and recommended by Baron and Kenny (1986) is used.14 The conclusion of the test is that the relationship between subjective norm and the behavior of underreporting income on tax returns in only partially mediated by the intention toward the behavior.
COMPARING RESULTS: REGRESSION VERSUS SEM The preceding regression and SEM analyses differ in several ways, but the most obvious is the partial versus complete mediation of the subjective norm–behavior relationship through intention. The regression analysis suggested this relationship was partially mediated, but the SEM analysis supported the complete mediation of this relationship. Unfortunately, a researcher is likely to only perform one of these analyses, either the regression or the SEM. For example, had a researcher used only the regression analysis, the conclusion that TRA is only partially applicable to the behavior of underreporting income on tax returns might be suggested and published. Further, modifications to TRA regarding the relationship between subjective norm and behavior might be hypothesized. Possible reasons for the inapplicability of TRA to this scenario might be suggested and debated in the literature. However, further analysis indicates that the conclusions from the SEM are more appropriate in this scenario because
Why You should Consider SEM
207
the SEM model accounts for the imperfect reliability of the latent constructs. So why do the results from the two analyses differ? Primarily because of the way they account for the measurement error in the latent constructs. A regression analysis with averaged questionnaire items or factor scores as variables, implicitly assumes that the variables are measured without error, or have perfect reliability. In contrast, an SEM model explicitly assumes that the latent constructs are measured with error or that they are not perfectly reliable. The error terms for the item indicators in the SEM model separate the portion of variability in the items that is due to measurement error, and the remaining variability is allocated to the latent construct. The use of the average of multiple indicators in a regression allows no way for the measurement error to be separated from the variable of interest. Thus, viewing the relationships among latent constructs in a regression is similar to viewing the coefficients through a cloudy window, with vision blurred by the unaccounted measurement error. However, in SEM the relationships among the latent constructs are crisply viewed because the measurement error has been filtered into the error terms. In the example detailed in the main section of this paper, measurement error was apparent in our latent constructs. Because the reliability coefficients for the latent constructs ranged from 0.72 to 0.84, the regression analysis was clouded by this measurement error. Of specific concern for this model is the measurement error in the mediator (intention), which has an alpha coefficient of 0.75. Baron and Kenny (1986, p. 1177) noted that the presence of measurement error in the mediator tends to produce an overestimation of the effect of the predictor on the outcome. This is exactly what happened in the regression analysis as the relationship between subjective norm and behavior remained statistically significant when intention was added to Model 6 (see Table 4), and the test for partial mediation supported this link.
This page intentionally left blank
208
A POSTMODERN STAKEHOLDER ANALYSIS OF TELEWORK Anita Reed, James E. Hunton and Carolyn Strand Norman ABSTRACT Telework is becoming a viable and appealing work option in the accounting profession (Hunton, 2005). Many accounting firms have implemented telework arrangements to provide flexibility and support for employees who seek an acceptable balance between career and family. This form of work also supports business sustainability in the event of acts of terrorism or natural disasters. Increased reliance on various forms of telework gives rise to questions of appropriate ethical treatment of affected workers. The objectives of the present study are to examine the ethical implications of telework and identify policies for telework that might help organizations implement this type of work arrangement for their employees in an ethically informed manner. Our analysis draws upon a framework proposed by Yuthas and Dillard (1999) that combines postmodern ethics with stakeholder theory. Although this framework was developed to study the ethical design of information technology systems, we maintain that this structure is equally useful to study the ethical issues inherent with telework. Legislators, regulators, unions, and employers can use the telework policy considerations presented herein as guidelines as they deliberate, design, and implement ethical telework strategies.
Advances in Accounting Behavioral Research, Volume 9, 209–235 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 1475-1488/doi:10.1016/S1475-1488(06)09008-9
209
210
ANITA REED ET AL.
INTRODUCTION Telework arrangements include any kind of work that is accomplished remotely from the employer (e.g., homes, satellite offices and call centers) on either a full- or part-time basis, often involving electronic information processing, but always using telecommunications (Gray, Hodson, & Gordon, 1993). Telework can be performed individually or in geographically distributed work teams, thereby creating virtual, ubiquitous business networks on national and international scales. Based on a survey conducted in August 2005, the Telework Advisory Group (ITAC)1 estimates that approximately 45.1 million US workers over the age of 18 are involved in some form of telework arrangements (ITAC, 2005). A similar study from the British Government2 reports the number of teleworkers in the UK at more than 2.4 million. Lomo-Dvid and Griffin’s (2001) survey results of business majors from 22 universities indicate that more than half (52 percent) would be interested in telecommuting upon graduation. Thus, telework is a growing phenomenon spanning nearly all industry sectors and affecting an array of organizational stakeholders. The impetus for telework work arrangements is compelling. Terrorist actions threaten the safety of employees, especially those concentrated in densely populated cities and high-rise office buildings. The number and severity of natural disasters makes business sustainability a growing challenge. The cost of gasoline and lengthy commutes are frustrating to workers, and the resulting pollution is an environmental concern. At the same time, legislators are demanding changes in the working environment. For example, House Representative Frank Wolf, R-VA, sponsored a bill that would force five Federal agencies to prove that the number of teleworkers in their agency is increasing or give up $5 million in funding (Pullium, 2005). The growing percentage of workers in various telework arrangements, coupled with recent, comprehensive studies of telework conditions (e.g., Hunton, 2005; Tietze, 2005; Tietze, 2002; West, 2002; Atkyns, Blazek, & Roitz, 2002), suggest a number of potential concerns for employees, employers, and other stakeholders. The purpose of this paper is to examine one of those concerns – the ethical ramifications of telework on affected employees. This issue is particularly relevant to the accounting profession as CPA firms craft a variety of flexible work arrangements to accommodate current employees and provide incentives to compete for the best-qualified future employees. To address this topic, the telework analysis herein draws upon the framework proposed by Yuthas and Dillard (1999) combining postmodern ethics
A Postmodern Stakeholder Analysis of Telework
211
with stakeholder theory. Although this framework was originally developed to study the ethical design of advanced information technology systems, this structure is equally useful to study the ethical propositions that are inherent in the diversity of telework arrangements. Thus, the objectives of the present study are to examine the ethical implications of telework and identify policies for telework that might be mutually beneficial to the stakeholders. Researchers can use the framework presented here to develop an ethical theory of telework; legislators, regulators, and unions can draw upon the implications of this paper to ensure that managers embed ethical considerations into their telework policies; and employers can incorporate the policy considerations suggested in this study into their telework arrangements. The remainder of the paper proceeds as follows. In the next section, we briefly review the literature on telework and various ethical concerns regarding telework. Afterward, we identify the postmodern stakeholder framework advanced by Yuthas and Dillard (1999). The succeeding section uses their framework to analyze a number of reported studies of telework in an attempt to identify the ethical implications on various stakeholders. Lastly, we conclude with a discussion of policy considerations for telework.
LITERATURE REVIEW Telework The 2005 ITAC survey identifies three factors leading to the growth of telework: (1) the Internet, (2) enabling technologies and telecommunications that link these technologies, and (3) employees who realize that telework might be a viable option (ITAC, 2005). The 45.1 million participants in the 2005 ITAC telework study indicated that they work from an average of 3.4 locations and that 60 percent more of them use broadband Internet connections to accomplish their work (ITAC, 2005). Hunton (2005) notes that advanced technologies provide the motivation for an increasing number of accountants to consider telework options or flexible work arrangements. Phelan (2002) illustrates how CPA firms use technology to extend the firm, includes a ‘‘getting started’’ checklist for CPA firms, and identifies a number of telework resources such as organizations, books, and software. A number of studies identify potential advantages and disadvantages of telework arrangements from the perspective of the employee, the employer, and society (e.g., Anonymous, 2003; Montreuil & Lippel, 2003; Daniels, Lamond, & Standen 2001; Fairweather, 1999). These advantages and
212
ANITA REED ET AL.
disadvantages are summarized in Table 1. From the individual’s perspective, advantages might include the opportunity for training and advancement because teleworkers must become adept at using information and communication technologies; additionally, telework can reduce personal anxiety and stress related to the process of preparing for and winding down from the external work environment. Yet, the isolation of workers from the organization can lead to a potential loss of cultural identity. From the organizational perspective, the advantages of offering telework to employees could include greater worker productivity, increased employee retention, enhanced staffing flexibility, reduced real estate costs, and improved business continuity. These advantages are offset by a number of possible disadvantages, which include heightened supervisory challenges, increased employee and technology costs, higher development costs for new performance measures, greater computer and network security issues, and increased safety concerns of employees working at alternate locations. As depicted in Table 1, there are also a number of advantages and disadvantages that might accrue to society when employees engage in telework arrangements. The advantages and disadvantages of telework, identified in Table 1, underscore conflicting interests among individuals, organizations and society, and provide insight into areas of concern. For example, a recent survey of British Telecom (BT) employees, one of Europe’s leading telecommunication companies, reports that 73 percent of respondents feel their work-life balance is good or very good (Anonymous, 2003); however, 69 percent said that their working hours increased, with almost half reporting working more than 15 hours extra each week. While BT would probably view this as an advantage, this apparent inconsistency could eventually lead to stress and dissatisfaction on the part of the employee. Labor unions, regulating bodies, and academic studies have already voiced concerns about employees in telework arrangements (Montreuil & Lippel, 2003; Wade, 1999; Fairweather, 1999), thereby indicating a perception that potential conflicts of interest exist and should be addressed. Telework presents a paradox – while perceived advantages exist for individuals, organizations, and society, adoption of telework simultaneously creates potential conflicts within and among the affected stakeholders. Bailey and Kurland’s (2002) review of more than 80 telework research studies goes beyond the simple identification of advantages and disadvantages to answer such questions as who participates in telework, why they participate, and what happens when they participate. The authors conclude that the telework population appears to be divided according to occupation
213
A Postmodern Stakeholder Analysis of Telework
Table 1. Potential Advantages and Disadvantages of Telework Arrangements. Constituents
Advantages
Individuals
1. Opportunity to work, despite family or physical limitations 2. More time for home/family 3. Reduced commuting 4. Greater job autonomy 5. Fewer interruptions 6. Flexible working hours
Organizations
1. 2. 3. 4. 5.
Society
1. Reduced pollution/urban congestion 2. Promote entrepreneurial activity
Increased productivity Increased employee retention Greater staffing flexibility Reduced real estate costs Greater resilience to disruption from natural disasters
Source: Adapted from Daniels et al. (2001).
Disadvantages 1.
Fewer chances for development/ promotion 2. Perception of not being valued by managers 3. Increased conflict between work/home 4. Limited contact with colleagues/ social isolation 5. Routine tasks 6. More time spent working 7. Lower job security 8. Weakened collective representation 9. Loss of privacy 10. Injury due to inadequate safety 1. Difficulty supervising/motivating employees 2. Increased selection, training, and support costs 3. Difficulty socializing new employees to the organization 4. Cost to develop planning/ performance measures 5. Technology costs 6. Computer security/unprotected networks 7. Cost of safety issues/OSHA regulations 1. Lost creativity due to withdrawal from group 2. Community instability created by call centers that can disrupt existing culture 3. Lost jobs due to outsourcing/ offshoring 4. Economic impact to related industries, e.g., gas stations, restaurants, etc.
214
ANITA REED ET AL.
and gender, with the professional group composed mostly of men and the clerical group primarily women. Factors that best predict who will choose to participate in telework include level of personal discipline, teamwork vs. individual work preference, self-perceived job suitability, tendency to overwork, availability of technology, manager’s willingness, family focus, and household distractions. Regarding why individuals choose to telework, Bailey and Kurland (2002) report that neither the time to commute nor the distance of the commute will predict an individual’s decision to telework or the frequency of telework. Further, work-life balance is not a primary motivator. Although women are more likely to identify work-life balance as a primary issue, they do not dominate telework populations. From the organizational perspective, managers express concerns about controlling employees and implementing costs. Finally, the outcomes most frequently identified (or what happens when people telework) usually include improved productivity, organizational loyalty, job satisfaction, and employee retention. However, Bailey and Kurland (2002) maintain that accounts of increased productivity are based primarily on self-reported data and that little clear evidence exists for the claims of increased job satisfaction. Belanger and Collins (1998) propose a framework to study distributed work arrangements that considers the influence of organizational, individual, work and technology characteristics on societal, organizational, and individual outcomes. The authors note that successful outcomes for one stakeholder (such as the individual) may be in conflict with successful outcomes for other stakeholders (such as other individuals, the organization, or society). These conflicts suggest that ethical dilemmas exist in the decision to adopt telework arrangements. Ethical Considerations The ethical aspects of telework are interesting to researchers in a number of disciplines. For example, Moon and Stanworth (1997) evaluate telework from both a teleological and a deontological approach, giving voice to the rights as well as the duties of employees and employers. The teleological approach is viewed as a cost–benefit analysis that places more emphasis on short-term (vs. long-term) consequences and ignores certain fundamental social rights of teleworkers. The deontological approach is viewed as an evaluation of the duties and obligations of the employer toward the employee. The authors suggest that employers are more interested in the cost effectiveness of telework than they are in developing employees, in
A Postmodern Stakeholder Analysis of Telework
215
essence, favoring the teleological approach over the deontological approach. Accordingly, the authors recommend adoption of an agreement between employers and teleworkers that clarifies the rights and responsibilities of each. Guthrie (1997) examines the ethics of telework by surveying the perceptions of 30 business people regarding a series of 18 ethical scenarios that cover such topics as work ethic, compensation equity, access to resources, work and family issues, and privacy and monitoring. She focuses on four previously identified issues of telework: privacy, access, property, and accuracy. The results indicate a diversity of reactions to these topics, which suggests little agreement regarding what constitutes appropriate ethical behavior for teleworkers and their managers. She predicts that, over time, attitudes and norms about telecommuting will converge and that posing ethical conflicts is an effective way to challenge existing beliefs about telework. Observing and controlling employee’s work is not a new concept; however, the capability of using technology to monitor and collect detailed information on work performance is new and expanding. Wilson (1995) notes that advanced computer technology coupled with performance-shadowing policies allows managers to closely monitor, reward, and discipline employees. Fairweather (1999) identifies a number of surveillance software solutions, known as electronic performance monitoring systems, which managers use to constantly monitor employees’ work. Employers’ incentives to use surveillance techniques to monitor teleworkers are similar to their motivations to track the performance of traditional office workers – maintain control over work output, document the basis for performance-based rewards and punishments, and enforce system security. Fairweather also points out the major downside of employee monitoring – potential invasion of privacy. Aside from legal and ethical implications, such invasion could create fear and resentment on the part of the teleworker, thereby affecting morale and productivity. Feminist theory views the boundary between public and private as a division between what is valued and what is not valued: public (masculine) is valued; private (feminine) is not valued. Mirchandani (1999) evaluates telework from this public–private dichotomy and finds evidence of masculine values having favor over feminine values. She also discovers that men and women are more likely to devalue their non-work activities in order to give legitimacy to their work activities. Further, women tend to express more concerns that their work will be devalued if it is commingled with non-work, and this may act as a barrier to women (and minorities) from entering into
216
ANITA REED ET AL.
or remaining in telework arrangements. Both men and women express difficulty in protecting the boundary between home-life and work-life, and indicate that it is (too) easy to overwork when they are at home. These findings provide insight into the gap between perceived benefits of telework and the realities of its implementation. Raines and Leathers (2001) express concern that telecommuting, and the widespread use of technology to conduct business away from the traditional workplace, might significantly affect production and social institutions. The authors suggest that telework might have profound and permanent effects on many of the social relations which exist in the US economy. It has the potential to change the arrangement of child care and educational institutions, revolutionize family relationships, radically alter the wage bargain, shift the distribution of income to the technologically literate, affect marital relations, and foster a social consciousness that is centered in individual independence and freedom. (p. 307)
The authors note that economic benefits are certainly important drivers that encourage widespread use of technology. However, management must sort out circumstances that give rise to the most efficient use of telework conditions. For example, employees must be computer literate, self-disciplined, and able to work independently. Under these conditions, workers should be more productive working at home and managers might be able to justify the costs of implementing telecommuting programs. Other studies address potential occupational health issues and risks that may be associated with telework. For example, Montreuil and Lippel (2003) focus on such issues as schedules and absences, layout of the work equipment in homes, musculoskeletal problems that can occur with frequent computer use, and psychosomatic stress. Almost all of the teleworkers in their study claim to work more hours, and many refuse to take sick leave when warranted to convince employers to continue the telework arrangement. While this work ethic is commendable, increased and prolonged use of the computer can give rise to increased musculoskeletal problems if workers do not have the self-discipline to take necessary breaks. In addition, while some teleworkers report less stress due to the perception of more control over their work and family life, others report increased stress due to the omnipresence of work. The preceding articles provide valuable insight into a number of the conflicts and ethical issues associated with telework. Although telework and the technologies that enable telecommuting are ethically neutral, the method of implementation can produce ethical biases; thus, telework is examined through the lens of the postmodern stakeholder perspective developed by
A Postmodern Stakeholder Analysis of Telework
217
Yuthas and Dillard (1999). We first describe their ethical reasoning framework and then apply the framework’s principles to a number of telework case studies as a way to analyze the impact of telework on the various stakeholders.
POSTMODERN STAKEHOLDER PERSPECTIVE Although the extant literature on postmodern ethics is rich and complex, this section is not intended to be a comprehensive review of the literature. Rather, our goal is to select a framework from this body of knowledge that might provide a better understanding of the ethical complexities that are an integral part of telework arrangements and offer useful policy considerations for the stakeholders involved. The Yuthas and Dillard (1999) Framework Yuthas and Dillard (1999) claim that postmodernism has become one of the most influential intellectual movements in the social sciences and that this enlightened perspective is particularly relevant as a lens for examining business ethics. The authors define postmodernism by comparing it to modernism. Postmodernists reject the modernist view that universal theories or laws can be developed based solely on reason and rationality, independent of the values and beliefs of the human agents. Instead, postmodernists embrace a view that theories and laws are embedded in historical and cultural contexts, interpreted through the beliefs, values, and attitudes of the human agent; thus, universal theories and laws cannot be constructed. A key element of the postmodernist view is that different subjective positions cannot be objectively compared; instead, each must be examined within the bounds of its historical and cultural complexities and local variation. Focusing specifically on postmodern ethics, Yuthas and Dillard (1999) rely on the work of Baumann (1993), perhaps the most renowned theorist in postmodern ethics. According to Bauman, modern organizations and institutions design moral codes and rules to improve human morality, but these codes have just the opposite effect because they usurp the moral responsibility of individuals. Bauman recommends abolishing these rules and returning moral responsibility (choices) to individuals. Bauman concedes that this action might cause some individuals to lose a sense of security about what they should do in a particular situation; but, in so doing, they would be forced to accept moral responsibility for their decisions. In the
218
ANITA REED ET AL.
words of Gustafson (2000, p. 645), ‘‘Ultimately, what postmodernism has to offer business is not rules, but questions that raise issues of responsibility.’’ Yuthas and Dillard (1999) identify conditions and related principles of postmodern ethics to analyze advanced information technology systems, and then combine this with stakeholder theory to develop the postmodern stakeholder enabling perspective. This view ‘‘recognizes the need for specifying processes directed toward arriving at shared values, as well as gaining consent and consensus among various interest groups’’ (p. 42). The present study adapts this analysis to telework arrangements, identifying conditions and principles that are related to the stakeholders involved in flexible work arrangements (see Table 2). The postmodern stakeholder perspective is typically presented as a sequential process. However, Yuthas and Dillard (1999) believe that different stages of the process are interrelated and interactive and should be viewed as a collaborative effort on the part of all stakeholders as events unfold and is described as follows: y the goal of the process is to break down traditional principles of systems development in order to allow for a process guided by the moral impulses of individuals leading to shared empathy and understanding y which leads to solidarity among various stakeholder groups y to identify moral issues and pursue projects that can be accepted by, and benefit to, the stakeholders as a group of mutually concerned, rather than competing, individuals (p. 43).
Analysis of Telework Using the Postmodern Stakeholder Perspective This section examines telework from two points of view and is based on the Yuthas and Dillard (1999) framework. First, a hypothetical example is proposed to illustrate a situation in which a company might use the postmodern stakeholder perspective to implement telework or flexible work arrangements for employees. Second, using the Hunton (2005) and Tietze (2002) case studies of telework, we examine whether the stakeholders considered each of the six principles of the framework when implementing telework arrangements (see Table 2). As our hypothetical example, consider the Monolith Company. This firm recently implemented a telework program for employees. The company has an employee structure that includes about 150 accounting personnel at its corporate headquarters and approximately 50 (or one-third) of these accounting personnel are teleworkers. Over time, owing to regional concentrations of sales and service activities, Monolith hired about 50 additional accounting personnel as teleworkers throughout the US. All of these teleworkers perform essentially the same work (A/R, A/P, inventory).
219
A Postmodern Stakeholder Analysis of Telework
Table 2.
Postmodern Ethics and Stakeholder-Oriented Telework Arrangements.
Stakeholder Enabling Approach Postmodern ethics Identify telework stakeholders
Form a telework stakeholder committee
Provide mechanism for stakeholder discourse
Identify and document stakeholder issues
Select appropriate telework arrangements
Postmodern Condition
Power consists in silencing issues – moral response cannot be exhibited with regard to a situation of which one is ignorant; authoritative agents create power by maintaining ignorance of moral issues; not allowing certain issues to be heard World of strangers – primary moral obligations are felt toward oneself and toward the individuals with whom one is familiar, such as family members or other intimates; no moral obligation is felt toward others Absence of solidarity – individuals care about and have moral impulses toward those individuals with whom they identify most closely and intimately and consider to be ‘‘part of us’’ A multitude of issues – there are too many issues to allow for each issue to receive proper moral response with little incentive for moral action, resulting in desensitization and anesthetization to the suffering of others No universal principles – no set of rules or universal moral code dictates an appropriate response to each moral dilemma encountered in life; ambiguity and uncertainty require unique response to each moral dilemma based on moral responsibility
Principles of Affirmative Postmodern Ethics
Exposing hierarchical, paternalistic power relationships through a recognition of a plurality of interests
Developing local opportunities for communication and collective action
Developing trust through relations jointly created and periodically negotiated
Shared appreciation of interdependence in a negotiated context
Action based on locally negotiated criteria
220
ANITA REED ET AL.
Table 2. (Continued ) Stakeholder Enabling Approach Monitor telework arrangement outcomes
Postmodern Condition
Unanticipated consequences – every action has unpredictable, unexpected outcomes and consequences that must be visualized and dealt with cautiously and responsibly
Principles of Affirmative Postmodern Ethics Each participant is exposed to the uncertainties and responsibilities surrounding purposeful action
Source: Adapted from Yuthas and Dillard (1999).
Several issues have arisen recently that Monolith must address, for instance – teleworkers feel they are not being treated equally in all respects when compared to on-site employees (as they were assured when hired); some on-site employees are requesting telework arrangements; managers and supervisors are concerned about adequate supervision of off-site employees; the human resource function is concerned with the safety of off-site workers; the information technology (IT) function is worried about the security of company data and equipment; and the CEO wants US telework policies to serve as a model for global implementation. Monolith faces the dilemma of trying to develop internal policies and structures related to telework that adequately address these issues as well as other potential future issues. At first glance, some of these issues do not appear to contain ethical elements. For instance, the IT function’s concerns about data and equipment safety could be resolved by establishing standards for the equipment and communications security. However, since employees would use IT equipment in a home environment, how can the CIO ensure that the equipment is being used properly and only as intended? Should an employee’s home be inspected, should the equipment be physically and logically secured when not in use, what mechanisms should be implemented to prevent children, spouses or others from using the equipment? Many questions require answers. Also, treating teleworkers equally with on-site workers would seemingly involve no more than giving them the same job classification, pay rates and benefits. But how do informal networking, personal relationships, and face-time affect performance evaluations, potential for promotion, and other rewards? Many of these issues are not amenable to universal standards; thus, by using the postmodern stakeholder perspective, Monolith has an opportunity
A Postmodern Stakeholder Analysis of Telework
221
to create policies that address both the face and substance of these issues. These concerns are examined in the framework provided by Yuthas and Dillard (1999); and, for clarity and guidance, the subheadings displayed in Table 2 are used below. Identify Telework Stakeholders The first step illustrated in Table 2 is to identify stakeholders. Yuthas and Dillard (1999) caution decision makers to consider the postmodern condition related to the power of management to silence certain moral issues by selectively ignoring their existence. The authors claim that an inclusive approach must be exercised in selecting stakeholders to allow a forum for all issues, not just those related to productivity or profitability. Dillard (2002) describes the extremes of stakeholder inclusivity as a spectrum. At one end is a very traditional, paternalistic level of inclusion that recognizes the interests of stakeholders other than managers/shareholders only to ‘‘the extent that their participation is instrumental in meeting shareholder objectives’’ (p. 185). At the other end of the spectrum, a ‘‘broad variety of stakeholders are included in organizational decision-making processes’’ (p. 185). The stakeholder enabling approach implies not only identifying stakeholders, but also including them in the decision-making process. The inclusion of the stakeholders must be done faithfully, guarding against the manipulation of the stakeholders merely to further organizational goals (Yuthas and Dillard, 1999). Owing to the distributed nature of telework, identifying potential stakeholders requires taking both a local and a global perspective. This may require the hypothetical company, Monolith, to institute new processes for identifying affected parties. The company may need to develop relationships with political and cultural leaders or consultants in the various locations where telework arrangements are implemented or planned. Although these other stakeholders may not be directly involved in the organization or the telework arrangement, they may have a political or social interest. Monolith needs to be careful, diligent, and inclusive in their identification of stakeholders. In Hunton’s (2005) study, 160 medical coders claimed that their work (primarily clerical) was suitable for telework and demanded the opportunity to work from home. The workers were so determined to achieve this result that they threatened to unionize. Accordingly, the CEO and Board of Directors for the health care company agreed to a six-month test of several different types of telework arrangements, but did not wish to commit to one course of action until they knew more about potential longterm effects of medical coders working exclusively from home. In this case,
222
ANITA REED ET AL.
the organization could not use its power to silence the coders because of the risk that the employees would seek union protection. Although the medical coders believed they were legitimate stakeholders in this discourse, management was reluctant to acknowledge them as such. Other studies (Tietze, 2005; Montreuil & Lippel, 2003; Phelan, 2002; Tietze, 2002) describe the experiences of high-salaried, professional employees who participate in telework. Tietze (2002) investigated the coping strategies of 25 managers who were college graduates. This study examined the impact of telework on family members, documenting the changes in social processes that occurred in the household as the teleworker constructed his/her ‘‘work’’ identity in the ‘‘home’’ environment. The important point, according to Tietze, was that the meeting and coexistence of the work discourse with the home discourse caused a degree of ambiguity and uncertainty about the traditional roles of the individuals involved. Thus, in the circumstances of the professionals, the stakeholders included not only the supervisor and the teleworkers, but also the other members of the household where the teleworker would be conducting business. Comparing the Hunton (2005) study with the Tietze (2002) study, it is apparent that the teleworkers in each study were treated differently based on their relative standing within the corporate structure. In the former study, the medical coders believed they had no voice in the decision to telework and therefore threatened the management of the health care company with unionization. As a result, management was forced to view these employees as legitimate stakeholders and include them in the decision process. On the other hand, the employees in the Tietze (2002) study were highly educated managers who were encouraged to develop unique telework arrangements that fit their personal needs. That is, based on the intellectual capital that these managers possessed, they negotiated whatever schedule they desired with their respective supervisors, and then were required to acknowledge (and negotiate with) their family members as stakeholders in the work environment (the home). Although the Tietze study describes in more detail the impact of the teleworker in the home environment, the medical coders in Hunton’s study also encountered a number of similar challenges in their telework experience. Form a Telework Stakeholder Committee Once potential stakeholder groups are identified, the next step is to select representatives of each stakeholder group to form stakeholder committees. Essential to the formation of these committees is the recognition of the postmodern condition related to the need to develop a sense of mutual
A Postmodern Stakeholder Analysis of Telework
223
moral obligation among the stakeholders through shared knowledge and reciprocal understanding of one another. As Monolith implements a telework arrangement, it should establish a stakeholder committee and bear in mind that the committee may act in several capacities, such as overseeing the development of telework policies, providing a structure that provides legitimacy to the stakeholder groups, and promoting communication, shared understanding, and recognition of commonalities among the diverse stakeholders. The medical coders in Hunton’s (2005) study were included as stakeholders in early focus groups. However, management representatives from the health care organization and the researcher decided the combination of possible telework arrangements, and coders were randomly ‘‘assigned’’ to a particular arrangement based on the location of the hospital (four metropolitan areas). A fifth location/hospital was used as a control group where no telework was allowed, which mirrored the current work policy. The arrangements included downtown office space (yes, no) and satellite office space (yes, no), and in all cases the coders could work exclusively at home, should they choose to do so. In contrast, as noted in the previous section, the employees in Tietze’s (2002) study were high-salaried employees at the professional core of their respective organizations, and as such, had a great deal of latitude in how they constructed their telework experience. Drawing on fundamentals from organizational behavior, employee commitment occurs when employees are allowed to actively participate in decision-making and to directly influence the duties of their job (Fracaro, 2005). Since the medical coders did not have this sort of control over their jobs, the employees in the first study would likely have benefited from the formation of a telework stakeholder committee, as they would have formed the perception (and hopefully the reality) of actively participating in the particulars of their telework arrangements. The organization could also benefit from the formation of a stakeholder committee since employee commitment is required to successfully meet organizational goals, plans, and objectives (Fracaro, 2005). In contrast, the managers in Tietze’s (2002) study enjoyed a significant amount of autonomy and were allowed to determine what days and how often they wished to telecommute. These individuals negotiated their own telework circumstances with their respective organizations; thus, they did not need the structure of a telework stakeholder committee. Note a disturbing difference between these two cases; that is, in the Hunton (2005) study, the affected teleworkers were clerical employees, while in the Tietze (2002) study, the teleworkers were high-paid professionals. The
224
ANITA REED ET AL.
way management treated the two groups suggests that the opinions, concerns, and opinions of clerical workers were devalued relative to professional employees. One of the authors herein interviewed managers of the health care organization involved in the Hunton (2005) study. There was a clear indication that the managers felt insecure in turning over any process- or decision-control to lower level employees, for such employees could not possibly have better ideas than the ‘‘all-knowing’’ managers. It is precisely this type of situation about which Yuthas and Dillard (1999) discuss, where managers selectively ignore certain stakeholders, especially stakeholder groups that might not be solely focused on organizational profits and growth. Such self-selection behavior by management reflects an area of caution for organizations that are considering developing and implementing telework policies; if all key-affected stakeholders are not identified, either intentionally or unintentionally, the policy setting process is seriously flawed from the onset. Provide Mechanisms for Stakeholder Discourse Stakeholder committees can also help develop solidarity among the diverse stakeholders. These committees can provide an important conduit of contact between authoritative individuals and diverse stakeholders, facilitating detailed description and interaction that leads to recognition of the validity of the needs and interests of all parties. This encourages a sense of familiarity and promotes caring attitudes that might lead to mutual moral responsibility and collective decision-making. Crowther and Hosking (2005) would call this ‘‘relational constructionism’’ where the knower and the known (self and other) are co-constructed. To encourage relational constructionism, committees should use mechanisms that allow open discussion and develop trust. A variety of communication mechanisms are suggested in the literature. Yuthas and Dillard (1999) recommend conditions that allow members equal access to the discussion, and Guthrie (1997) suggests the use of scenarios that are embedded with ethical dilemmas to challenge individuals to think about the issues. Monolith may find that it needs to employ multiple mechanisms to accommodate the different cultural and organizational contexts of telework, some of which may entail the use of electronic media due to temporal and geographic distance among the various stakeholders. Hunton’s (2005) study used the experience-sampling method (ESM) to collect data at multiple, random times during each day to obtain activities and record subjective experiences from each medical coder. This data collection technique allowed the researcher (and the administrators of the
A Postmodern Stakeholder Analysis of Telework
225
health care organization) to examine a considerable amount of information regarding the behaviors, attitudes, and beliefs of the employees (e.g., work/ home-life balance, satisfaction of personal responsibilities, and ability to concentrate on tasks). Note, however, that this was a one-way communication, where the employees were not granted the option to talk with or hear from the health care administrators. Adapting an illustration from Crowther and Hosking (2005) to fit this context, the researcher and the administrators constructed themselves as the ‘‘knowing’’ architects of possible telework arrangements and ‘‘others’’ (the medical coders) as uninformed, which limited the perceived need for input as well as what might be achieved through mutual dialogue. Furthermore, management had the power to continue collecting data from the medical coders regardless of the type of flexible work arrangements that were institutionalized. Wilson (1995) maintains that this sort of invisible surveillance of employees functions as a form of control – to monitor the quality and quantity of work performance, work behaviors, and compliance with instructions. In Wilson’s view, this asymmetrical power relationship will be most effective when employees believe they are participating and are able to develop group norms that regulate acceptable behavior. In contrast, the ‘‘stakeholder committees’’ in Tietze’s (2002) study were the teleworkers (professional managers) and their families. As discussed earlier, the purpose of these committees is to facilitate interaction that leads to recognition of the validity of the needs and interests of all parties, which promotes caring attitudes that might lead to mutual moral responsibility and collective decision-making (between the teleworker and family members). Thus, the teleworkers in Tietze’s study were the authoritative individuals. In this capacity, they were the ones who needed to recognize the desires and interests of the other household members and provide the opportunity for discussion. For example, Tietze described the mechanism for stakeholder discourse used by Max, one of the 25 managers in her study. Max used a former guest room and kept the door closed when he was working. He used a ‘‘flag system’’ to indicate when he was available to his family (white flag) and when he did not want to be interrupted (red flag). During his established working hours, he treated his family ‘‘professionally, briefly, courteously.’’ When he did not want to be involved in a family discussion, he said, ‘‘Pretend I’m not here.’’ Max and his family members jointly developed these household rules to facilitate Max’s professional productivity as well as to respect other family members who were sharing Max’s work environment (the home).
226
ANITA REED ET AL.
Identify and Document Stakeholder Issues Stakeholder committees can also act to sensitize various stakeholders to the diversity of needs as well as common threads among the various issues. Individually, stakeholders may feel isolated; however, as they work together, the needs of each stakeholder group are given voice in an atmosphere of mutual regard and moral concern. For example, both managers and teleworkers should be allowed to express their needs and address methods for accomplishing adequate supervision without excessive intrusiveness. Monolith might consider discussions between on-site workers, teleworkers, and managers to allow for greater understanding of the impacts of teleworking on productivity, performance evaluation, and promotion potential as well as the potential impact the teleworker’s absence creates for on-site workers. The stakeholders can then act together to identify issues that reflect the multiplicity of needs and represent the collective interests of each group. Ward and Shabha (2001) encourage teleworkers to become more involved, more proactive, and more engaged in user-initiated change. At the conclusion of Hunton’s (2005) study, debriefing sessions were conducted with the participants (coders) to identify and discuss their attitudes and opinions with respect to telework. Those coders who worked only at home said they experienced ‘‘a great deal of stress and conflict among household members, friends, relatives, and so on – to the point where they were unable to satisfactorily focus on their work tasks or personal activities’’ (p. 133). The coders in this treatment condition (working only from home) had the lowest number of work-related interruptions, the highest number of non-work interruptions, and significantly lower productivity and performance than all other groups in the study, including the control condition. On the other hand, coders who could willingly choose to work at home or a nearby satellite office documented the highest performance, satisfaction, and well-being. Perhaps had managers identified earlier the underlying problematic issues inherent in the ‘‘work at home only’’ group, they could have intervened, stopped the deteriorating situation, and implemented satellite offices for the medical coders. This did not happen because the coders were not included in a stakeholder committee and the data collection effort was a one-way avenue. The managers in Tietze’s (2002) study also acknowledged that they encountered some troublesome issues. For example, Max (discussed in the previous section) felt obligated to telework one to two days per week to support his employer’s efforts to expand this program in the organization. However, Max indicated that he would soon stop teleworking and return to full-time work at the corporate office for two reasons. First, he preferred the
A Postmodern Stakeholder Analysis of Telework
227
mental stimulation and social interactions of the workplace; second, he needed to prepare for a career move in the near future and felt that he would improve his chances by being more visible. Another manager in Tietze’s study made the opposite decision – to continue to telework and forego a promotion opportunity. While this manager (Tom) also missed the social interactions of his colleagues at the office, he instead decided to become involved in his local community. Based on the disparate group of teleworkers in each study, medical coders in the Hunton (2005) study and professional managers in the Tietze (2002) study, very different issues and concerns were recognized by each. In Hunton’s study, the medical coders identified a number of salient topics that helped management determine the most effective and efficient telework arrangements that would be used by the firm in the future. Identification and documentation of such issues should be acknowledged by the medical coders as a manifestation of management’s genuine concern for their wellbeing as well as an opportunity to impact their job situation, which should provide the type of organizational climate that encourages commitment on the part of the employees. Thus, sharing of information through the mechanism of a stakeholder committee can be mutually beneficial for employees and employers. The telework issues identified by the professional managers in Tietze’s study would most likely be discussed and resolved between the managers and their individual supervisors. In those cases where managers wanted to change their circumstances, they were in a position to make the change. Thus, identifying and documenting these concerns would most likely benefit supervisors throughout organizations that have implemented telework policies or those who are considering the possibility of such work arrangements. Such ‘‘corporate memory’’ could be useful for future negotiations between managers and supervisors as they attempted to craft work arrangements to avoid any particular problems in the future. Select Appropriate Telework Arrangements In the previous section, the issues that were identified by each of the two telework groups were very different. Similarly, no single set of policies and structures will be effective or appropriate in all contexts. Within the stakeholder enabling approach, the inclusion of diverse groups expands the scope of experience and perspectives, thereby contributing to development of appropriate policies based on localized context and values. While implementing the telework program, Monolith should establish policies to ensure that various standards for IT security and human resources are met, but also
228
ANITA REED ET AL.
allow each location to set the parameters deemed necessary for the successful implementation of telework subject to the unique conditions of the location. The overall results from Hunton’s (2005) study suggested that working exclusively at home (for medical coders) was not a good policy, because of the personal nature and lengthy duration of interruptions that occurred. Specifically, the participants in that study who had choices, such as working at home or a satellite location or downtown at the hospital, performed significantly better (in terms of quality and quantity) than the control group or the ‘‘home only’’ group. The managers in Tietze’s (2002) study had complete autonomy and could choose what projects they worked on and which days to work at home. They could also decide to continue to telework or return to the traditional office environment – the terms and conditions of which were negotiated individually by each manager and his or her supervisor. Taken together, these two studies of telework suggest important principles for organizations that are considering alternative work arrangements for employees. The type of employees and degree of autonomy to which the employees are accustomed will most likely determine the nature of the arrangement, degree of monitoring, flexibility of telework, and specific terms and conditions. Management of organizations that are contemplating telework for employees who are classified as routine workers (such as the medical coders) should consider the possibility of a limited test period to try various options prior to implementing policy changes, since the policy would presumably be uniformly administered to an entire group of employees. However, as the test period unfolds, it is imperative that management gather subjective input from the teleworkers, respond swiftly to feedback, and respect the opinions and feelings of the employees. Alternatively, for employees who are highly educated and normally enjoy a great deal of autonomy, the work arrangements will most likely be negotiated individually between the employee and the supervisor, thus reflecting the unique needs or desires of each teleworker. Monitor Telework Arrangement Outcomes In the spirit of continuous improvement, telework policies should be examined on a regular basis, as such quality assessment processes can be used to quickly identify any unforeseen, adverse impacts on individuals, organizations, and society. As Ward and Shabha’s (2001) findings suggest, individuals respond differently to various aspects of telework situations, such as social isolation, family interaction, and time management. Accordingly,
A Postmodern Stakeholder Analysis of Telework
229
Monolith should encourage its stakeholder committees to periodically monitor localized contexts and conditions of various telework implementations for environmental and cultural changes, and make adjustments to the policies and structures as conditions evolve. Organizational policies must make provision for necessary remedial actions, such as the cessation of telework or reversion to less distributive arrangements, to accommodate developing circumstances for the teleworker, organization, or local community. Hunton (2005) reported that management at the health care organization fully examined the wide variety of information obtained from the study. Based on: (1) all known costs of implementing, (2) employee performance and retention, (3) employees’ sense of personal achievement, and (4) managers’ desire to retain some downtown office space for employees, the company implemented the option that gave all employees at all hospitals and clinics the widest choice of locations from which to work (home, satellite location, downtown office). While this reflected the most liberal telework policy for the employees, management did not institutionalize an on-going policy of feedback from the teleworkers. This oversight suggests that future, unanticipated problems might go unnoticed. The managers in Tietze’s study periodically self-monitored their conditions and self-determined their decision to telework. As mentioned earlier, the managers individually considered their promotion opportunities and some decided to return to the traditional office environment at a strategic point in time, while others decided to give up the opportunity for promotion to maintain the particular lifestyle achieved by the flexibility of telework. The managers who decided to continue telework arrangements admitted that controlling the environment to maintain the boundaries between ‘‘work’’ and ‘‘home’’ was challenging and at times stressful for them and other household members. Nevertheless, they had the power to monitor their own trajectory and make course corrections as time unfolded. As noted in earlier sections, the type of employees and the degree of autonomy to which they are accustomed will likely set the stage for the telework policies management develops with respect to ongoing monitoring. For routine workers, management should consider the need for postimplementation stakeholder committees as a forum for monitoring telework arrangements. For knowledge workers, the respective organizations will most likely empower highly educated managers to self-monitor their situations and consult with their supervisors as the need arises. In the foregoing discussion, we identified the six principles of the postmodern stakeholder perspective, offered general guidelines for Monolith and then illustrated the principles by drawing on two recent studies of tele-
230
ANITA REED ET AL.
work – one of medical coders (Hunton, 2005) and one of managers (Tietze, 2002). While these principles were discussed in the order they are presented in Table 2, the postmodern stakeholder enabling process should be viewed as iterative and interactive, with the opportunity to identify and include additional stakeholders and issues at each step. We believe the combination of the hypothetical case, combined with details from actual studies of telework, provides a rich and fairly comprehensive analysis of the six principles in the Yuthas and Dillard (1999) framework. Policy Considerations for Telework There are many calls for ethical evaluations of business decisions, including the development of telework policies and structures (Hendry, 2001). Utilizing a postmodern stakeholder-enabling perspective provides a broad basis for establishing and evaluating telework policies. Postmodernism also offers an inclusive, supportive environment within which all affected parties can participate in identifying moral issues and developing consensual, mutually beneficial, and acceptable telework structures. Managers who are considering implementing telework for their organizations can develop thoughtful, comprehensive, and dynamic telework programs by using the guidelines discussed herein and summarized in Table 3.
CONCLUSION The postmodern stakeholder enabling perspective on telework articulated in this article is not a prescription for ethical telework arrangements. Instead, the framework presented herein can serve as a guide for developing organizational policies that takes advantage of the moral responsibility inherent in the human condition. Our comparison of the Hunton (2005) and Tietze (2002) studies indicates that a postmodern stakeholder perspective would be most beneficial for clerical worker groups, who are primarily comprised of women employees (Bailey & Kurland, 2002). The major finding from our analysis of telework is that the power–distance relationship between managers who set the telework policies and workers who are affected by such policies profoundly influence the quality of the telework experience. Historically, some organizational managers have maintained an attitude of benevolent patronage with regard to routine workers who are at a relatively low power–distance from upper management. While wellintended, such managers assumed that they knew what was ‘‘good’’ for these
231
A Postmodern Stakeholder Analysis of Telework
Table 3. Policy Considerations for Telework Within the Context of the Postmodern Stakeholder Perspective. Stakeholder Perspective Identify telework stakeholders
Form a telework stakeholder committee
Provide mechanism for stakeholder discourse
Identify and document stakeholder issues
Select appropriate telework arrangements
Policy Consideration Owing to the distributed nature of telework, identifying potential stakeholders requires taking both a local and a global perspective. This may require a company to institute new processes for identifying affected parties. The firm may wish to develop an outreach program to identify potential future employees who are unable to work in a traditional work environment. The firm may need to develop relationships with political and cultural leaders or consultants in various locations where telework arrangements are implemented or planned. The distributed nature of telework may require the formation of multiple stakeholder committees to facilitate face-to-face meetings and encourage the development of mutual moral regard. Members of stakeholder committees should maintain communication with their respective groups, ensuring the full participation of the members. Committees should encourage open discussion and the development of trust. An organization may need to employ multiple methods of communication to accommodate the different cultural and organizational contexts of telework. Developing a sense of solidarity in the stakeholder groups encourages participation in collective decisions. While the selection of issues will vary depending on the location, issues should be formally documented. Documentation provides validity of their importance and offers guidelines for continuous evaluation of telework policies and implementations. It is important to note that the issues identified in one telework context may not be reflective of critical issues for a different telework context, but may serve as a guide for the discovery of issues. Therefore, the documentation of issues in one context should not be viewed as the model for all implementations. Each telework implementation should be treated as a separate situation with its own set of stakeholders, issues, and guidelines. Study results suggest that telework arrangements should be flexible and offer employees as many choices as possible, such as a combination of working at home, or a satellite office, or the main office of the firm, etc. The analysis should take into consideration implementation costs, employees’ productivity, and managers’ desires.
232
ANITA REED ET AL.
Table 3. (Continued ) Stakeholder Perspective Monitor telework arrangement outcomes
Policy Consideration Individuals respond differently to various aspects of telework situations, such as social isolation, family interaction, and time management. Thus, organizational policies must make provision for necessary remedial actions, such as the cessation of telework or reversion to less distributive arrangements to accommodate teleworkers whose needs change.
employees with respect to flexible work arrangements, thereby tending to ignore the workers’ values, concerns, feelings, and opinions. We maintain that managers in the future, especially those who aspire to be among the ‘‘100 Best Companies to Work For,’’ will eagerly seek the opinions of their employees at all levels of the organization. The obvious theme that unites the firms on this prestigious list for 2006, firms such as Sherwin–Williams, Starbucks, IKEA, SAS, Men’s Wearhouse, and others is that they create superior work environments for their employees by nurturing creativity, promoting healthy lifestyles, and enhancing a sense of community by involving employees in the decisions that affect them most (www.fortune.com). Another reason for managers to embrace the need for telework stakeholder committees and develop on-going mechanisms for stakeholders to share information is that employees in the very near future – the millennial generation – will demand that their voices to be heard (Daniels, Norman, & Stewart, 2004). In their profile of the millennial generation, the authors identify a number of characteristics that should be of interest to employers. Based on research and surveys, Daniels et al. (2004) maintain that this generation is expected to be the largest generation in US history, growing to perhaps 100 million; they have been the target of marketing efforts more than any prior generation; they now expect and are rewarded with ‘‘products just for me’’ by manufacturers; and they are very accustomed to adults listening to them and respecting their thoughts and ideas. We suggest that managers seriously consider and fully impound the telework concerns of all affected stakeholders, neither because they want their companies to appear in the Fortune 100 Best Companies, nor because the millennial generation will demand such involvement; rather, in the postmodern era, comprehensive inclusion of all concerned parties is the ethically correct course of action. The complex nature of telework, coupled
A Postmodern Stakeholder Analysis of Telework
233
with a different workforce in the future, requires that each implementation be analyzed within the constraints imposed by and opportunities offered from its stakeholders, location and context. Many resources are available to organizations that might choose to implement or expand the use of telework, including information on various websites such as ITAC, government agencies, and unions.3 A constructive first step in the process is to become as well informed as possible. Telework researchers can use the postmodern stakeholder policy consideration framework (Table 3) to develop an ethical theory of telework. Legislators, regulators, and unions can incorporate the policy considerations in their deliberations to ensure that managers embed ethics into telework policies. Finally, employers can integrate the policy consideration framework into their telework arrangements, thereby optimizing the beneficial impact of telework on all stakeholders and improving the chances of achieving successful telework implementations.
NOTES 1. On January 1, 2005, ITAC became the Telework Advisory Group for World at Work. Prior to this time, ITAC was known as the International Telework Association and Council (ITAC). World at Work, founded in 1955, is a not-for-profit professional association dedicated to knowledge leadership in compensation, benefits and total rewards (http://www.workingfromanywhere.org). 2. Reported on October 6, 2005, at www.egovmonitor.com/node/2987. 3. Several websites that might be helpful: http://www.telework.gov/ (established by the Office of Personnel Management and the General Services Administration to provide information for employees who think they might like to telecommute, for managers and supervisors who supervise teleworkers, and for agency telework coordinators); http://www.att.com/telework/calculator.html (AT&T site that includes a telecommuting calculator – shows number of pounds of carbon dioxide that can be saved from polluting the atmosphere, based on the number of miles the employee would normally commute and the miles per gallon that the employee’s auto burns; http://www.telcoa.org/ (the Telework Coalition); http://www.tca.org.uk/ (Europe’s largest organization dedicated to the promotion of teleworking).
REFERENCES Anonymous (2003). BT employees highlight benefits and drawbacks of teleworking. Human Resource Management International Digest, 11(2), 17–20. Atkyns, R., Blazek, M., & Roitz, J. (2002). Measurement of environmental impacts of telework adoption amidst change in complex organizations: AT&T survey methodology and results. Resources Conservation and Recycling, 36(3), 267–285.
234
ANITA REED ET AL.
Bailey, D., & Kurland, N. (2002). A review of telework research: Findings, new directions, and lessons for the study of modern work. Journal of Organizational Behavior, 23(4), 383–400. Baumann, Z. (1993). Postmodern ethics. Oxford: Blackwell. Belanger, F., & Collins, R. (1998). Distributed work arrangements: A research framework. The Information Society, 14, 147–152. Crowther, D., & Hosking, D. (2005). Accounting in babel? Constructing social accounting as a multi-logical performance. Critical Perspectives on Accounting, 16(5), 535–550. Daniels, D., Norman, C., & Stewart, I. (2004). A profile of the millennial generation: Implications for teaching and learning in the accounting classroom. Journal of the Academy of Business Education, 5(Fall), 50–62. Daniels, K., Lamond, D., & Standen, P. (2001). Teleworking: Frameworks for organizational research. Journal of Management Studies, 38(8), 1151–1185. Dillard, J. (2002). Ethics research in AIS. In: V. Arnold & S. Sutton (Eds), Researching accounting as an information systems discipline. Sarasota, FL: American Accounting Association Information Systems Section. Fairweather, B. (1999). Surveillance in employment: The case of teleworking. Journal of Business Ethics, 22(1), 39–49. Fracaro, K. (2005). Commitment: Requirement for success. Contract Management, 45(10), 4–9. Gray, M., Hodson, N., & Gordon, G. (1993). Teleworking explained. Chichester, NY: Wiley. Gustafson, A. (2000). Making sense of postmodern business ethics. Business Ethics Quarterly, 19(3), 645–658. Guthrie, R. (1997). The ethics of telework. Information Systems Management, 14(4), 29–32. Hendry, J. (2001). After Durkheim: An agenda for the sociology of business ethics. Journal of Business Ethics, 34(3/4), 209–218. Hunton, J. (2005). Behavioral self-regulation of telework locations: Interrupting interruptions!. Journal of Information Systems, 19(2), 111–140. International Telework Association (ITAC) (2005). www.workingfromanywhere.org and www.tca.org.uk. Lomo-Dvid, E., & Griffin, F. (2001). Personality traits of white-collar telecommuters: Perceptions of graduating business students. Journal of Education for Business, 76(5), 257–261. Mirchandani, K. (1999). Legitimizing work: Telework and the gendered reification of the work– nonwork dichotomy. The Canadian Review of Sociology and Anthropology, 36(1), 87–107. Montreuil, S., & Lippel, K. (2003). Telework and occupational health: A Quebec empirical study and regulatory implications. Safety Science, 41, 339–358. Moon, C., & Stanworth, C. (1997). Ethical issues of teleworking. Business Ethics: A European Review, 6(1), 35–45. Phelan, S. (2002). Home is where the office is. Journal of Accountancy, 19(6), 37–45. Pullium, D. (2005). House panel again tells agencies: Increase telework or lose funds. www.govexec.com (June 10). Raines, J., & Leathers, C. (2001). Telecommuting: The new wave of workplace technology will create a flood of change in social institutions. Journal of Economic Issues, 35(2), 307–313. Tietze, S. (2002). When ‘‘work’’ comes ‘‘home’’: Coping strategies of teleworkers and their families. Journal of Business Ethics, 41(4), 385–396. Tietze, S. (2005). Discourse as strategic coping resource: Managing the interface between ‘‘home’’ and ‘‘work’’. Journal of Organizational Change Management, 18(1), 48–62.
A Postmodern Stakeholder Analysis of Telework
235
Wade, D. (1999). The expanding workplace: Telecommuting and legal liability under OSHA, ADA, and workers’ compensation. Labor Law Journal, 50(4), 252–263. Ward, N., & Shabha, G. (2001). Teleworking: An assessment of socio-psychological factors. Facilities, 19(1/2), 61–70. West, H. (2002). Teleworking in the public sector. The Public Manager, 31(2), 34–36. Wilson, F. (1995). Managerial control strategies within the networked organization. Information Technology and People, 8(3), 57–72. Yuthas, K., & Dillard, J. (1999). Ethical development of advanced technology: A postmodern stakeholder perspective. Journal of Business Ethics, 19(1), 35–49.
This page intentionally left blank
236