Advances in Accounting in Behavioural Research (Advances in Accounting Behavioral Research, Vol. 13)

ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH Series Editor: Vicky Arnold Recent Volumes: Volumes 1 4: Volumes 5 12: Edit...

Author: Vicky Arnold

31 downloads 1514 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH Series Editor: Vicky Arnold Recent Volumes: Volumes 1 4: Volumes 5 12:

Edited by James E. Hunton Edited by Vicky Arnold

ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH VOLUME 13

ADVANCES IN ACCOUNTING BEHAVIORAL RESEARCH EDITED BY

VICKY ARNOLD Kenneth G. Dixon School of Accounting, University of Central Florida, USA and Department of Accounting and Business Information Systems, The University of Melbourne, Australia ASSOCIATE EDITORS

B. DOUGLAS CLINTON Northern Illinois University, USA

ANNE LILLIS University of Melbourne, Australia

ROBIN ROBERTS University of Central Florida, USA

CHRIS WOLFE Texas A&M University, USA

SALLY WRIGHT University of Massachusetts Boston, USA

United Kingdom North America India Malaysia China

Japan

Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2010 Copyright r 2010 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. No responsibility is accepted for the accuracy of information contained in the text, illustrations or advertisements. The opinions expressed in these chapters are not necessarily those of the Editor or the publisher. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978 0 85724 137 5 ISSN: 1475 1488 (Series)

Awarded in recognition of Emerald’s production department’s adherence to quality systems and processes when preparing scholarly journals for print

LIST OF CONTRIBUTORS Vicky Arnold

Dixon School of Accounting, University of Central Florida, Orlando, FL, USA and Accounting and Business Information Systems, University of Melbourne, Australia

Charles Cullinan

Accounting Department, Bryant University, Smithﬁeld, RI, USA

Brian Daugherty

Sheldon B. Lubar School of Business, University of Wisconsin Milwaukee, Milwaukee, WI, USA

Patricia L. Derrick

Department of Accounting and Legal Studies, Salisbury University, Salisbury, MD, USA

Denise Dickins

Accounting Department, East Carolina University, Greenville, NC, USA

Jane Dillard-Eggers

College of Business Administration, Belmont University, Nashville, TN, USA

Paul M. Goldwater

Dixon School of Accounting, University of Central Florida, Orlando, FL, USA

Amy M. Hageman

Department of Accounting, Kansas State University, Manhattan, KS, USA

D. Kip Holderness Jr.

Department of Accountancy, Bentley University, Waltham, MA, USA

James E. Hunton

Department of Accountancy, Bentley University, Waltham, MA, USA and Erasmus University, Rotterdam School of Management, The Netherlands vii

viii

LIST OF CONTRIBUTORS

Michael R. Koval

Department of Accounting and Legal Studies, Salisbury University, Salisbury, MD, USA

Wei Li

Department of Accounting, Kent State University, Kent, OH, USA

Michael L. Roberts

Business School, University of Colorado Denver, Denver, CO, USA

Kenneth J. Smith

Department of Accounting and Legal Studies, Salisbury University, Salisbury, MD,USA

Steve G. Sutton

Dixon School of Accounting, University of Central Florida, Orlando, FL, USA

Kimberly A. Zahller

Dixon School of Accounting, University of Central Florida, Orlando, FL, USA

REVIEWER ACKNOWLEDGMENTS 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.

Elizabeth Almer Portland State University, USA

Ronny Daigle Sam Houston University, USA

Cindy Blanthorne University of North Carolina at Charlotte, USA

Stan Davis University of Tennessee at Chattanooga, USA

Dennis Bline Bryant University, USA

Naman Desai University of Central Florida, USA

Peter Booth University of Technology Sydney, Australia

Craig Emby Simon Fraser University, Canada Amy Hageman Kansas State University, USA

Gary Braun University of Texas El Paso, USA

Mary Hill Kennesaw State University, USA

Wayne Bremser Villanova University, USA

Karen L. Hooks Florida Atlantic University, USA

Richard Brody University of New Mexico, USA Erin Burrell University of Central Florida, USA

Susan Ivancevich University of North Carolina Wilmington, USA

Janie Chang San Diego State University, USA

Steve Kaplan Arizona State University, USA

Bryan Church Georgia State University, USA

Kip Krumweide Boise State University, USA ix

x

Lorraine Lee University of North Carolina Wilmington, USA Patrick Leung Hong Kong Polytechnic University, China Nace Magner Western Kentucky University, USA Linda Matuszewski Northern Illinois University, USA Joann Segovia Winona State University, USA Mike Shaub Texas A&M University, USA Georgia A. Smedley University of Missouri Kansas City, USA Steve Sutton University of Central Florida, USA

REVIEWER ACKNOWLEDGMENTS

John Sweeney Washington State University, USA Greg Trompeter Boston College, USA Sandra Vera-Munoz University of Notre Dame, USA Ralph Viator Texas Tech University, USA Chantal Viger University of Quebec at Montreal, Canada Rick Warne George Mason University, USA Tammy Waymire Northern Illinois University, USA Brett Wilkinson Baylor University, USA Kimberly A. Zahller University of Central Florida, 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, and is receptive to methodological examinations that can potentially inform future behavioral research. 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 reviewed by 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 Canada (2005).’ In the text, use the form Hageman 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. (Phillips, 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 FUNDAMENTAL ROLE OF TECHNOLOGY IN ACCOUNTING: RESEARCHING REALITY Steve G. Sutton ABSTRACT Behavioral accounting research has ﬂourished over the past 40 years and vastly improved our understanding of accounting judgment and decisionmaking, human behavior as it is affected by accounting information and processes, and inﬂuences on organizational and social structures. However, to increase the validity and reliability of the work, researchers have generally narrowed the area of study to exclude many of the environmental factors that can inﬂuence the resulting behaviors that are observed. One environmental factor that has largely been ignored by the broader accounting research community is the rapidly increasing impact of information technology (IT) on all aspects of accounting. The purpose of this chapter is to elaborate on the predominance of IT in all areas of accounting and to urge behavioral accounting researchers to integrate IT aspects into their research to enhance the value and relevance of our research. Each of the major areas of accounting disciplinary research is considered (i.e., ﬁnancial accounting, managerial accounting, auditing, and tax). This disciplinary focus is not intended to exclude the area of accounting information systems as is often the case in commentaries on behavioral accounting research but rather to focus on how accounting

Advances in Accounting Behavioral Research, Volume 13, 1 11 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013004

1

2

STEVE G. SUTTON

information systems are fundamentally integrated across the decision environments of every aspect of the accounting discipline.

INTRODUCTION Few eyebrows are raised when one exclaims that technology has inﬁltrated all parts of our life. Most people drive cars that are controlled by computers, surf the Internet to locate information, and use e-mail as a primary communications mode; some people will even read this article as an electronic download. Information technology (IT) has been adopted as an integral part of everyday life with little concern that IT has altered the manner in which people communicate, access entertainment, and work. A researcher would likely be scoffed at by her peers if she proposed to study letter writing among teenagers or VCR rentals by college students. Yet, accounting researchers rarely question studies of individual decision-making or interorganizational relationships that ignore the inﬂuence of IT. The negative side is that much of behavioral accounting research may lead to erroneous conclusions absent consideration of how IT changes the environment. The positive side is that there is great opportunity for researchers who are willing to tackle the complexities of IT in accounting environments. IT has radically changed the manner in which accounting information is produced, disseminated, and used. Most organizations’ accounting information is aggregated and stored in enterprise systems – IT-based systems that create a central database where accounting information is stored for use. This is the same information that must be audited for reliability and the same information that is extracted for budgeting, performance monitoring, and business reporting. The audit teams that must tackle these databases are armed with systems designed to facilitate data extraction and testing and operate within teams that are coordinated through electronic workpaper systems that increasingly automate the audit planning process and the identiﬁcation of appropriate audit tests. Under new Securities & Exchange Commission (SEC, 2008) mandates, this same information must be communicated for external reporting purposes using advanced tagging systems [i.e., eXtensible Business Reporting Language (XBRL)]. As a result, users of external ﬁnancial reports are also being continually shifted into IT-driven processes for retrieving and analyzing corporate information. One should not forget tax, where most individual ﬁlers not using a tax preparer are now using tax software to prepare and ﬁle federal and state tax returns. Behavioral accounting researchers need to

The Role of Technology in Accounting Research

3

seize the opportunities that exist to explore how IT alters accounting decisions and related behaviors. Bonner (2008) offers several concluding perspectives in her recent book summarizing the state of judgment and decision-making research. Of most interest here are the statements related to the inﬂuence of IT. As she notes, ‘‘[a] ﬁnal, and perhaps most important, realization that comes from stepping away from the framework is that [judgment and decision making] research in accounting traditionally has not focused signiﬁcantly on accounting systems’’ (p. 381). The call in this chapter is for a change that addresses this fundamental inﬂuence on accounting decision-making. The following sections focus on how IT is altering the environments in which each disciplinary area of accounting exists. The following section addresses the changes impacting ﬁnancial accounting, the next section addresses management accounting, the next section auditing, and the next section tax. The ﬁnal section provides some brief concluding comments.

IT AND FINANCIAL ACCOUNTING The most visible change from a ﬁnancial accounting perspective is the SEC’s new mandates requiring all public companies to ﬁle their annual ﬁnancial statements with the SEC using interactive tagged data (i.e., in its current format or XBRL). This mandate is being rolled out across SEC registered companies over a three-year period beginning July 2009. Although the general conclusion is that requiring interactive tagged data will make it easier for ﬁnancial statement users to access speciﬁc data and to compare speciﬁc data across companies, the beneﬁts or usefulness of XBRL-tagged data remains unknown. How will ﬁnancial statement users access and aggregate tagged ﬁnancial data? Will the tagging of data affect search strategies, data selection, or other aspects of information processing? One of the SEC’s major concerns is facilitating accessibility for retail (i.e., nonprofessional) investors. Will retail investors be advantaged or disadvantaged with the availability of XBRLtagged data? To date, a limited amount of research has been conducted, which sheds light on impacts of data tagging for investors. Hodge, Kennedy, and Maines (2004) ﬁnd that linking quantitative ﬁnancial statement information with associated footnotes improves decision-making. Arnold, Bedard, Phillips, and Sutton (2010) ﬁnd that retail investors become more directed in their data search and that data tagging can impact the saliency of certain key qualitative information. Much more research needs to be conducted to understand the

4

STEVE G. SUTTON

impact of data tagging on information biases, information aggregation, information processing, and investor decision outcomes. While the mandates are in place, the SEC is also concerned about use and usability of the data. The SEC has released a software tool that is freely available to users to facilitate the access and aggregation of XBRL-tagged data. However, the tool is not particularly easy for the novice to use. This leaves several concerns that are open to research. How do investors use the tool? Does the tool affect investor decision-making? Does the tool lead to certain biases in judgments? Perhaps even more importantly, can we study investor use of information to prescribe better designs or better information focus for such software tools? These software tools can be user-driven or they can provide greater guidance to the user to try and improve use and decision-making. Absent more research that helps us understand the links between information and investor decision-making, such prescriptions can only be provided through conjecture. There are many other ways that IT affects the ﬁnancial accounting process. However, for most behavioral accounting researchers, these investment decision processes that use ﬁnancial information are likely to be of most interest. The opportunities for research are fairly wide open; the challenge is in ﬁguring out how to study such a phenomena in an ex ante form.

IT AND MANAGERIAL ACCOUNTING Arguably, the biggest change in the managerial accounting function has come from the implementation of enterprise systems [also referred to as enterprise resource planning (ERP) systems]. As noted earlier, these systems integrate information from across the organization into a single repository – a central enterprise-wide database. As shown in Hannan, Rankin, and Towry (2006), taking data that has traditionally been more private through the use of disparate systems in decentralized organizations and making them centralized and integrated can affect managers’ behavior. By bringing information together into integrated systems, the data becomes more accessible and more visible across the organization – and thereby also less private. This has broad implications for budgeting behavior, which in most behavioral accounting research studies has been premised on information being somewhat private and unobservable by fellow managers and superiors (see Kren, 1997). This change in transparency has broad implications for research in management accounting. First, the availability of enterprise-wide data can

The Role of Technology in Accounting Research

5

impact the strategies and information used by managers in formulating budgets and performance measures. Second, the changes in the transparency of information used in budgeting processes can affect the manager’s budgeting ﬂexibility and ability to control performance measures. This has implications for role ambiguity, job satisfaction, turnover intentions, and an array of other social and psychological factors (Arnold, 2006). Third, the dynamics between managers and their peers, as well as subordinates and superiors, can all be altered by this transparency. How does this transparency impact budget negotiations, operational ﬂexibility, and performance incentives? For instance, in a case study, Scapens and Jazayeri (2003) observed that the implementation of an enterprise system required managers to develop networks of cooperative relationships that by necessity included identifying shared goals, sharing more information, building consensus, and promoting trust to maintain such relationships. Many of the assumptions for how enterprise systems reshape the management accounting function are premised on the belief that managers can and will access the available enterprise data. To date, however, little research has examined managers’ use of these systems, their ability to query information through the system, and their ability to leverage the now available data. How does the greater real-time nature of the data affect decision models and decision analyses? How proﬁcient are managers at aggregating data in dynamic models as opposed to continuing to rely on static report formats? Does the ability to dynamically analyze real-time data enhance decision performance or are there factors such as information load that have negative effects on decision-making? One response to the broad research questions posed here is that prior research has not demonstrated a radical change in the management accounting function post–enterprise system implementation (e.g., Granlund & Malmi, 2002; Dechow & Mouritsen, 2005; Quattrone & Hopper, 2005; Rom & Rohde, 2007; Chapman & Kihn, 2009). As Chapman and Kihn (2009) note, enterprise systems are simply a resource that is made available, and enhancements in the management accounting function will only accrue when the managers themselves effectively leverage the system. However, further extensions to enterprise systems are providing greater power to effectively utilize enterprise system data, making the process much simpler for users. These add-on systems include customer relationship management systems that facilitate analysis of sales and customer data, operational resource management systems that facilitate analysis of operational expenses, business intelligence systems that support detailed aggregation and analysis of enterprise data, and so forth.

6

STEVE G. SUTTON

These add-on systems improve data accessibility and analysis capability, but can also add substantial structure to managerial control processes. For instance, Elbashir, Collier, and Sutton (2011) examine the use of a speciﬁc business intelligence system used across a wide array of organizations. They note that the particular business intelligence system in their study is designed to integrate with the enterprise system database to make 200 prebuilt reports readily available as well as over 500 key performance indicators and analytics addressing over 2,900 business critical questions. These reports include a multitude of scorecard analyses. The net effect is a structuring of scorecards and other performance measures used to assess managers and their business units. While Elbashir et al. (2011) demonstrate a linkage at the organizational level between assimilation of business intelligence systems and organizational performance, very little is known about how managers use these broad arrays of performance measures and the multitude of metrics. Relatedly, very little is known about how the existence of these measures affects manager’s behavior. The management accounting function could be argued to be the most affected by IT integration. Yet, very little attention has been given to the emerging phenomena by behavioral accounting researchers. The transformation that is occurring has the potential to completely alter the nature of the management accounting function and the nature of the relationships that have been the subject of prior research on budgeting and performance management.

IT AND AUDITING From an audit perspective, two fundamental dimensions of IT change should be considered. First, IT systems are widely used to support the completion of audits and to drive the planning, execution, documentation, and review of the audit. Although on the face this may not seem like a major change, a closer look reveals how these systems drive and structure the audit process. Second, IT systems are increasingly automating audit functions and redeﬁning the role of the auditor. This is particularly true for SOX 404 compliance engagements, but particularly so in ERP environments. These dimensions are subsequently considered in greater detail. Dowling and Leech (2007) provide a review of the systems capabilities provided in the audit support systems of ﬁve international audit ﬁrms. Their discussion explores the capabilities of the systems across several dimensions: (1) determining client acceptance, (2) understanding the control environment,

The Role of Technology in Accounting Research

7

(3) designing tests of controls and substantive tests, and (4) reviewing audit work and formulating audit opinions. Perhaps more importantly, they explore the level of audit support system restrictiveness that represents both system restrictiveness and audit structure. Their analysis reveals a strong movement toward increasingly restrictive systems that place limits on auditor judgment in favor of greater audit process consistency. Increasingly, the nature of these systems is to control the audit process. However, Dowling (2009) reveals that auditors will attempt to appropriate the system toward their desired audit tests and processes to the degree the system allows them to do so. This appropriation effort is impacted, however, by the views and inﬂuences of the audit team, ﬁrm ofﬁce, and audit ﬁrm. Although more restrictive systems may enhance consistency in the audit process, the side effect of such systems may be greater technology dominance (see Arnold & Sutton, 1998), and this restrictiveness/dominance appears to have a potential de-skilling effect on auditors as they rise through the ﬁrm (Dowling, Leech, & Maroney, 2008). Whereas there may be other unintended consequences of the move to more restrictive systems, these changes at a minimum highlight the need for reconsidering much of the audit judgment research in terms of the inﬂuence of these systems. There are also implications for audit expertise and the effects on expertise development. Research that considers the knowledge transfer potential of various technologies has been conducted (see Smedley & Sutton, 2007; McCall, Arnold, & Sutton, 2008), but these technologies are different from those being used to support audit processes in the contemporary environment. Further research into the implications of these restrictive systems should be of great interest to both academics and practitioners. The other phenomenon that is slowly emerging within the audit environment is the increasing use of continuous auditing and continuous monitoring techniques (see Alles, Brennan, Kogan, & Vasarhelyi, 2006 for an example implementation in practice). Continuous auditing is primarily taking hold within internal audit groups where the automation of control monitoring and testing is increasingly viewed as a cost-effective way of meeting SOX 404 requirements. The automation of the audit process redeﬁnes the auditor’s role and the push behind the audit testing process. Traditionally, the auditor has designed and executed audit tests, perhaps using technology to conduct tests or analyze data. In a continuous audit environment, the system contains preprogrammed monitoring algorithms that execute tests. The auditor’s role becomes one of receiving alerts of potential errors from the system and addressing those ﬂagged events. Thus, the information processing of the auditor is radically different, and the

8

STEVE G. SUTTON

associated heuristics and biases that are part of that process are potentially quite different. Kuhn and Sutton (2010) explore a wide array of research questions that arise from the use of continuous auditing methods. A subset of these research questions should be of particular interest to audit researchers. As continuous auditing procedures are increasingly used by internal auditors, how does this affect the external auditor’s reliance on the internal auditor’s work? How willing are auditors to accept continuous audit applications as viable audit tools that meet expectations for testing and completeness? How willing is management to accept the use of embedded continuous audit techniques? Is this willingness affected by whether the manager or her subordinates’ actions are being monitored? How do auditors integrate continuous audit methods into their overall audit strategies? How do auditors receiving the ﬂood of alerts to potential problems that come with continuous auditing applications cope with the information overload? What is the general impact of ‘‘alert ﬂood’’ on auditors’ information processing and decision-making? What are the organizational characteristics that facilitate organizations’ successful adoption of continuous auditing systems? The auditor is not the only participant whose behavior and judgments are affected by such systems. Continuous monitoring of managers’ and other organizational members’ actions can also be expected to cause changes in behavior. Hunton, Mauldin, and Wheeler (2008) demonstrate one such scenario. In an experimental test of manager’s decision behavior, they observed a movement toward greater conservatism in decision-making when being monitored. This may or may not be a desirable behavior change. There is much still to be learned about the impacts of continuous auditing on managers’ behavior. The issue only becomes more critical as continuous auditing becomes more widely accepted and software support is provided for governance, risk management, and compliance activities.

IT AND TAX Many different angles could be used to approach the need for more behavioral accounting research as it applies to tax – an area long neglected by the research community. In particular, individual taxpayer behavior has received little attention by accounting researchers. Slightly more research related to the professionals who often help individual taxpayers complete compliance activities has been conducted. However, tax software is quickly becoming the tax professional who helps individual taxpayers complete their

The Role of Technology in Accounting Research

9

required tax returns and is an aid that professional tax preparers use to complete returns for others. Yet, we know little about taxpayer or tax preparer behavior, motivations, and the interacting effect of software. Research examining taxpayer behavior indicates that taxpayer attitudes have a strong inﬂuence over compliance behavior (Bobek & Hatﬁeld, 2003). One such attitude they highlight is perceived behavioral control that is related to the visibility of the taxpayer’s income and the likelihood of detection if the taxpayer misreports on her tax return. Masselli, Ricketts, Arnold, and Sutton (2002) similarly ﬁnd that the audit ﬂags raised in the TurboTax software alerting the user to the fact that certain numbers on their return appear outside of a normal range (and may trigger an audit by the IRS) will lead taxpayers to adjust their reported number even if the adjustment results in an overpayment of taxes. If you take the two results in tandem, the audit ﬂags would appear to inﬂuence the taxpayer’s perceived behavioral control. A small body of research on individual taxpayer behavior that investigates the impact of tax software on taxpayer behavior has been conducted. Masselli et al. (2002) demonstrates the impact of the audit ﬂags on individual taxpayers, but also shows that more knowledgeable tax professionals were not similarly intimidated by the audit ﬂags. Noga and Arnold (2002) also studied the impacts of software usage and found that taxpayers using the software to complete their tax returns became dependent on the software and did not acquire the knowledge to complete their tax return without the software. However, those users ﬁrst learning how to complete returns by hand easily transferred to the software to complete the returns. Finally, Hageman (2010) ﬁnds that taxpayers using tax software were much more conﬁdent in the accuracy of their tax returns than were taxpayers not using the software even though they were more prone to errors. Overall, we still know very little about individual taxpayer behavior and choices in meeting compliance requirements. Given the dominant position of tax software for both the individual taxpayer and the professional tax preparer, further research on the impact of such software within this context seems warranted.

CONCLUDING COMMENTS This chapter has provided a brief overview of how IT is impacting every facet of accounting and its main disciplines (e.g., ﬁnancial, managerial, audit, and tax). The intent is not to provide the full depth within each area

10

STEVE G. SUTTON

that a researcher would need to start designing studies but rather to highlight the changes that are taking place within each discipline. The reader should take away from this discussion an understanding of the profound role IT has assumed in the structuring of accounting tasks and in affecting human information processing. While many of us will not be surprised by the discussion presented here, behavioral accounting researchers should step back and reﬂect on how IT may have reshaped the phenomena that they are studying. IT fundamentally changes the nature of tasks, the methods by which they are undertaken, and the outcomes of task completion. Our research community must make a fundamental choice as to whether we wish to study reality or continue to assume we can ignore the dominance of IT in our world and its inﬂuence on all aspects of accounting.

REFERENCES Alles, M., Brennan, G., Kogan, A., & Vasarhelyi, M. (2006). Continuous monitoring of business process controls: A pilot implementation of a continuous auditing system at Siemens. International Journal of Accounting Information Systems, 7(2), 137 161. Arnold, V. (2006). Behavioral research opportunities: Understanding the impact of enterprise systems. International Journal of Accounting Information Systems, 7(1), 7 17. Arnold, V., Bedard, J., Phillips, J., & Sutton, S. G. (2010). The impact of information tagging in the MD&A on investor decision making: Implications for XBRL. Working Paper. University of Central Florida, Orlando, FL. Arnold, V., & Sutton, S. G. (1998). The theory of technology dominance: Understanding the impact of intelligent decision aids on decision makers’ judgments. Advances in Accounting Behavioral Research, 1, 175 194. Bobek, D. D., & Hatﬁeld, R. C. (2003). An investigation of the theory of planned behavior and the role of moral obligation in tax compliance. Behavioral Research in Accounting, 15, 13 38. Bonner, S. (2008). Judgment and decision making in accounting. Upper Saddle River, NJ: Pearson Education, Inc. Chapman, C. S., & Kihn, L. A. (2009). Information systems integration, enabling control and performance. Accounting Organizations and Society, 34, 151 169. Dechow, N., & Mouritsen, J. (2005). Enterprise resource planning systems, management control and the quest for integration. Accounting Organizations and Society, 30, 691 733. Dowling, C. (2009). Appropriate audit support system use: The inﬂuence of auditor, audit team, and ﬁrm factors. The Accounting Review, 84(3), 771 810. Dowling, C., & Leech, S. A. (2007). Audit support systems and decision aids: Current practice and opportunities for future research. International Journal of Accounting Information Systems, 8(2), 92 116. Dowling, C., Leech, S. A., & Maroney, R. (2008). Audit support system design and the declarative knowledge of long term users. Journal of Emerging Technologies in Accounting, 5, 99 108.

The Role of Technology in Accounting Research

11

Elbashir, M. Z., Collier, P. A., & Sutton, S. G. (2011). The role of organizational absorptive capacity in strategic use of business intelligence to support integrated management control systems. The Accounting Review (forthcoming). Granlund, M., & Malmi, T. (2002). Moderate impact of ERPS on management accounting: A lag or permanent outcome? Management Accounting Research, 13, 299 321. Hageman, A. M. (2010). The role of conﬁdence in tax return preparation using tax software. Advances in Accounting Behavioral Research, 13, 31 57. Hannan, R. L., Rankin, F. W., & Towry, K. L. (2006). The effect of information systems on honesty in managerial reporting: A behavioral perspective. Contemporary Accounting Research, 23(4), 885 918. Hodge, F., Kennedy, J., & Maines, L. (2004). Does information search technology improve the transparency of ﬁnancial reporting? The Accounting Review, 79(3), 687 703. Hunton, J. E., Mauldin, E. G., & Wheeler, P. R. (2008). Potential functional and dysfunctional effects of continuous monitoring. The Accounting Review, 83(6), 1551 1569. Kren, L. (1997). The role of accounting information in organizational control: The state of the art. In: V. Arnold & S. G. Sutton (Eds), Behavioral Accounting Research: Foundations and Frontiers (pp. 1 48). Sarasota, FL: American Accounting Association. Kuhn, J. R., & Sutton, S. G. (2010). Continuous auditing in ERP system environments: The current state and future directions. Journal of Information Systems, 24(1), 91 112. Masselli, J., Ricketts, R., Arnold, V., & Sutton, S. G. (2002). The impact of embedded intelligent agents on tax compliance decisions. Journal of the American Tax Association, 24(2), 60 78. McCall, H., Arnold, V., & Sutton, S. G. (2008). Use of knowledge management systems and the impact on declarative knowledge acquisition. Journal of Information Systems, 22(2), 77 101. Noga, T., & Arnold, V. (2002). Do tax decision support systems affect the accuracy of tax compliance decisions? International Journal of Accounting Information Systems, 3(3), 125 144. Quattrone, P., & Hopper, T. (2005). A ‘time space odyssey’: Management control systems in two multinational organisations. Accounting Organizations and Society, 30, 735 764. Rom, A., & Rohde, C. (2007). Management accounting and integrated information systems: A literature review. International Journal of Accounting Information Systems, 8(1), 40 68. Scapens, R. W., & Jazayeri, M. (2003). ERP systems and management accounting change: Opportunities or impacts? A research note. European Accounting Review, 12(1), 201 233. Securities & Exchange Commission. (2008). SEC approves interactive data for ﬁnancial reporting by public companies, mutual funds Press release 2008 300 (December 18). Washington, D.C.: Securities and Exchange Commission. Smedley, G. A., & Sutton, S. G. (2007). The effect of alternative procedural explanation types on procedural knowledge acquisition during knowledge based systems use. Journal of Information Systems, 21(1), 27 51.

TECHNOLOGY MONOCULTURE: ERP SYSTEMS, ‘‘TECHNO-PROCESS DIVERSITY’’ AND THE THREAT TO THE INFORMATION TECHNOLOGY ECOSYSTEM Charles Cullinan, Steve G. Sutton and Vicky Arnold ABSTRACT During the past decade, enterprise resource planning (ERP) system implementations have exponentially grown within ﬁrst large and then small- and medium-sized enterprises. Contemporary implementations, often through application service providers (ASPs), increase already existing pressures to adopt the embedded ‘‘best practices’’ that have been incorporated into the ERP software. The result is the rapid spread of generic business processes enabled through one of only a handful of leading ERP packages. This chapter focuses on the extant research on biodiversity and its focus on the negative effects of monoculture strategies – that is, the focus on a single crop (system) versus a diversity of crops (systems). The biodiversity research establishes a clear pattern of deleterious effects resulting from the vulnerabilities of monoculture strategies. These patterns

Advances in Accounting Behavioral Research, Volume 13, 13 30 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013005

13

14

CHARLES CULLINAN ET AL.

are mirrored in the ERP environment as vulnerabilities loom from the diminution of diverse business processes, limited adaptability to business environment changes given technology-driven/enabled processes, and increased susceptibility to widespread parasite damage through cyberattacks. The implications of the study raise questions as to the sustainability of accounting systems, the business environment, and society as a whole from the rapid implementation of sterilized business processes and uniformly vulnerable enterprise software.

INTRODUCTION Enterprise resource planning (ERP) systems have been readily adopted by a wide range of businesses. By the early 2000s, over 90 percent of U.S. Fortune 500 companies had implemented ERP systems with SAP alone accounting for 50 percent of adoptions (Leib, 2002). Companies of all sizes continue to adopt such software as a replacement for their homegrown, organizationwide information systems due to the perceived beneﬁts of integration and low cost in comparison with internally developing such integrated systems. For instance, SAP currently reports that it has 140,000 installations worldwide across some 75,000 customers, which include 12 million users on a daily basis (SAP, 2010). The successful implementation of ERP systems, however, requires companies to compromise their own business processes in favor of those supported by the software versus having the software conform to a company’s processes.1 The business processes previously used by these companies may have been developed over many years and presumably served to provide a competitive advantage for the organization. The conformance of the business to the software is premised on the idea that the software is based on ‘‘best practices’’; therefore, standardization of business processes should enhance business performance. However, these best practices are drawn from other companies and may not necessarily represent the ‘‘best practices’’ that could be used by the company implementing the ERP system. The end result is that widespread adoption of ERP systems tends to reduce the diversity of business process practices across the range of companies in an industry and across industries. In the natural world, many researchers have concluded that a diminution in the number of species present in an ecosystem makes the ecosystem less healthy and more vulnerable to signiﬁcant disruption (Naeem et al., 1999). At the limit, the decrease in the number of species to a single one creates a

Technology Monoculture

15

monoculture environment. In such an environment, an agent that threatens the single species has the potential to eliminate the entire ecosystem (because the ecosystem consists entirely of one species). Scott (1998) discusses criteria that have caused governmental agents and others to encourage decreasing diversity of plant culture. One of the main criteria for an agent’s successful push for greater standardization is a sense of crisis. In the IT context, the catalyst for widespread adoption of ERP systems was the initial concerns over potential year 2000 (Y2K) computer problems with legacy business information systems. The implementation of ERP subsequently became perceived as a necessary component to successfully competing, and, another wave of implementations continues to take place under desires to achieve competitive parity as the software has become more affordable for smaller organizations. Nonetheless, the motivation for rapid achievement of integrated information systems parity through implementation of standardized ERP systems comes at the sacriﬁce of diverse business processes that often provide alternative types of strategic and competitive advantage. The purpose of this chapter is to review the negative effects that have been experienced as a result of agricultural standardization and to relate these effects to the parallel practices in the business environment with the current movement toward diminution of diverse business processes in favor of ‘‘best practices’’ business processes incorporated in ERP systems. Two speciﬁc risks are examined in greater detail: (1) lack of adaptability to business environmental changes and (2) susceptibility to widespread parasite damage across multiple organizations’ systems through cyber-attack. The remainder of the chapter proceeds as follows. The next section presents a review of the basic ﬁndings of research examining biodiversity. Then, the following section examines various governmentally sponsored attempts to increase production through greater standardization of agriculture along with discussion of how these attempts often fail because they fail to fully consider the local environment within which implementation of the schemes occurs. The proliferation of ERP systems is then addressed in the next section, together with the process of sterilization of ﬁrms’ business processes as a result of business process reengineering methods used to create a match between a business’ processes and predetermined ‘‘best practices’’ enforced through software constraints. In the two subsequent sections of the chapter, two potential threats to mono-cropped systems are explored: (1) environmental shifts and the inherent limits on adaptability of existing ERPbased systems and (2) a parasite invasion as represented through susceptibility to cyber-attacks. The ﬁnal concluding section overviews the threats

16

CHARLES CULLINAN ET AL.

faced by businesses, which choose to mono-crop their information system requirements.

THE EFFECTS OF A LACK OF BIODIVERSITY About the time the ﬁrst wave of companies were rapidly completing adoptions of ERP systems in the face of the Y2K scares, the Ecological Society of America released a monograph (Naeem et al., 1999) presenting the results of an extensive review of literature on the effects of loss of biodiversity. The monograph indicates that a loss of biodiversity often has the following results (Naeem et al., 1999, p. 1): 1. ‘‘Plant production may decline as regional and local diversity declines.’’ 2. ‘‘Ecosystem resistance to environmental perturbations y may be lessened as biodiversity is reduced.’’ 3. ‘‘Ecosystem processes such as soil nitrogen levels, water use, plant productivity, and pest and disease cycles may become more variable as diversity decreases.’’ Evidence is available of historical instances in which a lack of diversity of plant species has caused signiﬁcant productivity loss and disruption to both economic systems and people. One of the more well-known cases of deleterious effects resulting from lack of biodiversity is the Irish potato famine of the mid-1800s (Scott, 1998). In the case of the potato famine, a material portion of the population depended on a single crop that was largely destroyed by a disease outbreak. Lacking other sources of nourishment, the potato failure caused disruption to large portions of Irish society. A more recent crop failure resulting from lack of biodiversity was the corn crop failure of 1970 in the United States. This instance led the United States National Research Council (1972, p. 21) to conclude that These encounters show clearly that crop mono culture and genetic uniformity invite epidemics. All that is needed is the arrival on the scene of a parasite that can take advantage of the vulnerability. If the crop is uniformly vulnerable, so much the better for the parasite.

The basic issue here is that if a threat to a plant is able to take advantage of some vulnerability in that particular plant, that same vulnerability can be exploited in all the other plants due to their genetic similarities. In a more diverse environment, ‘‘only a few individuals are likely to be susceptible to a given pathogen, and they are likely to be widely scattered. The mathematical

Technology Monoculture

17

logic of the epidemic is broken’’ (Scott, 1998, p. 269). Tilman (1999, p. 1467) also suggests that greater diversity should ‘‘decrease the chance that any given species would successfully invade y’’ Interestingly, Tilman (1999) refers to stabilizing effects of diversity as ‘‘the portfolio effect’’ (p. 1458) after the ‘‘long-standing principle in economics that more diverse portfolios are less volatile’’ (p. 1458). Marchado (2009) also discusses the negative effects of agricultural monoculture on sustainability. Three publications illustrate the productivity and resistance to threat aspects of biodiversity (Hector et al., 1999; Zhu, Chen, Fan, & Wang, 2000; and Armstrong, Albrecht, Lauer, & Riday, 2008). Hector et al. (1999) examined the effects of diversity on biomass (a measure of the total productivity of the plots) in European grasslands. They examined eight geographic sites that had between 1 and 32 distinct species of grasses on different plots at each site. Their results strongly indicate that the greater the variety of plant species that are present on each plot, the greater the productivity (biomass) of the plots. Zhu et al. (2000) examined both productivity and resistance to a fungus of rice ﬁelds in China. They manipulated levels of two different kinds of rice – a hybrid rice and a more glutinous rice. They estimated that it would take 1.18 hectares of a single species plot to produce as much rice as a 1-hectare plot planted with a combination of both varieties of rice. They also found that losses from rice blast caused by a fungus were signiﬁcantly lower in the more diverse plots than in the mono-crop plots. The results were particularly signiﬁcant for the glutinous rice, where losses from rice blast decreased from 20 percent of the grain in mono-crop plots to 1 percent in the mixed plots. Armstrong et al. (2008) examined the effects of biodiversity on corn yields. They compared corn monoculture, compared to corn intermixed with various types of beans. They found that corn ﬁelds mixed with various types of beans produced signiﬁcantly greater yields than mono-cropped corn.

WHY MONO-CULTURE REGIMES (AND ERP SYSTEMS) HAVE BEEN ADOPTED Iyegha (2000) presents a discussion of the importation of mechanized agricultural schemes into Nigeria and the failure of these systems to fulﬁll their promise. He focuses on the role of agricultural experts who, while well intentioned, often failed to recognize that the contexts from which they were adapting techniques, including mono-cropping, were not especially well suited to the climate, soil, and cultural conditions of Nigeria.

18

CHARLES CULLINAN ET AL.

Scott (1998) develops a general model for the conditions in which standardization schemes developed by governmental experts (including agricultural reforms) sometimes fail to fulﬁll their original objectives – a condition referred to as ‘‘failed state-initiated social engineering schemes’’ (Scott, 1998, p. 4). This model consists of four criteria, which are subsequently discussed along with the correlative conditions under which ERP systems have been implemented. The ﬁrst criterion of Scott’s model is the implementation of a standardization scheme. One of the principal ideas behind ERP systems, and especially the dominance of the market by a limited group of players, is that standardization of software and business processes will facilitate both intraorganizational and interorganizational communication between information systems. Additionally, ERP system implementers are generally encouraged both not to change the software and to modify their business practices to conform to the software’s notion of the most appropriate business processes. The second criterion of Scott’s model is what he terms a ‘‘high-modernist’’ ideology, which is based on a ‘‘strong y self conﬁdence about scientiﬁc and technical progress, the expansion of production, [and] the growing satisfaction of human needs’’ (Scott, 1998, p. 4). In particular, the notion seems to be that the institutionalized knowledge is superior to the local knowledge. The analogy to ERP adoptions is very tightly coupled when one considers the notion of ‘‘best practices.’’ ERP vendors have modeled their software to ﬁt their perceptions of the preferable business processes. As such, any adopter of the software is strenuously encouraged to model their processes on ‘‘best practices.’’ The assumption is that if adopters are using practices other than the ERP vendors’ designated ‘‘best practices,’’ their business processes have been compromised and must be inferior. However, as Richards notes, ‘‘The proper test for any practice was whether it worked in the environment concerned, not whether it looked ‘advanced’ or ‘backward’’’ (quoted in Scott, 1998, p. 4). The third criterion is the presence of an ‘‘authoritarian state that is willing and able to use the full weight of its coercive power to bring [the reform] into being’’ (Scott, 1998, p. 5). The adoption of ERP systems ﬁts less cohesively with this criterion. On the surface, there appears to be no actor with the ‘‘coercive’’ power equivalent to an authoritarian state driving ERP adoptions. A reasonable degree of pressure can be applied in business situations, however, by consultants encouraging certain techniques, by businesses following other ﬁrms in their industry, and by investor perceptions – no matter how awry they may be. Malmi (1999) provided evidence of the ﬁrst two factors interacting to encourage the adoption of ABC systems in Finland,

Technology Monoculture

19

even if such systems were neither cost-beneﬁcial nor well-suited to the particular company circumstances. Other evidence indicates that investors respond positively to announcements of ERP system implementation through increased share prices (Hunton, Lippincott, & Reck, 2003) suggesting that market pressures may inﬂuence companies’ decisions to adopt ERP systems. The fourth and ﬁnal criterion for a failed standardization process is a ‘‘civil society that lacks the ability to resist these plans’’ (Scott, 1998, p. 5). One of the main impetuses for the adoption of ERP systems in the mid- to late 1990s was fear of potential year 2000 (Y2K) computer problems. In fear of the alleged catastrophic consequences of the Y2K issue, many companies were willing to make signiﬁcant changes in their software – generally with little or no attention to the impact on their business processes. This fear has been replaced in the post-Y2K by response to market demands derived from a perspective that an integrated business information system is necessary to maintain strategic and competitive parity. As such, businesses have similarly demonstrated a diminished capacity to resist the ERP standardization schemes. Thus, all four coercive and mimetic forces apparently have come together to help drive ERP system implementation and adoption. This is of major concern given that the four factors in Scott’s model are perceived as having serious deleterious effects when they interact to drive mono-cropping strategies. These concerns, along with key risks that exude from such policies, are examined more closely in the following sections.

ERP-DRIVEN BUSINESS PROCESS REENGINEERING: THE PROCESS STERILIZATION CRAZE Two primary drivers of the move toward standardized business processes built on ‘‘best practices’’ as determined by ERP vendors are (1) the economics of using ERP vendors to achieve integrated information systems objectives and (2) the outsourcing movement with the concurrent focus on core competencies. These are not necessarily independent drivers in that the demand for integrated information systems are most easily and cheaply satisﬁed by ERP vendors; in turn, the major vendors are more than happy to host such systems for companies. In the remainder of this section, these two issues are examined more closely as they pertain to the business process sterilization phenomena.

20

CHARLES CULLINAN ET AL.

First, consider the evolution of ‘‘best practices’’ in ERP systems. The ﬁrst generations of ERP systems were designed as slightly modiﬁable, one process ﬁts all, integrated information systems. As early difﬁculties were encountered in the attempts to take one base system and ﬁt it to all companies, vendors began examining where systems successes were occurring in an effort to determine optimal system/process matches. A success in an industry became a ‘‘best practice’’ for that ERP system industry solution and the model for future implementations in that industry. Thus, the ‘‘best practices’’ for a client’s implementation became the consultant’s model for successful implementation with another company – not necessarily an optimal match for the client’s business. Of course, the other interesting side of this perspective is that if a company had a system with a competitive advantage, that advantage was probably lost when their consultant replicated the systems advantage as a ‘‘best practice’’ and applied it to multiple competitors’ implementations (Kay, 1996). The business press is replete with brief summaries of organizations that have struggled through business process incompatibilities in an effort to implement ERP systems. One analyst with B2B Analysts (a Cambridge, MA, company) noted that ‘‘Going back to an old application is always tempting, because it works better [than a packaged one]’’ (Songini, 2002b, p. 16). But companies make the sacriﬁce due to the promises afforded by an integrated information system that makes enterprise data available throughout the organization. The analyst noted further that, for instance, adaptation to high-end, commercial order–management systems are especially difﬁcult for consumer product companies. Songini (2002a) notes similar type problems with the materials supply and forecasting modules for a company with highly complex distribution processes. Slater (1999) discusses the breadth of such problems as he notes that ‘‘[c]ompanies buy multimillion-dollar software packages only to ﬁnd out they don’t work—or at least don’t work well—for one of their key business processes’’ (p. 30). The reason, Slater suggests, is that ERP software is so hot, the ﬂames fanned by consultants and the technical press cause companies to simply push forward without dealing with such key restrictions. Slater further notes that the problem is worsened because the mismatch with business functions is critical, and more subtle issues, such as corporate culture and management style, are also affected by such systems. He notes that marketdominant SAP is notoriously dictatorial in terms of forcing adherence to a preprescribed model for doing business. On the contrary, many of the major vendors promise avoidance of the long, tedious implementations (for which ERP systems are renown) if the

Technology Monoculture

21

implementing organization is willing to accept the standard set of industry ‘‘best practices’’ embedded in the vendors’ basic software. The contemporary benchmark for rapid implementation by the major ERP vendors is 30–60 days. But all have limitations similar to Oracle that mandates a time compression approach to business process analysis that includes (1) adoption of the business ﬂows in the software, (2) minimal (if any) modiﬁcations to the software, and (3) limited conversion of existing historical data (Karamouzis, 2001). The trade-off is that the organization acquiring the ERP system has reduced consulting costs for implementation and support as a result of vanilla processes. The outsourcing trend that includes a move to application service providers (ASPs) has signiﬁcantly exacerbated the business process sterilization problem. The cost-efﬁciencies again make such options appealing as illustrated by Lambert (vice president of information solutions) who noted that John I. Haas, Inc. reduced the cost of their Oracle applications by 20 percent through use of an Oracle ASP–hosted solution (Harreld, 2001). Operating costs are only part of the potential savings; savings due to reduced downtime can also be signiﬁcant. As organizations, by the nature of ERP systems, become completely dependent on the software, downtime becomes a signiﬁcant cost factor. Past surveys have indicated that software outages for implemented ERP software average 2.8 hours per week for ﬁnancial and manufacturing modules (Dryden, 1998). Vendors promise to greatly reduce this average downtime by hosting applications. Still, with ERP system downtime costs estimated on average to cost companies $6,400–$7,900 per minute, downtime costs can escalate quickly (BMC Software, 2006). The trade-off is an even greater reduction in business process customization in favor of strict adherence to software-prescribed processes. The limitations placed by Oracle are typical of the industry and include, among others, (1) organizations are not allowed to integrate any other software products with the Oracle ERP software, (2) a ‘‘zero-tolerance’’ approach to customizations (i.e., there will not be any), and (3) determination by Oracle as to when software will be updated, which may concurrently affect the nature of the embedded business processes (Karamouzis, 2001). Again, conformity in business processes across organizations is extended through generic implementation strategies. The orientation toward outsourcing ERP systems through ASPs continues to spread and is already at unprecedented levels. Offshore outsourcing to India alone is expected to reach $50 billion for the ﬁscal year ending March 31, 2010, and grow by another 10–15 percent in the following ﬁscal year (Ribiero, 2010). Presumably, the sterilization of business

22

CHARLES CULLINAN ET AL.

processes would grow tremendously with the steady movement toward outsourcing business processing, an effect that should only accelerate in light of the massive offshoring effort.

ENVIRONMENTAL SHIFTS IN BUSINESS: THE INABILITY TO REACT AND THRIVE What are the risks of the sterilization of business processes? Similar to the mono-cropping risks noted in agricultural studies, reducing diversity of business processes appears to increase a business’s susceptibility to changes in the business environment. Furthermore, even a series of optimal individual processes rarely combine to make up an overall optimal system (Schwartz, 2002). Yet, optimization becomes difﬁcult once a system has been implemented with all of its compromises. Many organizations will assess the initial trade-off between rapid, simpliﬁed implementation of enterprise software versus business needs and opt for quick and dirty customizations to speed implementation. Buxbaum (2002a) notes that such a strategy can ‘‘come back to bite you later’’ (p. 1). Recovering from the compromises that have been made can be very difﬁcult. If a business process is core to an organization’s business and the organization has compromised its business processes by not providing the full range of options that are needed, the organization will have put itself at a signiﬁcant competitive disadvantage (Buxbaum, 2002b). The problems are compounded by the fact that automated business processes are far more resistant to change than the systems they replace. As Finkelstein (1996, p. 5) claims, Organizations have buried themselves in concrete that has now set hard: computer systems introduced to improve organizational responsiveness now inhibit the very business changes needed to survive.

The problem is that organizations have traditionally strived for differentiation through unique business processes for speciﬁc functions that can provide competitive advantage (Karamouzis, 2001). Sutton, Arnold, and Hunton (1999) note that organizations that use fairly decentralized management structures provide employees with the chance to self-organize, and within loose structures, they will generally self-organize very effectively. In the process, they operate in an environment of mild chaos that promotes productivity, performance, and ingenuity. Furthermore, ERP systems can enforce a suboptimal structure that limits employee groups from operating in such an environment of mild chaos and limits productivity, performance

Technology Monoculture

23

and, most of all, ingenuity. Thus, an ERP-based organization uses processes that become ‘‘buried in concrete’’ and limit the ability of the organization to adapt to changes in the business environment. Recall from Naeem et al. (1999, p. 1) that one of the major risks from lack of biodiversity is that ‘‘Ecosystem resistance to environmental perturbations ... may be lessened y’’ The sterilization of ERP-driven business processes provides just such a susceptible environment. If the business environment shifts or new competitors with more efﬁcient (or effective) business processes enter the marketplace, an ERP-driven organization may be unable to modify business processes within an acceptable time frame and may lose market share – or even disappear from the market. Furthermore, given that a majority of the major players in most markets have implemented ERP systems based on very similar business processes, one or more competitors who are in a position to react to business environment and market changes may gain rapid competitive advantage to the detriment of virtually all organizations currently in the marketplace. Such shifts may cause marketplace instability, worker insecurity, and reduced partnering opportunities between organizations due to increased risk of rapid change in a market’s players. The net effect is that organizations appear to be taking a major risk through the use of ERP systems that lock in sterilized business processes that eliminate or restrict achievement of competitive advantage and may even make competitive parity difﬁcult to maintain. In an effort to attain quick solutions to the complex problem of integrating information systems across the organization, companies are likely choosing short-term cost/ effort minimization strategies as opposed to long-term, strategically sound investments in tailored information systems that maximize the efﬁciency and effectiveness of an organization’s own business processes. Furthermore, through control of their systems, an organization is better prepared to adjust such systems and respond to business environment changes.

INCREASED SUSCEPTIBILITY TO CYBER-PARASITE BLIGHT As noted earlier, business processes in an ERP environment are entirely dependent on the software, and software downtime is extremely costly (Dryden, 1998). While signiﬁcant risks from business processes being cemented into place through software-deﬁned processes exist, these business

24

CHARLES CULLINAN ET AL.

processes are at even greater risk if they are rendered nonfunctional for an extended period. This can occur simply by the underlying ERP software becoming nonfunctional. Cyber-attacks have become a common occurrence in the business environment. The FBI surveyed large corporations and government agencies and found that 90 percent of respondents reported that one or more computer security breaches were detected within the past 12 months. Of the respondents detecting such a breach, 74 percent cited their Internet connection as the most frequent attack point. The director of the FBI’s Computer Security Institute, Patricia Rapalus, noted that signiﬁcantly more illegal and unauthorized activity is taking place in cyberspace than organizations report to business partners, stockholders, customers, and law enforcement ofﬁcials. (Verton, 2002). In a three-year period in which ERP systems became commonplace in large organizations, the number of cyberattacks doubled annually (Pethia, 2001). Similarly, data breaches in the United States increased by 47 percent between 2007 and 2008 (Kirk, 2009), while the cost of security breaches rose 97 percent on average for Canadian companies between 2008 and 2009 (Rotman/TELUS, 2009). As an example of the risk to an ERP environment, consider the vulnerabilities found in Oracle’s database product – the database underlying roughly 90 percent of ERP implementations for the major ERP vendors. Some security ﬂaws were so signiﬁcant that it was reportedly possible for a hacker to gain access to Oracle’s database server without a user ID or password. The vulnerability that allowed a hacker to access the database also allowed the attacker to execute functions in the software from a remote location. This vulnerability was found in implementations across all operating systems. Additionally, when installed in certain operating environments, denial-of-service attacks were also made possible (Berger, 2002). In short, the failures in one database product could potentially make the vast majority of ERP system installations in major public companies vulnerable to attack – a problem with lack of software diversity as a function of broad ERP system adoption. The problems with Microsoft Corporation’s software have also been welldocumented with the viruses in the summer of 2003 providing a short snapshot of the potential widespread effects of software monoculture. Microsoft has a 95 percent share of the desktop software market where the end user of ERP systems is invariably operating (Moran, 2003). Furthermore, the small percentage of ERP systems that are not operated using Oracle databases frequently use Microsoft’s SQL Server database instead. SQL Server has also proven to be very susceptible to hackers as exempliﬁed

Technology Monoculture

25

by an identiﬁed vulnerability. The software is shipped with a default password login that if not changed by the database administrator leaves database access wide open (Rosencrance, 2002). The Oracle and the Microsoft examples are presented simply as a case in point that software vulnerabilities are possible and could be exploited. Perhaps even more alarming is a report on the Carnegie Mellon CERT Coordination Center Website (considered one of the best websites available for security information) that listed 178 vulnerabilities for Oracle software systems and 514 vulnerabilities for Microsoft products (Conrath, 2002). Clearly, these dominant software systems appear vulnerable to widespread attack. When this information is coupled with the number of cyber-attacks reported in the FBI survey, there is little question as to the risk and vulnerabilities of corporate IT systems to attack (Verton, 2002). Pethia (2001) notes the complexity with which these attacks can occur and the difﬁculty organizations can have in tracking, defending and recovering from attacks. He further notes that if the source code of software used in an attack is not discoverable, solution through decompiling binary code can take hours, days, or even weeks – potentially paralyzing an organization to the point of complete failure. The vulnerabilities to attack are heightened by the poor security procedures in place within many organizations. These vulnerabilities are compounded by software vendors’ inattention to security coupled with the onerous number of security patches that must subsequently be installed to upgrade software security. Pethia (2001) notes the increased vulnerability that comes from the difﬁculties associated with securely conﬁguring operating systems and application software. He notes that this software is frequently shipped to customers with security disabled, requiring the user to go through the technically complex and challenging process of properly enabling security. As a case in point, the authors are aware of one company that recently installed an ERP system. During the implementation and testing phase, everyone was assigned the same user ID and password – providing all users with access to all programs and data. When the system went online, no one reset the IDs and passwords for individual level security and access. An auditor detected the error not long after implementation, but clearly signiﬁcant risk exposure during the period before discovery was present. The bigger risk from a parasite blight perspective, however, is that a vulnerability can be identiﬁed in the software that might enable an attacker to breach virtually all in-common systems across all user organizations. This would appear to be the case in the Oracle database ﬂaw discussed earlier. The risk is not just at the database software level but also at the ERP system

26

CHARLES CULLINAN ET AL.

level. In discussions with a developer working on the supply chain module of one of the dominant ERP systems, the developer noted that a common password was in place across all implementations of the supply chain module – including Electronic Data Interchange (EDI) applications for funds transfer. Unless the company using the system made the actual changes to the password that should be done during implementation, the implemented system leaves a gateway into the system for anyone knowing or identifying the standard password. Such vulnerability at the database or ERP level would allow an intruder to develop exploit scripts that could be shared among multiple attackers targeting multiple systems. In today’s cyber-attacks, such scripts are often combined with other forms of technology that automatically scan networks for vulnerable systems, prepare attacks of such systems, and either compromise or take the systems down. Thus, future cyber-attacks have the capability to spread rapidly among like-type software systems across accessible networks (Pethia, 2001). The fear of cyber-attack escalated immediately after the destruction of the World Trade Center towers in New York City on September 11, 2001. Numerous reports suggested that immediately after the attack on the towers, there was increased activity on the Internet that was suggestive of a high risk for mass cyber-attacks. By the very nature of a broad array of organizations using a limited number of ERP systems in common database and operating system environments, the setting is primed for a parasite to attack a large number of systems through an identiﬁed vulnerability. By mono-cropping business information systems, the parasite has the potential to spread rapidly and cause widespread destruction. Thus, not only are individual companies at signiﬁcant risk of attack and shutdown, but potentially a major portion of the economy could be rendered inoperable – potentially crippling the Western world’s entire market economy.

CONCLUDING THOUGHTS Threats to businesses and their accounting systems from the adoption of standardized software and business processes come in various forms. First, there are resource allocation issues within a business. In the agricultural world, standardization of crops results in more intensive use of fertilizers, pesticides, and water to make the standardized crop grow in areas to which it may not be naturally adapted. In addition, this standardization may result in ‘‘the virtual elimination of local knowledge’’ (Scott, 1998, p. 302). Similar resource allocation issues may arise with ERP implementations as

Technology Monoculture

27

companies spend resources to match their business processes with the standardized ‘‘best practices.’’ Businesses may also lose local knowledge, as employees who developed the previous business processes are now encouraged to adopt the software-promoted ‘‘best practices’’ rather than develop their own practices that work best in their context. If the business environment changes or a more nimble competitor enters the marketplace, the ERP-adopting ﬁrm may ﬁnd itself locked into inefﬁcient or ineffective software-deﬁned business processes. The strategic threats to businesses and workers spring from the lack of diversity in business processes. This lack of diversity creates the opportunity for sterilization of the work environment, devaluation of employees’ skill set, and threats to businesses that become locked into certain business practices while competitor companies potentially develop better processes that ultimately transfer competitive advantage to these competitor companies. As Armstrong (2002) indicates, ‘‘it would be hard to ﬁnd a successful company which consists solely of a core function backed up by routine services’’ (p. 117). Another source of threats to business arising from ERP system standardization is cyber-attack threats to the software. Cyber-attacks on ERP systems arise from deliberate attempts to compromise system security and can be either passive or active attacks. A passive attack could result in theft of key strategic information, whereas an active attack could corrupt or disable the ERP software and data. As an example of the realistic potential for such an attack, many virus attacks have relied on the prevalence of certain pieces of software (such as Microsoft’s Outlook e-mail servers) in many different companies to spread and disable systems (Lanza, 2000). The commonality of systems across organizations simpliﬁes the hacker’s task in penetrating systems security to reach corporate data and systems. As such, the increasing use of one, or a limited group of ERP systems, may permit widespread intentional attacks on users of the prevailing system(s). Ironically, Scott (1998, p. 22) uses a business analogy early in his discussion to illustrate the fundamental risks that arise from standardization: a merchant who, not knowing what conditions her ships may face at sea, sends out scores of vessels with different designs, weights, sails, and navigational aids stands a better chance of having much of her ﬂeet make it to port, while a merchant who stakes everything on a single ship design and size runs a higher risk of losing everything y

In effect, the standardization of business processes encouraged by ERP system implementations place businesses in the position of the merchant with only one type of ship.

28

CHARLES CULLINAN ET AL.

NOTE 1. For example, when updating certain software packages (like SAP), modiﬁca tions to the existing software to customize for the particular business’ needs will not be carried forward, but will have to be completed again. Companies that attempt to force modiﬁcations into the software to support unique business processes frequently abandon these modiﬁcations during subsequent upgrades to eliminate on going supplemental development costs for current and future upgrades.

REFERENCES Armstrong, P. (2002). The costs of activity based management. Accounting, Organizations and Society, 27(1), 99 120. Armstrong, K. L., Albrecht, K. A., Lauer, J. G., & Riday, H. (2008). Intercropping corn with lablab bean, velvet bean, and scarlet runner bean for forage. Crop Science, 48, 371 379. Berger, M. (2002). Security holes found in Oracle software. Computerworld. Available at http:// www.computerworld.com.au/article/69992/security_holes_found_oracle_software/?fp ¼ &fpid ¼ &pf ¼ 1.Retrived on February 7. BMC Software. (2006). Maximizing DB2 performance and availability. BMC Software Inc. Available at http://documents.bmc.com/products/documents/42/57/64257/64257.pdf Buxbaum, P. (2002a). Bevcore takes a big gulp. Computerworld. Available at http://www. computerworld.com/s/article/69423/Bevcore_Takes_a_Big_Gulp. Retrieved on March 25. Buxbaum, P. (2002b). Striking a balance. Computerworld. Available at http://www. computerworld.com/s/article/print/69424/Striking_a_Balance?taxonomyName¼ Defense %2FAerospace&taxonomyId ¼ 128. Retrieved on March 25. Conrath, C. (2002). Opinion: Secure software? Don’t hold your breath. Computerworld. Available at http://www.computerworld.com/s/article/74764/Opinion_Secure_software_ Don_t_hold_your_breath. Retrieved on October 1. Dryden, P. (1998). ERP failures exact high price. Computerworld, 32(July), 16 17. Finkelstein, C. (1996). Business re engineering: Three steps to success. Information Engineering Services Pty Ltd. Available at http://www.ies.aust.com/papers/brepaper.htm Harreld, H. (2001). Companies report growing demand for hosted e biz apps. Computerworld. Available at http://www.computerworld.com.au/article/37355/companies_report_growing_ demand_hosted_e biz_apps/. Retrieved on November 13. Hector, A., Schmid, B., Beierkuhnlein, C., Caldeira, M., Diemer, M., Dimitrakopoulo, P., Finn, J., Freitas, H., Giller, P., Good, J., Harris, R., Hogberg, P., Huss Danell, K., Joshi, J., Jumpponen, A., Korner, C., Leadley, P., Loreau, M., Minns, A., Mulder, C., O’Donovan, G., Otway, S., Pereira, J., Prinz, A., Read, D., Scherer Lorenzan, M., Schulze, D., Siamantziouras, A., Spehn, E., Terry, A., Troumbis, A., Woodward, F., Yachi, S., & Lawton, J. (1999). Plant diversity and productivity experiments in European grasslands. Science, 286, 1123 1127. Hunton, J. E., Lippincott, B., & Reck, J. L. (2003). Enterprise resource planning systems: Comparing ﬁrm performance of adopters and non adopters. International Journal of Accounting Information Systems, 4(3), 165 184.

Technology Monoculture

29

Iyegha, D. A. (2000). The wisdom of traditional farming in tropical Africa: The Nigerian experience. Journal of Third World Studies, 17, 73 92. Karamouzis, F. (2001). Research note: Oracle’s vision for services. Gartner Research. Available at http://www.gartner.com/DisplayDocument?doc_cd ¼ 100269. Retrieved on August 17. Kay, E. (1996). Desperately seeking SAP support. Datamation, February 15, pp. 42 45. Kirk, J. (2009). Data breaches rose sharply in 2008 study says. Infoworld. Available at http:// www.infoworld.com/d/security central/data breaches rose sharply in 2008 says study 630. Retrieved on January 7. Lanza, R. C. (2000). The e mail virus is telling us things. New York Times, May 6, p. A14. Leib, S. (2002). Core values: Part II The lowdown on twelve ERP vendors who’s buying what, who’s making what, and why you need what they’re making. Available at http:// CFO.com. Retrieved on February 13. Malmi, T. (1999). Activity based costing diffusion across organizations: An empirical analysis in Finland. Accounting, Organizations and Society, 29, 649 672. Marchado, S. (2009). Does intercropping have a role in modern agriculture? Journal of Soil and Water Conservation, 64(2), 55A. Moran, N. (2003). Twin forces stir desktop debate. Financial Times, October 15, p. 4. Naeem, S., Chapin, F. S., III., Costanza, R., Ehrlich, P., Golley, F., Hooper, D., Lawton, J., O’Neil, R., Mooney, H., Sala, O., Symstad, A., & Tilman, D. (1999). Issues in ecology, No. 4: Biodiversity and ecosystem functioning: Maintaining natural life support process. Washington: Ecological Society of America. Pethia, R. (2001). Information technology Essential but vulnerable: How prepared are we for attacks? Hearings of the House Government Affairs Subcommittee on Government Efﬁciency, Financial Management and Intergovernmental Relations. Available at http:// www.nist.gov/hearings/2001/inftech.html. Retrieved on September 26. Ribiero, J. (2010). India’s outsourcing exports to hit $50 billion in 2010. Infoworld Available at http://www.infoworld.com/t/ﬁnancial results/indias outsourcing exports hit 50 billion in 2010 107. Retrieved on February 4. Rosencrance, L. (2002). Hacker duo says they hack for sake of national security. Computer world. Available at http://www.computerworld.com/s/article/70728/Hacker_duo_says_ they_hack_for_sake_of_national_security. Retrieved on May 2. Rotman/TELUS. (2009). 2009 Rotman TELUS joint study on Canadian IT security practices. Toronto: Rotman School of Management, University of Toronto. SAP. (2010). SAP history. Available at http://www.SAP.com. Retrieved on March 20. Schwartz, M. (2002). Corning Inc. ERP plan cuts costs at factories. Computerworld. Retrieved on March 11. Scott, J. C. (1998). Seeing like a state. New Haven: Yale University Press. Slater, D. (1999). An ERP package for you y and you y and even you. CIO Magazine, February 15, pp. 12, 30. Songini, M. L. (2002a). GM locomotive unit puts ERP rollout back on track. Computerworld. Available at http://www.computerworld.com/s/article/68169/GM_locomotive_unit_puts_ ERP_rollout_back_on_track. Retrieved on February 11. Songini, M. L. (2002b). Teddy bear maker prepares for second attempt at ERP rollout. Computerworld. Retrieved on February 4, p. 16. Sutton, S. G., Arnold, V., & Hunton, J. E. (1999). On the death and dying of originality in the workplace: A critical view of enterprise resource planning systems’ impact

30

CHARLES CULLINAN ET AL.

on workers and the work environment. Critical Perspectives on Accounting Symposium, New York. Tilman, D. (1999). The ecological consequences of changes in biodiversity: A search for general principles. Ecology, 80(5), 1455 1474. United States National Research Council, Committee on Genetic Vulnerability of Major Crops, Agricultural Board, Division of Biology and Agriculture. (1972). Genetic vulnerability of major crops. Washington, DC: National Academy of Sciences. Verton, D. (2002). New FBI survey ﬁnds cybercrimes up, reporting ﬂat. Computerworld. Available at http://www.computerworld.com/s/article/69952/New_FBI_survey_ﬁnds_ cybercrimes_up_reporting_ﬂat. Retrieved on April 8. Zhu, Y., Chen, H., Fan, J., & Wang, Y. (2000). Genetic diversity and disease control in rice. Nature, 406, 718 722.

THE ROLE OF CONFIDENCE IN TAX RETURN PREPARATION USING TAX SOFTWARE Amy M. Hageman ABSTRACT This chapter investigates the nature of tax preparers’ conﬁdence, as well as how the introduction of a tax decision support system (TDSS) affects tax preparers’ conﬁdence levels. Psychological theories of conﬁdence (e.g., Einhorn & Hogarth, 1978) are drawn upon to develop predictions regarding the role of process (ex ante) and outcome (ex post) conﬁdence in tax return preparation. An experimental methodology is used with 114 inexperienced and experienced participants who prepare an individual income tax return manually or with tax preparation software (a TDSS). Less-experienced tax preparers have lower levels of ex-ante conﬁdence and are more likely to be overconﬁdent in the accuracy of their performance. Furthermore, when examining only the participants who made errors in their tax return preparation task, those that prepare the return with the TDSS are signiﬁcantly more likely to be overconﬁdent in their performance. These results support the predictions of Noga and Arnold (2002) and suggest that inexperienced users’ over-reliance on a TDSS (Masselli, Ricketts, Arnold, & Sutton, 2002) may be due to individuals’ overconﬁdence in the accuracy of their performance with the software.

Advances in Accounting Behavioral Research, Volume 13, 31 57 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013006

31

32

AMY M. HAGEMAN

INTRODUCTION Tax return preparation software is considered to be an interactive, intelligent tax decision support system (TDSS) that assists tax preparers in their tax return preparation process (Masselli et al., 2002; Noga & Arnold, 2002). In recent years, the use of such software has skyrocketed. Only 8% of returns were ﬁled using personal tax preparation software in 1993, but by 2003, this number grew to 25% (Guyton, Korobow, Lee, & Toder, 2005; Toder, 2005). Even more dramatically, the Internal Revenue Service (IRS) recently reported that nearly 32 million taxpayers e-ﬁled from their home computers in the 2009 ﬁling season (IRS, 2009), nearly double the volume from 2006 (IRS, 2006). Similarly, tax professionals continue to rely on tax preparation software and e-ﬁling to prepare their clients’ returns (IRS, 2009). Part of the popularity of the use of TDSS by novice and experienced tax preparers alike is that such software is perceived to increase tax preparers’ accuracy in preparing their return and thus may increase users’ sense of conﬁdence in their tax preparation abilities. A past advertising campaign by one software manufacturer, Intuit, included a direct appeal to users’ conﬁdence in its press release, claiming that its product (TurboTax) ‘‘Gives You Conﬁdence Your Taxes Are Done Right’’ (Intuit, 2003). Such direct appeals to users’ conﬁdence levels have continued. During the beginning of the 2009 ﬁling season, TurboTax’s homepage included the claim that its software ‘‘lets you ﬁle with conﬁdence,’’ as well as guaranteeing ‘‘100% accuracy’’ for users (http://www.turbotax.com). Thus, tax preparation software such as TurboTax is marketed as giving users the conﬁdence to accurately prepare a tax return. This suggests that conﬁdence is an important part of the tax return preparation process and that a TDSS assists with the process by increasing tax preparers’ conﬁdence levels. Despite the potential importance of conﬁdence in a multi-step, semistructured, knowledge-based task such as tax return preparation, this factor remains an under-studied variable. Although some psychology research has investigated the role of conﬁdence in judgment and decision making (JDM), such research does not always clearly distinguish between ex-ante conﬁdence (before the actual performance of the task) and ex-post conﬁdence (subsequent to the actual task performance) (Bonner, 2008). Thus, it is unclear whether increases in ex-ante conﬁdence would actually inﬂuence tax preparers’ conﬁdence in their ability to prepare an accurate tax return, and whether this same condition would be true with the introduction of a TDSS. Even more importantly, little is known regarding whether tax preparers’

Conﬁdence in Tax Return Preparation Using Tax Software

33

levels of ex-post conﬁdence are appropriately calibrated. If the introduction of a TDSS indeed increases tax preparers’ conﬁdence levels, such increases in conﬁdence could be detrimental if individuals develop a sense of overconﬁdence (i.e., greater perceived accuracy on a given task than the actual accuracy of the performance; Lichtenstein, Fischhoff, & Phillips, 1982). The introduction of a TDSS may be particularly problematic because less-experienced tax preparers often fail to function at the same level as more experienced preparers (Noga & Arnold, 2002). The purpose of this study is two-fold. First, this study investigates the multi-faceted role of conﬁdence in the tax return preparation process for tax preparers of varying experience levels, and for unaided (without a TDSS) and aided (using a TDSS) preparers. Second, this study deepens and extends conceptualizations of conﬁdence by considering how this variable jointly functions as both an input and an output of the decision process. This study investigates questions regarding the role of conﬁdence in the tax return preparation process through an experimental methodology with 114 inexperienced and experienced tax preparers, each of whom prepare an income tax return either unaided (manually) or aided with a commercially available TDSS. Results indicate that levels of ex-ante conﬁdence inﬂuence tax preparers’ levels of accuracy in the aided (but not manual) condition. More experienced tax preparers had higher levels of ex-ante conﬁdence and were less likely to be overconﬁdent in the accuracy of their performance. Interestingly, of the tax preparers who made errors in their tax return preparation, individuals using a TDSS were much more likely to be overconﬁdent than manual preparers, supporting the predictions of Noga and Arnold (2002). This phenomenon appears to be driven by the fact that TDSS users were more likely to make input errors that affected multiple facets of the tax return. This chapter contributes to the literature by considering how the introduction of a TDSS cognitively inﬂuences the tax return preparation process and by extending existing theorizations of conﬁdence to develop a more complete picture of this variable. This research is important for several reasons. First, most prior studies have examined either ex-ante or ex-post conﬁdence, leaving open questions regarding the role of this variable in JDM processes; this study suggests that conﬁdence functions as both an input and an output of decisions. Second, overconﬁdence is one of the most problematic biases in JDM; results can be disastrous if individuals fail to realize the inaccuracy of their task performance (Plous, 1993). Although some evidence suggests that decision aids may enhance overconﬁdence in probabilistic tasks (e.g., Davis & Kottemann, 1994), this study indicates that

34

AMY M. HAGEMAN

the phenomenon of overconﬁdence continues in a complex, non-probabilistic task such as tax return preparation. Third, the explosion of the use of TDSS raises important questions regarding how such software inﬂuences the tax return preparation process, as well as whether such user perceptions of greater accuracy are beneﬁcial or detrimental. The results of this study suggest that the ﬁnding in Masselli et al. (2002) that inexperienced tax preparers tend to over-rely on a TDSS is due to users’ overconﬁdence in the accuracy of their performance when using tax preparation software. Thus, although the use of a TDSS does improve tax return preparation accuracy of both inexperienced and experienced tax preparers (Noga & Arnold, 2002), the results of this study indicate that overconﬁdence in performance with a TDSS may eventually lead to technology dominance for users who do not fully understand the task (Noga & Arnold, 2002; Arnold & Sutton, 1998). The remainder of this chapter is organized as follows. The next section presents prior literature and develops this study’s hypotheses. The third section presents the research method for this experimental study. The fourth section provides an analysis of the results. The ﬁnal section concludes, with a discussion of the study’s implications, limitations, and opportunities for future research.

PRIOR LITERATURE AND HYPOTHESES DEVELOPMENT This study draws upon psychological theories of conﬁdence in seeking to understand tax preparers’ conﬁdence in their tax preparation process. In general, conﬁdence may be conceptualized in two different ways: as a process variable that is treated as a determinant of performance or as an outcome variable that reﬂects individuals’ conﬁdence in their previous task performance (see Bonner, 2008). Process conﬁdence refers to an individual’s ex-ante (beforehand) conﬁdence, which represents an individual’s personal conﬁdence in his or her ability to perform a task (Pincus, 1991; Whitecotton, 1996). Outcome conﬁdence refers to an individual’s ex-post (after the fact) conﬁdence in performance (Einhorn & Hogarth, 1978; Lichtenstein et al., 1982). Exploring both of these roles is important in understanding the multifaceted function of conﬁdence. Prior literature on these conceptualizations of conﬁdence is discussed below.

Conﬁdence in Tax Return Preparation Using Tax Software

35

Process View of Conﬁdence According to Bonner (2008), the level of conﬁdence that an individual brings to a task (process conﬁdence) can have signiﬁcant consequences on the quality of decision making. However, Bonner (2008, p. 94) emphasizes that the effects of conﬁdence on JDM are not well understood, since most studies ‘‘tend not to examine any effects of this conﬁdence on subsequent JDM.’’ Thus, although some prior studies in accounting have examined conﬁdence as a process variable in individuals’ JDM, such studies still leave open questions regarding the inﬂuence of conﬁdence on JDM. The studies of Pincus (1991) and Whitecotton (1996) are examples of accounting research that have conceptually examined conﬁdence as a process variable. As conceptualized by Pincus (1991), process conﬁdence serves as a ‘‘stopping point’’ in auditor judgment, as auditors continue to gather necessary information until reaching a pre-established internal level of conﬁdence. However, Pincus (1991) measured individual conﬁdence in the decision subsequent to the actual decision, and did not speciﬁcally measure conﬁdence as an input into the judgment process itself; thus, the results are unclear whether ex-ante conﬁdence (before the task) would have inﬂuenced the task performance itself. Similarly, Whitecotton (1996) examined the relationship between an individual’s personal conﬁdence in the ability to perform a task (before engaging in the actual task) and reliance on a decision aid, ﬁnding a strong inverse relationship between personal conﬁdence and decision aid reliance. Whitecotton (1996) also did not ﬁnd a relationship between ex-ante conﬁdence and performance accuracy. Thus, although Whitecotton (1996) established that personal conﬁdence in ability level is an input in judgment tasks, results of the study suggest that process conﬁdence may not directly inﬂuence performance accuracy for a given task. Overall, few accounting studies have speciﬁcally examined the role of process conﬁdence in the performance of a task. However, motivation theory suggests that higher levels of process conﬁdence may lead to increased performance accuracy by increasing an individual’s motivation to succeed in a given task (Benabou & Tirole, 2002). Similarly, lower levels of conﬁdence may be detrimental to performance accuracy, because ‘‘when people expect to fail, they fail quite effectively’’ (Salancik, 1977; Benabou & Tirole, 2002, p. 873). While not examining performance accuracy per se, sports psychology literature has speciﬁcally studied the degree to which ‘‘self-conﬁdence’’ (i.e., process conﬁdence) may inﬂuence competitive performance. Woodman and Hardy (2003, p. 443) deﬁne ‘‘self-conﬁdence’’ as ‘‘one’s belief in meeting the challenge of the task to be performed,’’ and,

36

AMY M. HAGEMAN

in a meta-analysis of 48 studies on competitive performance, document a signiﬁcant, large mean effect size for the positive relationship between selfconﬁdence and performance. The relationship between process conﬁdence, motivation, and performance suggests that individuals with higher levels of ex-ante conﬁdence in their ability to perform a given task will have higher levels of task performance. It therefore follows that tax preparers with higher levels of ex-ante conﬁdence in their ability to accurately prepare a tax return will perform this task with greater accuracy. This leads to the ﬁrst hypothesis: H1. Tax preparers with higher levels of conﬁdence in their ability to accurately prepare a tax return will demonstrate higher levels of accuracy in their performance of the task. Individuals may differ in the levels of process conﬁdence that they bring to a given task. One reason for such differences may be due to varying levels of experience. The theory of the illusion of validity postulates that greater experience with a task of moderate or high difﬁculty may increase an individual’s conﬁdence in his or her performance on that particular task due to the wider base of knowledge acquired by the individual (Einhorn & Hogarth, 1978). Over time, greater task experience logically leads to higher levels of process conﬁdence. This suggests that tax preparers that have greater experience in preparing tax returns would be more likely to have greater ex-ante conﬁdence in their ability to perform such a task. This leads to the second hypothesis: H2. Tax preparers with greater tax return preparation experience will be more conﬁdent in their ability to accurately prepare a tax return than tax preparers with less tax return preparation experience.

Outcome View of Conﬁdence In addition to a process variable, conﬁdence may also be conceptualized as an outcome variable. Most studies in psychology that have examined conﬁdence have used this conceptualization, assessing an individual’s ex-post conﬁdence in the performance of a particular task. Under this perspective, conﬁdence is described as the degree of belief in the accuracy of task performance (Lichtenstein et al., 1982).

Conﬁdence in Tax Return Preparation Using Tax Software

37

Outcome Conﬁdence and the Role of Calibration An evaluation of the appropriateness of an individual’s level of ex-post conﬁdence in the performance of a task can be made by comparing the level of conﬁdence to the actual accuracy of a decision (Oskamp, 1965, 1982). Some psychologists label the correspondence between these constructs as the degree of an individual’s calibration. Perfect calibration exists when an individual’s self-assessed probability of the accuracy of his or her performance on a given task is precisely equal to the actual accuracy of the performance (Lichtenstein et al., 1982). Perfect calibration is rare (Einhorn & Hogarth, 1978), resulting in overconﬁdence or underconﬁdence in performance. Overconﬁdence exists when an individual’s self-assessed probability of the accuracy of performance on a task exceeds the actual accuracy, whereas underconﬁdence is characterized by greater accuracy in performance than self-assessed accuracy (Oskamp, 1965, 1982; Lichtenstein et al., 1982, p. 308). Evidence of overconﬁdence is pervasive throughout the psychology literature (see Lichtenstein et al., 1982; McGraw, Mellers, & Ritov, 2004). In general, although individuals may display underconﬁdence on easy tasks, overconﬁdence is the most pronounced for tasks of moderate or high difﬁculty (Brenner, Koehler, Liberman, & Tversky, 1996). Overconﬁdence also abounds due to representativeness, ability, and internal coherence biases (Hogarth, 1980). Taken together, these prior psychology studies suggest that individuals are likely to be overconﬁdent when assessing the accuracy of task performance of a task of moderate complexity, such as tax return preparation. The third hypothesis, a control hypothesis, is: H3. Tax preparers will demonstrate overconﬁdence in their self-assessed accuracy of their preparation of a tax return. Understanding the factors that can inﬂuence individuals’ levels of overconﬁdence is of critical importance, as, according to Plous (1993), ‘‘no problem in judgment and decision making is more prevalent and more potentially catastrophic than overconﬁdence’’ (Bonner, 2008, p. 93). Two such factors suggested by prior research that may inﬂuence the level of overconﬁdence are experience and the use of a decision aid, such as a TDSS. Outcome Conﬁdence and Experience Several psychological studies have explored the relationship between experience and overconﬁdence. Some prior literature has suggested that

38

AMY M. HAGEMAN

individuals with greater task experience are more prone to overconﬁdence. Over time, greater task experience may lead to increased levels of ex-post conﬁdence in performance that could exceed the corresponding improvements in performance accuracy (Einhorn & Hogarth, 1978; Grifﬁn & Varey, 1996). For instance, Arkes, Dawes, and Christensen (1986) found that individuals with a higher knowledge base in a particular domain were more prone to overconﬁdence in their performance than those with a moderate level of knowledge, as measured by the degree of miscalibration between the self-assessed accuracy of performance and actual performance.1 Alternatively, other studies have found that experienced individuals may be less prone to overconﬁdence than those with little task experience. Oskamp (1962) found that less-experienced decision makers had greater degrees of miscalibration than experienced decision makers. Individuals with greater experience may have higher self-awareness of the potential accuracy of their performance (Kahneman & Tversky, 1982). Others suggest that the degree of calibration, or the relationship between conﬁdence levels and accuracy, is strongest for experienced performers performing relatively easy tasks who receive complete feedback regarding the accuracy of their decisions (Fischer & Budescu, 2005). Thus, experts in a particular domain may be less prone to overconﬁdence than novices due to greater task experience (Keren, 1987). Closer analysis shows a common theme on the relationship between task experience and overconﬁdence. The theory of the illusion of validity states that overconﬁdence is the most pronounced for tasks of moderate or high difﬁculty, whereas individuals performing easier tasks generally suffer from underconﬁdence (Einhorn & Hogarth, 1978). As an individual’s performance on a given task generally determines whether the task is difﬁcult or easy, it follows that an individual that performs well on a particular task is less likely to be overconﬁdent due to the task’s ease (Brenner et al., 1996; Fischer & Budescu, 2005). Individuals with greater task experience are likely to demonstrate improved calibration as compared to individuals with less task experience due to the reduced difﬁculty level of the task. Likewise, tax preparers with greater task experience are likely to demonstrate improved calibration as compared to less-experienced tax preparers. This leads to the fourth hypothesis: H4. Tax preparers with less tax return preparation experience will demonstrate higher levels of overconﬁdence in their self-assessed accuracy of their tax return preparation than tax preparers with more tax return preparation experience.

Conﬁdence in Tax Return Preparation Using Tax Software

39

Outcome Conﬁdence and Decision Aids An additional variable that may be associated with overconﬁdence is the use of a decision aid. Evidence is mixed regarding whether participants are more or less likely to exhibit overconﬁdence when using a decision aid to perform a task. Indeed, Rose (2000) calls for future studies to investigate alternative theories and settings to understand the conditions in which overconﬁdence when using a decision aid may be present. On the one hand, the interaction between individuals and computerized machines, including decision aids, has been characterized as simulating group interaction (Woods & Roth, 1988; Kasper, 1996). Others have suggested that an intelligent decision aid approximates an ‘‘electronic colleague’’ and that collaboration between such an aid and the user is similar to the type of interaction between a two-person group (Arnold & Sutton, 1998). Thus, the interaction of multi-person groups is a useful analogy in describing the interaction between individuals and collaborative decision aids.2 In general, the interactive nature of groups leads to group decisions that are typically even more conﬁdent and accurate in judgments than individual decisions (Sniezek & Henry, 1989). Other studies have demonstrated that groups tend to be better calibrated in the conﬁdence of their decisions than individual decision makers (e.g., Ahlawat, 1999). Thus, some evidence suggests that groups may be less prone to overconﬁdence than individuals, implying that overconﬁdence may be less prevalent among individuals using decision aids to perform a task. On the other hand, several studies in the systems literature have documented that the use of such aids may increase users’ overconﬁdence in the accuracy or quality of such decisions (e.g., Davis & Kottemann, 1994; Kottemann, Davis, & Remus, 1994; Kahai, Solieri, & Felo, 1998). This overconﬁdence may be due to the illusion of control (Langer, 1975) that individuals exhibit when overvaluing the usefulness of a decision aid to perform a probabilistic task. Furthermore, as suggested by Kasper (1996), interaction with a decision aid may increase user overconﬁdence due to the publicized expertise of such a system, particularly in regard to the system’s inquirability (cuing the user to particular alternatives or decisions). Thus, compared to unaided users, decision aid users may overweight the extent to which their performance increases with such an aid and may be more likely to be overconﬁdent in their task performance. Overall, using a decision aid tends to improve the accuracy of performance, but also increases users’ ex-post conﬁdence in the accuracy of their performance. Whether the higher degree of conﬁdence from using a decision aid such as a TDSS would exceed the increased accuracy resulting

40

AMY M. HAGEMAN

from the use of such an aid is unclear. Speciﬁcally, tax preparers may demonstrate increased accuracy in performance when using a TDSS (Noga & Arnold, 2002), but the increased conﬁdence brought about by the use of the TDSS may outweigh improvements in performance. Thus, the following two-sided hypothesis is proposed: H5. Tax preparers will have differing levels of overconﬁdence in the accuracy of their tax return preparation based on whether they prepare the return manually or with tax preparation software.

RESEARCH METHOD To examine the ﬁve hypotheses, an experiment was conducted using a between-subjects design with two treatment conditions. The section below details the experimental task, participants, experimental procedures, and the operational measures of variables.

Experimental Task The experimental task consisted of the completion of an individual tax return for a hypothetical family either with a commercially available TDSS (TurboTax, a type of tax preparation software) or manually with paperbased forms. Tax preparation software represents a TDSS with high external validity due to its widespread commercial availability (Masselli et al., 2002; Noga & Arnold, 2002). This experimental task itself was based on the complex return developed in Noga and Arnold (2002) and was modiﬁed to include a married couple ﬁling jointly with two dependents, W-2 wages, itemized deductions, a child tax credit, Schedule C income, and self-employment tax. All participants were randomly assigned to two treatment groups: the aided group (i.e., with the TDSS) and the unaided group (i.e., the manual group). Participants in both aided and unaided condition received the same basic taxpayer information and supporting documentation. Following Noga and Arnold (2002), participants in the unaided condition also received the following forms and instructions: 1040, 1040-A, 1040-EZ, Schedule A, Schedule B, Schedule C, Schedule C-EZ, Schedule D, Self-Employment Tax, and Earned Income Credit. As with Noga and Arnold (2002), additional forms were provided to this group to determine whether participants were

Conﬁdence in Tax Return Preparation Using Tax Software

41

capable of selecting the appropriate forms. The instructions for this task were adapted from Noga and Arnold (2002) and Masselli et al. (2002).

Participants The participants in this research study were students from a large southeastern university. Two groups of students participated in the study: students currently enrolled in the ﬁrst undergraduate course in taxation and students currently enrolled in a graduate-level taxation course. A total of 69 students were enrolled in the graduate taxation course, and 114 students were enrolled in the undergraduate taxation course. Originally, 132 participants volunteered to participate in the study; the responses of 114 participants could be used in the ﬁnal analysis. Of the 18 participants that could not be included, ﬁve volunteers did not report to the experiment, six participants did not complete all required experimental materials, three participants either failed the manipulation check or did not answer the question, three responses had corrupted electronic ﬁles that could not be read at a later date, and one participant completed the experimental materials out of order. Of these 114 participants, 39 were students currently enrolled in the graduate-level taxation course, whereas 75 of the participants were currently enrolled in the undergraduate-level course. Among participants from the graduate level course, 41% were master’s of accounting students, 33.3% were upper-level undergraduate accounting students, 23% were master’s of taxation students, and 2.6% were MBA students. By comparison, 80% of the participants from the undergraduate course were accounting majors, 13% were ﬁnance majors, and the remainder were other business majors. Table 1 summarizes information regarding the two groups of participants. The more experienced group (students in the graduate-level taxation course) was signiﬁcantly (at po0.05) older, had a higher grade point average (GPA), and had completed more taxation courses. Participants in the more experienced group were also more likely to prepare both their own tax return and returns for third parties and had completed a higher volume of tax returns in the past than the inexperienced group. Although the more experienced group reported greater familiarity with TurboTax software, groups did not differ (at po0.05 signiﬁcance) in whether they had used tax software in the past or in their familiarity with the current TurboTax advertising campaign. Groups also did not display any signiﬁcant difference in prior tax work experience. As a result, these groups serve as proxies for tax preparers with varying levels of experience.3

42

AMY M. HAGEMAN

Demographic Information.

Table 1.

Inexperienced (n ¼ 75)

Experienced (n ¼ 39)

24.23 years 56% 3.33 1.19 45% 31% 2.53 Months 1.29 49% 1.95 2.29

27.50 years 67% 3.48 2.44 82% 51% 5.56 Months 2.18 67% 2.74 2.21

Average age Gender (females) Average self reported GPA Average tax classes taken Prepare tax return own Prepare tax return other parties Average tax work experience Average tax returns prepareda Previously used tax software Familiarity with TurboTaxb Familiarity with TurboTax advertisingb

Signiﬁcantly different between the two groups at po.05 (two tailed). a

6 point scale, measured with 0 ¼ none; 1 ¼ 1 5 returns; 2 ¼ 6 10 returns; 3 ¼ 11 15 returns; 4 ¼ 16 20 returns; and 5 ¼ more than 20 returns. b 7 point scale, measured with 1 ¼ not at all and 7 ¼ very.

Table 2.

Inexperienced Experienced Overall

Number of Participants.

Unaided (Manual) Group

Aided (TDSS) Group

Overall

39 19 58

36 20 56

75 39 114

Experimental Procedures Each participant in the inexperienced or experienced group was randomly assigned to one of the treatment conditions (aided or unaided) and was informed of the location of their assigned session. To maximize participation, ﬁve different experimental sessions were held. Each session consisted of both an aided and an unaided group, held in separate locations. Refer to Table 2 for a breakdown of the participants.4 Participants in the unaided condition completed the experiment in a traditional classroom environment. The researcher or an assistant provided a general overview of the study and distributed a set of four packets to each participant that were to be completed in a pre-speciﬁed order. Each packet contained an experimental task and a legal-sized envelope. Participants were advised to open each packet, complete the experimental task, and seal the completed task inside the legal-sized envelope before proceeding to the next

Conﬁdence in Tax Return Preparation Using Tax Software

43

packet. At the end, the participants turned in the four, sealed, legal-sized envelopes and all of the other experimental materials. The ﬁrst packet contained a pre-experimental questionnaire to obtain (1) information regarding participants’ background experience in tax preparation and (2) an assessment of participants’ conﬁdence in their ability to prepare a tax return either with or without tax preparation software, as well as their conﬁdence in the accuracy of the tax preparation software itself. The second packet contained the experimental task itself and included taxpayer information, supporting documentation, forms, and instructions. This task was designed to measure participants’ accuracy in their tax preparation decisions. The third packet contained a postexperimental questionnaire gathering additional demographic information, participants’ conﬁdence in their performance of the task, and their conﬁdence in the accuracy of tax preparation software. The fourth packet contained a copy of the correctly prepared tax return (determined through the consensus of the researcher and two taxation professors). This ﬁnal packet also contained a short questionnaire with questions measuring participants’ future conﬁdence in their ability to complete a task both with and without tax preparation software and the manipulation check. Participants in the aided condition completed the experiment in a computer-laboratory setting. Their packets were identical to those in the unaided group, with the exception of the material in the second packet. The participants in the aided group did not have hard copies of the forms or instructions, but instead received a ﬂoppy diskette and copies of brief instructions of how to launch TurboTax in the laboratory. Furthermore, these participants completed the experimental task using TurboTax and then sealed the ﬂoppy diskette with a saved copy of the prepared return in one of the legal-sized envelopes. The experiment was pre-tested by nine graduate students. On the basis of feedback received by pre-testers, some changes in wording were made to questionnaire items.

Measurement of the Variables One dependent variable of interest was performance accuracy (H1). This was measured as: (1) the number of errors made in the completion of the experimental task and (2) the absolute value of the dollar amount of the errors (Noga & Arnold, 2002).

44

AMY M. HAGEMAN

Conﬁdence is the multi-dimensional construct of interest. Both aspects of ‘‘conﬁdence’’ were measured on a 100 point-scale (Whitecotton, 1993; Whitecotton, 1996) ranging between ‘‘not at all’’ (0%), ‘‘moderately’’ (50%), and ‘‘extremely’’ (100%). Participants used the following scale to answer the conﬁdence questions: Not at All

0%

Moderately

10%

20%

30%

40%

50%

Extremely

60%

70%

80%

90%

100%

Measures of ex-ante conﬁdence in performance (H1 and H2) were obtained before the experimental task itself and were measured based on responses to the following questions: ‘‘How conﬁdent are you in your ability to accurately prepare an individual income tax return manually without the assistance of tax software?’’ and ‘‘How conﬁdent are you in your ability to accurately prepare an individual income tax return with the assistance of tax software?’’ The measurement of outcome conﬁdence, or ex-post overconﬁdence (H3, H4, H5) concerns the miscalibration of errors, which is operationalized as the difference between the number of errors participants believed they made in the preparation of the tax return and the actual number of errors (see Brenner et al., 1996). Participants may be over- or under-conﬁdent in this measure. However, the scale used to measure this item gave participants the option to select ‘‘ﬁve or more errors,’’ which biases the measure toward calibration since participants who made extensive errors could select this option and still be classiﬁed as accurately calibrated. Concerns about this bias, however, may be partially remedied by the fact that the measures of participants’ ex-post conﬁdence in their accuracy of performance and the self-assessed number of errors are very highly correlated (r ¼ .779, po0.001). Finally, taxation course enrollment serves as a proxy for tax return preparation experience (H2, H4). Students currently enrolled in the undergraduate taxation course are considered inexperienced tax preparers, whereas students currently enrolled in the graduate taxation course are considered experienced tax preparers. Supplemental analyses use the number of tax returns previously prepared as an alternative measure of tax return preparation experience.

Conﬁdence in Tax Return Preparation Using Tax Software

45

RESULTS Data collected from the experimental procedures are used to test the ﬁve hypotheses pertaining to process conﬁdence and outcome conﬁdence in tax return preparation. Descriptive statistics are presented in Table 3, Panels A through D. Speciﬁcally, this table presents means and standard deviations for measures of ex-ante conﬁdence, performance accuracy as measured by the number of errors and absolute value of the dollar amount of errors, and miscalibration in conﬁdence (i.e., underconﬁdence or overconﬁdence). The table details the breakdown of these variables at both the group level (experienced or inexperienced) and the treatment condition (aided or unaided). Examination of these descriptive statistics demonstrates that the experienced group had higher levels of ex-ante conﬁdence in their ability to prepare a tax return manually, had higher levels of performance accuracy in the task itself, and was less likely to be overconﬁdent in the accuracy of their tax return preparation than the inexperienced group. Both inexperienced and experienced groups had higher levels of ex-ante conﬁdence in their tax return preparation abilities under aided rather than unaided conditions. Furthermore, for the inexperienced taxpayers, the aided group had higher levels of performance accuracy in obtaining an accurate tax liability than the unaided group.

Process View of Conﬁdence – H1 and H2 Hypothesis 1 predicts that tax preparers’ ex-ante conﬁdence in their ability to accurately prepare a tax return (both with and without a TDSS) will be positively associated with performance accuracy (as measured by both the number and the value of errors). Thus, four regressions are run to separately assess the aided and unaided groups and the two measures of performance accuracy. Control variables for the number of tax returns previously prepared, tax work experience, and taxation classes are also included. As shown in Table 4, results differ between the unaided and aided groups. For the unaided (manual) group, the relationships between ex-ante conﬁdence in performance without software and both the number and dollar value of errors are insigniﬁcant. Thus, increased conﬁdence does not improve the manual preparation of tax returns. However, the control variable for the number of tax returns previously prepared is statistically signiﬁcant (po0.01) in the analysis of the number of errors made by the unaided group; participants who have prepared more returns in the past make fewer errors.

46

AMY M. HAGEMAN

Table 3. Group

Descriptive Statistics.

Unaided Mean (SD)

Aided Mean (SD)

Overall Mean (SD)

Panel A: process conﬁdence (ex ante) prior to using tax preparation softwarea Inexperienced 52.99, 76.65 64.82 (26.30) (22.05) (21.84) 81.46 75.80 Experienced 70.13, (18.23) (20.26) (16.86) Overall 58.85 78.30 68.57 (25.13) (21.49) (20.87) Panel B: performance accuracy of participants (absolute value of dollar amount of errors) 1260.00 3999.65 Inexperienced 6258.56, (11592.22) (2415.33) (8876.19) 596.95 756.46 Experienced 924.37 (893.11) (1078.20) (1078.20) Overall 4692.71 1023.20 2890.14 (9842.65) (2082.87) (7373.88) Panel C: performance accuracy of participants (number of errors) 4.89 Inexperienced 5.82 (3.76) (5.36) 1.50 Experienced 2.47 (2.53) (3.41) Overall 4.72 3.68 (3.73) (4.99)

5.37 (4.59) 1.97 (3.01) 4.21 (4.41)

Panel D: overconﬁdence (underconﬁdence) of participantsb Inexperienced 1.67, 2.53, (2.90) (2.08) .70 Experienced .26 (2.31) (3.03) 1.21 1.88 Overall H3 (2.78) (1.54)

2.08 (3.56) .49 (2.67) 1.54 (3.36)

Signiﬁcantly different between inexperienced and experienced groups at po.05. Signiﬁcantly different between aided (with TDSS) and unaided (without TDSS) conditions at

po.05. Level of miscalibration of the participants (the difference between actual and perceived errors) is signiﬁcantly different from 0 (i.e., perfect calibration) at po.001. Positive (negative) values represent overconﬁdence (underconﬁdence). a All participants answered questions regarding process conﬁdence in tax return preparation under both aided and unaided conditions. b Measured as the miscalibration of participants or the difference between actual errors and self perceived errors in the tax return preparation task.

47

Conﬁdence in Tax Return Preparation Using Tax Software

Table 4.

Inﬂuence of Ex-Ante Conﬁdence on Performance Accuracy (H1). Unaided Group Number of Errorsa

Constant Ex ante conﬁdence (H1) Number of returns prepared Tax work experience Taxation classes Familiarity with software Adjusted R squared F statistic

Unaided Group Dollar Value of Errorsa

6.267 (1.246) .012 (.022) .933 (.334) .052 (.045) .396 (.461)

12102 (3582) 90.83 (64.36) 742.04 (961.5) 24.88 (128.2) 710.77 (1325)

.193 4.40

.041 1.610

Aided Group Number of Errorsa 11.454 (2.842) .072 (.036) .420 (.397) .040 (.069) .696 (.755) .036 (.368) .142 2.817

Aided Group Dollar Value of Errorsa 4221 (1238) 33.88 (15.5) 11.78 (173.1) 8.742 (30.13) 285.9 (328.8) 2.859 (160.3) .063 1.735

Notes: For the OLS regressions for the unaided group, ex ante conﬁdence refers to participants’ conﬁdence in their ability to accurately prepare a tax return without tax preparation software (manually). For the OLS regressions for the aided group, ex ante conﬁdence refers to participants’ conﬁdence in their ability to accurately prepare a tax return with tax preparation software (with a TDSS). po.10; po.05; po.01; two tailed tests, except for H , which is directional (one tailed). 1 a Coefﬁcients (standard errors) for OLS regression models, by dependent variable.

In the analysis of the aided condition, an additional control variable is included for participants’ familiarity with TurboTax (the TDSS in the study); ex-ante conﬁdence in performance with software is the independent variable in question. For participants completing a tax return with a TDSS, higher levels of ex-ante conﬁdence in performance result in fewer errors in both magnitude and dollar value (i.e., higher levels of performance accuracy) at a statistically signiﬁcant level (po0.05). Participants’ levels of conﬁdence appear to be a particularly important inﬂuence on performance accuracy when using a TDSS. None of the control variables are statistically signiﬁcant. Thus, Hypothesis 1 is supported for participants in the aided condition. Hypothesis 2 examines whether tax preparers with greater tax return preparation experience will be more conﬁdent in their ability to accurately prepare a tax return. To test this hypothesis, two separate ANOVAs are used with the dependent variables of ex-ante conﬁdence in performance

48

AMY M. HAGEMAN

without a TDSS (manually) and ex-ante conﬁdence in performance with a TDSS; the independent variable for both analyses is the inexperienced or experienced group.5 As shown in Table 5, Panel A, tax preparers in the more experienced group were signiﬁcantly more conﬁdent in their ability to prepare a tax return manually (po0.001). Interestingly, there were no such differences between the groups in their levels of ex-ante conﬁdence in their ability to prepare an accurate tax return with software (TDSS) as shown in Table 5, Panel B. This seems to suggest that inexperienced tax preparers believe that they can use a TDSS to perform at a higher level of accuracy, equivalent to a more experienced tax preparer. An alternative reason for the difference in the effect of experience on conﬁdence levels is that both groups demonstrated low levels of familiarity with tax preparation software. A supplemental analysis uses the number

Table 5.

Inﬂuence of Experience on Ex-Ante Conﬁdence (H2). F Statistic

Signiﬁcance Level (One Tailed)

Panel A: ANOVA, ex ante conﬁdence without TDSS (manually) 13.232 Experience level (H2) Adjusted R2 ¼ .098

o.001

Panel B: ANOVA, ex ante conﬁdence with TDSS (tax preparation software) 1.288 .259 Experience level (H2) Adjusted R2 ¼ .003

Panel C: regression analyses Constant Number of returns prepared (H2) Tax work experience Taxation classes

Ex ante conﬁdence without TDSS (manually)a

Ex ante conﬁdence with TDSS (software)a

45.00 (4.04) 6.446 (1.304) .137 (.215) 1.90 (2.23)

67.24 (4.05) 3.131 (1.235) .246 (.192) .361 (1.99) 2.59 (1.03) .166 6.617

Familiarity with software Adjusted R squared F statistic

.229 12.171

po.10; po.05; po.01; two tailed tests, except for H , which is directional (one tailed). 2 a

Coefﬁcients (standard errors) for OLS regression models, by dependent variable.

Conﬁdence in Tax Return Preparation Using Tax Software

49

of tax returns previously prepared as an alternative measure of tax return preparation experience, while controlling for tax work experience and taxation classes; the analysis of ex-ante conﬁdence when using a TDSS also controls for familiarity with TurboTax. Table 5, Panel C, shows the regression analyses that demonstrate that prior tax return preparation experience (i.e., the number of returns previously prepared) signiﬁcantly inﬂuences ex-ante conﬁdence both with and without the TDSS (both po0.01, one-tailed). Greater familiarity with the software also increases conﬁdence in the ability to prepare a return with a TDSS (po0.05). Overall, these results generally support H2 by indicating that tax preparers with greater tax return preparation experience are more likely to have greater degrees of ex-ante conﬁdence in tax return preparation; direct task experience is particularly important.

Outcome View of Conﬁdence – H3, H4, and H5 Hypothesis 3 is a control hypothesis used to establish that tax preparers will demonstrate overconﬁdence in their self-assessed accuracy of their preparation of a tax return. An examination of the descriptive statistics in Table 3, Panel D, demonstrates that on average, participants are overconﬁdent in their assessment of their performance, in that the actual number of errors in preparation exceeds their self-estimated number of errors. A t-test demonstrates that the overall measure of overconﬁdence is signiﬁcantly different from zero (po0.001). Thus, Hypothesis 3 is supported, in that participants are overconﬁdent in their self-assessed tax return preparation accuracy. Hypothesis 4 predicts that tax preparers with less tax return preparation experience will demonstrate higher levels of overconﬁdence in their selfassessed performance accuracy, whereas Hypothesis 5 predicts that tax preparers’ level of overconﬁdence in tax return preparation will vary based on whether they prepare the return manually or with software. Thus, these hypotheses examine the differences in participants’ estimated number of errors and the actual number of errors (i.e., overconﬁdence) to determine whether any of the conditions were more prone to overconﬁdence. These hypotheses are tested with an ANOVA. Results shown in Table 6, Panel A, indicate that participants in the inexperienced group are signiﬁcantly more overconﬁdent than those in the experienced group (po0.01), supporting the predictions of H4.6 However, no signiﬁcant differences emerge between participants in the aided and unaided conditions.

50

AMY M. HAGEMAN

Table 6.

Factors Inﬂuencing Overconﬁdence (H4 and H5).

Panel A: ANOVA, all participants Model Experience level (H4) Aided/unaided condition (H5) Experience aid interaction Adjusted R2 ¼ .038 Panel B: ANOVA, participants with errors Aided/unaided condition (H5) Adjusted R2 ¼ .189

F Statistic

Signiﬁcance Level

2.483 6.167 .995 .106

.065 o.01 .321 .745

16.807

o.001

Signiﬁcance level is one tailed for directional tests (H ) and two tailed for non directional 4

tests (H5).

Further analysis is performed to determine whether participants that completed the task accurately (i.e., without any errors) may be confounding the results for H5. Using a TDSS does appear to improve tax preparer accuracy, as 79% of tax preparers made errors on the return when preparing it manually, compared to only 41% of tax preparers that used a TDSS (signiﬁcantly different at po0.001). This suggests that overconﬁdence may be particularly problematic for tax preparers who were unable to perform the task accurately. Thus, additional analysis eliminates the participants that completed the task without any errors. Eliminating the participants who completed the task accurately results in 46 participants in the unaided condition and 23 participants in the aided condition that made errors in their preparation. A supplemental ANOVA considers whether there are any differences in overconﬁdence between aided and unaided users that made errors in their performance. Considering only the participants who made errors, those who completed the tax return using a TDSS have signiﬁcantly higher levels of overconﬁdence than those who completed the tax return manually (po0.001) as shown in Table 6, Panel B. Thus, using a TDSS appears to result in a false sense of overconﬁdence among users who make errors in their tax return preparation. Hypothesis 5 is therefore partially supported. Additional Analysis Additional analysis further explores the consequences of tax preparers’ overconﬁdence. One such area pertains to the type of errors made by

51

Conﬁdence in Tax Return Preparation Using Tax Software

participants in the aided and unaided conditions. For participants who made errors in their tax return preparation, no statistically signiﬁcant differences between aided and unaided participants emerged regarding the net magnitude of such errors. However, closer investigation reveals that certain types of errors were more prevalent for paper-based returns than ones completed with software. Table 7 provides a summary of the differences in errors made by participants in the aided and unaided conditions. Overall, tax preparers using a TDSS were signiﬁcantly more likely to omit certain items from the tax return at a statistically signiﬁcant level (all po0.05, two-tailed), including information about dependents, wages, and itemized deduction items on Schedule A. Such omissions for participants in the TDSS condition were likely a result of answering a question incorrectly in the electronic ‘‘interview’’ in the TDSS, which likely affects multiple items on the return. For example, a taxpayer who omits information regarding a dependent will pay more tax, due to the impact of this error on both dependency exemptions and the child tax credit. Thus, errors when using tax preparation software may be magniﬁed across the return. Other types of errors were more prevalent between the aided and unaided conditions. Tax preparers using the TDSS were more likely to calculate an incorrect amount for business expenses on the Schedule C-EZ, in large part due to the common tendency to treat mortgage interest (a personal expense) as a business expense. This tendency may have been more likely due to the electronic prompting of the TDSS interview, causing participants to misclassify the item to a much larger degree than in the manual condition. On the contrary, participants in the manual condition were much more likely to make errors in the calculation of tax amounts. In fact, of the participants Table 7.

Examples of Common Errors for Participants with Errors.

Type of Error

Dependents omitted from return Wages Child tax credit Schedule A sales tax Schedule A real estate tax Schedule A mortgage interest Schedule A charitable contribution Schedule C EZ business expenses po.10; po.05; po.01; two tailed tests.

% Error Without TDSS (Unaided)

% Error With TDSS (Aided)

0% 7% 76% 15% 11% 13% 13% 43%

13% 26% 22% 52% 52% 30% 35% 74%

52

AMY M. HAGEMAN

making an error in the manual condition, 76% either neglected to compute or made an error in the calculation of the child tax credit, resulting in an overpayment of tax. Overall, although participants were much less likely to make errors in their tax return preparation when using the TDSS, those who did make errors tended to have mistakes whose consequences were magniﬁed across the return or were a consequence of the nature of the interactive questionnaire in the TDSS. This helps to explain why tax preparers may develop a false sense of conﬁdence when using a TDSS. Another area of further investigation concerns the factors associated with overconﬁdence. Although this study focuses on the associations between inexperience and a TDSS and overconﬁdence, other variables may also be important in explaining this tendency. One such important factor is gender, as some evidence exists that men may be more prone to overconﬁdence, particularly in traditionally ‘‘masculine’’ ﬁelds (Bonner, 2008). Supplemental analysis (not tabulated) shows that this gender difference is also prevalent in the realm of tax return preparation, with male participants signiﬁcantly more likely to be overconﬁdent than female participants (po0.05, two-tailed). Further investigation shows that of the participants who made errors in their tax return preparation, male participants using the TDSS were particularly prone to overconﬁdence (p ¼ 0.06). Thus, addressing the potential effects of overconﬁdence is particularly important for male tax preparers.

DISCUSSION AND CONCLUSION This study investigates the role of conﬁdence in tax return preparation, and examines tax preparers’ process (ex-ante) and outcome (ex-post) conﬁdence in preparing a tax return both manually (without a TDSS) and with tax preparation software (with a TDSS). Hypothesis 1 investigates whether tax preparers with higher levels of ex-ante conﬁdence have higher levels of accuracy in preparing a tax return. Results demonstrate that this relationship is present for tax preparers using a TDSS, but not for those performing the task manually. Hypothesis 2 examines whether greater experience increases levels of ex-ante conﬁdence; results using prior tax return preparation experience support this prediction. Thus, this study adds to the literature by examining how conﬁdence may affect subsequent judgments and decisions (Bonner, 2008), by demonstrating that ex-ante conﬁdence is particularly important when preparing a tax return with software.

Conﬁdence in Tax Return Preparation Using Tax Software

53

Hypothesis 3 tests whether tax preparers are subject to miscalibration in the form of overconﬁdence; results support this prediction, indicating that tax return preparation is another ﬁeld where the ‘‘potentially catastrophic’’ problem of overconﬁdence arises (Plous, 1993). Whereas prior studies have tended to focus on domains whose outcome was at least partially probabilistic in nature (e.g., Davis & Kottemann, 1994), these results show that overconﬁdence is a problem even in non-probabilistic domains such as tax return preparation. Hypothesis 4 examines whether experience is related to levels of overconﬁdence. Results indicate that overconﬁdence abounds among individuals with less prior tax return preparation experience, supporting the view that when a task is of greater difﬁculty for an individual, overconﬁdence is more likely (Einhorn & Hogarth, 1978). Finally, Hypothesis 5 addresses whether overconﬁdence is more likely in tax return preparation when using a TDSS. Although there were no differences in overconﬁdence levels for the whole sample between the aided and the unaided conditions, an interesting result is that of the participants that made errors in their tax return preparation, participants using software were signiﬁcantly more likely to be overconﬁdent in the perceived accuracy of their return. Thus, although conﬁdence and accuracy both tend to increase in an interactive, computerized environment (Sniezek & Henry, 1989), tax preparers who are inaccurate in preparing returns using software are more likely to be overconﬁdent in performance and not perceive the full extent of their errors. This also conﬁrms the speculation of Noga and Arnold (2002) that tax preparation software may give the tax preparer a false sense of conﬁdence in the accuracy of the results, which may make them even less likely to be aware of any potential errors. These ﬁndings have both theoretical and practical implications. The pervasive nature of overconﬁdence (Lichtenstein et al., 1982; McGraw et al., 2004) is shown to exist in a complex, multi-step task such as task return preparation. That overconﬁdence in inaccurate performance is more likely when preparing a tax return using software suggests that the ‘‘illusion of control’’ phenomenon present when using a decision support system (e.g., Davis & Kottemann 1994) may also be present in the performance of a non-probabilistic based task. Finally, the results of this study suggest that the over-reliance on tax preparation software demonstrated by novice taxpayers in Masselli et al. (2002) may be due to individuals’ overconﬁdence in their accuracy of performance with the software. Over time, this overconﬁdence may lead to technology dominance for users that do not fully understand the task (Arnold & Sutton, 1998) and explains why

54

AMY M. HAGEMAN

individuals that are not accustomed to performing a task in a manual environment may be unable to do so (Noga & Arnold, 2002). Several practical implications also arise, particularly in highlighting the differences in tax return preparation in a manual or computerized TDSS environment. That overconﬁdence in inaccurate performance is more likely when a TDSS is used suggests that tax preparers may not fully understand the impact of errors. For instance, in the case of tax preparation software, incorrectly answering an item in the software’s questionnaire (such as information on dependents) may affect multiple items on the resulting tax return (such as dependency exemptions and the child tax credit, resulting in the tax preparer overstating the tax liability). Thus, errors in preparation may be magniﬁed when tax preparation software is used. This deﬁciency suggests an opportunity for educators to provide extensive training on the use of tax preparation software, particularly on the impact of errors on the ﬁnal return. Furthermore, because increased conﬁdence in performance with software was linked with more accurate performance, training on the use of such software in order to properly calibrate users’ conﬁdence levels is critical. Overall, tax preparers should not abandon the use of TDSS, but should be adequately trained regarding the systems’ limitations and the potential for pervasive errors to reduce overconﬁdence. The implications of the current study should be interpreted in light of its limitations. First, this study used student subjects as a proxy for inexperienced and experienced tax preparers. However, these groups still serve as proxies for tax preparers with varying levels of tax return preparation experience; furthermore, when applicable, alternative proxies for ‘‘experience’’ were also employed. Second, the scale that measured participants’ estimated number of errors made in the experimental task contained an option for ‘‘ﬁve or more errors,’’ which biases the measure toward calibration (i.e., against ﬁnding overconﬁdence). Despite this limitation, all experimental conditions still displayed a tendency toward overconﬁdence, and participants’ estimates of their predicted number of errors were signiﬁcantly correlated with ex-post conﬁdence in performance. Third, some variables are operationalized with single self-reported measurements. The limitations and implications of the current study suggest several avenues for future research. Future studies could determine the effects of training in the use of tax preparation software on tax preparers’ conﬁdence levels. Another possibility is to determine whether offering participants incentives for appropriate calibration could reduce overconﬁdence. Finally, researchers could examine the constructs of trust or face validity in relation

Conﬁdence in Tax Return Preparation Using Tax Software

55

to the TDSS to determine if such facets are necessary precursors of conﬁdence. Such future research can begin to examine some of the issues raised in this study.

NOTES 1. The ﬁndings of this particular study, however, should be interpreted with caution since the group with higher domain knowledge also was much less likely to rely on a decision aid with high predictive accuracy. It is possible that the greater knowledge base of the expert group was not the sole determinant of the group’s overconﬁdence (Arkes et al., 1986). 2. Although the interaction between an individual user and a computerized decision aid is a type of group processing (Woods & Roth, 1988), judgments regarding conﬁdence in the performance of a task will reﬂect the judgment of only one member of the group: the individual user. Nevertheless, the analogy of the group is a useful tool in understanding this phenomenon. 3. To motivate participants to attend to the experimental task, students in both the undergraduate and the graduate taxation classes were awarded 10 extra credit points (approximately 2% of the ﬁnal course grade) for making a good faith effort to complete the study without obvious or blatant carelessness. Nearly all participants appeared to be highly motivated and focused on the task, and instructors in each of the classes reported that the participants were interested in the outcome. 4. There were no statistically signiﬁcant differences in any demographic or tax experience questions between subjects assigned to the aided or unaided conditions (all pW0.10). 5. Separate ANOVAs, rather than a MANOVA, are used because the two dependent variables are very highly correlated (correlation coefﬁcient 0.601). 6. Results are robust to an alternative speciﬁcation of ‘‘experience,’’ as participants who had previously prepared more (fewer) tax returns in the past were also less (more) prone to overconﬁdence (po0.01).

ACKNOWLEDGMENTS I am very grateful for the helpful comments and suggestions from Vicky Arnold, Donna Bobek Schmitt, Charles Cho, Clark Hampton, Charlie Kelliher, Steve Sutton, and participants at the 2007 AAA Information Section JIS Research Development Workshop. I am also grateful for the assistance in carrying out the experimental procedures provided by Joe Canada and Will Hageman.

56

AMY M. HAGEMAN

REFERENCES Ahlawat, S. S. (1999). Order effects and memory for evidence in individual versus group decision making in auditing. Journal of Behavioral Decision Making, 12, 71 88. Arkes, H. R., Dawes, R. M., & Christensen, C. (1986). Factors inﬂuencing the use of a decision rule in a probabilistic task. Organizational Behavior and Human Decision Processes, 37, 93 110. Arnold, V., & Sutton, S. (1998). The theory of technology dominance. Advances in Accounting Behavioral Research, 1, 175 194. Benabou, R., & Tirole, J. (2002). Self conﬁdence and personal motivation. Quarterly Journal of Economics, 117(3), 871 915. Bonner, S. (2008). Judgment and decision making in accounting. Upper Saddle River, NJ: Pearson Prentice Hall. Brenner, L. A., Koehler, D. J., Liberman, V., & Tversky, A. (1996). Overconﬁdence in probability and frequency judgments: A critical examination. Organizational Behavior and Human Decision Processes, 65(3), 212 219. Davis, F. D., & Kottemann, J. E. (1994). User perceptions of decision support effectiveness: Two production planning experiments. Decision Sciences, 25(1), 57 78. Einhorn, H. J., & Hogarth, R. M. (1978). Conﬁdence in judgment: Persistence of the illusion of validity. Psychological Review, 85(5), 395 416. Fischer, I., & Budescu, D. V. (2005). When do those who know more also know more about how much they know? The development of conﬁdence and performance in categorical decision tasks. Organizational Behavior and Human Decision Processes, 98, 39 53. Grifﬁn, D. W., & Varey, C. A. (1996). Towards a consensus on overconﬁdence. Organizational Behavior and Human Decision Processes, 65(3), 227 231. Guyton, J. L., Korobow, A. K., Lee, P. S., & Toder, E. J. (2005). The effects of tax software and paid preparers on compliance costs. National Tax Journal, 58(3), 436 448. Hogarth, R. (1980). Judgment and choice. New York: Wiley. Internal Revenue Service (IRS). (2006). On line ﬁling sets new record; More than 17 million returns e ﬁled by home computer users. IR 2006 64 (April 14). Available at http:// www.irs.gov/newsroom/article/0,,id ¼ 156222,00.html Internal Revenue Service (IRS). (2009). Two out of three individuals now using IRS E File. IR 2009 99 (November 2). Available at http://www.irs.gov/newsroom/article/0,,id ¼ 214853,00.html Intuit. (2003). TurboTax for 2003 gives you conﬁdence your taxes are done right. Press release (December 2). Available at http://web.intuit.com/about_intuit/press_releases/ 2003/12 02a.html Kahai, S. S., Solieri, S. A., & Felo, A. J. (1998). Active involvement, familiarity, framing, and the illusion of control during decision support system use. Decision Support Systems, 23(2), 133 148. Kahneman, D., & Tversky, A. (1982). Intuitive prediction: Biases and corrective procedures. In: D. Kahneman, P. Slovic & A. Tversky (Eds), Judgments under uncertainty: Heuristics and biases (pp. 414 421). New York: Cambridge University Press. Kasper, G. M. (1996). A theory of decision support system design for user calibration. Information Systems Research, 7(2), 215 232. Keren, G. (1987). Facing uncertainty in the game of bridge: A calibration study. Organizational Behavior and Human Decision Processes, 39(1), 98 114.

Conﬁdence in Tax Return Preparation Using Tax Software

57

Kottemann, J. E., Davis, F. D., & Remus, W. E. (1994). Computer assisted decision making: Performance, beliefs, and the illusion of control. Organizational Behavior and Human Decision Processes, 57(1), 26 37. Langer, E. J. (1975). The illusion of control. Journal of Personality and Social Psychology, 32, 311 328. Lichtenstein, S., Fischhoff, B., & Phillips, L. D. (1982). Calibration of probabilities: The state of the art to 1980. In: D. Kahneman, P. Slovic & A. Tversky (Eds), Judgment under uncertainty: Heuristics and biases (pp. 306 334). Cambridge: Cambridge University Press. Masselli, J. J., Ricketts, R. C., Arnold, V., & Sutton, S. G. (2002). The impact of embedded intelligent agents on tax reporting decisions. Journal of the American Taxation Association, 24(2), 60 78. McGraw, P. A., Mellers, B. A., & Ritov, I. (2004). The affective costs of overconﬁdence. Journal of Behavioral Decision Making, 17, 281 295. Noga, T., & Arnold, V. (2002). Do tax decision support systems affect the accuracy of tax compliance decisions? International Journal of Accounting Information Systems, 3, 125 144. Oskamp, S. (1962). The relationship of clinical experience and training methods to several criteria of clinical prediction. Psychological Monographs, 76(28), 547. Oskamp, S. (1965). Overconﬁdence in case study judgments. The Journal of Consulting Psychology, 29, 261 265. Oskamp, S. (1982). Overconﬁdence in case study judgment. In: D. Kahneman, P. Slovic & A. Tversky (Eds), Judgments under uncertainty: Heuristics and biases (pp. 287 289). New York: Cambridge University Press. Pincus, K. V. (1991). Audit judgment conﬁdence. Behavioral Research in Accounting, 3, 39 65. Plous, S. (1993). The psychology of judgment and decision making. Philadelphia: Temple University Press. Rose, J. (2000). Behavioral decision aid research: Decision aid use and effects. In: V. Arnold & S. Sutton (Eds), Researching accounting as an information systems discipline (pp. 111 133). Sarasota, FL: American Accounting Association. Salancik, G. (1977). Commitment and the control of organizational behavior and belief. In: B. Staw & G. Salancik (Eds), New directions in organizational behavior (pp. 1 54). Chicago: St. Clair Press. Sniezek, J. A., & Henry, R. A. (1989). Accuracy and conﬁdence in group judgment. Organizational Behavior and Human Decision Processes, 43(1), 1 28. Toder, E. J. (2005). Changes in tax preparation methods, 1993 2003. Tax Notes, 100(6), 759. Whitecotton, S. M. (1993). Decision aid reliance as a determinant of accuracy in earnings forecasting judgments: The impact of contextual and decision maker characteristics. Unpublished Ph.D. Dissertation, University of Oklahoma. Whitecotton, S. M. (1996). The effects of experience and conﬁdence on decision aid reliance: A causal model. Behavioral Research in Accounting, 8, 194 216. Woodman, T., & Hardy, L. (2003). The relative impact of cognitive anxiety and self conﬁdence upon sport performance: A meta analysis. Journal of Sports Sciences, 21, 443 457. Woods, D. D., & Roth, E. (1988). Cognitive engineering: Human problem solving with tools. Human Factors, 30(4), 415 430.

NOVICE LEVEL KNOWLEDGE ACQUISITION USING A TECHNOLOGY-BASED EDUCATIONAL DELIVERY SYSTEM: THE ROLE OF EXPERIENTIAL PRACTICE Paul M. Goldwater and Kimberly A. Zahller ABSTRACT Increasing constraints on personnel and resources have led to a focus on alternative methods of transmitting knowledge to novices, whether university students or newly hired staff. This chapter focuses on one such alternative through the use of a technology-based educational delivery system (TBEDS). Prior research has addressed individual components of technology-supported systems or performed experiments of limited time and direct external relevance to the participants, but has not addressed the effect of a holistic approach to technology-based learning on users. This study capitalizes on a unique, holistically designed TBEDS to longitudinally examine the impact of systems on novices’ procedural knowledge acquisition under conditions of actual usage. The longitudinal data also illustrates the role of user-determined experiential practice on achievement as moderated by comfort with technology. The ﬁndings Advances in Accounting Behavioral Research, Volume 13, 59 88 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013007

59

60

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

indicate a strong relationship between the use of a TBEDS for repeated experiential practice and procedural knowledge acquisition. Individual components of effort (quantity of problems attempted, frequency of practice sessions, and quality of practice) are examined, with quantity being signiﬁcantly positively related to performance, as is quality when the user is accountable to an external authority for that quality.

INTRODUCTION Increasing enrollment, a rising shortage of academically qualiﬁed accounting professors, and drastic reductions in university budgets have led to burgeoning class sizes, often of more than 100 students, and greater teaching loads. In such conditions, instructors can no longer assign meaningful amounts of graded practice problems with rapid feedback, features that are both necessary for effective learning. To address this situation, universities are turning to various means of system-mediated delivery of homework and practice opportunities. The ultimate concern is whether users of this technology actually acquire the requisite knowledge. Research has consistently demonstrated that mastery and increased performance occurs when information is transferred from short-term into long-term memory through repeated practice of relevant, related-but-different problems (Bonner, 1990; Sweller, 1988). Experience from practice assists in the formation of mental schemas that speed both efﬁciency (recognition of ‘‘classes’’ of problems or solutions) and effectiveness (knowledge of how to appropriately apply the correct methodology) (Anderson, 1982; Anderson & Schooler, 1991; Sweller, 1993). Universities are not the only context in which this dilemma occurs: the effectiveness of technology-based educational delivery systems (TBEDS) in fostering novice knowledge acquisition, especially of problem-solving skills (procedural knowledge), is also directly relevant for practitioners. The accounting ﬁeld relies heavily on its members’ ability to make good judgments in domains that are highly unstructured and require the application of expert judgment. The accountant or auditor must be able to acquire the requisite declarative and procedural knowledge, organize it into appropriate schemas, and recall and apply it appropriately to varied decisions. As time and resource constraints have reduced the availability of senior staff to provide mentoring, training, or one-on-one explanations to junior staff, computerized training systems, knowledge management systems

Knowledge Acquisition Using a TBEDS

61

(KMS), and decision aids1 (DA) have been used to leverage organizational knowledge. Firms have found these systems to be at least as effective as conventional training and at a fraction of the cost (Gal & Steinbart, 1992). Thus, organizations increasingly rely on technology-based training to deliver task- and domain-speciﬁc knowledge, and junior staff accountants are expected to be self-motivated in using these systems. Concerns have been raised regarding a failure to develop independent problem-solving skills from the use of a DA (Arnold & Sutton, 1998), and these concerns are also applicable for TBEDS, which may provide too many ‘‘clues,’’ overly-limited or easily determined answers, or too extensive a framework for problem-solving. In such cases, technology may actually inhibit the acquisition of procedural knowledge. McCall, Arnold, and Sutton (2008) found that users of a KMS in an educational environment outperformed others who had used traditional knowledge acquisition methods in declarative knowledge and interpretive problem solving. However, the differential performance between the two groups disappeared when the support of the KMS was removed. These results are unsurprising as the actual time using the KMS was limited; the construction of independent, easily retrieved schemas in long-term memory requires repeated, effortful practice over time. Furthermore, the use of DAs predominantly designed for intermediate and expert users creates an inherent mismatch for novice users causing overreliance on the DA. According to the theory of Task-Technology Fit (Goodhue & Thompson, 1995) and the Theory of Technology Dominance (TTD) (Arnold & Sutton, 1998), systems that are not appropriately matched to the user’s knowledge level, the relevant task at hand, and the overall environment will not lead to user perceptions of utility and ease of use; consequently, the user will fail to (appropriately) use the system. Thus, the design of a TBEDS and the appropriate match of task, technology, and user are critical in fostering knowledge acquisition and problem-solving ability in novice users. System-based instructional support assists in adapting to a resourceconstrained environment that simultaneously demands a high level of problem-solving skill and specialized knowledge. In allowing users to move at their own pace, to return to subjects not yet mastered, and to receive rapid feedback on accuracy and progress, systems provide the beneﬁts of individual attention from an instructor that are no longer feasible for contemporary organizations. The purpose of this study is to examine how the use of a holistically designed TBEDS will contribute to procedural knowledge acquisition through the manner in which the user chooses to use the system for repeated, experiential practice.

62

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Prior research in the accounting literature has tended to examine individual components of system design facilitating knowledge acquisition, and some research has distinguished between basic, declarative knowledge, and the more advanced, problem-solving skills of procedural knowledge. Behavioral and cognition theories have been concerned with identifying the factors that support or limit the acquisition, retention, and application of knowledge (Anderson, 1982, 1987, 1995; Anderson & Fincham, 1994; Anderson, Fincham, & Douglass, 1997; Anderson & Schooler, 1991; Owen & Sweller, 1985; Sweller, 1988, 1989, 1993; Sweller, Chandler, Tierney, & Cooper, 1990; Sweller & Chandler, 1991; Sweller & Cooper, 1985; Sweller & Levine, 1982; Sweller, Mawer, & Ward, 1983; Sweller, van Merrienboer, & Paas, 1998). System design theory has focused on designing one or more of these functions into DAs, with the concurrent risk of overreliance by novice users and an emphasis on system-provided results versus the development of user problem-solving skills (Arnold & Sutton, 1998). On the contrary, very little research has been conducted into how the use of a holistically designed TBEDS may affect either the amount or the quality of practice and how use of the practice function may inﬂuence knowledge acquisition. Nor has research examined extended use of a TBEDS under actual conditions, where the effective use of the system and successful learning has personal consequences to the user. Finally, prior research tends to focus on the negative aspects of system use for learning, rather than the positive aspects. This study utilizes a longitudinal ﬁeld experiment to directly address these gaps and combines both system design and behavioral/cognitive learning theory. This study is also able to make a unique contribution by tracking procedural learning acquired over time through the extended, repeated use of a practice system with difﬁcult, extensive problems using a TBEDS under conditions providing signiﬁcant incentives to motivate user effort in achievement. In comparing performance of students using traditional methods for study and practice to those using a TBEDS, this study ﬁnds that a TBEDS results in greater procedural knowledge acquisition, a better rate of improvement, and less variance in performance than the use of traditional methods. The driving factor behind the difference in procedural knowledge acquisition is the student’s ability to use the system for selfdetermined, repeated, and meaningful practice. Examination of the various factors comprising the effort expended in practice (number of questions attempted, frequency of practice sessions, and accuracy of practice) indicates that the amount of practice determines overall mastery of procedural knowledge, while accuracy of practice was determined by the frequency of

Knowledge Acquisition Using a TBEDS

63

practice. Practice accuracy, however, was not signiﬁcantly related to subject mastery; possibly due to nonaccountability for practice. Finally, comparison of average practice scores before and following the user’s obtaining a basic level of system familiarity and comfort provide support for the importance of Task-Technology Fit, with user performance increasing once the usefulness of the TBEDS and its ease of use were established. The remainder of the chapter is divided into four sections. The next section presents the model, examines cognitive and learning theory and the link between effort and knowledge acquisition, situates the study within the IS environment, and develops the hypotheses. The following section details the research methodology, including a description of the unique, web-based TBEDS used. The section following that presents the results, and the ﬁnal section concludes with a discussion of limitations and implications for future research.

THEORY REVIEW AND HYPOTHESES DEVELOPMENT Knowledge Acquisition and Learning Theory Anderson’s development of ACT-R theory and its antecedents, ACT and ACT, has been seminal in understanding knowledge acquisition (Anderson, 1982, 1987, 1995; Anderson & Fincham, 1994; Anderson et al., 1997; Anderson & Schooler, 1991). Knowledge is subdivided into two categories: declarative and procedural. Declarative knowledge may be thought of as ‘‘what,’’ or rules, deﬁnitions, or examples, while procedural knowledge may be thought of as ‘‘how’’ or the direct application of declarative knowledge to analysis or problem solving. ACT-R theory states that knowledge is grouped into ‘‘chunks’’ of memory (also known as production rules) deﬁning the steps in which a problem is solved. The declarative information encoded is derived from instruction, reading descriptions, and studying examples and rules. Once acquired, declarative knowledge must then be converted to procedural knowledge – the more advanced application of theoretical knowledge to judgment and problemsolving tasks – through repeated practice. Practice forms production rules into schemas or mental representations of classes of problems. This, in turn, frees up working memory by transferring the knowledge acquired to long-term memory and results in strengthened

64

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

productions that are automatic, much faster, and nearly unconscious (Anderson, 1982, 1987; Anderson & Fincham, 1994). Greater procedural knowledge acquisition is positively correlated with recent, frequent, and extensive practice and leads to more efﬁcient and effective problem solving and judgments (Anderson et al., 1997; Anderson & Schooler, 1991). Schema formation through repeated practice is critical in moving knowledge and the ability to process that knowledge from working memory to longterm memory. Human working memory is limited, whereas long-term memory is not; thus, transferring knowledge and skills to long-term memory leads to much greater and deeper learning, improved performance, and the development of expertise through the ability to recognize the applicability of existing schemas to new problems (Sweller, 1988). Cognitive load theory indicates that reducing the demand on working memory required in constructing schemas will increase knowledge acquisition (Sweller & Chandler, 1991). Intrinsic (material difﬁculty), extraneous (instructional design), and germane (mental effort required) cognitive load combine to determine overall cognitive load and the resultant level of knowledge acquisition (Sweller et al., 1998). It is not simply the amount of practice that matters; to develop the requisite schemas, learners must be able to recognize similar problems that respond to similar treatments. When the learner realizes that procedural knowledge applicable to one domain may be relevant in whole or in part to another domain, transfer learning occurs. Varied but similar problems involve problems in the same domain, but with different characteristics that the user must learn to recognize as belonging to the same schema. The varied but similar format allows new information to be presented in a similar format to previously acquired knowledge, strengthening existing mental schemas (Pei & Reneau, 1990). The use of worked examples as either instructional method or feedback allows the learner to study the correct procedure and independently arrive at the procedural rules necessary through deductive reasoning and germane cognitive load. Thus, practice utilizing worked examples, varied but related problems, and transfer learning are critical in knowledge acquisition (Sweller & Cooper, 1985; Sweller et al., 1998).

Knowledge Acquisition and the Use of Technology The relationship between knowledge acquisition and performance is especially close, and the end goal of any system designed to enhance

65

Knowledge Acquisition Using a TBEDS

Technology (Environment)

Knowledge Acquisition Knowledge

Motivation

Effort

Ability

Feedback on performance

Fig. 1.

Model of Knowledge Acquisition Using a Technology Based Educational Delivery System.

learning is to improve the user’s ability to effectively and efﬁciently apply knowledge acquired across a range of situations and problems. Fig. 1 proposes a model of the effect of a TBEDS in moderating knowledge acquisition. Knowledge acquisition – consisting of both the overall knowledge and the ability to use that knowledge effectively – is a function of motivation, effort, the prior level of knowledge acquisition, and feedback as to performance. Performance is both an absolute measure (mastery of subject) and a relative measure (improvement). Environmental factors interact with the knowledge acquisition process through the effort needed to achieve and apply the combination of knowledge and ability. Technology is a major component of the learner’s environment and will affect the level of subject mastery. Technology can substitute to a degree for deﬁciencies in the user’s knowledge base (Cloyd, 1997) by providing targeted, task-speciﬁc knowledge, guidance in problem solving through explanations and feedback, and repeated opportunities for experiential learning with rapid feedback that are not available in traditional, text-based materials, thus helping novice users develop the necessary declarative and procedural knowledge to become experienced decision makers (Cloyd, 1997; Libby & Luft, 1993; Sweller & Cooper, 1985; Sweller et al., 1998). Knowledge acquisition is not a discrete event; feedback regarding performance and associated explanations of any errors are crucial in the

66

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

continuous, iterative nature of the learning process. Explanations improve performance and learning, increase user motivation, and result in more positive user perceptions of the system. Again, the effort involved in accessing the feedback system requires a careful balance. If the user must exert too much effort to retrieve explanations, they will not seek explanations (Gregor & Benbasat, 1999; Rose & Wolfe, 2000). However, requiring a minimal amount of effort to access feedback does result in greater knowledge transfer (Hornik & Ruf, 1997), possibly due to the user’s active involvement in the learning process (Gal & Steinbart, 1992) as this may assist the formation of memory ‘‘chunks’’ and their conversion to production rules and mental structures necessary for schema acquisition (Smedley & Sutton, 2007). Explanatory feedback in the form of worked solutions has been shown to be effective (Sweller & Cooper, 1985) as this both provides a mental cue enabling the access of relevant schemas and proper task classiﬁcation and forces the user to actively think about the problem by working backwards from the presented solutions (Pei, Steinbart, & Reneau, 1994). Other research (Steinbart & Accola, 1994) found evidence that detailed explanations were no more effective than simple rule trace explanations, emphasizing the role cognitive load plays in user preferences for the minimal effort necessary. Excessive feedback (in terms of either quantity or complexity of explanations) may actually decrease knowledge acquisition through information (and cognitive) overload (Odom & Dorr, 1995). Feedback, then, is a recursive element resulting from the objective evaluation of knowledge acquired and indirectly affecting future knowledge acquisition through its inﬂuence on motivation and effort. Rapid, accurate, and complete explanatory feedback regarding current performance affects the user’s motivation to continue the learning process and his or her assessment of the necessary amount of subsequent effort. Knowledge and ability are highly related concepts and are both antecedents and consequents of knowledge acquisition. Knowledge, as shown in Fig. 1, is speciﬁcally information stored in memory and includes the schemas and mental constructions used to organize and classify this knowledge from cognitive and learning theory. In general, knowledge may be conceptualized as declarative knowledge (the ‘‘what’’) and procedural knowledge (‘‘the how’’) and measured by overall subject mastery and the ability to apply rules, deﬁnitions, and examples in effective problem solving. Ability is the individual’s capacity to complete problem-solving tasks and is a function of both internal characteristics (innate aptitude in problem solving) and external characteristics (prior procedural knowledge).

Knowledge Acquisition Using a TBEDS

67

Holding effort levels equal, as ability increases, the rate of knowledge acquisition should increase. Both ability and prior knowledge will affect subsequent knowledge acquisition. Novices will ﬁrst store declarative knowledge in their knowledge base. Over time and with the exertion of signiﬁcant effort in experiential practice in applying problem-solving rules and relational deﬁnitions to declarative knowledge, novices will develop procedural knowledge that will then be added to the knowledge base and increase ability in an iterative effect. Research has shown that technology-based systems can facilitate knowledge acquisition. Users of DAs have been shown to outperform traditional method learners in acquiring declarative knowledge (McCall et al., 2008), have shown strong support for declarative-based KMS explanations for procedural knowledge transfer (Smedley & Sutton, 2004), and have had improved performance with procedural knowledge acquisition as they continue to develop generalized mental models (Fedorowicz, Oz, & Berger, 1992; Pei & Reneau, 1990; Smedley & Sutton, 2007). Systems research has shown the repeated use of a DA facilitates the formation of schemas through experiential learning and, thus, the development of expertise (Fedorowicz et al., 1992). Experience allows the novice to obtain a greater and more accurate knowledge base, form more detailed and stronger links among related schemas, and identify atypical or false results (Fedorowicz et al., 1992; Pei & Reneau, 1990; Tubbs, 1992). Repeated and extensive practice using a TBEDS allows a greater accumulation of this experience, in less time, and with more rapid and accurate feedback signiﬁcantly decreasing the amount of time required to ‘‘unlearn’’ errors in declarative knowledge and mistakes in procedural knowledge. Schema formation is the key to increased performance and the transitions from novice to experienced user and from experienced user to expert. Each subsequent stage contains more concise schemas that can be used to quickly identify salient problem characteristics and ignore associated environmental ‘‘noise’’ to process relevant information appropriately through pattern recognition (Lehmann & Norman, 2006). Consequently, an effective TBEDS must enable the transfer of simple declarative knowledge and allow that knowledge to be encoded in schemas through repeated practice and meaningful feedback, which will then result in procedural knowledge acquisition. TBEDSs are similar in design to DAs, especially in the provision of feedback that allows the user to self-correct errors immediately. However, DAs generally focus on guiding decisions of users who already have a certain basic level of procedural knowledge. The focus of TBEDS, on the

68

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

contrary, is the development of the basic knowledge base and associated mental schemas through repeated, challenging practice of applying deﬁnitions, rules, and procedures to problem solving. Traditional, textbased methods can only provide a limited number of practice problems, cannot repeat the same problem using new variables until the student has mastered the process without inadvertently memorizing the answers, and cannot provide instant performance feedback with worked examples to allow solving a problem by working backwards. These features are possible and are present to varying degrees, in properly designed TBEDSs. Increased practice with the TBEDS produces the increased experience that allows the user to encode the required knowledge in the necessary long-term memory structures, and the feedback provided by the TBEDS prevents the encoding of incorrect knowledge and problem-solving procedures. Consequently, novices who use a TBEDS for practice should experience both greater overall performance and rate of improvement over time than novices who do not have access to a TBEDS for practice. H1. Users of a TBEDS will show better performance than users of traditional methods alone. Effort is synonymous with experiential practice. The experience construct in the Libby and Luft (1993) model was speciﬁcally deﬁned as a ‘‘wide variety of ﬁrst- and second-hand task-related encounters that provide opportunities for learning’’ (Libby, 1995, p. 179). The relationship between experiential practice and knowledge acquisition is echoed in learning theory and expertise development research demonstrating that knowledge acquisition and expertise development result from the use of varied problems (Sweller et al., 1998) and repeated and extensive practice (Camerer & Johnson, 1991; Ericsson & Smith, 1991; Sloboda, 1991). Experience is not the same thing as exposure; to reap the beneﬁts of a TBEDS allowing for repeated, meaningful experiential practice, the user must engage in actual system use for practice. Thus, all things being equal, between two users who both have access to a TBEDS, the one exerting more effort in practice should achieve greater knowledge acquisition. However, over time, the required amount of effort to maintain a given level of knowledge acquisition should decrease due to increasing procedural knowledge and ability to apply that procedural knowledge. The effort an individual chooses to exert in practice and the subsequent level of knowledge acquisition (which then affects subsequent practice and knowledge acquisition) are the drivers behind increased performance using a TBEDS. Increased practice (effort) drives both the increased ability to

Knowledge Acquisition Using a TBEDS

69

solve problems accurately (effectiveness) as procedural knowledge increases and the reduced time required for problem solving (efﬁciency) as schemas expand and strengthen. Furthermore, as procedural knowledge continues to increase and schema formation continues to develop, the amount of practice necessary to maintain performance levels may actually decrease as effectiveness and efﬁciency increase.2 The subsequent amount of effort an individual chooses to exert is then affected by his or her problem-solving efﬁciency and effectiveness (Rose & Wolfe, 2000). Consequently, the greater the amount of effort, the better will be performance. The use of a DA generally reduces cognitive effort (Todd & Benbasat, 1994), which is important in avoiding extraneous cognitive load that might demotivate the user and lead to negative learning. However, maintaining a meaningful level of cognitive effort is critical for practice to result in increased knowledge acquisition. A DA that does not require active cognitive effort (and hence signiﬁcant time in practice) in applying the underlying task domain and operational concepts can actually impede knowledge acquisition (Glover, Prawitt, & Spilker, 1997). Practice opportunities must be varied but similar, relevant, and challenging enough to produce the necessary germane cognitive load. The consequent cognitive effort required then results in the formation of schemas and the transfer of knowledge to long-term memory. Knowledge acquisition is most effective when the user is actively involved and challenged by the type and difﬁculty of the problems to be solved. Motivation is a critical factor in the decision to use the TBEDS, especially when the amount of practice and the level of desired knowledge acquisition are self-determined; users with low motivation may choose satisﬁcing behavior and prefer lower levels of knowledge acquisition and less effort. Motivation may be internal (self-actualization, ambition, etc.) or external (job requirements, direct supervision or monitoring of performance, etc.), with internal motivation being stronger and producing greater efforts. As the user receives feedback on practice performance, demonstrating the usefulness of practice with the TBEDS in achieving overall subject mastery, internal motivation should increase to engage in greater amounts of practice (and enjoy the consequent increase in knowledge acquisition). In discussing practice effort, differentiating between amount of practice and duration of practice is important. Duration of practice, a simple measure of time spent in practice, may reﬂect the individual’s learning style, competing distracters, or other environmental constraints and may not represent meaningful effort (Reneau & Grabski, 1987). For example, the user may be called away from the problem, leaving the TBEDS running, or may simply sit and stare at the

70

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

practice problem without exerting an effort to ﬁnd the solution. Effort in using a TBEDS for experiential practice, then, consists of the amount of problems examined, the frequency of practice sessions, and the degree to which the user attempts to achieve accuracy in problem solution. Thus, a TBEDS that leaves the amount, frequency, and quality of practice to the user’s discretion should produce greater knowledge acquisition and subject mastery for those users who are more internally motivated and thus choose to expend greater efforts at practice. H2A. Users of a TBEDS who engage in a greater amount of practice will show better performance than TBEDS users engaging in less practice. H2B. Users of a TBEDS who engage in more frequent practice will show better performance than TBEDS users who engage in less frequent practice. H2C. Users of a TBEDS with more accurate practice will show better performance than TBEDS users with less accurate practice. For practice to be effective when using a TBEDS, the user must ﬁrst achieve a level of comfort and familiarity with the system. Reneau and Grabski’s (1987) review of computer–human interaction research emphasizes that for someone to use a system to achieve a goal, they must understand both how the system works (in broad terms) to achieve its purpose and how to use the system to perform the required tasks. In so doing, a mental model is constructed of the inputs, processes, and outcomes necessary, much as mental schemas are formed when learning procedural knowledge. The theory of Task-Technology Fit explicitly states that for technology to have a positive impact on performance, it (a) must actually be used and (b) must be a good ﬁt with the tasks it supports. User attitude develops from perceptions of system usefulness and satisfaction with ease of use; attitude then predicts the actual level of usage (Eining & Dorr, 1991; Gal & Steinbart, 1992; Goodhue & Thompson, 1995). Explanatory feedback is also important in determining system usage. Feedback in DAs has been shown to improve performance effectiveness through perceptions of increased usability (Arnold, Clark, Collier, Leech, & Sutton, 2004). Positive experiences with a system (ease of use and adequacy of explanation) produce increased familiarity and skill, which then leads to increased usage for the critical experiential practice. Experience with the system also allows users to evaluate the actual impact of system usage on performance, affecting their expectations and attitudes that subsequently

Knowledge Acquisition Using a TBEDS

71

affect future usage (Goodhue & Thompson, 1995). A TBEDS exhibiting good Task-Technology Fit will reduce problem complexity, minimize unnecessary cognitive workload for the novice user (Benford & Hunton, 2000), and provide appropriate feedback, thus increasing usage of the TBEDS for practice and subsequent performance. The user will accept a TBEDS when he or she sees that it is a useful part of the learning process – or in other words, when he or she sees exam scores improving or problem solving becoming easier with the use of the TBEDS for meaningful practice. The user will, however, also compare the amount of input effort necessary to produce the beneﬁts, and the decision to use the system will depend on the effort not being perceived as too onerous (Arnold & Sutton, 1998) – in this case, students will choose to use the TBEDS for meaningful practice if the difﬁculty involved is a function of the problems and not of using the system. Arnold and Sutton (1998), in their TTD, posit that once use of the system is established, overreliance on the system can occur, especially in cases of low task knowledge or great task complexity, leading to technology dominance and skill loss. Thus, a critical component of the TBEDS is that it provides neither easy (or invariant) problems nor too detailed of problem-solving structure or explanations. Users must exert cognitive effort in learning to solve these problems independently, without the crutch of the TBEDS, or the knowledge will not transfer to long-term memory and form the necessary schemas. This chapter contends that TTD occurs when there is a mismatch between the user’s knowledge level and the goals of the system being utilized. Novices using a DA for decision making will exhibit TTD and fail to develop independent skills because they are simply learning input/output procedures (Hampton, 2005). Consequently, they will not necessarily acquire knowledge that they are capable of accessing and using without the system. Novices using a TBEDS designed to develop procedural knowledge through repeated, challenging practice, however, will acquire mental models incorporating input, output, and the necessary procedures to evaluate and analyze the information independently and will display knowledge acquisition. The TBEDS must not only have appropriate design features, but it also must be matched appropriately to the task and the user; a TBEDS designed for beginning college algebra practice will differ radically in terms of explanatory feedback, structure of problems, and degree of embedded structure from a TBEDS designed for the advanced application of physics to engineering problems. Individuals will choose to use the TBEDS when they perceive it as useful, easy to use, and not requiring an unreasonable level of effort to produce results (Arnold & Sutton, 1998;

72

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Eining & Dorr, 1991; Gal & Steinbart, 1992; Goodhue & Thompson, 1995). Such positive attitudes only develop after the user has a certain basic level of experience with the TBEDS; therefore, once familiarity and comfort are achieved, user attitudes should lead to greater usage and subsequently greater knowledge acquisition. H3. Users of a TBEDS for practice will show greater knowledge acquisition after achieving basic system familiarity.

METHODOLOGY AND DESIGN Participants Study participants were solicited from an upper-level undergraduate accounting class that had recently adopted the holistic TBEDS used in this study, Practice4Performance (P4P). This course, at a large public university in the southeast United States, is required for accounting majors, a recommended elective for business majors, and a prerequisite for admission to graduate programs in accounting. To test H1, results from the experimental group were compared with results from a control group. The control group consisted of 55 students who had taken the same class, with the same instructor and text, the previous semester, but who had not had access to the TBEDS. The experimental group consisted of 40 participants,3 of whom 13 were male, 25 were female, and 3 did not provide gender data. Thirty-two (78 percent) were accounting majors, with the remainder majoring in either ﬁnance (nine participants) or other business disciplines (three participants). Thirty-seven participants were undergraduates (90 percent); the remaining participants consisted of a graduate student taking prerequisites and second degree seekers.4 Description of the TBEDS: Practice4Performance A TBEDS designed to encourage users to actively use the system to manage their own learning will enhance procedural knowledge acquisition; the challenge is to design such an interface (Gal & Steinbart, 1992). The TBEDS used in this research project, P4P, incorporates the critical features of repeated varied but related problems and worked examples to support subject mastery in novice users. P4P is a web-delivered practice and testing system, which unobtrusively tracks detailed information on use of the

Knowledge Acquisition Using a TBEDS

73

system and resulting performance. Because it is web-delivered, students can access the TBEDS at the time and place of their choosing. The system has been developed over many years’ extensive use in an advanced undergraduate accounting course and effectively provides a near-inﬁnite number of problems for practice. P4P presents randomized questions, variables, and solutions from the test bank; all potential questions are completely renewed every two hours. Questions and solutions are generated using stochastic simulation, artiﬁcial intelligence, Cholesky’s algorithm, and other advanced methods and result in a constantly changing, practically limitless bank of substantively difﬁcult but contextually similar problems for practice. Students may choose the number of questions to receive, the level of difﬁculty, and topics for each practice session, or they may accept defaults (20 questions and a mix of topics and difﬁculty for each chapter). P4P provides immediate outcome feedback (right/wrong), allows multiple attempts to solve the problem (with decreasing points awarded), and once the practice session is completed with a minimum score,5 exemplary feedback with correctly worked solutions is accessible to the student. The system provides real-time feedback, including overall performance and current class ranking, which allows each student to judge his or her performance and adjust practice efforts accordingly. P4P also administers exams from a reserved databank. The instructor may choose number of questions, speciﬁc questions (or allow random questions), length of time allowed, whether repeats will be permitted, and when the exam will be available. Students may be permitted to take the exam wherever they wish, as the randomization of questions and individual variables within questions results in each student effectively receiving an individual exam.6 However, students may also be constrained to testing labs should the instructor choose speciﬁc questions or so desire. Exams are graded immediately upon completion, and the score and responses are available for both student and instructor (correct responses are not displayed until the instructor releases them, although he or she may access them at any time). Both declarative and procedural knowledge questions are administered by the system. The procedural knowledge questions require calculations and application of relevant rules and deﬁnitions; the majority of these questions are also advanced, with multiple steps and procedures needed. Students are encouraged to simultaneously open Microsoft Excel and build the required formulas and models within a spreadsheet, thus reinforcing actual business conditions. Answers in P4P are presented in a multiple-choice format, but within a very tight range of values so that students must have an exact answer and cannot simply eliminate outliers or ‘‘estimate’’ the most likely answer.

74

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Following the ﬁrst exam for the experimental group, P4P was introduced into the class. Use of the system was mandatory for the remaining two exams and for a case-based homework assignment; use of the system was purely voluntary for practice. The assigned homework was available for two weeks, there were no time limits for practice, and students could leave and return to both as necessary. Students were informed that practice questions were direct variants of questions they would receive on exams and were encouraged to practice using the system, as the ability to pass the course requires considerable procedural knowledge and mastery of complex accounting concepts. To encourage intentional learning, and to parallel the instructor’s previous practice of encouraging but not requiring students to practice exercises from the text, the amount, timing, and subjects of practice were left to the discretion of the individual and reﬂect motivation. Students were encouraged to work together to learn and could repeat the single assigned homework case, which was due shortly after the introduction of the TBEDS. As a result, the majority of students received 100 percent on homework. Consequently, the homework score is not included as a representative performance measure but proxies for a signal of system familiarity and acceptance. Exams were each open for a two-week window, with retake opportunities available during the same period. Students were allowed two retakes, at their discretion, with a ﬁve-question ‘‘penalty’’ added for each attempt, to ensure serious effort. A 24-hour waiting period was also enforced between exam attempts to encourage additional study and practice. The exams were each two hours long and consisted of 30 questions chosen by the instructor from the database. These questions would look quite similar to previously practiced questions, but all the variables would have changed, producing completely new problems. Thus, students who had put in signiﬁcant effort in terms of number of questions meaningfully practiced should have been quite comfortable with the concepts and necessary problem-solving skills.

Variables Performance consists both of absolute and relative measures of knowledge acquisition. The absolute performance, which represents the level of subject mastery, is operationalized by exam scores for knowledge acquired at distinct points throughout the course. The ﬁnal exam (exam 3) was comprehensive, thus operationalizing subject mastery over the entire course. The relative performance, representing the change in efﬁciency and

Knowledge Acquisition Using a TBEDS

75

effectiveness of learning and which proxies for ability to apply prior knowledge, is measured by improvement rates between exams. H1 uses both measures of performance to compare performance between the experimental and the control group. The aggregated components of H2 use the ﬁnal exam score for the experimental group to examine the effect of differential use of the TBEDS on overall knowledge acquisition. H3 examines the effect on practice of becoming comfortable or familiar with the TBEDS; students were required to complete a TBEDS-based homework assignment following system introduction. As the homework was mandatory, all students were forced to use P4P and would become familiar with system features, how to input the data, and the feedback provided. Thus, the dependent variable in this case is the average practice score before and after the mandatory casebased homework was due. The independent variables for H1 are group (experimental vs. control) and the testing event (exam 1, exam 2, and exam 3). The overall question addressed by the components of H2 is concerned with how individual differences in using the system would affect individual performance. The predictors for this hypothesis include the exam 1 (pre-TBEDS) score as a proxy for ability and initial knowledge and the exam 2 score as an indicator of sustained learning, demographic variables (class standing, major, and gender), and measures of the quality and quantity of practice efforts. Quality of practice is operationalized as the average practice session score for all practice sessions attempted. Quantity of practice is divided into both amount and frequency of practice and operationalized as the total number of practice questions examined (regardless of the number of sessions into which they were divided) and the number of practice sessions (regardless of how many questions were seen in each session). H3, as a simple test of performance before and after a given point in the course, simply uses an indicator for practice performed prior to and following the mandatory case homework due date.

RESULTS This investigation is concerned with the ability of a TBEDS to impact procedural learning for novice users, primarily through its ability to provide the opportunity for repeated, meaningful experiential practice with prompt feedback. For H1, the experimental group and the control group are assumed to be comparable in terms of knowledge acquisition before the introduction of the TBEDS. As the TBEDS was not introduced to the

76

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

experimental group until after completion of the ﬁrst exam, there should be no statistically signiﬁcant difference between exam 1 scores for the experimental and the control groups. Any subsequent difference may be attributed to the presence or absence of the TBEDS. Table 1 (Panel A) provides the scores for exam 1. The control group had an average exam score of 74.93, whereas the experimental group had an average score of 77.56. The difference between the two groups is not statistically signiﬁcant, which suggests that the assumption of equivalence between the two groups is supported [F (1, 94) ¼ .855, p ¼ .358]. Thus, it is appropriate to proceed with testing H1. Table 1 (Panel A) also presents the descriptive statistics using absolute measures of performance at discrete points during the semester and clearly shows the difference in knowledge acquisition with the use of the TBEDS. For exam 3, there is a 10.54 percentage point difference in class means for the experimental group using P4P – an entire letter grade improvement – over the control group. In terms of overall performance across the entire

Table 1. Panel A: Mean Scores on Exams

Exam 1 (before TBEDS) Exam 2 Exam 3

Panel B: Improvement Rates

Exam 1 to exam 2 improvement Exam 2 to exam 3 improvement Exam 1 to exam 3 improvement

Descriptive Statistics. Control group (n ¼ 55)

Experimental group (n ¼ 40)

Mean (SD)

Mean (SD)

74.93 (12.22) 72.45 (13.41) 69.62 (14.93)

77.56 (15.55) 80.96 (10.96) 80.16 (8.74)

Difference in means

2.63 8.51 10.54

Control group (n ¼ 55)

Experimental group (n ¼ 40)

Mean (std error)

Mean (std error)

2.52% (.1634) 2.30% (.2021) 6.46% (.1565)

9.06% (.3162) 9.00% (.1194) 8.22% (.3013)

Note: Cell values are means calculated from all individual values for each group and not simple difference scores from values in Panel A.

Knowledge Acquisition Using a TBEDS

77

16-week period, the exam 3 score (the ﬁnal exam) for the control group is 5.31 percentage points lower than exam 1, whereas exam 3 score increases an average of 2.60 percentage points from exam 1 for the experimental group. Interestingly, there is a signiﬁcant drop in exam performance of 2.83 percentage points between exam 2 and exam 3 for the control group, whereas the experimental group maintains their performance level overall, with only a minor decrease in performance (0.8 percent). Furthermore, the variation in scores is much different between the two groups. The control group not only shows greater variance on both exam 2 and exam 3 but also shows increasing variance between exams. The experimental group, despite exhibiting greater variance than the control group on exam 1, has signiﬁcantly less variance on exam 2 and exam 3, and their overall variance over time is markedly less. As knowledge acquisition is reﬂected not only in performance at a speciﬁc point in time (e.g., an exam score) but also in performance across time (improvement), the two groups’ improvement rates are compared. Table 1 (Panel B) presents the descriptive statistics for this test of H1 using relative measures of performance and also clearly illustrates a difference between groups. Improvement rates are calculated by the difference between exam scores over the prior exam score for exam 1 to exam 2, exam 2 to exam 3, and exam 1 to exam 3. For example, the improvement rate between exam 2 and exam 3 is measured by (exam 3 scoreexam 2 score)/exam 2 score. An overall improvement rate for the semester is also calculated between the ﬁnal exam (exam 3) and the initial exam (before TBEDS introduction). The descriptive statistics in Table 1 indicate a difference in improvement rates between the experimental and the control groups. The control group worsens over time, with a fairly steady decline between each testing event (2.52 percent decline from exam 1 to exam 2 and 2.30 percent from exam 2 to exam 3). The experimental group improves steadily and at a much higher rate than the control group (9.06 percent from exam 1 to exam 2 and 9.00 percent from exam 2 to exam 3). Overall, the control group shows a decline in relative performance across the entire semester of 6.46 percent, whereas the experimental group shows an improvement in relative performance for the same period of 8.22 percent. Thus, repeated practice using a TBEDS appears to shift the performance level upward and leads to higher rates of improvement. H1 postulates that engaging in frequent practice of challenging, similar but different problems followed by feedback with worked examples by using P4P will lead to greater performance for the experimental group when compared to the control group, which had access only to traditional

78

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

text-based materials. To verify that signiﬁcant differences exist between the experimental and the control groups for both absolute performance (exam 3 score) and relative performance (percentage improvement between exam 1 and exam 3), a multiple analysis of variance (MANOVA) is conducted. As given in Table 2 (Panel A), the MANOVA indicates that the model exhibits signiﬁcant differences between groups [Wilks’ l ¼ .834, F (2, 92) ¼ 9.188, po0.001]. As given in Table 1 (Panel B), the differences between groups for overall knowledge acquisition and overall percentage improvement are signiﬁcant (po0.001 and p ¼ 0.003, respectively). Overall, the ability to use the TBEDS for practice accounts for 16.6 percent (Z2) of the variance between the two groups. These results demonstrate a signiﬁcant difference in performance between the control and the experimental groups, for both measures of performance and consistently across time. Thus, H1 is supported for better performance for the experimental group using the TBEDS for experiential practice. Having established that the use of the TBEDS for practice does lead to a signiﬁcant difference in performance for the experimental group, the next task is to examine the elements of system use. H2 postulates that the amount of practice (H2A), the frequency of practice (H2B), and accuracy of practice (H2C) with a TBEDS will lead to better performance. The ﬁrst step is to examine the correlations between these three predictors (amount, frequency, and accuracy), absolute performance (exam 3 score), and relative performance (exam 1 to exam 3 improvement) to understand the impact of practice (the various measures of effort) on performance. Table 3 presents the Pearson partial correlations, controlling for exam 1, which serves as a proxy for knowledge acquired before the introduction of P4P. While some correlations are obvious (amount to frequency, r ¼ .882, po0.01; absolute to relative performance, r ¼ .830, po0.01), some unexpected relationships appear, which were not apparent in the initial analysis, primarily due to small sample size and corresponding low power. The most interesting correlations are associated with accuracy (average practice score). Accuracy (mean 80.17, SD 8.75) is not signiﬁcantly correlated with either absolute performance (r ¼ .151, n.s.) or relative performance (r ¼ .046, n.s.). Although accuracy is signiﬁcantly correlated with frequency (r ¼ .342, po0.05), it is not signiﬁcantly correlated with amount (r ¼ .217, n.s.). These ﬁndings may indicate support for the spacing effect (Anderson & Schooler, 1991), whereby knowledge is best acquired through spaced learning, with intervals of rest between each session, and not by the overall amount of problems examined. Additionally, these

Panel A: Multivariate Analysis Effect Value Wilks’ l

.834

F

Hypothesis df

Error df

p value

Partial Z2

9.188

2

92

.000

.166

Panel B: Tests of Between Subject Effects Source DV Corrected model Intercept Group (experimental vs. control) Error Total Corrected total

Absolute performancea Relative performanceb Absolute performance Relative performance Absolute performance Relative performance Absolute performance Relative performance Absolute performance Relative performance Absolute performance Relative performance

Type III SS

df

Mean square

F

p value

2575.977 .499 519547.404 .007 2575.977 .499 15027.413 4.864 538652.520 5.363 17603.390 5.363

1 1 1 1 1 1 93 93 95 95 94 94

2575.977 .499 519547.404 .007 2575.977 .499 161.585 .052

15.942 9.541 3215.318 .138 15.942 9.541

.000 .003 .000 .711 .000 .003

Knowledge Acquisition Using a TBEDS

Table 2. MANOVA Test for Differences in Performance between Experimental and Control Groups.

a ¼ .05; R2 for absolute performance ¼ .146, adjusted R2 ¼ .137; R2 for percentage improvement ¼ .093, adjusted R2 ¼ .083. a Absolute performance is measured using the score for exam 3. b Relative performance is measured using the improvement from exam 1 to exam 3.

79

80

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Table 3. Partial Correlation Table for Effort Predictors of Performance Using a Technology-Based Educational Delivery System. Variable

Amount of practice Frequency of practice Accuracy of practice Absolute performance Relative performance

Amount of Practice

Frequency of Practice

Accuracy of Practice

Absolute Performance

Relative Performance

1.000 .882

1.000

.217

.342

1.000

.462

.288

.151

.318

.194

.046

1.000 .830

1.000

Signiﬁcant at po.05. Signiﬁcant at po.01.

correlations tend to support accepted wisdom that cramming is not as effective a study method as steady, regular, purposeful practice. In a further surprising relationship, while frequency (mean 41, SD 24.49) is signiﬁcantly and positively correlated with accuracy, it is not correlated with either performance measure. Instead, amount of practice (mean 548, SD 386.01) correlates signiﬁcantly with relative performance (r ¼ .318, po0.05) and absolute performance (r ¼ .462, po0.01). Intuitively, the high amount of practice reﬂects increased motivation, with more motivated students engaging in greater amounts of experiential practice and thereby experiencing greater achievement. The frequency of practice may not be signiﬁcantly correlated with either performance measure, due to the high variability in student choice of how many questions to examine per session (ranging from 5 to 25, in steps of ﬁve), which produces signiﬁcant differences between individual sessions. H2 is formally tested using linear regression. As absolute performance and relative performance are highly correlated, only the absolute measure of knowledge acquisition is used as the dependent variable. An initial model tests the three effort predictors (amount, frequency, and accuracy of practice) and includes several other potential measures of variance. As previous analysis indicated a sharp increase in achievement upon the initial introduction of the TBEDS, exam 2 scores are included as a variable to help capture sustained effort across the semester. Exam 1 scores are included as a

Knowledge Acquisition Using a TBEDS

81

covariate to proxy for initial knowledge and several demographic variables (gender, year in school, and major) are tested for explanatory power. The initial model indicates that at least one of the predictors probably affect performance (F ¼ 6.326, po.01). However, an examination of the tscores for the predictors, given in Table 4, indicates that only frequency, amount, and exam 2 score are signiﬁcant (p ¼ 0.019, p ¼ 0.005, and po0.001, respectively). Before dropping the other variables, tests for heteroskedasticity and serial correlation were performed and were not found to be present.7 None of the VIF factors have a value greater than 10, indicating no severe multicollinearity that would need to be addressed, and a Wald test of joint signiﬁcance indicates that the restricted model is correct. Therefore, the following respeciﬁed model is estimated: Y ¼ a þ b1 Frequency þ b2 Amount þ b3 Exam2 þ As given in Table 5, the resulting model is signiﬁcant (F ¼ 15.515, po0.001) and all independent variables pass the t-test (amount, p ¼ 0.034; frequency, p ¼ 0.006; and exam 2, p ¼ 0.000, respectively), indicating that each of them probably affects absolute performance. The adjusted R2 indicates that 52.8 percent of variation in absolute performance is explained by variation in the frequency, amount, and exam 2 score. Overall, this is a fair model with a good adjusted R2 value for cross-sectional data. Although the model could be improved by adding other factors explaining overall performance through internal attributes or motivation, over half of the variation is attributed to increased experiential practice with a TBEDS – a signiﬁcant indicator of the effectiveness of the technology in contributing to knowledge acquisition. Consistent with expectations, amount of practice is positively correlated with absolute performance, such that, all other things being equal, approximately 10 extra practice questions worked would improve exam 3 performance by .16 percentage points. The coefﬁcient on frequency of practice is negative, however, contrary to expectations. This would indicate that, ceteris paribus, for every additional session, the user’s ﬁnal exam score will decrease by .183 percentage points. This does make intuitive sense, as the total number of meaningful practice problems worked is probably more directly determinative of ﬁnal subject mastery than the number of study sessions in which they occur. Therefore, H2A (greater amount of practice leads to better performance) is supported while H2B (greater frequency of practice leads to better performance) is not supported. The insigniﬁcance of accuracy in practice may be due to high multicollinearity with amount and frequency of practice; more motivated

82

Test of Initial Regression Model of Absolute Performance.

Unstandardized Coefﬁcients

Constant Amount of practice Frequency of practice Accuracy of practice Exam 1 score Exam 2 score Year in school Major Gender

B

Std error

31.94 .017 .221 .146 .006 .442 1.233 1.463 3.309

11.261 .006 .089 .131 .067 .103 2.550 2.422 1.962

Standardized Coefﬁcients

t

p value

Beta

.742 .619 .142 .010 .554 .057 .068 .188

2.836 3.030 2.486 1.122 .085 4.310 .483 .604 1.687

0.008 0.005 0.019 0.271 0.933 0.000 0.632 0.550 0.102

95% Conﬁdence Interval for B Lower bound

Upper bound

8.972 .005 .403 .120 .131 .233 3.969 3.477 7.311

54.908 .028 .040 .413 .143 .651 6.434 6.403 .692

Collinearity Statistics

Tolerance

VIF

.204 .198 .763 .862 .741 .879 .974 .990

4.90 5.06 1.31 1.16 1.35 1.14 1.03 1.01

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Table 4.

83

Knowledge Acquisition Using a TBEDS

Table 5.

Test of Respeciﬁed Final Regression Model of Absolute Performance.

Panel A: Test of Model

Sum of squares

df

Mean square

F

Signiﬁcance

Regression Residual Total

1682.832 1301.599 2984.431

3 36 39

560.944 36.156

15.515

.000

Panel B: Coefﬁcient Values

Unstandardized coefﬁcients

Constant Amount Frequency Exam 2

B

Std error

40.869 .016 .183 .472

7.335 .005 .083 .094

Standardized coefﬁcients

t

p value

VIF

.000 .006 .034 .000

4.610 4.467 1.157

Beta

.686 .513 .592

5.572 2.905 2.206 4.998

R2 ¼ .564; Adjusted R2 ¼ .528

students will practice more questions and will also have potentially higher scores. Exam 2 scores in the model reﬂect student usage of the TBEDS for practice over an extended period and may be a better indicator of true quality of sustained practice. Additionally, lower practice scores (which were not used to determine the ﬁnal course grade) may reﬂect freedom of students to experiment more in problem solving and to risk incorrect answers in testing their understanding and working out rules and processes for themselves. Such behavior may then in turn lead to increased ability in problem-solving, as reﬂected in increased problem-solving ‘‘when it really matters’’ and the student is held accountable for performance (i.e., an exam score that does determine ﬁnal course grade). In other words, students may practice many questions, but it is not readily apparent whether they are exerting great effort at accuracy or simply trying to complete the questions to obtain the worked examples from the feedback function. Given this situation, exam 2 becomes a qualitative measure of which users are using the system for meaningful practice versus those who are not, even if accuracy, frequency, and amount of practice are identical between them. Therefore, H2C is partially supported, only when accuracy is measured as exam 2 score, but not for average score across practice sessions. Finally, H3 posits that there will be an increase in performance once users have achieved a certain level of comfort and familiarity in using a TBEDS for repeated, experiential practice. The independent variable in this test is

84

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Table 6.

Paired Samples Test of Accuracy Before and After System Familiarity. Paired Differences Mean

Average score before familiarity average score after familiarity

4.088

t

df

p value (Two Tailed)

SD 1.289

20.053 39

.000

the due date for the case-based homework assignment delivered through P4P; as the assignment was mandatory, all students had to use the TBEDS to complete and submit the homework, even if they had chosen previously to not utilize the system for practice. A simple paired-samples t-test of the average practice scores before and after the homework due date indicates a statistically signiﬁcant difference between the scores. Once the student has achieved a basic level of familiarity with the TBEDS through using P4P to complete the assigned homework, the mean average practice score increases 4.087 percentage points (po0.001). These results, summarized in Table 6, provide evidence that the theory of Task-Technology Fit and the ﬁrst part of the TTD (user acceptance of technology) are correct in their explanation of the effectiveness of a technology being dependent on user familiarity and comfort with that technology. Thus, H3 (system familiarity produces better performance) is supported.

CONCLUSION This chapter examines the ability of a TBEDS to transmit knowledge to novice users, allowing them to build expertise in problem solving. The limited sample available for testing the adoption of the TBEDS tended to bias against ﬁnding signiﬁcant results, due to potential concerns with low power in the statistical tests. Nevertheless, the primary purpose of this experiment is to determine whether the use of TBEDS for repeated practice results in a signiﬁcant improvement in knowledge acquisition. The ﬁndings indicate a signiﬁcant increase in absolute performance for the group using P4P over the group using traditional methods and a decrease in performance variance, indicating practice with a TBEDS was more effective in reliably transmitting knowledge and problem-solving skills. Furthermore, a comparison of improvement rates for both groups indicates signiﬁcantly better results for the group using P4P to support increased practice.

Knowledge Acquisition Using a TBEDS

85

Next, individual measures of effort are examined to determine which factors contribute the most to overall performance. Tests of the amount of practice clearly indicate that greater effort in terms of the total number of practice questions worked leads to improved performance. Results for accuracy are mixed, with evaluation scores only being signiﬁcant when the user is accountable to an external authority. Finally, by examining the utilization and effectiveness of the TBEDS over an extended time frame, the study provides support for Task-Technology Fit and the TTD, both of which state that individual usage of and reliance on the system will only occur once the user achieves a basic level of familiarity and comfort with the technology. Findings indicate a statistically signiﬁcant improvement once a base level of competence and practice is reached. There are areas of concern due to the nature of the data available. The sample size is very limited for tests involving only the group adopting the TBEDS, and future efforts should focus on expanding research with larger populations. However, to be comparable, this research must also concentrate on populations using a TBEDS over extended periods under conditions of external accountability and motivation. Additional examination of the effects of consistency, spacing of practice, and accuracy in practice are also warranted. Future directions for research include speciﬁcally testing the effect of iterative feedback on motivation and effort over time, the extent to which internal motivation and inherent ability serve as moderators of performance, and a further examination of under what circumstances and to what limits effort is able to substitute for inherent knowledge or ability.

NOTES 1. The term decision aid, as used in this study, encompasses expert systems, decision support systems, and other types of technology that are designed to support decision making. 2. Interestingly, one researcher (Cloyd, 1997) found that increases in effort duration improved effectiveness regardless of the level of prior knowledge, which suggests that effort can substitute to a certain degree for knowledge in overall performance. This validates countless instructor exhortations that hard work pays off and explains why a mediocre student who consistently and determinedly plugs away at the material may outperform the gifted student who chooses not to exert any effort in meaningful practice. 3. Forty one students out of 65 (63%) enrolled gave their consent to participate; one student was later dropped due to insufﬁcient participation. 4. Due to IRB restrictions, only summary information was available for the control group from the previous semester.

86

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

5. The minimum score is determined by the instructor and is designed to encourage honest effort at solving problems. 6. The instructor is always able to retrieve the exact question and the speciﬁc variables that the student received to address any questions. 7. White’s test was used to test for heteroskedasticity and the Durbin Watson test was used for possible serial correlation as the use of the TBEDS over time might conceivably possess characteristics of panel data.

ACKNOWLEDGMENTS The authors gratefully acknowledge the generous advice and suggestions of Vicky Arnold, Steve Kaplan, Paul Steinbart, Steve Sutton, and the participants in the Artiﬁcial Intelligence/Emerging Technologies workshop at the 2008 American Accounting Association meeting in Anaheim, CA, and the research roundtables at the 2009 IS Midyear Meeting in Charleston, SC. We are also grateful to Pam Graybeal for facilitating the study.

REFERENCES Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological Review, 89(4), 369 406. Anderson, J. R. (1987). Skill acquisition: Compilation of weak method problem solutions. Psychological Review, 94(2), 192 210. Anderson, J. R. (1995). Learning and memory: An integrated approach. New York: Wiley. Anderson, J. R., & Fincham, J. M. (1994). Acquisition of procedural skills from examples. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(6), 1322 1340. Anderson, J. R., Fincham, J. M., & Douglass, S. (1997). The role of examples and rules in the acquisition of a cognitive skill. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23(4), 932 945. Anderson, J. R., & Schooler, L. J. (1991). Reﬂections of the environment in memory. Psychological Science, 2(6), 396 408. Arnold, V., Clark, N., Collier, P. A., Leech, S. A., & Sutton, S. G. (2004). Explanation provision and use in an intelligent decision aid. Intelligent Systems in Accounting Finance and Management, 1, 5 27. Arnold, V., & Sutton, S. G. (1998). The theory of technology dominance: Understanding the impact of intelligent decisions aids on decision makers’ judgments. Advances in Accounting Behavioral Research, 1, 175 194. Benford, T. L., & Hunton, J. E. (2000). Incorporating information technology considerations into an expanded model of judgment and decision making in accounting. International Journal of Accounting Information Systems, 1, 54 65. Bonner, S. E. (1990). Experience effects in auditing: The role of task speciﬁc knowledge. The Accounting Review, 65(1), 72 92.

Knowledge Acquisition Using a TBEDS

87

Camerer, C. F., & Johnson, E. J. (1991). The process performance paradox in expert judgment. In: K. A. Ericsson & J. Smith (Eds), Toward a general theory of expertise (pp. 195 217). Cambridge, UK: Cambridge University Press. Cloyd, C. B. (1997). Performance in tax research tasks: The joint effects of knowledge and accountability. The Accounting Review, 72(1), 111 131. Eining, M. M., & Dorr, P. B. (1991). The impact of expert system usage on experiential learning in an auditing setting. Journal of Information Systems, 5(1), 1 16. Ericsson, K. A., & Smith, J. (1991). In: Toward a general theory of expertise (pp. 1 38). Cambridge, UK: Cambridge University Press. Fedorowicz, J., Oz, E., & Berger, P. D. (1992). A learning curve analysis of expert system use. Decision Sciences, 23(4), 797 818. Gal, G., & Steinbart, P. J. (1992). Interface style and training task difﬁculty as determinants of effective computer assisted knowledge transfer. Decision Sciences, 23(1), 128 143. Glover, S. M., Prawitt, D. F., & Spilker, B. C. (1997). The inﬂuence of decision aids on user behavior: Implications for knowledge acquisition and inappropriate reliance. Organiza tional Behavior and Human Decision Processes, 72(2), 232 255. Goodhue, D. L., & Thompson, R. L. (1995). Task technology ﬁt and individual performance. MIS Quarterly, 19(2), 213 236. Gregor, S., & Benbasat, I. (1999). Explanations from intelligent systems: Theoretical foundations and implications for practice. MIS Quarterly, 23(4), 497 530. Hampton, C. (2005). Determinants of reliance: An empirical test of the theory of technology dominance. International Journal of Accounting Information Systems, 6, 217. Hornik, S., & Ruf, B. M. (1997). Expert systems usage and knowledge acquisition: An empirical assessment of analogical reasoning in the evaluation of internal controls. Journal of Information Systems, 11(2), 57 74. Lehmann, C. M., & Norman, C. S. (2006). The effects of experience on complex problem representation and judgment in auditing: An experimental investigation. Behavioral Research in Accounting, 18, 65 83. Libby, R. (1995). The role of knowledge and memory in audit judgment. In: R. H. Ashton & A. H. Ashton (Eds), Judgment and decision making research in accounting and auditing (pp. 176 206). Cambridge, UK: Cambridge University Press. Libby, R., & Luft, J. L. (1993). Determinants of judgment performance in accounting settings: Ability, knowledge, motivation, and environment. Accounting, Organizations and Society, 18(5), 425 450. McCall, H., Arnold, V., & Sutton, S. G. (2008). Use of knowledge management systems and the impact on the acquisition of explicit knowledge. Journal of Information Systems, 22(2), 77 101. Odom, M. D., & Dorr, P. B. (1995). The impact of elaboration based expert system interfaces on de skilling: An epistemological issue. Journal of Information Systems, 9(1), 1 17. Owen, E., & Sweller, J. (1985). What do students learn while solving mathematics problems? Journal of Educational Psychology, 77(3), 272. Pei, B. K. W., & Reneau, J. H. (1990). The effects of memory structure on using rule based expert systems for training: A framework and an empirical test. Decision Sciences, 21(2), 263 286. Pei, B. K. W., Steinbart, P. J., & Reneau, J. H. (1994). The effects of judgment strategy and prompting on using rule based expert systems for knowledge transfer. Journal of Information Systems, 8(1), 21 42.

88

PAUL M. GOLDWATER AND KIMBERLY A. ZAHLLER

Reneau, J. H., & Grabski, S. V. (1987). A review of research in computer human interaction and individual differences within a model for research in accounting information systems. Journal of Information Systems, 2(1), 33 53. Rose, J. M., & Wolfe, C. J. (2000). The effects of system design alternatives on the acquisition of tax knowledge from a computerized tax decision aid. Accounting, Organizations and Society, 25, 285 306. Sloboda, J. (1991). Musical expertise. In: K. A. Ericsson & J. Smith (Eds), Toward a general theory of expertise (pp. 153 171). Cambridge, UK: Cambridge University Press. Smedley, G. A., & Sutton, S. G. (2004). Explanation provision in knowledge based systems: A theory driven approach for knowledge transfer designs. Journal of Emerging Technologies in Accounting, 1, 41 61. Smedley, G. A., & Sutton, S. G. (2007). The effect of alternative procedural explanation types on procedural knowledge acquisition during knowledge based systems use. Journal of Information Systems, 21(1), 27 51. Steinbart, P. J., & Accola, W. L. (1994). The effects of explanation type and user involvement on learning from and satisfaction with expert systems. Journal of Information Systems, 8(1), 1 17. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12, 257 285. Sweller, J. (1989). Cognitive technology: Some procedures for facilitating learning and problem solving in mathematics and science. Journal of Educational Psychology, 81(4), 457 466. Sweller, J. (1993). Some cognitive processes and their consequences for the organisation and presentation of information. Australian Journal of Psychology, 45, 1 8. Sweller, J., & Chandler, P. (1991). Evidence for cognitive load theory. Cognition and Instruction, 8(4), 351 362. Sweller, J., Chandler, P., Tierney, P., & Cooper, M. (1990). Cognitive load as a factor in the structuring of technical material. Journal of Experimental Psychology: General, 119(2), 176 192. Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59 89. Sweller, J., & Levine, M. (1982). Effects of goal speciﬁcity on means ends analysis and learning. Journal of Experimental Psychology, 8(5), 463 474. Sweller, J., Mawer, R. F., & Ward, M. R. (1983). Development of expertise in mathematical problem solving. Journal of Experimental Psychology: General, 112(4), 639 661. Sweller, J., van Merrienboer, J. J. G., & Paas, F. G. W. C. (1998). Cognitive architecture and instructional design. Educational Psychology Review, 10(3), 251 296. Todd, P., & Benbasat, I. (1994). The inﬂuence of decision aids on choice strategies: An experimental analysis of the role of cognitive effort. Organizational Behavior and Human Decision Processes, 60, 36 74. Tubbs, R. M. (1992). The effect of experience on the auditor’s organization and amount of knowledge. The Accounting Review, 67(4), 783 801.

AN EFFICIENT METHOD FOR ACQUIRING AUDITING PROCEDURAL KNOWLEDGE Jane Dillard-Eggers and Michael L. Roberts ABSTRACT In light of advances in the theory of cognition (Anderson, 1996, 2000; Anderson & Fincham, 1994; Anderson & Lebiere, 1998) and research on learning from worked examples (Atkinson et al., 2000; Cooper & Sweller, 1987; Sweller & Cooper, 1985), this study extends earlier research ﬁndings that auditors need practice and certain kinds of feedback to acquire procedural knowledge to identify causes of variations between expected and actual ﬁnancial ratios. We test an alternative form of instruction: worked examples. As predicted by Anderson’s ACT-R 4.0 theory, the results indicate individuals’ pre-test declarative knowledge interacts signiﬁcantly with learning method (with or without examples) on procedural knowledge acquisition. In contrast to prior ﬁndings, this study shows that improvements in auditing procedural knowledge can be achieved by passive instruction in worked examples, a potentially more efﬁcient (cost-effective) method than practice and feedback for auditor training.

Advances in Accounting Behavioral Research, Volume 13, 89 111 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013008

89

90

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

INTRODUCTION Auditors play a critical role in the capital markets and in the overall economy by examining ﬁnancial reports and rendering opinions on the fairness of the reported information. Auditors must exercise professional judgment in conducting audit procedures and rendering audit opinions based on those procedures. Developing procedural knowledge is believed to be a key component in developing expertise (Ericsson & Smith, 1991) – the expertise that auditors must have to evaluate ﬁnancial reports and form expert judgments. Thus, understanding how auditors can better acquire the procedural knowledge essential to form professional judgments is important to the profession. Prior research has concluded experience (i.e., practice and feedback) is necessary in order for auditors to develop procedural knowledge and that instruction alone is insufﬁcient to develop procedural knowledge (Bonner & Walker, 1994). Earley (2001) found novice auditors could develop procedural knowledge in a different setting merely from being prompted to generate their own explanations for audit data; that is, that explanatory feedback is not always necessary. On the basis of advances in the theory of cognition (Anderson, 1996, 2000; Anderson & Fincham, 1994; Anderson & Lebiere, 1998) and research on learning from worked examples (Atkinson, Derry, Renkl, & Wortham, 2000; Cooper & Sweller, 1987; Sweller & Cooper, 1985), this study attempts to extend the work of Bonner and Walker (1994) and Earley (2001) by investigating the efﬁcacy of an alternative method of instruction: worked examples. This study examines whether instruction alone using worked examples can be sufﬁcient for less-experienced auditors to develop auditing procedural knowledge. The results indicate novice auditors are able to acquire analytical review procedural knowledge as a result of studying declarative information in the form of worked examples. In addition, as predicted by Anderson’s model, the results indicate a signiﬁcant interaction between the individual’s pre-test declarative knowledge and learning method (with examples or without) on the individual’s post-test procedural knowledge. The results of this research contribute to our general understanding of how people learn and could be helpful for developing training aids to enhance learning in audit practice. The ﬁndings demonstrate that costly practice and feedback are not always necessary to acquire procedural knowledge. Furthermore, we identify an alternative, more cost-effective method for auditor training and the conditions in which this alternative is/is not likely to be effective.

Acquiring Auditing Procedural Knowledge

91

THEORY AND HYPOTHESES Necessary Conditions for Acquiring Knowledge Prior research has identiﬁed numerous conditions under which new knowledge is more likely to be understood or comprehended (i.e., as cognition moves from level 1, Knowledge, to level 2, Comprehension, in Bloom’s Taxonomy) (Bloom, Engelhart, Frost, Hill, & Krathwohl, 1956). These conditions include: 1. the immediate goal is to comprehend, not to perform higher-level cognitive tasks (Anderson, 1990), 2. the new information involves a small amount of new knowledge that can easily be managed by working memory (Anderson, 1990; Peterson & Peterson, 1959), 3. the new knowledge consists of conceptually simple declarative statements, as opposed to pieces of a more complex network or conditional statements (a corollary is that newly presented knowledge that is abstract or vague is clariﬁed for the learner by concrete examples) (Baddeley, Thomson, & Buchanan, 1975), 4. the learner possesses pre-existing knowledge that allows new terminology and relationships to be easily recognized (Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977), and 5. the learner is either provided with a rationale for why the new information is important or such rationale is clearly discernible (i.e., the learner is sufﬁciently motivated) (Maslow, 1970; Stipek, 1993). Note that conditions 1, 4, and 5 relate primarily to the learner rather than the new information, that is, to the learner’s (1) goal or process orientation, (4) possession of pre-existing knowledge, and (5) understanding of why the new information is important. Conditions 2 and 3 relate primarily to the new information, that is, (2) the amount and (3) the complexity.

Conditions for Acquiring Procedural Knowledge Some cognitive psychologists make the distinction between declarative and procedural knowledge (Anderson, 1976, 1996; Matlin, 1998). Declarative knowledge is knowledge of facts or knowledge that; it consists of speciﬁc facts, generalities, and rules that can be stated or declared (Anderson & Lebiere, 1998; Woolfolk, 1998). In contrast, procedural knowledge

92

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

addresses the question how; it requires action, either mental or physical, and often is not verbally available. Earlier cognitive theories (e.g., Anderson, 1980; Fitts & Posner, 1967) proposed a three-stage mechanism for developing procedural knowledge: (1) a cognitive stage, during which declarative knowledge is acquired; (2) an associative stage, during which individuals practice using the declarative knowledge in a task; and (3) an automatic stage, during which task performance increases in speed (and, perhaps, accuracy) and the ability to explain (declare) how one performs the task is ‘‘lost.’’ According to this view, declarative knowledge should be relatively more important during the cognitive stage, as comprehension is developing, and during the early part of the associative stage. Practice and feedback should be relatively more important, and increase in importance, throughout the associative stage.1 These earlier theories served as the basis for Bonner and Walker’s (1994) research that stressed the importance of practice and feedback to the acquisition of procedural knowledge. More recent cognitive theories, however, emphasize the continuing interaction of declarative knowledge and procedural knowledge when solving problems or performing tasks that are not overly repetitive (Anderson, 1996, 2000; Anderson & Lebiere, 1998; Woolfolk, 1998). Research has suggested procedural knowledge is acquired through a ‘‘modest’’ mechanism of inferring rules needed to transform declarative knowledge to achieve a goal (Anderson, 1996; Anderson, Fincham, & Douglass, 1997; Anderson & Lebiere, 1998). This ‘‘learning by imitation’’ involves reasoning by analogy, hypothesizing, and/or mimicking examples (Anderson, 1996; Cheng, Holyoak, Nisbett, & Oliver, 1986; Reed & Actor, 1991). According to Anderson and Fincham (1994) what is initially retrieved from declarative memory during this process of acquiring procedural knowledge are examples of how the procedure should be executed. The initial use of examples entails reasoning by analogy and gradually new rules (or schemas) are compiled that summarize the analogy process. Anderson and Fincham (1994) acknowledge that this process is not the only way to acquire procedural knowledge. However, they believe that their results indicate that it is a ‘‘major avenue’’ for the acquisition of procedural knowledge. (McCall, Arnold, & Sutton, 2008 present a clear, concise summary of this process.) Novick and Holyoak (1991) note that schemas do not initially or automatically replace the speciﬁc examples in memory, so either may be used to solve later problems. Caplan and Schooler (1990) ﬁnd individuals perform best when the same type of processing [either episode-based

93

Acquiring Auditing Procedural Knowledge

(i.e., examples) or rules-based (i.e., schemas)] used in the initial encoding in memory is also used to retrieve information to solve current problems. For logically complex tasks, rule-based processing provides greater beneﬁts to problem-solving; for less complex tasks, episode-based processing provides greater beneﬁts. Fig. 1 depicts both the older and the newer views of the roles of declarative and procedural knowledge in procedural knowledge acquisition. The older view, shown in panel A, nicely explains the common ﬁnding that Panel A: Older View Declarative Knowledge is a Temporal Phase in Procedural Knowledge Acquisition Declarative Knowledge

Procedural Knowledge

Perceive/ → Comprehend/ Attend Retain

→

Compile → Transition → Automatic

Panel B: Newer View Declarative Knowledge Interacts with Procedural Knowledge in Procedural Knowledge Acquisition New Problem

Retrieve Declarative Knowledge

Production Rule Acquired by Analogy

Complete Production Rule

Success? No Retry

Yes Continue

Notes: The older view of the roles of declarative and procedural knowledge held that declarative knowledge preceded the development of procedural knowledge, that declarative knowledge was transformed (compiled) into automatized, nondeclarable procedural knowledge during practice and feedback, and that procedural knowledge rather than declarative knowledge determined performance of problem-solving tasks (Anderson, 1980; Bonner & Walker, 1994). The newer view is that procedural knowledge continues to interact with, and depends on, declarative knowledge during the performance of problem-solving tasks through the mechanism of trial-and-error production rules consisting of if-then hypotheses (Anderson & Lebiere, 1998).

Fig. 1.

Changing Relations between Declarative and Procedural Knowledge.

94

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

experts are often unable to explain how their expert knowledge guides their actions (Ericsson & Smith, 1991). In contrast, the newer view, shown in panel B, maintains declarative knowledge (often in the form of examples or previously acquired schemas related to those examples) is actively referenced throughout the reasoning process whenever existing procedural knowledge must be adapted to new contexts.

The Role of Feedback Feedback can be useful for acquiring both declarative and procedural knowledge – for example, to test one’s comprehension of declarative knowledge of deﬁnitions and procedural rules and, if necessary, correct misunderstandings. Leung and Trotman (2008) found that task properties (explanatory) feedback can increase the extent of conﬁgural cue processing which is positively associated with performance on an audit risk assessment task. In addition, because task properties (explanatory) feedback explicitly states the rules of the task, it can facilitate the transfer of learning to other tasks or other contexts (Leung & Trotman, 2005). Problem-solving practice is also important when the goal is to attain automatic performance (Atkinson et al., 2000). Bonner and Walker (1994) tested different methods of instruction and feedback for acquiring procedural knowledge about ratio analysis in audit planning. The three methods of instruction examined were as follows: none (i.e., no instruction2), how-to rules (a high-level list3 of seven steps to perform), and understanding rules (detailed explanations of how over- and understatements in numerators and denominators affect mathematical ratios). Four methods of feedback were examined: none (no practice), none (practice, but no feedback), outcome feedback, and explanatory feedback (i.e., an explanation of how a recording error affected account balances and how the changes in those account balances affected each of four ﬁnancial ratios) (Table 1). Bonner and Walker’s (1994) overriding hypothesis was that procedural knowledge could not be acquired through instruction alone; experience (both practice and feedback) was a necessary condition. They hypothesized that in only four of their 12 experimental conditions would procedural knowledge be acquired: by all three instruction groups that received explanatory feedback and by the one group that received outcome feedback along with instruction composed of understanding (math) rules. Their results supported the hypotheses with two exceptions: (1) the group that

95

Acquiring Auditing Procedural Knowledge

Table 1.

Bonner and Walker’s (1994) Experimental Design and Selected Results. Instruction

Practice and feedback None Practice only

None

How to rules

Math rules

þ1.14 (0.12)

þ1.50 (0.02)

þ0.50 (0.56)

þ2.13 (0.01)

þ1.87 (0.01) þ1.75 (0.06)

no feedback

Practice with outcome feedback Practice with explanatory feedback

þ1.50 (0.03)

Notes: Signed changes in procedural knowledge for a ratio analysis task, from pre to posttest, as reported in Bonner and Walker, Table 1, along with t test probabilities. We follow Bonner and Walker’s designations, except for clarity, we use the term math rules rather than the more generic term understanding rules. The ‘‘None’’ instruction treatment group, as well as all other treatment groups, received passive instruction concerning the accounts used in the numerators and denominators of the four relevant ﬁnancial ratios: current ratio, accounts receivables turnover, inventory turnover, and gross margin ratio.

should have improved the most in procedural knowledge (bottom/right cell in Table 1) was only marginally signiﬁcant (p ¼ .06) and (2) one of the control groups that received no experience (neither practice nor feedback) (top/middle cell in Table 1) improved signiﬁcantly in procedural knowledge (p ¼ .02). Bonner and Walker (1994) did not investigate the use of worked examples as an instruction method. However, they did suggest that future research investigate other forms of instruction, speciﬁcally examples (1994). Although the beneﬁts of feedback to learning have been demonstrated particularly in complex learning environments, practice and feedback may be unnecessary when comprehension of either declarative or procedural knowledge can be achieved by other means. For example, Earley (2001) demonstrates novice auditors are able to learn how to recognize patterns (evidence of procedural knowledge) when prompted to look for patterns and provide plausible explanations. ‘‘If a trainer makes the underlying structure of a problem obvious by requiring participants to abstract the relevant underlying features from examples, the participants are then better able to solve unrelated problems’’ (Earley, 2001, p. 84). Although explanatory

96

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

feedback is effective in improving learning, it did not improve learning beyond that achieved through self-explanation (Earley, 2001, 2003). Similarly, Mascha (2001) found feedback provided in an expert system assists in procedural knowledge acquisition only when the task is complex; no beneﬁt to knowledge acquisition is derived from feedback when the task is simple. Thus, the important condition for learners/decision makers to achieve is comprehension of new information. If feedback facilitates comprehension, it is useful and may be necessary. However, if comprehension can be achieved without feedback, then feedback is not necessary to the design of a learning or decision support system.

Research on Knowledge Acquisition in Knowledge-Based Systems Research on the development and use of knowledge-based systems (KBS), for example, intelligent systems, intelligent decision aids (IDAs), and/or expert systems, has investigated a number of the conditions for knowledge acquisition. Gregor and Benbasat (1999) reviewed the extant literature on use of explanations by intelligent systems, using three prominent cognitive theories (cognitive effort perspective, cognitive learning theory, and Toulmin’s model of argumentation) to develop a unifying theoretical framework. They developed a list of nine propositions based on these theories that relate to who will use explanations within an intelligent system and for what purpose, the impact of explanations on perceptions of beneﬁt, and the nonuse of explanations because of cognitive effort issues.4 Gregor and Benbasat (1999) propose explanations will be used more when the user has a goal of long-term learning (learning that transfers to a non-KBS system) (the same as necessary condition number one, discussed earlier). They also propose explanations that require less-cognitive effort to access and assimilate will be used more frequently (similarly, necessary condition number 2, earlier, states information is more likely to be comprehended when it involves only a small amount of new knowledge that can be easily managed in working memory). Furthermore, they propose expert users have the pre-existing knowledge to understand what information is needed and why the information is important (conditions 4 and 5, discussed earlier). McCall et al. (2008) cite mixed prior ﬁndings indicating users of intelligent systems may acquire less explicit knowledge than users of traditional resources; mixed results of this nature would be expected if prior studies did

Acquiring Auditing Procedural Knowledge

97

not control for goal orientation of users as suggested by necessary condition one above and by Gregor and Benbasat (1999). McCall et al. (2008) also note that a user may be easily overwhelmed by the amount of information available (violating necessary condition 2). In addition, the easy access of explicit knowledge (which may even appear automatically) decreases effort normally required in interpretive problem solving, which can lead to failure to encode the new knowledge, especially for ‘‘a user focused on problem solving’’ (McCall et al., 2008, p. 83). Arnold, Clark, Collier, Leech, and Sutton (2004) found experts and novices had different expectations regarding an IDA and perceived different beneﬁts. Experts reported receiving less beneﬁt than expected; novices reported more beneﬁt than expected (possibly because the explanations provided new information to the novices, but not to the experts). Novices (compared to experts) preferred feedforward explanations (generic explanations without case-speciﬁc information) to understand the underlying knowledge structures of the problem. Arnold, Clark, Collier, Leech, and Sutton (2006) found novices preferred explicit knowledge support, and experts preferred tacit knowledge support. Thus, each participant type preferred the type of knowledge support that provided the best match for their pre-existing knowledge (supporting necessary condition 2). Smedley and Sutton (2007) applied Anderson’s ACT-R 4.0 theory to identify two techniques (abstraction and goal structuring) theorized to provide learners with a simpliﬁed and situation-responsive set of production rules to use in a problem-solving context. Abstraction (the process of making production rules more general in nature) and goal structuring (the process of forming production rules that encompass different outcomes) explanations were provided to users in a KBS environment. Participants receiving goal-structuring explanations exhibited better problem-solving performance. Joint presentation of goal-structuring and abstraction explanations led to even greater problem-solving improvements. However, presentation of abstraction explanations alone did not lead to improved performance. These ﬁndings support the importance of necessary condition one related to the importance of goal orientation during problem solving for knowledge acquisition. A potential hazard of embedding KBS with knowledge acquisition assistance is the system could become a ‘‘crutch’’ and lead users to depend on the guidance in the system instead of enriching their own comprehension, knowledge structures, and expertise. McCall et al. (2008) found users with a knowledge management system outperformed problem-solvers who had access only to traditional resources. However, the advantage disappeared

98

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

when the knowledge management system was removed. Interestingly, although the traditional group acquired more rule-based knowledge, the knowledge management system group acquired more higher-level knowledge related to problem-solving skills. Thus, use of traditional resources versus knowledge management system resources produced different kinds of knowledge acquisition.

Learning from Worked Examples Since the 1950s, researchers have studied how the use of worked examples can facilitate learning in different ways. Worked examples are instructional devices that ‘‘provide an expert’s problem-solving model for a learner to study and emulate’’ (Atkinson et al., 2000). Experts focus on deeper structural aspects of the problem and use previously acquired schemas to classify problems and identify solutions (Paas & van Merrienboer, 1994; Sweller & Cooper, 1985; Sweller, van Merrienboer, & Paas, 1998; Ward & Sweller, 1990). Therefore, a major element in an individual’s movement from novice to expert is the acquisition of domain-speciﬁc knowledge in the form of schemas. How can a novice auditor acquire these schemas? In nonaudit settings, training using worked examples has been shown to be very effective in schema development (Carroll, 1994; Cooper & Sweller, 1987; Paas & van Merrienboer, 1994; Sweller & Cooper, 1985). Carroll (1994, p. 365) states that the use of worked examples ‘‘provides a scaffolding for learning,’’ illustrating important concepts and relationships. Students who are presented worked examples tend to focus more on the structural aspects of problems (Sweller & Cooper, 1985; Zhu & Simon, 1987). Earley (2001) documents the effectiveness of worked examples in cognitive psychology studies and concludes these results should apply to the ‘‘ill-structured’’ auditing domain. She ﬁrst required participants to practice by solving four case problems and then incorporated the essence of worked examples as ‘‘explanatory feedback’’ in the practice (treatment) session. We examine whether the use of worked examples as part of instruction (rather than explanatory feedback) is also effective; that it is not necessary to practice (solve the problem) before receiving the explanations contained in the worked examples. So, our research goes one step further than Earley (2001) and investigates conditions where procedural knowledge can be acquired through instruction only, that is, without practice plus prompted, self-generated explanations or practice and feedback.

Acquiring Auditing Procedural Knowledge

99

Hypotheses On the basis of prior research showing procedural knowledge gains from studying worked examples and theoretical advancements in Anderson’s theory of cognition emphasizing the continuing importance of declarative knowledge as well as the role of worked examples in acquiring procedural knowledge, we test the following hypothesis: H1. Auditing procedural knowledge can be acquired as a result of exposure to declarative knowledge contained in worked examples (i.e., passive instruction without practice with either outcome or explanatory feedback). Although prior research suggests this hypothesis, whether worked examples will be effective in procedural knowledge acquisition is not a foregone conclusion. As discussed in the methodology section later, the materials for this study were adapted from materials developed by Bonner and Walker (1994). As such, content of our worked examples is limited to the verbatim wording of Bonner and Walker’s (1994) eight practice problems with explanatory feedback. This necessitated violating two of the necessary conditions for knowledge acquisition described earlier: conditions 2 (a small amount of new information) and 3 (new information consisting of conceptually simple declarative statements). These departures provide tension for H1: will observance of conditions 1, 4, and 5 be sufﬁcient for the worked examples to be effective? This study also tests a second hypothesis derived from Anderson’s ACT-R 4.0 theory of knowledge acquisition that declarative knowledge is actively engaged throughout the reasoning process whenever existing procedural knowledge must be adapted to new contexts. Therefore, we test the following hypothesis: H2. Pre-existing declarative knowledge and exposure to worked examples will interact to produce auditing procedural knowledge.

METHODOLOGY Task The tasks used in this study were originally created and used by Bonner and Walker (1994). These tasks include examining differences between actual and expected ratios for four different ﬁnancial ratios at a time – current

100

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

ratio, gross margin ratio, accounts receivables turnover, and inventory turnover. Participants are required to select which one of four possible transaction-recording errors could have produced the pattern of changes in all four ratios. Ratio analysis is a routine part of audit planning, performed to allocate hours to areas that are more likely to require the auditors’ attention and testing. Bonner and Walker (1994) identiﬁed four types of declarative knowledge needed to perform ratio analysis: (1) classiﬁcation of ﬁnancial statement accounts by account type, (2) deﬁnitions of summary items on ﬁnancial statements, (3) journal entries for transactions, and (4) deﬁnitions of ﬁnancial ratios. The four types of procedural knowledge Bonner and Walker (1994) identiﬁed were: (1) rules on the mathematics of ratios (math rules), (2) rules for the effects of ﬁnancial statement errors on account balances (forward reasoning of the effects of errors on account balances), (3) rules showing what types of errors could have caused accounts to be misstated (backward reasoning of the same information in type 2), and (4) rules showing what types of errors could have caused ratios to be misstated. Like Bonner and Walker (1994), our experiment was organized in four phases: (1) knowledge pretest, (2) study phase, (3) test phase, and (4) knowledge posttest.5 This study employed two groups, and each group performed the same tasks in phases 1, 3, and 4; only phase 2 varied between groups. In phase 2, both the control group and the experimental (examples) group received declarative knowledge in the form of deﬁnitions of the four ﬁnancial ratios and math rules in the study phase, as in Bonner and Walker (1994). The experimental group received additional instruction consisting of eight ratio worked examples containing exactly the same information as in Bonner and Walker’s (1994) practice with explanatory feedback treatment (i.e., our participants received the same information content but in worked example format, without the practice and feedback). Research in the use of worked examples indicates that multiple examples are required when individuals are asked to learn complex concepts during instruction (Atkinson et al., 2000; Cooper & Sweller, 1987; Sweller & Cooper, 1985). The research design for this study, shown in Table 2, uses multiple examples that are integrated (each example can be understood in isolation); characteristics that Atkinson et al. (2000) identiﬁed as facilitating knowledge acquisition. All phases of the experiment were conducted using personal computers, as in Bonner and Walker (1994). Bonner and Walker (1994) imposed time limits on each question in each phase of their experiment. This study

101

Acquiring Auditing Procedural Knowledge

Table 2.

Experimental Design and Materials per Participant-Group.

Group Control group: ratio deﬁnitionsþmath rules Experimental group: ratio deﬁnitionsþmath rulesþtransaction knowledge through examples

Phase 1

Phase 2

Knowledge pretest Knowledge pretest

Phase 3

Phase 4

Ratio deﬁnitionsþmath Test Knowledge rules problems posttest Ratio deﬁnitionsþmath Test Knowledge rulesþratio analysis problems posttest examples

employed the same time limits on each question. The mean total times spent on the experiment for the control and experimental groups, respectively, were 66 and 63 minutes.6

Participants Since these experiments involve knowledge acquisition, it is important that participants not already possess the knowledge that is the focus of the knowledge acquisition task. Auditing students are ideal for this task because they have received instruction in introductory and intermediate accounting courses about ﬁnancial accounts, recording transactions, ﬁnancial statement classiﬁcation of accounts, and ﬁnancial ratios, the ﬁrst two of which comprise knowledge that is assumed for this experiment (the third and fourth are measured in the experiment). Additionally, they have not yet been exposed to deducing recording errors from analysis of ﬁnancial ratios, which is the critical knowledge to be acquired. Therefore, similar to Bonner and Walker (1994), this study employed undergraduate auditing students who had not yet studied ﬁnancial statement error detection using ratio analysis.7 Auditing students should be well motivated to learn the new information presented in these experiments because it is highly relevant to auditing and ﬁnancial statement analysis. To further enhance motivation, Bonner and Walker (1994) employed volunteer student participants and paid them $7.50 per hour for their participation and awarded prizes to top performers in each experimental session. This study differs slightly from Bonner and Walker’s (1994) approach. Volunteer auditing students were also used, but they were rewarded with extra credit points for participating in the study; thus, the students were motivated to participate, but not

102

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

to outperform their peers. To control for differences in motivation between groups, each participant self-rated their concentration and motivation levels on a 1 to 100 scale. Between group differences were not signiﬁcant (pW.10). Thirty-three auditing students participated in the experiment. Participants were randomly assigned to control or experimental groups using a random number generator within the computerized experimental instrument, producing a distribution of 14 and 19 participants, respectively, per cell. This compares favorably to cell sizes of seven and eight participants in Bonner and Walker’s (1994) study.

Variables Knowledge acquisition was measured using the pretest to posttest improvement on the same 12-item procedural knowledge test developed and used by Bonner and Walker (1994). The procedural knowledge test includes four sets of three questions testing procedural knowledge of (1) the mathematics of ratios, (2) rules for the effects of ﬁnancial statement errors on account balances, (3) rules showing what types of errors could have caused accounts to be misstated, and (4) rules showing what types of errors could have caused ratios to be misstated. The Appendix contains all 12 questions. To summarize, the dependent variable is procedural knowledge acquisition, as measured by posttest minus pretest differences on Bonner and Walker’s (1994) 12-item procedural knowledge test. Our categorical independent variable is the presence or absence of worked examples instruction.

RESULTS Table 3 presents results, including t-tests (two-tailed, .05 signiﬁcance level), for procedural knowledge acquisition measured by Bonner and Walker’s (1994) procedural knowledge test for the control and experimental groups.8 Participants in the control group improved on average .50 of 12 possible points (insigniﬁcant improvement at p ¼ .30), whereas participants in the experimental group, who studied worked examples, improved on average 1.37 points of 12 possible points (signiﬁcant improvement at p ¼ .003). The results for the control group (no improvement) are identical to Bonner and Walker’s (1994) results. However, the signiﬁcant improvement shown for the experimental (examples) group is inconsistent with Bonner

103

Acquiring Auditing Procedural Knowledge

Table 3. Procedural Knowledge Acquisition by Group on Bonner and Walker’s (1994) Procedural Knowledge Test. Group

Control group: ratio deﬁnitionsþmath rules Experimental group: ratio deﬁnitionsþmath rulesþexamples

Variable

Procedural knowledge Procedural knowledge

N

Improvement

t test

Mean

Standard error

t value

p value

14

0.50

0.45

1.1

0.300

19

1.37

0.39

3.5

0.003

and Walker’s (1994) conclusion that (1) practice with explanatory feedback, regardless of instruction, or (2) instruction consisting of understanding rules plus practice with outcome feedback are necessary for acquiring procedural knowledge. These results extend Earley’s (2001) ﬁnding in a different decision context that practice combined with self-explanations can produce improvement in procedural knowledge. Our results demonstrate auditing procedural knowledge as measured by Bonner and Walker’s (1994) procedural knowledge test can be achieved by studying worked examples, a form of passive instruction that emphasizes the continuing importance of declarative knowledge in developing procedural knowledge as predicted by Anderson’s ACT-R 4.0 theory of cognition. Results in Table 3 indicate the experimental group improved signiﬁcantly, whereas the control group did not. However, the control group’s post-test procedural knowledge was .50 higher than their pre-test score. To examine whether the improvement by the experimental group (1.37) is signiﬁcantly greater than the improvement shown by the control group (.50), a twogroup t-test was conducted using the between-group difference in improvement. The difference (.87) is signiﬁcant at po.10 (one-tailed) (t ¼ 1.45, p ¼ .079). Thus, between-group differences as well as univariate, pre-test versus post-test differences for each group examined separately, support our hypothesis.9 According to the ACT-R 4.0 model, procedural knowledge develops as an individual adapts prior declarative knowledge to new situations. To test whether participants’ pre-existing declarative knowledge interacts with learning method to produce increases in procedural knowledge, an interaction term was created consisting of GROUP (participation in either

104

Table 4.

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

Interaction of Declarative Knowledge and Learning Method on Posttest Procedural Knowledge.

Source

DF

Sum of Squares

Mean Square

F Value

pWF

GROUP DECLARE PRE Error Corrected total

2 30 32

11.37 84.63 96.00

5.68 2.82

2.01

0.076

the control group or the experimental group) DECLARE-PRE (pre-test declarative knowledge, measured, as suggested by Bonner & Walker, 1994, by participants’ correct answers on knowledge of the relevant accounting ratio deﬁnitions). The dependent variable was each participant’s post-test score on the 12-item procedural knowledge test. Results of the general linear model (PROC GLM in SAS), shown in Table 4, indicate the model and interaction term are signiﬁcant at po.10 (one-tailed test). The possible association between participants’ individual self-reported concentration and motivation measures and improvements in procedural knowledge was also examined. The results show a marginally signiﬁcant correlation between procedural knowledge acquisition and self-reported concentration at p ¼ .10 (one-tailed). There was no signiﬁcant correlation between procedural knowledge acquisition and motivation (p ¼ .16, onetailed test). Self-reported knowledge was highly signiﬁcant at p ¼ .0014.

SUMMARY, IMPLICATIONS, AND CONTRIBUTIONS This study uses the same experimental procedures, the same analytical review tasks, and the same procedural knowledge test created and used by Bonner and Walker (1994). When an alternative form of instruction, worked examples, is examined, we ﬁnd instruction alone can result in the acquisition of procedural knowledge. Thus, we extend Bonner and Walker’s (1994) ﬁnding that practice and feedback are always necessary for procedural knowledge acquisition and Earley’s (2001) ﬁnding that practice and selfexplanation increase procedural knowledge. Studying worked examples, a form of passive instruction, can also be effective in achieving procedural knowledge of ratio analysis in auditing. It is important to note the purpose of this research is not a comparison of the beneﬁt to learning from studying worked examples compared to practice and feedback. This research does not contend that practice and feedback

Acquiring Auditing Procedural Knowledge

105

does not contribute to learning; nor does it propose that studying worked examples contributes to learning more than practice and feedback. Further research that examines the relative beneﬁts of these two avenues to acquire procedural knowledge would provide useful information. However, this research has important implications for practice because it supports a more efﬁcient, less costly alternative for auditor training. Training utilizing practice and feedback is more costly than training using instruction only (Earley, 2001). Thus, these results suggest guidelines for developing knowledge acquisition training utilizing instruction alone. If both practice with feedback and passive instruction training methods are effective to enhance learning and build expertise, the least costly alternative would provide the greater cost-savings beneﬁt. Atkinson et al. (2000, p. 185) states that worked examples literature is ‘‘particularly relevant to programs of instruction that seek to promote skills acquisition, a goal of many workplace training environments.’’ In addition, the beneﬁts of worked examples in computer-aided instruction have been recognized for a number of years (Zhu & Simon, 1987). The declarative knowledge contained in worked examples could readily be incorporated into the distributed learning systems that are commonly used by accounting professionals (cf. Smedley & Sutton, 2004). Our research has important implications in other areas as well. The use of worked examples to aid novice auditors in acquiring procedural knowledge is particularly well suited for use in applying Ericsson and Smith’s (1991) expertise approach to systematically identify and study the development of expertise in judgment/decision making. The expertise approach involves three stages of research design: (1) identiﬁcation of representative tasks by which superior performance can be reproduced, (2) analysis of knowledge differences and cognitive processes mediating that performance and design of experimental tasks to elicit critical aspects of performance, and (3) theoretical and empirical accounts of how the identiﬁed aspects can be acquired (p. 32). The focus of the expertise approach is the development of expert schemas in nonexperts. Thus, as evidenced in our study, whenever audit expertise needs to be developed in nonexperts to meet new auditing demands, the use of worked examples that facilitate the formation of schemas could be beneﬁcial.

NOTES 1. Task performance is known to improve through the Power Law of Practice, but this relation describes speed, not developing comprehension/understanding.

106

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

2. However, all 12 treatment groups received instruction on how to calculate relevant ﬁnancial ratios. Since the participants should have been exposed to this instruction earlier, the ratio deﬁnitions should have served to refresh the memories of participants in each treatment group equally. 3. For example, step 5 is ‘‘hypothesize possible accounting errors or irregularities that could have caused the pattern of ratio differences you have.’’ However, participants were not given any training in how to ‘‘hypothesize possible accounting errors or irregularities.’’ Thus, the how to rules are ‘‘high level’’ in the sense of providing abstract procedures without concrete examples. 4. Proposition 1: Explanations will be used when the user experiences an expectation failure or perceives an anomaly. Proposition 2: Explanations will be used more when the user has a goal of long term learning (i.e., learning that transfers to a non KBS context). Proposition 3: Explanations will be used when the user lacks knowledge needed so he or she can contribute to problem solving. Proposition 4: Explanations that require less cognitive effort to access and assimilate will be used more and will be more effective with respect to performance, learning, or user perceptions. Proposition 5: Use of explanations improves the performance achieved with a KBS as an aid. Proposition 6: Use of explanations aids learning (transfer of knowledge to non KBS contexts). Proposition 7: Novices will use explanations more for learning (short and long term) than experts. Proposition 8: Experts will use explanations more for resolving anomalies (disagreement) and for veriﬁcation than novices. Proposition 9: Use of explanations conforming to Toulmin’s model (justiﬁcation explanations) will give rise to more positive user perceptions of a KBS than other explanations (trace and strategic explanations). 5. Our focus in this chapter is performance on the 12 item test of procedural knowledge developed by Bonner and Walker. Their pre test and post tests of knowledge included 24 items 12 declarative items and 12 procedural items. We developed an expanded knowledge test consisting of 61 test questions. Thirteen test items were needed to test for classiﬁcation of accounts by type and knowledge of the components of the ﬁnancial ratios. Each of the 16 transactions used in the study and test phases required two knowledge test items to identify effects on accounts (i.e., both debits and credits). Two of the transactions required four test items because perpetual inventory accounting was used. Eliminating two redundancies produced a total of 32 test items to capture the knowledge needed for effects of errors on account balances. Another 16 test items were needed to test for knowledge of effects of the transaction recording errors on math ratios. 6. Bonner and Walker reported only that their participants ‘‘spent approximately 80 140 minutes completing the study.’’ 7. For similar reasons McCall et al. (2008) employed undergraduate managerial accounting students in their study of the effects of traditional versus knowledge based systems on knowledge acquisition. 8. Wilcoxon signed ranks tests produced the same results as the parametric t tests. 9. Analysis of variance results [performed using PROC GLM in the Statistical Analysis System (SAS) to control for unequal cell sizes] produces identical results to the between group t test.

Acquiring Auditing Procedural Knowledge

107

ACKNOWLEDGMENTS This research was supported by grants from the University of Colorado Denver and the University of Alabama Culverhouse College of Commerce and Business Administration and Culverhouse School of Accountancy. We gratefully acknowledge the cooperation and assistance of Rich Houston and John Mason, who encouraged their students to participate in this research; Sarah Bonner, who provided copies of materials used in Bonner and Walker (1994); Theresa Roberts, who wrote the computer program used in this research; and Lorraine Magrath, who suggested additional ways to analyze knowledge. We also appreciate comments from Christine Earley, Tom Kozloski, Todd DeZoort, Rob Ingram, Bob Michaelsen, and workshop participants at the University of Alabama, University of Colorado Denver, North Texas University, Florida State University, the American Accounting Association Annual Meeting, and the International Symposium on Audit Research.

REFERENCES Anderson, J. R. (1976). Language, memory, and thought. Hillsdale, NJ: Lawrence Erlbaum. Anderson, J. R. (1980). Cognitive psychology and its implications. New York: W. H. Freeman and Company. Anderson, J. R. (1990). Cognitive psychology and its implications (3rd ed.). New York: W. H. Freeman and Company. Anderson, J. R. (1996). ACT: A simple theory of complex cognition. American Psychologist, 51(4), 355 365. Anderson, J. R. (2000). Learning and memory. New York: Wiley. Anderson, J. R., & Fincham, J. M. (1994). Acquisition of procedural skills from examples. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(6), 1322 1340. Anderson, J. R., Fincham, J. M., & Douglass, S. (1997). The role of examples and rules in the acquisition of a cognitive skill. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23(4), 932 945. Anderson, J. R., & Lebiere, C. (1998). Learning. In: J. R. Anderson & C. Lebiere (Eds), The atomic components of thought. Mahwah, NJ: Lawrence Erlbaum Associates. Arnold, V., Clark, N., Collier, P. A., Leech, S. A., & Sutton, S. G. (2004). Explanation provision and use in an intelligent decision aid. Intelligent Systems in Accounting, Finance and Management, 12, 5 27. Arnold, V., Clark, N., Collier, P. A., Leech, S. A., & Sutton, S. G. (2006). The differential use and effect of knowledge based system explanations in novice and expert judgment decisions. MIS Quarterly, 30(1), 79 97.

108

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

Atkinson, R. K., Derry, S. J., Renkl, A., & Wortham, D. (2000). Learning from examples: Instructional principles from worked examples research. Review of Educational Research, 70(2), 181 214. Baddeley, A. D., Thomson, N., & Buchanan, M. (1975). Word length and the structure of short term memory. Journal of Verbal Learning and Verbal Behavior, 14, 575 589. Bloom, B. S., Engelhart, M. D., Frost, E. J., Hill, W. H., & Krathwohl, D. R. (1956). Taxonomy of educational objectives. Handbook 1: Cognitive domain. New York: David McKay. Bonner, S. E., & Walker, P. L. (1994). The effects of instruction and experience on the acquisition of auditing knowledge. The Accounting Review (January), 157 178. Caplan, L. J., & Schooler, C. (1990). Problem solving by reference to rules or previous episodes: The effects of organized training, analogical models, and subsequent complexity of experience. Memory and Cognition, 18(2), 215 227. Carroll, W. M. (1994). Using worked examples as an instructional support in the algebra classroom. Journal of Educational Psychology, 86(3), 360 367. Cheng, P. W., Holyoak, K. J., Nisbett, R. E., & Oliver, L. M. (1986). Pragmatic versus syntactic approaches to training deductive reasoning. Cognitive Psychology, 18, 293 328. Cooper, G., & Sweller, J. (1987). Effects of schema acquisition and rule automation on mathematical problem solving transfer. Journal of Educational Psychology, 79(4), 347 362. Earley, C. E. (2001). Knowledge acquisition in auditing: Training novice auditors to recognize cue relationships in real estate valuation. The Accounting Review, 76(January), 81 97. Earley, C. E. (2003). A note on the use of self explanation as a training tool for novice auditors: The effects of feedback timing and level of reasoning on performance. Behavioral Research in Accounting, 15, 111 124. Ericsson, K. A., & Smith, J. (1991). Prospects and limits of the empirical study of expertise: An introduction. In: K. A. Ericsson & J. Smith (Eds), Toward a general theory of expertise: Prospects and limits. Cambridge: Cambridge University Press. Fitts, P. M., & Posner, M. I. (1967). Human performance. Englewood Cliffs, NJ: Prentice Hall. Gregor, S., & Benbasat, I. (1999). Explanations from intelligent systems: Theoretical foundations and implications for practice. MIS Quarterly, 23(4), 497 530. Leung, P. W., & Trotman, K. T. (2005). The effects of feedback type on auditor judgment performance for conﬁgural and non conﬁgural tasks. Accounting, Organizations and Society (30), 537 553. Leung, P. W., & Trotman, K. T. (2008). Effect of different types of feedback on the level of auditors’ conﬁgural information processing. Accounting and Finance (48), 301 318. Mascha, M. F. (2001). The effect of task complexity and expert system type on the acquisition of procedural knowledge: Some new evidence. International Journal of Accounting Information Systems, 2(2), 103 124. Maslow, A. H. (1970). Motivation and personality (2nd ed.). New York: Harper and Row. Matlin, M. W. (1998). Cognition (4th ed.). Fort Worth, TX: Harcourt Brace. McCall, H., Arnold, V., & Sutton, S. G. (2008). Use of knowledge management systems and the impact on declarative knowledge acquisition. Journal of Information Systems, 22(2), 77 101. Novick, L. R., & Holyoak, K. J. (1991). Mathematical problem solving by analogy. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17(3), 398 413. Paas, F. G. W. C., & van Merrienboer, J. J. G. (1994). Variability of worked examples and transfer of geometrical problem solving skills: A cognitive load approach. Journal of Education Psychology, 86(1), 122 133.

Acquiring Auditing Procedural Knowledge

109

Peterson, L. R., & Peterson, M. (1959). Short term retention of information items. Journal of Experimental Psychology, 58, 193 198. Reed, S. K., & Actor, C. A. (1991). Use of examples and procedures in problem solving. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 753 766. Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information processing: I. Detection, search, and attention. Psychological Review, 84, 1 66. Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and a general theory. Psychological Review, 84, 127 190. Smedley, G., & Sutton, S. G. (2004). The effect of alternative explanation types on procedural knowledge acquisition during knowledge based system use. Journal of Information Systems, 21(1), 27 51. Smedley, G., & Sutton, S. G. (2007). Explanation provision in knowledge based systems: A theory driven approach for knowledge transfer designs. Journal of Emerging Technologies in Accounting, 1, 41 61. Stipek, D. J. (1993). Motivation to learn (2nd ed.). Boston: Allyn and Bacon. Sweller, J., & Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59 89. Sweller, J., van Merrienboer, J. J. G., & Paas, F. G. W. C. (1998). Cognitive architecture and instructional design. Educational Psychology Review, 10(3), 251 296. Ward, M., & Sweller, J. (1990). Structuring effective worked examples. Cognition and Instruction, 7(1), 1 39. Woolfolk, A. (1998). Educational psychology (7th ed.). Boston: Allyn and Bacon. Zhu, X., & Simon, H. A. (1987). Learning mathematics from examples and by doing. Cognition and Instruction, 4(3), 137 166.

APPENDIX. BONNER AND WALKER’S (1994) ANALYTICAL REVIEW PROCEDURAL KNOWLEDGE TEST Item 1. Assume that the ratio of X to Y is 1.5 to 1. What effect would an equal INCREASE in X and Y have on the ratio? (a) Increase the ratio. (b) Decrease the ratio. (c) No effect on the ratio. (d) Answer cannot be determined from the information given. 2. Assume that the ratio of K to M is 8 to 1. What effect would an INCREASE in M have on the ratio? (a) Increase the ratio. (b) Decrease the ratio.

110

JANE DILLARD EGGERS AND MICHAEL L. ROBERTS

(c) No effect on the ratio. (d) Answer cannot be determined from the information given. 3. Assume that the ratio of P to Q is 6 to 2. What effect would P DECREASING by twice as much as Q have on the ratio? (a) Increase the ratio. (b) Decrease the ratio. (c) No effect on the ratio. (d) Answer cannot be determined from the information given. 4. What would be the result of recording sales at year-end even through the goods were not shipped? (a) Overstate accounts receivable. (b) Overstate inventory. (c) Understate cash. (d) Understate sales. 5. What would be the result of adjusting perpetual records incorrectly due to failure to count all goods in the warehouse? (a) Overstate accounts receivable. (b) Understate inventory. (c) Overstate inventory. (d) Understate accounts payable. 6. What would be the result of failing to record any entries related to returned goods bought on account? (a) Overstate accounts payable. (b) Overstate accounts receivable. (c) Overstate inventory. (d) Understate sales. 7. Which of the following could have caused sales to be OVERSTATED? (a) Recording payments on accounts payable twice. (b) Recording a sales that occurred in the current period in the next period. (c) Recording more hours than actually worked to manufacture a ﬁnished product that was later sold. (d) Using too high a price when invoicing and recording a sale. 8. Which of the following could have caused inventory to be UNDERSTATED? (a) Failing to record collections from receivables. (b) Goods returned by customers not counted in inventory.

Acquiring Auditing Procedural Knowledge

111

(c) Labor and overhead recorded in excess of actual amounts throughout the year. (d) Expensing items that should be recorded as prepaids. 9. Which of the following could have caused cash to be UNDERSTATED? (a) Cash receipts recorded twice. (b) Using too high a price when invoicing and recording a sale. (c) Recording tax accruals twice. (d) Failing to record collections from receivables. 10. Which of the following could have caused the current ratio to be OVERSTATED? (a) Recording materials purchases on account in current year’s inventory when they should have been recorded in next year’s inventory. (b) Recording payments on accounts payable twice. (c) Failing to record collections from receivables. (d) Expensing items that should be recorded as prepaids. 11. Which of the following could have caused inventory turnover to be UNDERSTATED? (a) Adjusting perpetual records incorrectly due to failure to count all goods in warehouse (cost of goods sold correct). (b) Recording collections from receivables twice. (c) Including consigned out items in inventory twice. (d) Failing to record any entries related to returned goods bought on account. 12. Which of the following could have caused the gross margin ratio to be OVERSTATED? (a) Recording too many items on the sales invoice and in sales while all shipping documents and inventory entries done correctly. (b) Recording tax accruals twice. (c) Recording more hours than actually worked to manufacture a ﬁnished product later sold. (d) Expensing items that should be recorded as prepaids.

STRESS AND ITS ANTECEDENTS AND CONSEQUENCES IN ACCOUNTING SETTINGS: AN EMPIRICALLY DERIVED THEORETICAL MODEL Kenneth J. Smith, Patricia L. Derrick and Michael R. Koval ABSTRACT Considerable progress has been made over the past 20 years toward the construction of a global stress paradigm for accountants in the workplace. Over this time period, a number of antecedents and consequences of personal and organizational stress have been identiﬁed and empirically veriﬁed. These efforts have provided the foundation for future investigations, which will likely provide additional guidance to those seeking to implement strategies aimed at enhancing individual well-being and organizational efﬁciency. This chapter synthesizes the ﬁndings of these studies to construct a model of the stress dynamic among accountants aimed at guiding future efforts designed to reﬁne our understanding of this critical phenomenon.

Advances in Accounting Behavioral Research, Volume 13, 113 142 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013009

113

114

KENNETH J. SMITH ET AL.

INTRODUCTION The accounting profession has maintained an interest in the phenomenology and prevalence of job-related and personal stress among its members dating back to the pioneering work of Friedman, Rosenman, and Carroll (1958), which measured serum cholesterol levels and blood clotting times of tax accountants during and after the peak tax season. However, Weick’s (1983) proposition that ‘‘stress’’ represents a unifying construct for exploring issues relating to performance and individual well-being among accountants in the workplace arguably served as the primary impetus for a large body of research over the past quarter century into the antecedents and consequences of stress among accountants. In a graphic interpretation of Weick’s propositions, Libby (1983) illustrated a model of the antecedents and consequences of stress in which (1) the quantity and quality of task demands (i.e., work-related stressors), mediated by predictability and control, cause stress; (2) in turn, stress causes increments or decrements in cognitive performance in the short-term, as well as long-run behavioral and physiological consequences. This model has arguably served as the foundation for a substantial body of subsequent research. Early studies subsequent to Weick’s (1983) treatise examined: (1) generalized measures of job stress; (2) relations of organizational stressors such as role conﬂict, role ambiguity, and overload with key job outcomes such as job satisfaction, turnover intentions, and performance; and (3) direct measures of physiological or psychological stress. These studies often produced mixed results (Fogarty, Singh, Rhoads, & Moore, 2000; LePine, Podsakoff, & LePine, 2005). Smith (1990) trichotimized the extant accounting stress literature into studies that examined generalized measures of job stress, those that incorporated job stressor measures, and others that actually measured an actual physiological or psychological stress response. However, Smith (1990, p. 522) notes the failure of those studies to undertake a global approach to examine the stress dynamic, that is, to incorporate stress antecedents, stress measures, and stress consequences into models designed to measure interactions among these constructs. In the two decades that have transpired since Smith’s (1990) review, there has been an ongoing effort on the part of accounting researchers to better understand various aspects of the stress dynamic in accounting settings. The nature and scope of these efforts has varied, yet they appear to have had the common goal of enhancing our understanding of the complex relationships between speciﬁc ‘‘stress’’ measures and their antecedents and/or consequences within the accounting

Stress in Accounting Settings

115

work environment. Libby (1983) emphasized the signiﬁcance of this dynamic after considering Weick’s (1983) statement that stress is an important accompaniment of accounting systems by noting, ‘‘the importance of this statement was not completely apparent until I realized that such a diverse group of important accounting issues could be encompassed under a single unifying concept – stress’’ (Libby, 1983, p. 370). Research over the past two decades that has examined various aspects of the stress phenomenon among accountants has added to our knowledge of the stressor-to-outcome dynamic. Moreover, it provides the basis for the development of a unifying theoretical structure to guide future research efforts as well as interventions designed to enhance performance and wellbeing of accountants in the workplace. The purpose of this chapter is to put forth a comprehensive model of the Antecedents and Consequences of Stress among Accountants based on an examination of the extant research during this period. This model addresses organizational and extra-organizational (i.e., interpersonal) antecedents of stress and their purported impact on noted personal and organizational outcomes. It also focuses on noted stress measures and their direct inﬂuence on the aforementioned outcomes as well as their role as mediators in the relations between stress antecedents and those outcomes. Fig. 1 illustrates the revised theoretical model. Although there is empirical veriﬁcation in prior research for many of the relationships depicted in Fig. 1, no single study to date has simultaneously tested and veriﬁed all the relations depicted therein. Rather, this study represents a synthesis of prior research to build the posited theoretical model using an inductive grounded theory approach as proposed by Glaser and Strauss (1967). In essence, by examining and comparing the ﬁndings from relevant prior studies, we posit relations among the noted constructs, which in turn may be used as a foundation for future research. Following Glaser and Strauss’s Elements of Theory approach (1967), we ﬁrst generate conceptual categories and their properties, then posit generalized relations between speciﬁc properties. For example, Fig. 1 illustrates mediating mechanisms as one of three major categories. Stress arousal represents one of the properties under this category, and we posit relations between stress arousal and a number of other speciﬁc properties (e.g., burnout) as indicated in the model development section. The relations purported herein are naturally subject to further veriﬁcation or refutation in future empirical investigations. We present the theory development underlying the revised conceptual model in four sections. First, we deﬁne stress to distinguish this construct from its antecedents and consequences. Next, we examine the direct

116

The Antecedents and Consequences of Stress in Accounting Settings1.

Notes: 1 Paths marked 7 represent exploratory relations due to conﬂicting (path X7) or inconclusive prior research results (paths EE, FF, GG) or multiple relations between constructs (path VV). 2 Exploratory path yet to be tested using latent variable structural equations (LVSE) modeling analysis, yet intuitively appealing and reported in prior non LVSE research.

KENNETH J. SMITH ET AL.

Fig. 1.

Stress in Accounting Settings

117

inﬂuence of environmental stressors (i.e., stress antecedents) on key organizational and personal outcomes. The Mediating Mechanisms section explores the role of noted mediating inﬂuences in the stressor-to-outcome dynamic as well as noted relations among the key outcomes themselves. The Discussion section puts forth suggestions for future research aimed at further reﬁning the theoretical model, and the Conclusion section presents concluding comments.

STRESS DEFINED Hans Selye (1956) deﬁned stress as the non-speciﬁc response of the body to any demand placed upon it, that is, the wear and tear on an individual. Similarly, Girdano and Everly (1986) refer to stress arousal as a fairly predictable arousal of mind–body systems that if prolonged may fatigue or damage to the point of malfunction or disease. A review of the organizational behavior and accounting literature reveals alternative theoretical orientations regarding the exact nature of the stress process. One school of thought views stress as the external stimuli that tend to cause behavioral or physiological change in a person, ‘‘thus asserting the ascendancy of environmental conditions in the stress process’’ (Smith, Everly, & Johns, 1993, p. 434). However, subsequent transactional stress models emphasize the cognitive aspects of the stress process wherein stress is viewed as part of a series of dynamic and complex interactions between an individual and the environment, in which events must be appraised as stressful before they can inﬂuence psychological well-being (see Daniels & Guppy, 1997 for a review). That is, an individual’s emotional, that is, stress, arousal is the result of interpreting and assigning meaning to environmental stressors (Everly & Sobelman, 1987). In turn, emotional arousal is a precursor to actual physical stress manifestations. Excessive stress activation in duration and/or intensity will lead to stress-related disease and dysfunction, which can manifest both physically and psychologically. Conversely, utilization of an effective coping strategy or a decrease in stress activation will restore an individual to physical and psychological balance, that is, a return to homeostasis. From an organizational perspective, Smith, Davy, and Everly (2007) note the importance of distinguishing stress from its antecedents and consequences. That is, environmental factors that cause excessive stress for one person might have a negligible inﬂuence on another. Again, the perception of a stressor as threatening is what initiates the stress process (LePine et al., 2005). In turn, stress can also be viewed as a mediating mechanism between

118

KENNETH J. SMITH ET AL.

speciﬁc environmental stressors and (1) key job-related outcomes such as job satisfaction, performance, and turnover intentions and/or (2) the aforementioned personal consequences. Table 1 provides a summary of the studies utilized to derive the revised theoretical model, and Table 2 provides a glossary of key terms used in these studies to help clarify the ensuing discussion. The criteria that we used to determine the relevant studies to include in Table 1 deserve comment. First, accountants are the target sample in all the studies to examine the stress dynamic in accounting systems. This is not to suggest that the postulated dynamic is unique to accounting systems. Rather, as suggested by Libby (1983, pp. 373–374), the stress construct provides ‘‘a useful structure for analyzing a wide variety of accounting issues.’’ Although the subjects in a majority of the studies are auditors working in public accounting, there is signiﬁcant ﬁrm size and geographic diversity as indicated in Table 1. Moreover, accountants working in private sector companies of various sizes in differing geographic locations are also represented. Accounting researchers over the past few decades have investigated many relations depicted in Fig. 1. A majority have utilized correlation, multiple regression, and analysis of variance techniques. Models derived using these techniques have been criticized for their susceptibility to the biasing effects of method variance and random measurement error (Williams & Hazer, 1986, p. 221), which can ‘‘attenuate estimates of coefﬁcients, make the estimate of zero coefﬁcients nonzero, or yield coefﬁcients with the wrong sign.’’ However, an increasing number of studies have employed latent variable structural equations (LVSE) analysis to examine various aspects of the stress dynamic among accountants. LVSE has the advantage of simultaneously testing structural parameters, that is, relations between constructs, while accounting for random and systematic errors (Anderson & Gerbing, 1988; Bollen, 1989; Williams & Podsakoff, 1989; Bentler, 1992; Byrne, 1994). In recognition of the purported superiority of this method for model development, and the sufﬁcient relevant extant research using this technique, all but three of the studies incorporated into the development of the Fig. 1 model utilize LVSE modeling. The other three studies utilize path analysis that is a subset of LVSE that deals with only measured (i.e., single indicator) variables as opposed to latent (multiple indicator) variables. Although not quite as sophisticated or robust as LVSE, path analysis is nonetheless a more direct approach to specify a model and relationships between variables than are the above-referenced traditional methods of analysis, thus motivating the inclusion of these three studies.

Study/Analysis

Summary of Studies Utilized for Model Development.

Subjects/Setting/ Distribution

Exogenous Predictor Variables

Mediating Variables

Rebele and Michaels (1990)/path analysis

155 auditors from four large international accounting ﬁrms/ onsite distribution

(a) Boundary spanning activity (BSA)

(b) Perceived environmental uncertainty (PEU) (c) Role conﬂict (d) Role ambiguity (Note: organizational level and need for achievement were tested as moderating variables)

Haskins, Baglioni, and Cooper (1990)/latent variable structural equations (LVSE) modeling

168 audit seniors/on Big Eight accounting ﬁrm/ onsite distribution

(a) Gender (b) Role stressors (c) Workload stressors (d) Interpersonal stressors (e) Type A behavior (also examined alternatively as a mediator)

Smith et al. (1993)/ path analysis

1,618 AICPA members/public accounting, industry, education, government/ randomized national mail sample

(a) Role ambiguity/ lack of control (b) Work overload

Outcome Variables

Findings

(e) Job satisfaction (f) Job-related tension (g) Performance

(a) Lysonski (1985); (b) Duncan (1972); (c & d) Rizzo, House, and Lirtzman (1970); (e)Hoppock (1935); (f) Lysonski (1985); (g) self-generated

(1) BSA to PEU (); (2) PEU to role conﬂict (þ), role ambiguity (þ), job satisfaction (), performance (), and job-related tension (þ); (3) role conﬂict to job satisfaction (), job-related tension (þ); (4) role ambiguity to job satisfaction () and performance () (Note: The hypothesized moderator effects of organizational level and need for achievement were not observed)

(e) Type A behavior (f) Problem-focused coping (g) Denial coping

(h) Psychological health symptoms (PHS)

(b–d) self-generated; (e) Bortner (1969); (f & g) Folkman, Lazarus, DunkelSchetter, DeLongis, and Gruen (1986); (h) Crown and Crisp (1979)

Mediated effects model (most parsimonious): (1) Gender (males) to PHS (þ); (2) Workload to PHE (þ);(3) Interpersonal stressors to denial coping (þ) and PHS (þ); (4) Type A to denial coping (þ) and PHS (þ); (5) problem-focused coping to PHS (); (6) Denial coping to PHS (þ)

(c) Stress arousal (d) Relaxation (e) Adaptive coping (f) Maladaptive coping

(g) Physiological health symptoms

(a & b) Kahn, Wolfe, Quinn, and Snoek (1964); (c & d) Everly, Sherman, and Smith (1989); (e & f) Everly and Newman (1982); (g) Everly et al. (1989)

(1) Role ambiguity/lack of control to stress arousal (þ) and relaxation (þ); (2) work overload to stress arousal (þ), relaxation () and maladaptive coping (þ); (3) stress arousal to adaptive coping (), maladaptive coping (þ), and physiological health symptoms (þ); (4) relaxation to adaptive coping (), maladaptive coping (þ), and physiological health symptoms (þ); (5) maladaptive coping to physiological health symptoms (þ)

119

Variable Measure Source

Stress in Accounting Settings

Table 1.

Table 1. (Continued ) Subjects/Setting/ Distribution

Exogenous Predictor Variables

Mediating Variables

Outcome Variables

Variable Measure Source

Findings

236 AICPA members/public accounting/ randomized national mail sample

(a) Type A personality (b) Work overload

(c) Role conﬂict (d) Role ambiguity (e) Stress arousal (f) Job satisfaction

(g) Turnover intentions

(a) Sales (1969); (b) Kahn et al. (1964); (c & d) Rizzo et al. (1970); (e) Everly et al. (1989); (f) Smith, Kendall, and Hulin (1969); (g) two author-generated questions

(1) Type A to job satisfaction (þ); (2) overload to role conﬂict (þ), role ambiguity (þ), and stress arousal (þ); (3) role ambiguity to stress arousal (þ) and job satisfaction (); (4) role conﬂict to job satisfaction (); (5) stress arousal to job satisfaction (); (5) job satisfaction to turnover intentions ()

Fogarty (1996)/ LVSE

462 auditors/Big Six accounting ﬁrms/ mail sample and onsite distribution

(a) Role conﬂict (b) Role ambiguity (c) Role overload

(d) Coping ability (e) Job tension

(f) Job satisfaction (g) Turnover intentions (h) Performance (i) Organizational commitment

(a & b) Rizzo et al. (1970); (c) self-designed sevenitem scale; (d) Hall (1972); (e) Lyons (1971); (f) Brayﬁeld and Rothe (1951); (g) authorgenerated questions; (h) self-designed scale; (i) Meyer and Allen (1984), Mowday, Steers, and Porter (1979)

(1) Role conﬂict to job tension (þ) and performance (); (2) role ambiguity to job tension (þ); (3) coping ability to job tension (); (4) job tension to job satisfaction (), turnover intentions (þ), performance (), and organizational commitment ()

Smith, Davy, and Stewart (1998)/ multi-sample LVSE, men and women (M, W)

199 AICPA and 114 AWSCPA members/public accounting/ randomized national mail sample

(a) Type A personality (b) Work overload

(c) Role conﬂict (d) Role ambiguity (e) Stress arousal (f) Job satisfaction

(g) Turnover intentions

(a) Sales (1969); (b) Kahn et al. (1964); (c & d) Rizzo et al. (1970); (e) Everly et al. (1989); (f) Smith et al. (1969); (g) two author-generated questions

(1) Type A to role ambiguity (þ) and job satisfaction (þ, M); (2) overload to role conﬂict (þ), role ambiguity (þ), and stress arousal (þ); (3) role ambiguity to stress arousal (þ), job satisfaction (), and turnover intentions (); (4) stress arousal to job satisfaction (); (5) job satisfaction to turnover intentions ()

Fogarty et al. (2000)/ LVSE

188 AICPA members/public accounting/ randomized mail sample – ﬁve states

(a) Role conﬂict (b) Role ambiguity (c) Role overload

(d) Burnout

(e) Job satisfaction (f) Turnover intentions (g) Job performance

(a & b) Rizzo, et al. (1970); (c) Beehr, Walsh, and Taber (1976); (d) Leiter and Maslach (1988); (e) Churchill, Ford, and Walker (1976); (f) Donnelly and Ivancevich (1975); (g) Dubinsky and Mattson (1979)

(1) Role overload to burnout (þ), job satisfaction (þ), and performance (þ); (2) role conﬂict to burnout (þ); (3) role ambiguity to burnout (þ) and job satisfaction (); (4) burnout to job satisfaction (), turnover intentions (þ), and performance ()

KENNETH J. SMITH ET AL.

Smith, Davy, and Everly (1995)/ LVSE

120

Study/Analysis

(a) Mentoring relationships

(b) Role ambiguity (c) Role conﬂict (d) PEU (Note: Employee organizational level and employee gender were tested as moderating variables)

(e) Job performance (f) Turnover intentions

(a) Chao, Waltz, and Gardner (1992), Dreher and Ash (1990); (b & c) Rizzo et al. (1970); (d) Rebele and Michaels (1990), Otley and Pierce (1995); (e) Kalbers and Fogarty (1995), Gregson et al. (1994); (f) selfgenerated from prior research

Full model: (1) informal mentoring to role ambiguity (), PEU (), and job performance (þ); (2) role ambiguity to job performance () and turnover intentions (þ); (3) role conﬂict to turnover intentions (þ); PEU to performance (); performance to turnover intentions () (Note: Mentor type and prote´ge´ organizational level modiﬁed several of the abovereferenced relations)

Sweeney and Summers (2002)/ LVSE

141 at single national accounting ﬁrm/ public accounting/ mailings. Respondents surveyed at two points in time

(a) Workload

(b) Role conﬂict (c) Role overload

(d) Burnout

(a) Self-reported; (b) Rizzo et al. (1970); (c) Beehr et al. (1976); (d) Maslach and Jackson (1981, 1986)

Pre-busy season – (1) workload to role conﬂict (þ) and role overload (þ); (2) role conﬂict and role overload to burnout (þ) Busy season – (1) workload to role conﬂict, role overload (þ), and burnout (þ); (2) role conﬂict and role overload to burnout (þ)

Kalbers and Fogarty (2005)/LVSE

291 Midwest IIA members/Internal auditors – diverse organizations/ onsite distribution

(a) Organizational trust

(b) Job skills (c) Locus of control

(d) Burnout

(a) Ashford, Lee, and Bobko (1989); author generated; (b) author generated; (c) Spector (1988); (d) Leiter and Maslach (1988)

(1) Locus of control to emotional exhaustion (EE) (), depersonalization (), and reduced personal accomplishment (); (2) job skills to reduced personal accomplishment (); (3) organizational trust to EE () and depersonalization (); (4) organizational trust to job skills (þ) and locus of control (þ); (5) job skills to locus of control (þ); (6) EE to depersonalization (þ); (7) depersonalization to reduced personal accomplishment (þ)

Fogarty and Kalbers (2006)/LVSE

298 Midwest IIA members/internal auditors/diverse organizations/ onsite distribution

(a) Burnout [EE; reduced personal accomplishment (RDA); depersonalization (DEP)]

(b) Job satisfaction (c) Organizational commitment: affective (OCA) (d) Organizational commitment: continuance (OCC) (e) Job performance

(f) External turnover intentions (g) Internal turnover intentions

(a) Singh, Goolsby, and Rhoads (1994); (b) Brayﬁeld and Rothe (1951); (c & d) Meyer and Allen (1991, 1997); (e) Fogarty et al. (2000); (f & g) Kalbers and Fogarty (1995)

(1) EE to job satisfaction (), external turnover intentions (þ), and OCA (þ); (2) RPA to job satisfaction (), external turnover intentions (), OCA (), and OCC (þ); (3) DEP to OCA (); (4) job satisfaction to external turnover intentions (); (5) OCA to external turnover intentions () and internal turnover intentions (þ); (6) OCC to internal turnover intentions ()

121

794 AICPA members of large accounting ﬁrms/ public accounting/ randomized national mail sample

Stress in Accounting Settings

Viator (2001)/LVSE

122

Table 1. (Continued ) Study/Analysis

Subjects/Setting/ Distribution

Exogenous Predictor Variables

Mediating Variables

Outcome Variables

Variable Measure Source

Findings

563 AICPA members/public accounting/ randomized national mail sample

(a) Role conﬂict (b) Role ambiguity (c) Role overload

(d) Stress arousal (e) Burnout

(f) Job satisfaction (g) Turnover intentions (h) Performance

(a & b) Rizzo et al. (1970); (c) Beehr et al. (1976); (d) Everly et al. (1989); (e) Singh et al. (1994); (f) Churchill, Ford, Hartley, and Walker (1985); (g) Donnelly and Ivancevich (1975); h) Dubinski & Mattson (1979)

(1) Role overload to stress arousal (þ), job satisfaction (þ), turnover intentions (þ), and performance (þ); (2) role conﬂict to stress arousal (þ), burnout (þ), and job satisfaction (); (3) role ambiguity to burnout (þ) and job satisfaction (); (4) stress arousal to burnout (þ), job satisfaction (), and performance (); (5) burnout to job satisfaction () and turnover intentions (þ); (6) job satisfaction to turnover intentions ()

Law, Sweeney, and Summers (2008)/ path analysis

112 public accountants/ Regional CPA Firm (72); National CPA ﬁrm 40)/nonrandomized mail sample

(a) Hardiness (b) Type-A behavior (c) Role overload (d) Role conﬂict (e) Role ambiguity (f) Neuroticism (g) Workaholism

(h) Workload

(i) Exhaustion

(a) Bartone, Ursano, Wright, and Ingraham (1989); (b) Bortner (1969); (c) Beehr et al. (1976); (d & e) Rizzo et al. (1970); (f) Saucier (1994); (g) Robinson (1999); (h) Sweeney and Summers (2002); (i) Maslach, Jackson, and Leiter (1996)

(1) Role overload to exhaustion (þ);(2) role conﬂict to exhaustion (þ);(3) role ambiguity to exhaustion (þ);(4) hardiness to exhaustion ();(5) workload to exhaustion ().

KENNETH J. SMITH ET AL.

Smith et al. (2007)/ LVSE

Construct

Glossary of Selected Terms. Deﬁnition

Boundary spanning

Intra- and extra-organizational interactions often undertaken to acquire information for decision making

Burnout

A negative psychological response to stressors with three separate dimensions: emotional exhaustion, i.e., lack of energy and a depletion of emotional resources; reduced personal accomplishment, i.e., reduced motivation and selfesteem; depersonalization, i.e., detachment and callousness towards others ‘‘Efforts, both action-oriented and intrapsychic, to manage (that is master, tolerate, reduce, minimize) environmental and internal demands, and conﬂicts among them, which tax or exceed a person’s resources. Coping can occur prior to a stressful confrontation, in which case it is called anticipatory coping, as well as in reaction to a present or past confrontation with harm’’ ‘‘A general sense that the environment is satisfying, which leads a person to approach life experiences with curiosity and enthusiasm or commitment’’ The extent to which individuals are bothered by stressful conditions at work

Coping

Hardiness Job tension Locus of control

Mentoring Neuroticism Organizational commitment

Organizational trust

Aldrich and Herker (1977); Leifer and Delbecq (1978) Cordes and Dougherty (1993, p. 623)

Cohen and Lazarus (1979, p. 219)

Maddi and Kobasa (1984, p. 50) Bohan (1990); Steffy, Jones, Noe, and Wiggans (1990) Rotter (1966)

Viator (2001, p. 73) Mayes, Johnson, and Sadri (2000) Smith, Organ, and Near (1984) Fogarty and Kalbers (2006) Kalbers and Fogarty (2005, p. 106) Beehr et al. (1976, p.42)

123

Overload

The extent to which a person believes that they can inﬂuence their environment ranging from little or no control (external locus of control) to a sense of strong control (internal locus of control) Providing prote´ge´s with role clarifying information and alternatives for dealing with role demands, including differing role expectations that may conﬂict One of the ﬁve major dimensions of personality (De Vries & van Heck, 2000), the tendency of an individual to have high emotional reactivity 1. Affective commitment willingness of an individual to work diligently on behalf of their employer 2. Continuance commitment saying with one’s employer because of the transaction costs of leaving An individual’s conﬁdence and trust in their employer in general and top management in particular Having too much work to ﬁnish in the time available

Source

Stress in Accounting Settings

Table 2.

124

Table 2. (Continued ) Construct Perceived environmental uncertainty Role ambiguity Role conﬂict Stress arousal Type A personality (behavior) Workaholism

Deﬁnition

Source

A perceptual state where one engages in directed behaviors based on incomplete knowledge of their relationship with the environment

Ferris (1977, pp. 23 24)

The absence of adequate information required for satisfactory accomplishment of one’s role The simultaneous occurrence of two or more sets of pressures such that the compliance with one would make difﬁcult or impossible with the other A fairly predictable arousal of psycho-physiological systems, which if prolonged can fatigue or damage the system to the point of malfunction or disease Aggressive, ambitious, impatient, sense of time urgency, fatigue suppression, excessive need for control An addiction where one feels compelled to work because of inner pressures that make that person feel guilty or distressed by not working as opposed to external demands or pleasure derived from work

Kahn, Wolfe, Quinn, and Snoek (1964, pp. 21 23) Wolfe and Snoek (1962, p. 103) Girdano and Everly (1986, p. 5) Maslach, Schaufeli, and Leiter (2001) Spence and Robbins (1992, p. 161)

KENNETH J. SMITH ET AL.

Stress in Accounting Settings

125

ENVIRONMENTAL EVENTS (STRESSORS) The accounting literature has examined the inﬂuence of both organizational and extra-organizational stressors on key organizational and personal outcomes. Organizational stressors represent the multiple inﬂuences within the work environment itself that can result in deleterious organizational and personal consequences (Smith, 1990). Oft-cited examples are role conﬂict, role ambiguity, time pressure, ﬁrm structure, leadership style, and so on. The origins of extra-organizational stressors such as type A personality extend beyond the immediate work environment, but like organizational stressors they have been linked to numerous organizational and personal consequences.

Organizational Stressors There is a signiﬁcant body of research linking organizational stressors to key outcomes in the accounting work environment. Although most of these studies have focused on the direct relations between stressors and outcomes, an increasing number have also investigated the relations between organizational antecedents, stress measures, and outcomes. A review of the underlying research reveals a number of inﬂuential organizational stressors. Boundary Spanning and Perceived Environmental Uncertainty Rebele and Michaels (1990) show that boundary spanning activity has a signiﬁcant negative inﬂuence on perceived environmental uncertainty, as predicted by and supportive of the proposition that the more information gathered, the less uncertainty there is regarding decisions to be taken (path A). Perceived environmental uncertainty also has positive relations to role conﬂict, role ambiguity, and job-related tension (paths B–D) and negative relations to job satisfaction and performance (paths E and F). Overload Work overload is also a job condition that can precede and inﬂuence role stress (Schuler, 1980; Locke, 1984; Jackson & Schuler, 1985; Schaubroeck, Cotton, & Jennings, 1989). In fact, it is an often-cited stressor in the accounting work environment (Smith et al., 1995, 1998). Schaubroeck et al. (1989, pp. 41–42) citing Kahn et al. (1964) note that overload makes it more difﬁcult for individuals in the ﬁrm to ‘‘manage the role agreement process’’ and that overload may inﬂuence tension both directly and indirectly,

126

KENNETH J. SMITH ET AL.

because of its relations with role conﬂict and role ambiguity (see also Parasuraman & Alutto, 1984). Despite the disparate measures utilized to measure overload, LVSE research in various accounting milieu shows fairly consistent results for this construct. As an antecedent condition to role stress, overload has a signiﬁcant positive relation with both role conﬂict and role ambiguity (paths G and H) (Smith et al., 1995, 1998; Sweeney & Summers, 2002). Overload and both job satisfaction and performance (paths I and J) are positively related (Fogarty et al., 2000; Smith et al., 2007). In explanation of these counterintuitive ﬁndings, Fogarty et al. (2000) propose that overload includes an ‘‘eustress’’1 component that is unmediated, and Smith et al. (2007) speculate that these relations may have resulted from individuals evaluating overload as a challenge rather than a threat, thus giving it the potential to promote personal gain and growth (LePine et al., 2005). Finally, as predicted, overload has a signiﬁcant positive relation to turnover intentions (path K) (Smith et al., 2007). From a health perspective, overload has signiﬁcant positive direct relations with stress arousal (Smith et al., 1993, 1995, 1998, 2007), maladaptive coping (Smith et al., 1993), and psychological health symptoms (Haskins et al., 1990), motivating paths L–N. Moreover, overload has a signiﬁcant positive relation to burnout (Fogarty et al., 2000), particularly during the January through April ‘‘busy season’’ (Sweeney & Summers, 2002). However, although Smith et al.’s (2007) initial replication of Fogarty et al.’s (2000) model also measures a signiﬁcant positive direct relation between overload and burnout, this relation is not signiﬁcant when stress arousal is added to the model as antecedent to burnout, leading to the proposition that a direct path between the former two constructs may not be warranted. Role Conﬂict and Role Ambiguity The potential inﬂuence of these stressors has been of considerable interest to accounting researchers over the years. As noted earlier, some theorize that they are consequences of other environmental inﬂuences. However, another stream of research positions them as independent (i.e., exogenous) predictors of stress and its consequences. Regardless, their inﬂuence as stress antecedents is well documented. Role conﬂict has a signiﬁcant positive relation to stress arousal (Smith et al., 2007) and job tension (Rebele & Michaels, 1990; Fogarty, 1996), burnout (Fogarty et al., 2000; Sweeney & Summers, 2002; Smith et al., 2007; Law et al., 2008), and turnover intentions (Viator, 2001), prompting paths

Stress in Accounting Settings

127

O–Q. Conversely, role conﬂict has negative relations to job satisfaction (Rebele & Michaels, 1990; Smith et al., 1995, 2007) and performance (Fogarty, 1996), motivating paths R and S. Role ambiguity has a signiﬁcant positive relation to stress arousal in studies where ambiguity is measured using items loading from a factor analysis of Kahn et al. (1964) and Rizzo et al. (1970) scales (Smith et al., 1993, 1995, 1998), as well as job tension (Fogarty, 1996), motivating path T.2 Role ambiguity is a signiﬁcant positive factor in burnout as shown in path U (Fogarty et al., 2000; Smith et al., 2007). These studies also support Rebele and Michaels (1990) and Smith et al. (1995), which ﬁnd that role ambiguity is a signiﬁcant negative factor in job satisfaction (path V). Viator (2001) ﬁnds a signiﬁcant negative relation between role ambiguity and performance (path W). Finally, mixed results have been reported with respect to the relation between role ambiguity and turnover intentions. Viator (2001) ﬁnds a positive relation between these constructs, which was intuitively expected. However, Smith et al. (1998) measure a negative relation and provide two possible explanations for this counterintuitive ﬁnding: (1) role ambiguity enhances an individual’s insecurity, which in turn attenuates one’s inclination to consider a job change; and (2) individuals experiencing high role ambiguity do not perceive that alternative job opportunities offer lower levels of ambiguity (Mackay & Cooper, 1987). On the basis of these mixed results, additional research appears warranted to determine whether a consensus can be reached on the sign for path X.

Extra-organizational Stressors/Personality Characteristics Locus of Control and Organizational Trust Evidence shows that auditors with an internal locus of control have a more positive perception of stress than those with an external locus of control (Bernardi, 1997). Using LVSE analysis, Kalbers and Fogarty (2005, p. 111) ﬁnd internal locus of control to ‘‘have a signiﬁcant negative inﬂuence on all three dimensions of burnout,’’ thus motivating path Y. With respect to organizational trust, the authors ﬁnd a signiﬁcant negative relation with both the emotional exhaustion and depersonalization component of burnout, suggesting that ‘‘organizations that do not build trust in the workplace build the groundwork for burnout’’ (p. 113). These ﬁndings motivate path Z. Internal locus of control is also associated with greater job satisfaction and reduced turnover intentions (Reed, Kratchman, & Strawser, 1994),

128

KENNETH J. SMITH ET AL.

as well as enhanced performance (Hyatt & Prawitt, 2001). Although these relations have yet to be tested using LVSE modeling, they are consistent with the proposition that when individuals perceive that they are in control of a situation, the less likely the situation will be perceived as threatening or stress-inducing (Chan, 1977), and the belief that locus of control orientation and ability to adapt to stress are related (Ivancevich & Matteson, 1980). Moreover, in a theoretical treatise based on a review of the locus of control literature, perceived internal locus of control is proposed to have a negative association with job-related stress and turnover (paths AA and BB), and a positive association with job satisfaction (path CC) (Bernardi, 1999). Type A Personality Characterized as highly competitive, hard-driving, achievement-oriented, and time-urgent (Friedman & Rosenman, 1974), type A personality has been employed in occupational stress research as a ‘‘measure of personality predisposition to stress’’ (Haskins et al., 1990, pp. 367–368). Type A personality and organizational stress have a direct relation seemingly due to ‘‘the overly competitive and fast lifestyle of type A’s (which) tends to place them in a constant state of anxiety when dealing with their daily work environment’’ (Choo, 1986). Numerous studies report signiﬁcant positive relations between type A behavior and self-report measures of stress and tension (see Gamble & Matteson, 1992 for a review). More recently, Fisher (2001) ﬁnds a signiﬁcant positive correlation between type A behavior and both job satisfaction and performance. Among the four LVSE studies that examined the potential inﬂuence of type A personality, Smith et al. (1998) ﬁnd a signiﬁcant positive relation with job satisfaction for males, partially supporting Smith et al.’s (1995) ﬁnding of a signiﬁcant positive relation between type A and job satisfaction, thus prompting path DD. However, contrary to expectation, Smith et al. (1998) ﬁnd type A to have a signiﬁcant negative relation with role ambiguity. They conjecture that their measure, which captured the goal-oriented, achievement, and task-oriented aspects of this construct, did not measure the hostility/aggression component of type A personality associated with increased stress and deleterious health consequences (Spence, Helmreich, & Pred, 1987). Further explanation may lie in the signiﬁcant positive covariance measured in both studies between type A personality and overload, the latter of which has signiﬁcant positive relations with role stress and stress arousal posited and discussed below. Overload may possibly account for sufﬁcient variance to negate the effects

129

Stress in Accounting Settings

of type A. These issues argue for additional research that examines the inﬂuence of the hostility/aggression component of type A personality on job stressors and stress arousal, thus prompting the inclusion of posited paths EE–GG. Hardiness As the deﬁnition in Table 2 implies, hardiness is a personal characteristic that may attenuate stress rather than exacerbate it. Hardiness integrates three closely related factors, that is, control, commitment, and challenge (Law et al., 2008; Sheard & Golby, 2007). A hardy person is able to perceive potentially stressful circumstances as meaningful and interesting (commitment), stressors as manageable (control), and change as an opportunity for personal growth instead of a threat (challenge). The commitment and challenge components of hardiness have a signiﬁcant negative relation with exhaustion, thus motivating path HH (Law et al., 2008).

MEDIATING MECHANISMS Stress and Tension Accounting researchers have found stress to play a pivotal role as both a direct and mediating inﬂuence on various personal and organizational consequences. Two ‘‘stress’’ measures emerge from a review of the relevant extant literature: stress arousal and job tension. Although job tension more directly links the measured level of dissonance to conditions at work, both measures appear to tap an individual’s level of discord that result from exposure to environmental events. Thus, for model development purposes, we posit that the constructs are synonymous. From a health and well-being perspective, stress arousal has a direct positive relation with maladaptive coping and physiological health symptoms (paths II and JJ) and a negative relation to adaptive coping (KK), while mediating the inﬂuence of role ambiguity/lack of control and overload on those outcomes (Smith et al., 1993). With respect to organizational outcomes, stress arousal has a positive relation to burnout (Smith et al., 2007) and negative relation to job satisfaction (Smith et al., 1995, 1998, 2007), and performance (Smith et al., 2007), while mediating the inﬂuence of (study speciﬁc) stressors on those outcomes. Similarly, job tension also has a negative relation to job satisfaction (and organizational commitment) and performance, while mediating the inﬂuence of role

130

KENNETH J. SMITH ET AL.

conﬂict, role ambiguity, and coping on those outcomes (Fogarty, 1996). These ﬁndings motivate paths LL–NN. Finally, although Fogarty (1996) ﬁnd a direct positive relation between job tension and turnover intentions, this ﬁnding was not replicated by Smith et al. (2007) when burnout was added as a direct stress consequence and potential mediating inﬂuence between stress and turnover.

Coping As denoted in Table 2, coping can occur before or after the onset of a psychological or physical stress response. Coping behaviors can be dichotomized as either adaptive or maladaptive in nature (Smith et al., 1993). Adaptive behavior is that which manages demands and reduces stress while enhancing health, for example, exercise or relaxation techniques. Maladaptive behaviors, while employed toward the same ends, are selfdebilitating as they create other demands and prolonged stress. Typical examples are substance abuse, compensatory eating, and smoking. Coping can also be dichotomized as either problem focused or denial based (Haskins et al., 1990). Problem-focused coping represents efforts to change the status quo (e.g., the circumstances, others, oneself) in a positive and meaningful way. Denial coping, on the other hand, represents psychological and/or physical efforts to distance one from their current circumstances and can include some of the deleterious behaviors noted earlier. As a stress antecedent, adaptive coping ability has a negative relation to job tension (Fogarty, 1996), and problem-focused coping and denial coping have signiﬁcant negative and positive relations, respectively, to psychological health symptoms (paths OO–QQ) (Haskins et al., 1990). As a stress consequence, maladaptive coping has a signiﬁcant positive relation to stress related illness (path RR) (Smith et al., 1993).3

Burnout Burnout has been studied by accounting researchers as an exogenous predictor of various outcomes (Fogarty & Kalbers, 2006), a mediator in the relations between environmental factors and those outcomes (Fogarty et al., 2000; Smith et al., 2007), and as an outcome (Kalbers & Fogarty, 2005). However, in the stress dynamic proposed herein, burnout appears best positioned as a mediating construct. Positioned as such, burnout mediates

Stress in Accounting Settings

131

the inﬂuence of role stressors (Fogarty et al., 2000; Smith et al., 2007) and stress arousal (Smith et al., 2007) on job satisfaction, performance, and turnover intentions. Moreover, burnout has a direct negative inﬂuence on job satisfaction and positive inﬂuence on turnover intentions even after considering the inﬂuence of stress arousal (paths SS–UU) (Smith et al., 2007). Finally, although not positioned as mediators, the emotional exhaustion and reduced personal accomplishment components of burnout have a signiﬁcant positive inﬂuence on the continuance component of organizational commitment, and the reduced personal accomplishment and depersonalization components of burnout have a signiﬁcant negative inﬂuence on the affective component of organizational commitment (Fogarty & Kalbers, 2006). These ﬁndings prompt exploratory path VV.

Organizational Commitment The relationship between organizational commitment and turnover in public accounting has been extensively studied (for a review, see Law, 2005). Ironically, although affective commitment is the most salient predictor of intent to turnover in non-accounting studies, results with subjects in public accounting have been mixed (Law, 2005). To illustrate, although Kalbers and Fogarty (1995) ﬁnd a signiﬁcant inverse relationship between continuance commitment and intent to turnover, other studies show a signiﬁcant negative relationship between affective commitment and intent to turnover (Ketchand & Strawser, 1998; Stallworth, 2003; Law, 2005). Within an LVSE context, affective commitment has a signiﬁcant negative relation to external turnover intentions, and continuance commitment has a signiﬁcant negative relationship to internal turnover intentions, prompting path WW (Fogarty & Kalbers, 2006).

Job Satisfaction, Performance, and Turnover Intentions These constructs have been primarily examined within LVSE studies as potential stress outcomes. However, a few studies have examined and uncovered relations among these constructs. Job satisfaction (Smith et al., 1998, 2007) and performance (Viator, 2001) are negatively related to turnover intentions, motivating paths XX and YY.

132

KENNETH J. SMITH ET AL.

DISCUSSION Smith (1990) challenged future researchers to take an integrated approach to examine stress and its antecedents and consequences as the basis for constructing a global stress paradigm, which in turn might serve as the focal point for interventions aimed at enhancing individual well-being and organizational efﬁciency. Research has made signiﬁcant progress over the past two decades toward realizing this goal as indicated by the growing number of relations that have been empirically validated in the research efforts denoted earlier. However, although Fig. 1 illustrates the underpinnings of the model that emerges from this analysis, there are many avenues for additional research. Consistent with transactional models of occupational stress such as that depicted in Fig. 1 is the notion of a circular process wherein psychological well-being and other key outcomes have the potential to be inﬂuenced by, and have an inﬂuence upon, various stressors and mediating factors. In a non-LVSE context, Daniels and Guppy (1997) tested this proposition by examining the inﬂuence of affective psychological well-being on stressors, locus of control, and social support among a sample of British accountants. Although the authors ﬁnd equivocal support for a causative relationship between affective well-being and subsequent stressor levels, they also ﬁnd that higher depression levels are associated with subsequently lower external locus of control, and higher levels of contentment are associated with subsequently higher levels of help support, prompting them to conclude that there is support for the viability of reciprocal relationships in the stress process. Although no speciﬁc LVSE study to date has formally attempted to examine reciprocal relationships, their potential nonetheless exists as evidenced by paths KK and OO in Fig. 1 wherein adaptive coping has been (in separate studies) reportedly inﬂuenced by, and had an inﬂuence on, stress and tension. A review of the transactional stress model literature and a cursory view of Fig. 1 reveal numerous, potentially viable, yet untested, reciprocal relationships. As noted previously, better clariﬁcation of the role of type A personality in the stress dynamic is needed. Speciﬁcally, future research should aim to better measure the hostility/aggression component of type A personality as well as its inﬂuence on job stressors, stress and tension, and various stress consequences. Adams and John (1997) cite numerous sources that report that trait hostility has been implicated in unhealthy behaviors and a range of negative health outcomes.

Stress in Accounting Settings

133

The Cook–Medley Hostility Index (Cook & Medley, 1954) is the most popular measure of hostility aggression (Seaward, 2006); however, other hostility measurement scales have also been developed and validated (Adams & John, 1997; Arthur, Garﬁnkel, & Irvine, 1999). The existence of valid and reliable hostility measures should facilitate studies that attempt to examine the inﬂuence of hostility aggression on the aforementioned outcomes and ultimately reduce the number of ambiguous relationships depicted in Fig. 1 with respect to type A behavior and its consequences. There is limited extant research within accounting settings of the roles played by locus of control, hardiness, organizational trust, and other key constructs illustrated in Fig. 1, thus providing considerable opportunity for those seeking to establish and conﬁrm theoretically justiﬁable relations among these constructs. To illustrate, as noted earlier, Daniels and Guppy (1997) ﬁnd a relation between depression and subsequently lower external locus of control. This ﬁnding points to the possibility that locus of control, depicted in Fig. 1 as an exogenous predictor, may be inﬂuenced by other constructs as well. In addition, as noted earlier, non-LVSE research indicates that there are additional potential relations between locus of control and other constructs, for example, stress and tension. Although not depicted in Fig. 1, a similar case could very well be made for examining potential relations between both hardiness and organizational trust, and stress and its consequences. The majority of the constructs illustrated in Fig. 1 and tested in prior research are each constructed using multiple indicators. Multiple indicators for each latent construct allow for better estimates of the random error associated with the respective construct. In turn, random error is taken into account when estimating paths from each construct to its indicator variables as well as within the structural model itself. Ironically, this aspect of the latent variable construction process may potentially mask theoretically meaningful and statistically signiﬁcant relations between speciﬁc indicators of various constructs and other constructs, or one or more of the indicators of those constructs, particularly with respect to multi-dimensional constructs such as stress arousal, burnout, organizational commitment, and job satisfaction.4,5 Interestingly, the aforementioned study by Fogarty and Kalbers (2006) illustrates signiﬁcant relations between speciﬁc burnout and organizational commitment indicators, that is, subscales. Their ﬁndings thus highlight the potential insight to be gained from future investigations that examine additional posited relations among theoretically meaningful subscales of the multi-dimensional constructs that comprise the posited model.

134

KENNETH J. SMITH ET AL.

A number of the depicted relations in Fig. 1 have been established using disparate measures of speciﬁc constructs calling into question the generalizability of the results. For example, alternate studies have tapped different variables from the Rizzo et al. (1970) scale to measure role conﬂict and role ambiguity. Moreover, a review of Table 1 indicates that disparate measures have been used for job tension, job satisfaction, and so on. Therefore, it is unclear how the results of each of these studies may have been impacted had alternative constructions of speciﬁc constructs been incorporated to examine their inter-relations with the other constructs under investigation. This observation seems to support the argument of Stout and Rebele (1996) for studies that replicate and extend prior research. Although not intended to limit or restrict future efforts to reﬁne the postulated stress model, it would appear propitious for investigations that attempt to conﬁrm or refute depicted relations in the model to incorporate the same measures (where applicable) as those utilized in the original studies. Questions surrounding the construct distinctiveness of certain measures utilized in prior research present an opportunity to clarify a number of the relations depicted in Fig. 1. For example, as noted earlier, stress arousal and job-related tension are posited to be synonymous constructs. However, although Table 1 indicates that a single measure of stress arousal was utilized in all the studies incorporated in the present effort to develop the proposed theoretical model, three separate job-related tension measures were utilized. Furthermore, one of the measures, that is, the Job Tension Index (JTI) (Kahn et al., 1964), was found by Smith et al. (1993) to be multifactorial in nature. Their factor analysis of the JTI and Stress Arousal Scale (SAS) (Everly et al., 1989) items revealed the JTI to load on two distinct factors, role ambiguity/lack of control and overload, both of which were separate and distinct from the two measured SAS factors. In the Smith et al. (1993) study, the JTI factors were incorporated as exogenous predictors of stress arousal and its consequences, a positioning theoretically justiﬁed by prior research. Visual inspection of the items that comprise the job tension measures incorporated in the other two studies (Rebele & Michaels, 1990; Fogarty, 1996) motivated positioning job tension as a stress arousal correlate in this theoretical presentation. However, conﬁrmation of this positioning awaits future efforts that establish the construct distinctiveness of the measures taken before attempting to establish structural linkages. This two-step procedure, recommended by Anderson and Gerbing (1988), is the generally accepted process among LVSE researchers, yet it is not clear that this procedure has been followed in all the aforementioned studies.

135

Stress in Accounting Settings

For this reason, many of the depicted relations in Fig. 1 should be considered tentative and subject to future investigation. Although two of the studies incorporated into this model formally attempted to examine the moderating inﬂuence of speciﬁc factors on the speciﬁc dynamics under investigation (Rebele & Michaels, 1990; Viator, 2001), the remaining studies examined models with mediating effects only. However, a number of intuitively appealing potential moderators appear to exist and should be considered in future efforts to reﬁne the posited model. For example, organizational level could have a viable moderating inﬂuence on both role conﬂict and role ambiguity and their proposed consequences. Logically, one would expect that staff accountants would be more susceptible to role stress than would partners. Organizational size, ﬁrm structure, type of organization, years of service, and numerous other constructs should be considered as having a potential impact on one or more the relations postulated in Fig. 1. From a methodological standpoint, even the most sophisticated of the aforementioned studies are subject to self-containment concerns. That is, to the extent that unexplained variance in each model is due to non-random inﬂuences (i.e., correlated omitted variables), there is bias in the structural parameter estimates relating the measured ‘‘stress’’ constructs with their antecedents and consequences. Future efforts should consider the identiﬁcation of additional inﬂuences on the key ‘‘stress’’ outcomes to lessen this bias and further reﬁne the present model. Also, until a number of the depicted relations are empirically veriﬁed through longitudinal designs, causality inferences must be deferred.

CONCLUSION Jelinek and Jelinek (2008), in echoing Weick’s sentiments from a quarter century earlier, provide a cogent argument and compelling evidence that supports the extent to which stress can serve as a focal point for examining numerous issues related to accountants’ workplace performance and individual well-being. Speciﬁcally, based on their research into workplace deviance at Big 4 accounting ﬁrms, Jelinek and Jelinek (2008) present a model which illustrates how the Sarbanes–Oxley Act auditing compliance procedures, the market shortage of auditors, and the lingering cloud over the accounting profession from the well-documented accounting and ﬁnancial scandals over the past decade, have combined to increase job stress. In turn, job stress has resulted in deviant auditor behavior that has

136

KENNETH J. SMITH ET AL.

had a deleterious impact on organizational efﬁciency and effectiveness. As noted in their ﬁnal observations, ‘‘Unfortunately for the accounting profession, recent developments in the ﬁeld have ramped up a known driver of deviance-workplace stress (emphasis added) – and audit managers and partners must respond’’ (Jelinek & Jelinek, 2008, p. 232). Clearly, this statement underscores the practical beneﬁts that may accrue from systematic efforts to better understand the stress dynamic within the accounting work environment. The accounting stress literature over the past two decades has uncovered a number of interesting ﬁndings that have helped to shape the occupational stress in accountancy paradigm as it is presently conceptualized. Of equal importance, the paradigm provides insights to those charged with mitigating the deleterious consequences of excessive stress to both accounting professionals and employers alike. However, much work remains if we are to develop a more complete understanding of the stress dynamic to serve as the foundation for mitigation strategies designed to enhance performance and individual well-being of accountants in the workplace.

NOTES 1. Eustress or ‘‘good stress’’ is a positive form of stress that is healthful, gives one a feeling of fulﬁllment, and enhances one’s performance. 2. Smith et al. (2007) failed to measure a signiﬁcant direct relation between role ambiguity and stress arousal using solely Rizzo et al. (1970, p. 150) items to measure role ambiguity (and role conﬂict). Noting the signiﬁcant correlation measured between role ambiguity and role conﬂict (.43), they postulated that their ﬁndings for role ambiguity supported the proposition by Schaubroeck et al. (1989) that the signiﬁcant correlation between these stressors may have attenuated an otherwise signiﬁcant ﬁnding. 3. Failing to measure a signiﬁcant negative relation between adaptive coping and stress related illness, Smith et al. (1993) speculated that their aggregation of individual coping behaviors into a single index might have obscured relations between individual behaviors and reported illness. However, as noted earlier with respect to Smith et al.’s (2007) failure to ﬁnd a signiﬁcant relation between role ambiguity and stress arousal, in this case the signiﬁcant negative covariance between adaptive and maladaptive coping behavior may have resulted in the latter subsuming the explanatory power of the former. 4. ‘‘Speciﬁc indicators’’ represent unidimensional subscales associated with a speciﬁc factor as determined by exploratory factor analysis of a particular scale in prior research. For example, Smith et al. (1993, 1995, 1998, 2007) ﬁnd that stress arousal as measured on the Stress Arousal Scale (Everly et al., 1989) consists of two subscales, that is, psychological discord (13 items) and relaxation (4 items).

137

Stress in Accounting Settings

Similarly, Fogarty et al. (2000) and Smith et al. (2007) ﬁnd, in line with previous research, that the 24 item multi dimensional role speciﬁc version of the Maslach Burnout Inventory consists of three distinctive 8 item dimensions: depersonalization, reduced personal accomplishment, and emotional exhaustion. 5. For diagramming ease and interpretability, Fig. 1 does not illustrate the underlying indicators for each construct.

REFERENCES Adams, S. H., & John, O. P. (1997). A hostility scale for the California psychological inventory: MMPI, observer Q sort, and big ﬁve correlates. Journal of Personality Assessment, 69(2), 408 424. Aldrich, H., & Herker, D. (1977). Boundary spanning roles and organizational structure. Academy of Management Review, 2(2), 217 230. Anderson, J., & Gerbing, D. (1988). Structural equation modeling in practice: A review and recommended two step approach. Psychological Bulletin, 103(3), 411 423. Arthur, H. M., Garﬁnkel, P. E., & Irvine, J. (1999). Development and testing of a new hostility scale. Canadian Journal of Cardiology, 15(5), 539 544. Ashford, S., Lee, C., & Bobko, P. (1989). Content, causes and consequences of job insecurity: A theory based measure and substantive test. Academy of Management Journal, 32, 803 829. Bartone, P. T., Ursano, R. J., Wright, K. M., & Ingraham, L. H. (1989). The impact of a military air disaster on the health of assistance workers. The Journal of Nervous and Mental Disease, 177(6), 317 328. Beehr, T. A., Walsh, J. T., & Taber, T. D. (1976). Relationship of stress to individually and organizationally valued states: Higher order needs as a moderator. Journal of Applied Psychology, 61(1), 41 47. Bentler, P. M. (1992). EQS structural equations program manual. Los Angeles: BMDP Statistical Software. Bernardi, R. A. (1997). The relationships among LOC, perceptions of stress, and performance. Journal of Applied Business Research, 13(4), 1 8. Bernardi, R. A. (1999). Stress, locus of control, and longevity. Business Forum, 26(3 4), 27 33. Bohan, G. (1990). Solutions to all your problems: Build, don’t battle. Training and Development Journal, 44(2), 15 18. Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley. Bortner, R. W. (1969). A short rating scale as a potential measure of pattern a behavior. Journal of Chronic Diseases, 12, 87 91. Brayﬁeld, A., & Rothe, H. (1951). An index of job satisfaction. Journal of Applied Psychology, 35, 307 311. Byrne, B. M. (1994). Structural equation modeling with EQS and EQS/Windows. Thousand Oaks, CA: Sage Publications, Inc. Chan, K. B. (1977). Individual differences in reactions to stress and their personality and situational determinants: Some implications for community mental health. Journal of Science and Medicine, 11, 11 28.

138

KENNETH J. SMITH ET AL.

Chao, G. T., Waltz, P. M., & Gardner, P. D. (1992). Formal and informal mentorships: A comparisons on mentoring functions and contrast with nonmentored counterparts. Personnel Psychology, 45, 619 636. Choo, F. (1986). Job stress, job performance, and auditor personality characteristics. Auditing: A Journal of Theory and Practice (Spring), 17 34. Churchill, G., Ford, N., Hartley, S., & Walker, O. (1985). The determinants of salesperson performance: A meta analysis. Journal of Marketing Research, 22, 103 118. Churchill, G. A., Ford, N. M., & Walker, O. C. (1976). Organizational climate and job satisfaction in the sales force. Journal of Marketing Research, 13, 323 332. Cohen, J., & Lazarus, R. S. (1979). Coping with the stresses of illness. In: G. Stone, F. Cohen & N. Adler (Eds), Health psychology (pp. 217 254). San Francisco: Jossey Bass. Cook, W. W., & Medley, D. M. (1954). Proposed hostility and pharisaic virtue scales for the MMPI. Journal of Applied Psychology, 38, 414 418. Cordes, C. L., & Dougherty, T. W. (1993). A review and an integration of research on job burnout. Academy of Management Review, 18(4), 621 656. Crown, S., & Crisp, A. H. (1979). Manual of the crown crisp experiential index. London: Hodder & Stoughton. Daniels, K., & Guppy, A. (1997). Stressors, locus of control, and social support as consequences of affective psychological well being. Journal of Occupational Health Psychology, 2(2), 156 174. De Vries, J., & van Heck, G. L. (2000). Persoonlijkheid en emotionele uitputting. Een overzicht van de literatuur. Gedrag & Gezondheid: Tijdschrift voor Psychologie & Gezondheid, 28(3), 90 105. Donnelly, J., & Ivancevich, J. (1975). Role clarity and the salesman. Journal of Marketing, 39, 71 74. Dreher, G. F., & Ash, R. A. (1990). A comparative study of mentoring among men and women in managerial, professional and technical positions. Journal of Applied Psychology, 75, 539 546. Dubinsky, A., & Mattson, B. (1979). Consequences of role conﬂict and ambiguity experienced by retail salespeople. Journal of Marketing, 55, 70 86. Duncan, R. B. (1972). Characteristics of organizational environments and perceived environmental uncertainty. Administrative Science Quarterly, 17, 313 327. Everly, G. S., & Newman, E. C. (1982). The development of a tool to assess adaptive and maladaptive health related coping behaviors: preliminary ﬁndings. Paper presented to the Health Education Risk Reduction Conference, Newark Delaware, 1982. Everly, G. S., Sherman, M., & Smith, K. J. (1989). The development of a scale to assess behavioral health factors: The Everly stress and symptom inventory. In: R. H. Feldman & J. Humphrey (Eds), Health education: current selected research (Vol. II, pp. 71 86). New York: AMS Press. Everly, G. S., & Sobelman, S. A. (1987). The assessment of the human stress response. New York, NY: AMS Press. Ferris, K. R. (1977). Perceived uncertainty and job satisfaction in the accounting. Environment. Accounting, Organizations and Society, 2, 23 28. Fisher, R. T. (2001). Role stress, the type A behavior pattern, and external auditor job satisfaction and performance. Behavioral Research in Accounting, 13, 143 170. Fogarty, T. (1996). An examination of job tension and coping in the relationship between stressors and outcomes in public accounting. Journal of Managerial Issues, 8(3), 269 285.

Stress in Accounting Settings

139

Fogarty, T., & Kalbers, L. (2006). Internal auditor burnout: An examination of behavioral consequences. Advances in Accounting Behavioral Research, 9, 51 86. 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. Folkman, S., Lazarus, R. S., Dunkel Schetter, C., DeLongis, C., & Gruen, R. (1986). Dynamics of a stressful encounter: Cognitive appraisal, coping, and encounter outcomes. Journal of Personality and Social Psychology, 50(5), 992 1003. Friedman, M., & Rosenman, R. (1974). Type A behavior and your heart. New York: Knopf. Friedman, M., Rosenman, R., & Carroll, V. (1958). Changes in the serum cholesterol and blood clotting time in men subjected to cyclic variation of occupational stress. Circulation, 18, 852 861. Gamble, G., & Matteson, M. T. (1992). Type A behavior, job satisfaction, and stress among black professionals. Psychological Reports, 70, 43 50. Girdano, D. E., & Everly, G. S. (1986). Controlling stress and tension (2nd ed.). Englewood Cliffs, NJ: Prentice Hall, Inc. Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory: Strategies for qualitative research. Chicago: Aldine Publishing Company. Gregson, T., Wendell, J., & Aono, J. (1994). Role ambiguity, role conﬂict, and perceived environmental uncertainty: Are the scales measuring separate constructs for accoun tants? Behavioral Research in Accounting, 6(6), 144 159. Hall, D. (1972). A model of coping with role conﬂict: The role behavior of college educated women. Administrative Science Quarterly, 17(3), 471 486. Haskins, M. E., Baglioni, A. J., & Cooper, C. L. (1990). An investigation of the sources, moderators, and psychological symptoms of stress among audit seniors. Contemporary Accounting Research, 6(2), 361 385. Hoppock, R. (1935). Job satisfaction. New York: Harper and Row. Hyatt, T. A., & Prawitt, D. F. (2001). Does congruence between audit structure and auditors’ locus of control affect job performance? The Accounting Review, 76(2), 263 274. Ivancevich, J. M., & Matteson, M. T. (1980). Stress and work: A managerial perspective. Glenview, IL: Scott Foresman. Jackson, S. E., & Schuler, R. S. (1985). A meta analysis and conceptual critique of research on role ambiguity and role conﬂict in work settings. Organizational Behavioral and Human Decision Processes, 36, 16 78. Jelinek, R., & Jelinek, K. (2008). Auditors gone wild: The ‘‘other’’ problem in public accounting. Business Horizons, 51, 223 233. Kahn, R. L., Wolfe, D. M., Quinn, R. P., & Snoek, J. D. (1964). Organizational stress: Studies in role conﬂict and ambiguity. New York: Wiley. Kalbers, L. P., & Fogarty, T. J. (1995). Professionalism and its consequences: A study of internal auditors. Auditing: A Journal of Practice and Theory, 14(1), 64 86. Kalbers, L. P., & Fogarty, T. J. (2005). Antecedents to internal auditor burnout. Journal of Management Issues, 17(1), 101 118. Ketchand, A. A., & Strawser, J. R. (1998). The existence of multiple measures of organizational commitment and experience related differences in a public accounting setting. Behavioral Research in Accounting, 10, 109 137. Law, D. W. (2005). Interactive organizational commitment and hardiness in public accountants’ turnover. Managerial Auditing Journal, 20(4), 383 393.

140

KENNETH J. SMITH ET AL.

Law, D. W., Sweeney, J. T., & Summers, S. L. (2008). An examination of the inﬂuence of contextual and individual variables on public accountants’ exhaustion. Advances in Accounting Behavioral Research, 11, 129 153. Leifer, R., & Delbecq, A. (1978). Organizational/environmental interchanges: A model of boundary spanning activity. Academy of Management Review, 3(1), 40 50. Leiter, M., & Maslach, C. (1988). The impact of interpersonal environment on burnout and organizational commitment. Journal of Organizational Behavior, 9(3), 297 308. LePine, J., Podsakoff, N., & LePine, M. (2005). A meta analytic test of the challenge stressor hindrance stressor framework: An explanation for inconsistent relationships among stressors and performance. Academy of Management Journal, 48(5), 764 775. Libby, R. (1983). Comments on Weick. The Accounting Review, 58(2), 370 374. Locke, E. A. (1984). Job satisfaction. In: M. Gainesberg & T. Wall (Eds), Social psychology and organizational behavior. New York: Wiley. Lyons, T. (1971). Role clarity, need for clarity, satisfaction, tension and withdrawal. Organizational Behavior and Human Performance, 5(1), 99 110. Lysonski, S. (1985). A boundary theory investigation of the product manager’s role. Journal of Marketing, 49, 26 40. Mackay, C., & Cooper, C. (1987). Occupational stress and health: Some current issues. In: C. Cooper & I. Robertson (Eds), International review of industrial and organizational psychology. New York: Wiley. Maddi, S. R., & Kobasa, S. C. (1984). The hardy executive. Homewood, IL: Dow Jones Irwin. Maslach, C., & Jackson, S. E. (1981). The measurement of experienced burnout. Journal of Occupational Behavior, 2, 99 113. Maslach, C., & Jackson, S. E. (1986). The Maslach burnout inventory. Palo Alto, CA: Consulting Psychologists Press. Maslach, C., Jackson, S. E., & Leiter, M. M. (1996). The Maslach burnout inventory manual (3rd ed.). Palo Alto, CA: Consulting Psychologists Press. Maslach, C., Schaufeli, W., & Leiter, M. M. (2001). Job burnout. Annual Review of Psychology, 52, 397 422. Mayes, B. T., Johnson, T. W., & Sadri, G. (2000). Personality, job levels, job stressors, and their interaction as predictors of coping behavior. Psychological Reports, 87, 61 81. Meyer, J., & Allen, N. (1984). Testing the ‘‘Side Bet Theory’’ of organizational commitment: Some methodological considerations. Journal of Applied Psychology, 69(3), 372 378. 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. Mowday, R., Steers, R., & Porter, L. W. (1979). The measurement of organizational commitment. Journal of Vocational Behavior, 14(2), 244 247. Otley, D. T., & Pierce, B. J. (1995). The control problem in public accounting ﬁrms: An empirical study of the impact of leadership style. Accounting, Organizations and Society, 17, 405 420. Parasuraman, S., & Alutto, J. A. (1984). Sources and outcomes of stress in organizational settings: Toward the development of a structural model. Academy of Management Journal, 27, 330 350. Rebele, J. E., & Michaels, R. E. (1990). Independent auditors’ role stress: Antecedent, outcome, and moderating variables. Behavioral Research in Accounting, 2, 124 153.

Stress in Accounting Settings

141

Reed, S. A., Kratchman, S. H., & Strawser, R. H. (1994). Job satisfaction, organizational commitment, and turnover intentions of United States accountants. Accounting, Auditing and Accountability Journal, 7(1), 31 58. Rizzo, J. R., House, R. J., & Lirtzman, S. I. (1970). Role conﬂict and role ambiguity in complex organizations. Administrative Science Quarterly, 15, 150 163. Robinson, B. E. (1999). The word addiction risk test: Development of a tentative measure of workaholism. Psychological Reports, 76, 1226. Rotter, J. B. (1966). Generalized expectations for internal versus external control of reinforcement. Psychological Monographs, 80(1), 1 28. Sales, S. M. (1969). Differences among individuals in affective, behavioral, biochemical, and psychological responses to variations in workload. Dissertation Abstracts International, 30, 2407B. Saucier, G. (1994). Mini markers: A brief version of Goldberg’s unipolar big ﬁve markers. Journal of Personality Assessment, 63(3), 506 516. Schaubroeck, J., Cotton, J. L., & Jennings, K. R. (1989). Antecedents and consequences of role stress: A covariance structure analysis. Journal of Organizational Behavior, 10, 35 58. Schuler, R. S. (1980). Deﬁnition and conceptualization of stress in organizations. Organizational Behavior and Human Performance, 23, 331 340. Seaward, B. L. (2006). Managing stress: Principles and strategies for health and wellbeing (5th ed.). Sudbury, MA: Jones & Bartlett. Selye, H. (1956). The stress of life. New York: McGraw Hill. Sheard, M., & Golby, J. (2007). Hardiness and undergraduate academic study: The moderating role of commitment. Personality and Individual Differences, 43(3), 579 588. 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. Smith, C., Organ, D., & Near, J. (1984). Organizational citizenship behavior: Its nature and antecedents. Journal of Applied Psychology, 68(4), 653 663. Smith, K. J. (1990). Occupational stress in accountancy: A review. Journal of Business and Psychology, 4(4), 511 524. Smith, K. J., Davy, J. A., & Everly, G. S. (1995). An examination of the antecedents of job dissatisfaction and turnover intentions among CPAs in public accounting. Accounting Enquiries, 5(1), 99 142. Smith, K. J., Davy, J. A., & Everly, G. S. (2007). An assessment of the contribution of stress arousal to the beyond the role stress model. Advances in Accounting Behavioral Research, 10, 127 158. Smith, K. J., Davy, J. A., & Stewart, B. R. (1998). A comparative study of the antecedents and consequences of job dissatisfaction and turnover intentions among women and men in public accounting ﬁrms. Advances in Public Interest Accounting, 7, 193 227. Smith, K. J., Everly, G. S., & Johns, T. (1993). The role of stress arousal in the dynamics of the stressor to illness process among accountants. Contemporary Accounting Research, 9(2), 432 449. Smith, P., Kendall, L., & Hulin, C. (1969). The measurement of satisfaction in work and retirement. Chicago, IL: Rand McNally. Spector, P. (1988). Development of the work locus of control scale. Journal of Occupational Psychology, 61, 335 340.

142

KENNETH J. SMITH ET AL.

Spence, J. T., Helmreich, R. L., & Pred, R. S. (1987). Impatience versus achievement strivings in type A pattern: Differential effects on students’ health and academic achievement. Journal of Applied Psychology, 72, 528 552. Spence, J. T., & Robbins, A. A. (1992). Workaholism: Deﬁnition, measurement, and preliminary results. Journal of Personality Assessment, 58(1), 160 178. Stallworth, H. L. (2003). Mentoring, organizational commitment and intentions to leave public accounting. Managerial Accounting Journal, 18(5), 405 419. Steffy, B., Jones, J., Noe, J., & Wiggans, A. (1990). The impact of health habits and lifestyle on the stressor strain relationship: An evaluation of three industries. Journal of Occupa tional Psychology, 63(3), 217 229. Stout, D. E., & Rebele, J. E. (1996). Establishing a research agenda for accounting education. Accounting Education A Journal of Theory, Practice and Research, 1(1), 1 18. Sweeney, J. T., & Summers, S. L. (2002). The effect of the busy season workload on public accountants’ job burnout. Behavioral Research in Accounting, 14, 223 245. Viator, R. E. (2001). The association of formal and information public accounting mentoring with role stress and related job outcomes. Accounting, Organizations and Society, 26(1), 73 93. Weick, K. E. (1983). Stress in accounting systems. The Accounting Review, 58(2), 350 369. Williams, L., & Hazer, J. (1986). Antecedents and consequences of satisfaction and commitment in turnover models: A reanalysis using latent variable structural equation methods. Journal of Applied Psychology, 71(2), 219 231. Williams, L., & Podsakoff, P. (1989). Longitudinal ﬁeld methods for studying reciprocal relationships in organizational behavior research: Toward improved causal analysis. Research in Organizational Behavior, 11, 247 292. Wolfe, D. M., & Snoek, J. D. (1962). A study of tensions and adjustment under role conﬂict. Journal of Social Issues (July), 102 121.

EXAMINING THE ANTECEDENTS AND CONSEQUENCES OF REGULAR EXERCISE IN THE AUDIT PROFESSION: HOW CPA FIRMS CAN PROMOTE AUDITORS’ PSYCHOLOGICAL AND PHYSICAL HEALTHINESS D. Kip Holderness Jr. and James E. Hunton ABSTRACT This study relies on the theory of planned behavior (Ajzen, 1985) to examine the antecedents of regular exercise in the audit profession; in addition, the research model tested herein includes two key consequences of exercise: physical healthiness and subjective vitality (one dimension of psychological healthiness). A total of 490 auditors (154 from a large regional CPA ﬁrm and 336 from a Big-4 CPA ﬁrm) participated in the survey. The results indicate that the antecedents of exercise, as articulated by the theory of planned behavior (attitudes, social norms, and perceived behavioral control), are signiﬁcantly and positively related to actual exercise behavior. As a consequence of exercising, the auditors indicated improved physical and psychological healthiness. From a theoretical Advances in Accounting Behavioral Research, Volume 13, 143 168 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013010

143

144

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

perspective, this is the ﬁrst study to our knowledge to test both antecedents and consequences of exercise in a single model. Practically, the results suggest that CPA ﬁrms should create a culture where engaging in regular exercise is expected, accepted, and encouraged; additionally, ﬁrms should ensure that auditors have the opportunity and means to exercise on a regular basis, particularly when they are on the road working at client sites. Rising health care costs are a concern for all employers and employees. A greater understanding of how to improve the physical and psychological healthiness of employees will beneﬁt individuals, organizations, and societies.

INTRODUCTION One objective of this chapter is to examine the antecedents of regular exercise for auditors within the framework of the theory of planned behavior (Ajzen, 1985). Another objective is to examine the extent to which exercise behavior is associated with improved physical and psychological healthiness. While public accounting ﬁrms typically offer wellness programs to promote healthy lifestyles, which include exercise, the theory of planned behavior provides a useful framework for better understanding the determinants of regular exercise. Through such understanding, CPA ﬁrms can learn ways to encourage and facilitate exercise behavior, thereby improving the overall healthiness of auditors and other employees. To our knowledge, no prior research has studied the determinants and consequences of regular exercise in the auditing profession. In a related study, Jones, Norman, and Wier (2010) explore how a healthy lifestyle, as mediated by subjective vitality and improved psychological well-being, can affect self-reported job satisfaction, job performance, and turnover intention. Among other ﬁndings, their results indicate that a healthy lifestyle is signiﬁcantly correlated with subjective vitality, which is deﬁned as ‘‘one’s conscious experience of possessing energy and aliveness’’ (Ryan & Frederick, 1997, p. 530). In their model, the construct of healthy lifestyle is reﬂected by attitude toward exercise, sleep habits, diet, and the excessive use of tobacco and alcohol. They do not, however, measure actual exercise behavior or physical healthiness. We extend their study by more precisely examining antecedents to regular exercise and linking actual exercise behavior to subjective vitality and self-reported indicators of physical healthiness.

Exercise in the Audit Profession

145

This is a timely and relevant issue for CPA ﬁrms, as health care costs continue to skyrocket. National health spending is projected to surpass $2.5 trillion dollars in 2009, comprising 17.6% of gross domestic product (Sisko et al., 2009). According to a recent report from the PricewaterhouseCoopers’ (PwC) Health Research Institute, health insurance premiums have increased four times faster than wages during the past ﬁve years (PwC, 2009). Companies are also placing a greater emphasis on employee wellness programs to lower healthcare expenses. Aon Consulting surveyed 1,313 employers nationwide and found that more than half of employers are planning to introduce or expand employee wellness programs in 2010 (Aon Corporation, 2009). Although the majority of employers are implementing wellness programs in the workplace, few are convinced that these programs are reducing healthcare costs (PwC, 2009). Recent research seems to suggest that wellness programs can have an impact on healthcare expenditures. According to a large study of Arkansas State and Public School Employees Health Plan members, personal health habits are a good indicator of healthcare costs. Researchers ﬁnd that healthcare costs are 45% higher for obese people and 33% higher for people who are not physically active ﬁve or more days per week (Hill, Thompson, Shaw, Pinidiya, & Card-Higginson, 2009). Insurance companies have found similar results. According to an article published by the American Medical Association, more insurers are developing group health plans with lower premiums for healthier employees (Dolan, 2007). Unhealthy employees also add costs to employers in the form of absenteeism. According to the 2007 Mercer National Survey of EmployerSponsored Health Plans, unplanned absence accounts for about 9% of total salary costs in the United States (Nichols, 2008). A report from the U.K.’s 2008 Chartered Institute of Personnel and Development survey ﬁnds that the average per-employee cost of absenteeism is d666 (approximately $1,235) per year (Chartered Institute of Personnel & Development [CIPD], 2008). According to the CCH Inc. 2000 Unscheduled Absence Survey, sickness absence accounts for 40% of all unscheduled absenteeism (CCH Inc., 2002). The evidence suggests that employers beneﬁt from having healthy employees. In this backdrop, this study examines the exercise behavior of auditors who, due to their travel schedules and workload demands, may ﬁnd regular exercise less convenient than do many other professionals. Improving our understanding of the antecedents can help CPA ﬁrms to institutionalize strategies aimed at encouraging and facilitating regular exercise. Furthermore, gaining a better understanding of the physical and psychological

146

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

consequences of exercise can shed light on potential beneﬁts for auditors and their employers. Our sample consists of 490 auditors (154 from a large regional CPA ﬁrm and 336 from a Big-4 CPA ﬁrm). The auditors responded to survey questions designed to measure the three antecedents of planned behavior: attitude toward exercise, perceived social norms toward exercise, and perceived behavioral control over exercise. They also answered questions regarding their level of regular exercise, subjective vitality, and overall physical healthiness, in addition to demographic variables. The results suggest that the theory of planned behavior adequately captures the essential determinates of exercise behavior. The model indicates that if auditors hold positive attitudes toward exercise and if they believe that referent others would view their exercise behavior in a positive light, they are more likely to exercise on a regular basis. The ﬁndings also indicate that auditors’ perceived behavioral control over exercise is predictive of actual exercise behavior. Furthermore, the research ﬁndings suggest that more exercise is signiﬁcantly associated with higher levels of subjective vitality and physical healthiness. This study contributes to extant literature in several ways. First, this study examines the utility of the theory of planned behavior in determining the antecedents of regular exercise for auditors. Second, a structural equation model is used to combine the antecedents and consequences of actual exercise behavior. From a practical perspective, the results suggest that CPA ﬁrms can encourage a healthy lifestyle by promoting positive attitudes toward regular exercise, creating a corporate culture where engaging in regular exercise is an expected social norm, and providing time and resources to encourage auditors to exercise regularly while they are in the ofﬁce or on the road. The remainder of the chapter is organized as follows. The following section summarizes relevant background literature and presents the research model that will be tested in this study. Then, the next section describes the research method. Afterward, the results of statistical analysis are presented in the following section and the ﬁnal section discusses our ﬁndings.

BACKGROUND LITERATURE AND RESEARCH MODEL Theory of Planned Behavior and Exercise The theory of planned behavior (Ajzen, 1985) has long been used to predict behavior based on an individual’s attitudes and beliefs. According to the

Exercise in the Audit Profession

147

theory, actual behavior is largely inﬂuenced by one’s intention to perform a given behavior. Intention is a function of three independent variables: (1) attitude toward the behavior, (2) the subjective norm regarding the behavior, and (3) perceived behavioral control over the behavior. Attitude reﬂects the degree to which an individual positively or negatively values the self-performance of a particular behavior. According to expectancy-value theory (Eccles, 1983), individuals assign weights to the possible outcomes of a behavior, and the aggregation of the weighted outcomes determines their attitude. Subjective norm reﬂects an individual’s perceived social pressure to engage in or abstain from a particular behavior. This pressure is determined by the perceived opinions of important referent individuals and groups, including family, friends, and coworkers. Perceived behavioral control refers to an individual’s perception of his/her ability to perform a particular behavior, which is determined by an individual’s beliefs about the presence of factors that may facilitate or impede performance of a behavior. One of the earliest applications of the theory of planned behavior was a study that used the theory to predict weight loss (Schifter & Ajzen, 1985). They found that all three independent variables of the theory of planned behavior (attitudes, social norms, and perceived behavioral control) were signiﬁcantly associated with the intention to lose weight. However, they also found that the theory of planned behavior was only moderately successful in predicting actual weight loss, as the intention to lose weight had a low, but signiﬁcant, correlation with actual weight loss. The authors suggested that the weaker-than-expected link between intention to lose weight and actual weight loss may be due to the fact that weight loss ‘‘is not a behavior but, rather, an outcome over which individuals have only limited behavioral control’’ (p. 850). While weight loss is contingent on numerous factors, some of which are not controllable, most people have volitional control over their exercise habits, within the conﬁnes of physical capabilities. Accordingly, numerous studies have examined exercise habits in the context of the theory of planned behavior. Godin and Kok (1996) found that the theory of planned behavior was a better predictor of exercise intention and behavior than other health-related behaviors. Hausenblas, Carron, and Mack (1997) conducted a metaanalytical study of the theory of planned behavior as it relates to physical activity. Their study reviewed previous literature to examine the relationship between the three independent variables of the theory of planned behavior with intention to exercise and exercise behavior. Their results supported the utility of the theory of planned behavior to predict exercise intention and

148

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

behavior. They reported that while attitude and perceived behavioral control had a large effect on exercise behavior, the relationship between subjective norm and exercise behavior was nonsigniﬁcant. They singularly tested the individual links between each antecedent of exercise behavior included in the theory of planned behavior but did not use statistical techniques to establish the validity of the model as a whole. In another study, Hagger, Chatzisarantis, and Biddle (2002) conducted a more comprehensive examination of the theory of planned behavior as it pertains to physical activity by using regression and path analysis. They used meta-analytic techniques to examine the theory of planned behavior constructs from 72 previous studies. Additionally, they used regression and path analysis to determine the validity of the model in the context of exercise. Their results supported the utility of the theory of planned behavior in predicting exercise intention and behavior. A review of the relevant literature demonstrates that the theory of planned behavior has been used to predict exercise intention and behavior in numerous populations including college students (Courneya & McAuley, 1994; Norman, Conner, & Bell, 2000), medical patients (Courneya & Friedenreich, 1999; Courneya, Friedenreich, Arthur, & Bobick, 1999; Godin, Valois, Jobin, & Ross, 1991), health club members (Kerner & Grossman, 1998; Theodorakis, 1994), university employees (Godin, Valois, Shephard, & Desharnais, 1987), and company employees (Chatzisarantis & Biddle, 1998; Godin & Gionet, 1991; VanRyn, Lytle, & Kirscht, 1996). While the theory of planned behavior has not been used in the context of auditors, Kimiecik (1992) ﬁnds the theory of planned behavior to be useful in explaining the exercise behavior of corporate employees, which are arguably more similar to auditors than most subjects used in other research of this nature. There is some evidence that the effect sizes of various dimensions of the theory planned behavior vary according to the sample used in each study. For example, Kwan, Bray, and Ginis (2009) examined the relationship between the theory of planned behavior and exercise in ﬁrst-year university students and found that the relationship of social norms on exercise intention was substantially stronger than previously reported meta-analytic results. Understanding the relative effect sizes of the variables of the theory of planned behavior on different populations facilitates the creation of effective exerciseencouraging initiatives at individual, organizational, and societal levels. The results can potentially beneﬁt the accounting profession as this is the ﬁrst study to use the theory of planned behavior to determine the antecedents of exercise for auditors, who have unique and demanding work

Exercise in the Audit Profession

149

schedules that typically include long hours at numerous client locations, which often results in less time and opportunity for engaging in regular exercise. While previous literature has not examined exercise behavior in auditors, we nevertheless expect the theory of planned behavior to adequately reﬂect the determinants of actual exercise behavior based on prior literature. This study is also unique in that it combines the antecedents of exercise with self-reports of actual exercise behavior, physical healthiness, and subjective vitality into a single model, as next discussed.

Regular Exercise, Physical Healthiness, and Subjective Vitality Regular exercise has long been established to have numerous positive outcomes with regard to personal health. In 1995, the Centers for Disease Control and Prevention and the American College of Sports Medicine issued a public health recommendation on the importance of exercise. A panel of experts, chosen for their expertise on the health implications of physical activity, reviewed the pertinent literature and concluded that every U.S. adult should exercise at least 30 minutes on most days of the week (Pate et al., 1995). The 1996 Surgeon General’s Report issued similar advice. The report recommended that all people regardless of age obtain ‘‘a minimum of 30 minutes of physical activity of moderate intensity (such as brisk walking) on most, if not all, days of the week’’ (CDC, 1996, p. 6). Many other studies support the health beneﬁts of regular exercise (Haskell et al., 2007; Luepker et al., 1996; Warburton, Nicol, & Bredin, 2006). On the basis of the results of prior studies, auditors who exercise regularly should report better physical healthiness than those who do not exercise regularly. The concept of subjective vitality, which reﬂects one dimension of psychological healthiness, was ﬁrst studied by Ryan and Frederick (1997). They describe vitality as the sense of being full of energy and alive. The authors designed a seven-item scale to measure subjective vitality. Subjective vitality reﬂects a person’s self-reported feeling of vitality. Through a series of six studies, they ﬁnd subjective vitality to be inversely related to psychological/ somatic distress, anxiety, and depression and positively associated with mental health (Ryan & Frederick, 1997). Bostic, Rubio, and Hood (2000) use principal component factor analysis to assess the construct validity and utility of the subjective vitality scale. Their results indicate that the seven-item scale ‘‘does in fact measure one construct, vitality’’ (p. 321). Several studies have found that higher levels of physical activity are associated with increased vitality. Stewart et al. (2003), in a cross-sectional

150

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

investigation, found that physically active older people reported higher levels of vitality than inactive older people. They also reported that aerobic ﬁtness was a strong predictor of higher vitality. Medical studies have examined the effect of physical activity counseling on patients. In randomized controlled trials, subjects who received physical activity counseling exhibited higher levels of exercise and reported increased levels of vitality (Elley, Kerse, Arroll, & Robinson, 2003; Kerse, Elley, Robinson, & Arroll, 2005). Longitudinal studies have also suggested a positive association between physical activity and vitality (Tessier et al., 2007; Vuillemin et al., 2005; Wendel-Vos, Schuit, Tijhuis, & Krombout, 2004). In the accounting literature, only Jones et al. (2010) have examined the link between a healthy lifestyle and subjective vitality. The authors collected survey data from 1,704 partners and employees at a large public accounting ﬁrm to examine the effects of a healthy lifestyle as a coping mechanism for role stress. They hypothesized that a healthy lifestyle, as mediated by vitality and improved psychological well-being, would be positively associated with job satisfaction and job performance and negatively associated with turnover intentions. In their study, a healthy lifestyle scale was used, which included factors such as diet, sleep habits, and tobacco/alcohol use, and attitudes toward exercise. Survey participants indicated the extent to which they agreed with statements reﬂecting their attitudes toward exercise (i.e., I am an active person; exercise puts me more in control). Participants did not indicate their actual exercise behavior. As a measure of vitality, Jones et al. used the six-question scale developed by Ryan and Frederick (1997) and modiﬁed by Bostic et al. (2000). Their results indicated that a healthy lifestyle was positively associated with subjective vitality. We believe that our study will yield results on the effect of exercise on vitality similar to Jones et al. (2010). This study extends Jones et al. (2009) by examining the effect of regular exercise on physical healthiness. Furthermore, while Jones et al. recommends the use of healthy lifestyle programs in public accounting organizations, obtaining a better understanding of the determinants of exercise can help to increase the effectiveness of these programs. Research Model This chapter is the ﬁrst to use the theory of planned behavior to incorporate both the antecedents and the consequences of exercise behavior into a single model, as shown in Fig. 1. Using structural equation modeling, the model is tested to determine the extent to which attitudes toward exercise behavior, social norms toward exercise behavior, and perceived behavioral control

151

Exercise in the Audit Profession

Attitude Toward Exercise Behavior

Physical Health Social Norms Toward Exercise Behavior

Exercise Behavior

Subjective Vitality Perceived Behavioral Control Over Exercise Behavior

Fig. 1.

Research Model.

over exercise behavior can predict actual exercise behavior, and the degree to which exercise behavior is associated with indicators of physical healthiness and subjective vitality.

RESEARCH METHOD Survey Construction Subjective Vitality The subjective vitality scale from Ryan and Frederick (1997), as tested and validated by Bostic et al. (2000), is used in this study. The scale includes seven items, which appear as items one through seven in the appendix.

152

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

Theory of Planned Behavior The theory of planned behavior (Ajzen, 1985) includes three antecedent constructs aimed at the target behavior, in this case, exercise behavior: attitude toward exercise, social norms about exercise, and perceived behavioral over exercise. A focus group, composed of 16 auditors from a Big-4 CPA ﬁrm who did not participate in the survey, helped the researchers to develop the survey response items. One of the issues was how to assess the extent to which the survey respondents actually exercise. Ajzen (2006) urges that the target behavior be as clearly deﬁned as possible. In this study, the behavior of interest is exercise, and the participants need to have a clear idea of what is meant by exercise with regard to the amount of contiguous time that represents exercise and how many days per week constitutes an exercise pattern or routine. The researcher asked the focus group the period each day that would constitute exercise and how many days per month of exercise would reﬂect a reasonable but rigorous routine. After some deliberation, the group decided that a period of at least 30 minutes was the minimum and that exercising for at least 30 minutes per day every day would be reasonable but rigorous. The next part of the issue dealt with the time horizon to assess. The group felt that respondents could remember how much they exercised over the past month, but any longer period would be difﬁcult to recall with precision; yet, there could have been uncontrollable circumstances over the past month that prevented them from exercising as much as normal (e.g., work and family demands). After discussing the time horizon, the group agreed that the survey should ask participants about the number of exercise days over the past month, 3 months, 6 months, and 12 months to rule out abnormalities that might be associated with assessing only a single period. Accordingly, as shown in the appendix, items 8 through 11 ask about exercise behavior from the past month to the past year. According to Ajzen (2006), measurement of the attitude, social norm, and perceived behavioral control constructs should be compatible with the target behavior. With help of the focus group and guidance by Ajzen (2006), survey items reﬂecting attitude toward exercise are indicated in items 12 through 16, survey items reﬂecting social norms are indicated in items 17 through 19, and survey items reﬂecting perceived behavioral control are indicated in items 20 through 22 (see appendix). Healthiness Assessments The participating CPA ﬁrms were also interested in understanding the relationships among subjective vitality, exercise behavior, and physical

153

Exercise in the Audit Profession

healthiness. Accordingly, to evaluate healthiness, the survey included three self-assessment items regarding the participants’ weight relative to other people in their age range (item 23), overall medical health relative to other people in their age range (item 24), and the approximate number of sick days they have taken during the past 12 months (item 25) (see appendix). Demographics Finally, the survey asked for the participants’ gender (item 26), age (item 27), and position title (item 28) (see appendix).

Administration One Big-4 CPA ﬁrm and one large regional CPA ﬁrm participated in the study. One of the researchers was conducting 90-minute educational sessions on the topic of business information technology continuity planning for audit clients. At the beginning of each session (four sessions for the Big-4 ﬁrm and three sessions for the regional ﬁrm), the researchers asked the participating auditors if they would volunteer to complete a survey. All attendees in each session agreed and completed the survey. The average time to ﬁnish the survey was about 10 minutes.

Participants A total of 490 participants completed the survey, of which 224 were male and 266 were female; 154 worked for a large regional CPA ﬁrm and 336 were employed by a Big-4 CPA ﬁrm. There were 159 staff auditors, 218 senior auditors, 99 managers, and 14 partners in the sample. The average (standard deviation) age was 32.65 years (6.07). Statistical testing indicated no signiﬁcant (pW.10) difference between the two ﬁrms on mean age, position level distribution, or gender distribution.

RESULTS As reﬂected in Fig. 1, there are six latent constructs in the model: attitude toward exercise behavior, social norms toward exercise behavior, perceived behavioral control over exercise behavior, exercise behavior, physical healthiness, and subjective vitality. Because these constructs were articulated

154

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

theoretically a priori, the data analysis begins with a conﬁrmatory factor analysis that includes all of the response items. Using varimax orthogonal rotation, only factors with eigenvalues equal to or greater than 1 were retained. The result is summarized in Table 1. The conﬁrmatory factor analysis (Table 1) resulted in six distinct constructs with eigenvalues greater than 1. The model accounted for 75.20% of the variance in the data. Without exception, each survey question loaded high (W.7) on its predicted construct, which suggests sufﬁcient convergent validity. Additionally, no survey question loads high on any other predicted construct, which suggests sufﬁcient divergent validity.

Table 1.

Conﬁrmatory Factor Analysis Results (Varimax Rotation). Eigenvalues (Percent of Variance)a

Survey Response Items (See Appendix) 5.16 (20.63) Subjective vitality Subjective vitality Subjective vitality Subjective vitality Subjective vitality Subjective vitality Subjective vitality Exercise behavior Exercise behavior Exercise behavior Exercise behavior Attitude Attitude Attitude Attitude Attitude Social norm Social norm Social norm Perceived control Perceived control Perceived control Weight Relative health Sick days past year a

3.64 (14.58)

3.30 (13.21)

2.46 (9.83)

2.36 (9.45)

1.88 (7.50)

.954 .824 .838 .843 .820 .807 .835 .747 .894 .939 .908 .932 .798 .811 .805 .784 .852 .761 .702 .923 .878 .863 .848 .880 .810

Only factor loadings Z.50 are shown. Cumulative Percent of Variance ¼ 75.20

155

Exercise in the Audit Profession

Table 2.

Descriptive Statistics (Sample Size ¼ 490). Mean

Panel A: Construct indicesa Subjective vitality Exercise behavior Attitudes Social norms Perceived behavioral control Physical healthiness Panel B: Detailed physical healthiness indicatorsb Relatively overweight or underweightc Physical health, relative to referent othersd Number of sick days in past yeare

Standard Deviation

3.95 3.44 4.13 3.87 4.26 0.68

1.62 1.23 1.51 1.34 1.69 4.92

1.30 3.62 3.00

0.87 1.69 2.95

a

Indices comprise the mean values of all survey items reﬂecting the latent constructs. The self reported physical healthiness indicators were averaged into a single index; yet, because each item is measured on a different scale and two of the three items are reverse coded for analysis purposes, the composite mean and standard deviation are not easily interpretable. Thus, Panel B presents a detailed breakdown of the three physical healthiness survey items. c A response of ‘‘average’’ (in the middle of the scale) was coded as 0. The overweight observations were coded to be consistent with the notion that more overweight is less healthy, thus to the right of ‘‘average,’’ responses were coded as 1, 2, and 3 (very overweight). There were only eight underweight observations. One could argue that being underweight is as unhealthy as being overweight. Hence, to be conservative, the eight underweight observations were coded with a 1, just as the overweight observations were coded with a 1. d Scale: 1 (much worse), 4 (average), 7 (much better). e Scale: Number of missed days due to sickness or illness. The data were reverse coded to be consistent with the notion that more missed days reﬂects less healthiness. b

Table 2 presents descriptive statistics for the six latent constructs of our model. The indices were created by averaging all of the survey questions reﬂecting a particular construct. Panel A (Table 2) presents means and standard deviations of the latent constructs. Some of the data were reverse coded for the physical healthiness construct so that there would be an interpretable relationship between physical healthiness and exercise behavior, as next discussed. Panel B (Table 2) presents the three response items that underlie the physical healthiness construct. The ﬁrst item reﬂects the participants’ assessments of their weight, ranging from very underweight to average to very overweight with ‘‘average’’ coded as zero (0). Next, because being more overweight is suggestive of poorer healthiness, all overweight responses were coded 1, 2, and 3 (very overweight). Only eight respondents indicated that they were underweight and all of them checked the ﬁrst blank line to the

156

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

left of average (0), suggesting that they were slightly underweight. One could argue that being underweight is just as unhealthy as being overweight; yet, being slightly underweight might not indicate unhealthiness. Nevertheless, we decided to take the conservative route – the one that biases us against ﬁnding a positive relationship between exercise behavior and weight – by coding the eight underweight response as 1 (rather than coding them as þ1), which equates slightly underweight (1) with slightly overweight (1). The second response item, physical health relative to other people in the participants’ age range, ranged from 1 (much worse) to 4 (average) to much better (7). Starting from the left side of the scale, the responses were coded from þ1 through þ7, because this approach directionally associates better relative health with more exercise behavior. The third response item asks how many days of work they missed due to sickness or illness. These responses were reverse coded, suggesting that more missed days reﬂect less physical healthiness.

Preliminary Testing Before testing the research model, training session (there were seven different sessions), ﬁrm size (Big-4 vs. regional), survey version (there were three versions wherein the survey questions were randomized), age, gender, and position level (staff, senior, manager, and partner) were examined to determine whether they were associated with self-reported exercise behavior. An omnibus regression model was used, where attitudes toward exercise, social norms toward exercise, behavioral control over exercise, physical healthiness, and subjective vitality were included as independent variables, and exercise behavior served as the dependent variable; additionally, training session, ﬁrm size and survey version, age, gender, and position level were included as potential covariates. The overall regression model was signiﬁcant (F ¼ 17.40, pW.01, adjusted R2 ¼ .27). The signiﬁcance level of each independent variable was as follows: attitudes (t ¼ 5.85, po.01), social norms (t ¼ 2.04, p ¼ .04), perceived behavioral control (t ¼ 1.33, p ¼ .19), subjective vitality (t ¼ 4.92, po.01), and physical healthiness (t ¼ 5.88, po.01). While perceived behavioral control was nonsigniﬁcant in the omnibus regression model, the standardized path coefﬁcient from perceived behavioral control to exercise behavior nevertheless could be signiﬁcant in the structural equation model because the path coefﬁcients are determined by a multistage regression modeling process.

157

Exercise in the Audit Profession

Gender (coded as 0 ¼ female and 1 ¼ male) was signiﬁcant (t ¼ 4.24, po.01), indicating that female auditors are engaging in a higher mean level of regular exercise than male auditors. None of the remaining potential covariates were signiﬁcant: training session (t ¼ 0.43, p ¼ .66), ﬁrm size (t ¼ 1.22, p ¼ .22), survey version (t ¼ 0.75, p ¼ .46), age (t ¼ 1.00, p ¼ .32), and position level (t ¼ 0.02, p ¼ .99). After testing the full structural equation model (below), sensitivity testing on gender will be conducted.

Model Testing Table 3 presents correlations, covariances, and interitem reliability estimates (Cronbach’s alpha). All of the constructs have a reliability of .68 or higher, which suggests satisfactory interitem reliability, as supported by the earlier factor analysis. All correlations are signiﬁcant at .01 or .05, except for the correlation between subjective vitality and social norms (.03), social norms and attitude (.04), and social norms and physical healthiness (.09). All signiﬁcant correlations and covariances are positive, as expected. Results of the structural equation model, using AMOS (version 17.0), are shown in Fig. 2.1 The goodness of ﬁt indices suggest that the model adequately ﬁts the data (GFI ¼ 0.98; AGFI ¼ 0.98; CFI ¼ 0.91; RMSEA ¼ 0.06). Ajzen (1985, 2006) suggests that attitudes, social norms, and perceived behavioral control might be intercorrelated; thus, the possibility of such intercorrelation in the structural equation model were included, which minimizes the effect of multicollinearity among the Table 3.

Subjective vitality (SV) Exercise behavior (EB) Attitude (AT) Social norms (SN) Perceived control (PC) Physical healthiness (PH)

Correlation–Covariance Matrix. SV

EB

.94 .26 .13 .03 .13 .15

.51 .94 .34 .13 .16 .37

AT

SN

.31 .62 .89 .04 .09 .27

.06 .22 .09 .68 .26 .09

PC

PH

.37 .33 .23 .58 .88 .18

1.22 2.25 1.97 0.58 1.50 .74

Notes: Correlations are shown below the diagonal, Cronbach’s alpha values are shown on the diagonal, and covariances are shown above the diagonal. Signiﬁcant at .01 (two tailed). Signiﬁcant at .05 (two tailed).

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

158

Attitude Toward Exercise Behavior

.43*

Physical Healthiness

.05 .55*

.11*

Social Norms Toward Exercise Behavior

.28* Exercise Behavior .40*

.35*

.15*

Subjective Vitality

Perceived Behavioral Control Over Exercise Behavior

Fig. 2.

Tested Model.

exogenous constructs on exercise behavior. Indeed, the correlation between attitude toward exercise behavior and perceived behavioral control over exercise behavior is signiﬁcant (r ¼ .11, po.01), as is the correlation between social norms toward exercise behavior and perceived behavioral control over exercise behavior (r ¼ .35, po.01). The correlation between attitude toward exercise behavior and social norms toward exercise behavior (r ¼ .05), however, is not signiﬁcant (p W.10). As expected, attitudes, social norms, and perceived behavior control all have a signiﬁcant impact on exercise behavior (po.01). The link between attitude toward exercise behavior and actual exercise behavior is the strongest of the three antecedents (standardized path coefﬁcient ¼ .43). Although past research suggests that subjective norms toward exercise behavior is a signiﬁcantly weaker indicator of actual exercise behavior than

Exercise in the Audit Profession

159

either attitude or perceived behavioral control (Kwan et al., 2009; Godin & Kok, 1996; Hagger et al., 2002), these results show that subjective norms toward exercise behavior has a stronger impact on exercise behavior (standardized path coefﬁcient ¼ .28) than perceived behavior control (standardized path coefﬁcient ¼ .15). This ﬁnding highlights the importance of creating a culture within the public accounting profession that fosters, expects, and respects exercise as a way to encourage such behavior. The association between exercise behavior and one dimension of psychological healthiness (subjective vitality) is signiﬁcant (standardized path coefﬁcient ¼ .40, po.01). The path between exercise behavior and physical healthiness is also signiﬁcant (standardized path coefﬁcient ¼ .55, po.01). These results indicate that more exercise is positively associated with improved psychological and physical healthiness, which suggests that regular exercise appears to be beneﬁcial for employees as well as employers. Sensitivity Analysis of Gender Differences As noted in preliminary testing, the variable ‘‘gender’’ was found to be a signiﬁcant covariate. To understand the nature of the gender effect, mean differences among the latent constructs between male and female auditors were ﬁrst examined (Table 4, Panel A). As indicated by the previously reported omnibus regression model, the reported degree to which female auditors engage in regular exercise (mean ¼ 3.69) is signiﬁcantly higher than male auditors (mean ¼ 3.15). The means for attitudes toward exercise and perceived behavioral control over exercise are not signiﬁcantly different; however, the mean for social norms for female auditors (3.98) is signiﬁcantly higher than male auditors (3.73). This indicates that female auditors, relative to male auditors, perceive a higher level of support and encouragement from referent others to engage in regular exercise. The reported mean levels of subjective vitality are the same for female (3.92) and male (3.99) auditors. The remaining signiﬁcant mean difference is physical healthiness, where female auditors reﬂect a higher level of healthiness (0.12) than male auditors (1.33). Next, we attempted to run a multigroup analysis of the full-research model, subdividing the data by gender; however, the two sample sizes (male ¼ 224 and female ¼ 266) were too small for the structural equation model to converge to a solution. Thus, the model was ﬁrst reduced to reﬂect the theory of planned behavior constructs (attitude, social norm, perceived behavioral control, and exercise behavior), and then multigroup analysis was performed, which indicated that the male and the female auditor subsamples yield signiﬁcantly

160

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

Table 4. Sensitivity Analysis of Gender Differences. Female Auditors

Panel A: Construct indices Exercise behavior Attitudes Perceived behavioral control Social norms Subjective vitality Physical healthiness

Male Auditors

Mean

SD

Mean

SD

3.69 4.23 4.26 3.98 3.92 0.12

1.06 1.31 1.69 1.31 1.51 4.78

3.15 4.01 4.27 3.73 3.99 1.33

1.34 1.71 1.68 1.36 1.75 5.02

t statistic

4.87 1.56 0.07 1.98 0.51 2.73

Female Auditors

p value

.01 .12 .94 .05 .61 .01 Male

Auditors

Z score

p value

.33 .33 .15

5.86 4.36 2.32

.01 .01 .02

a

Panel B: Parameter estimates of exercise determinants Attitudes to exercise behavior .23 Social norms to exercise behavior .26 Behavioral control to exercise behavior .09

Female Auditors Panel C: Parameter estimates of exercise consequences Exercise behavior to subjective vitality .35 Exercise behavior to physical healthiness .40

Male

Auditors

Z score

p value

.51 .21

3.82 8.53

.01 .01

a

Multigroup analysis of path coefﬁcients compares parameter estimates because standardized coefﬁcients may differ across groups only because the variances differ across groups.

different results (chi-square ¼ 453.6, po.01).2 Subsample analysis suggests that path coefﬁcients from the exogenous constructs to exercise behavior are signiﬁcantly different between genders (Table 4, Panel B). The parameter estimates3 between attitudes and behavior are lower for female auditors (.23) than male auditors (.33), suggesting that the attitude that female auditors hold toward exercising is less inﬂuential on their actual exercise behavior than the attitude held by male auditors. The parameter estimates between social norms and exercise behavior is also less for female auditors (.26) than male auditors (.33), indicating that female auditors are less inﬂuenced by social norms than male auditors. With regard to the link between perceived behavioral control and exercise behavior, the parameter estimate for female auditors (.09) is signiﬁcantly smaller than for male

161

Exercise in the Audit Profession

auditors (.15); thus, the inﬂuence of perceived behavioral control on exercise is less for female than male auditors. Overall, the determinants of exercise behavior appear to be less predictive of actual exercise for female, relative to male, auditors. Finally, one more reduced model was used to reﬂect the paths from exercise behavior to subjective vitality and from exercise behavior to physical healthiness, and multigroup analysis indicated that the male and female auditor subsamples yield different results (chi-square ¼ 314.1, po.01).4 The analysis suggests that path coefﬁcients are signiﬁcantly different between the subsamples (Table 4, Panel C); speciﬁcally, the parameter estimate reﬂecting the path from exercise behavior to subjective vitality is smaller for female (.35) than male (.51) auditors, suggesting a weaker association, and the parameter estimate from exercise behavior to physical healthiness is greater for female (.40) than male (.21) auditors, indicating a stronger association.

DISCUSSION This study relies on the theory of planned behavior to link the determinants of exercise with self-reported exercise behavior in the audit profession. As a baseline, the extent to which attitudes, social norms, and perceived behavior control are predictive of actual exercise behavior was tested, as has been indicated in previous academic literature in different contexts and professions. The model and extant research was then expanded by adding indicators of physical healthiness and subjective vitality (one dimension of psychological healthiness) as consequences of exercise behavior. A total of 490 auditors participated in the survey. The ﬁndings indicate that attitudes toward exercise behavior, subjective norms toward exercise behavior, and perceived behavioral control over exercise are all instrumental in determining an auditor’s actual exercise behavior; furthermore, increased exercise is positively correlated with improved physical healthiness and psychological feelings of vitality. Subanalyses of gender effects indicate the following: female auditors, relative to the male auditors, (1) exercise more often, (2) perceive a higher level of support from referent others to exercise regularly, and (3) are physically healthier; yet, the theory of planned behavior is less predictive of actual exercise behavior for female auditors, relative to male auditors. From a practical perspective, the ﬁndings indicate that CPA ﬁrms should create a corporate culture where regular exercise is an expected, accepted,

162

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

and desired social norm. Firms should also ensure that auditors believe they have behavioral control over engaging in regular exercise by allowing them the opportunity to exercise. Firms may consider booking their employees into hotels with adequate exercise facilities and possibly providing them with membership in a national or international health club/gym so that they have the means to exercise while on the road. Such measures should facilitate increased exercise behavior, particularly when auditors are traveling to various client locations. These ﬁrm-level initiatives should improve the auditors’ attitude toward exercise. On the whole, our results suggest that if attitudes, social norms, and perceived behavioral control are properly aligned toward exercise, the likelihood that auditors will actually engage in a regular exercise program will increase. The results make several contributions to research. First, the theory of planned behavior appears to be useful in determining the antecedents of auditors’ actual exercise behavior, which supports prior research. The ﬁndings extend prior research by further linking regular exercise to improved psychological and physical healthiness. Consistent with Jones et al. (2010), we recommend the implementation of employee wellness programs within ﬁrms, with a particular emphasis on regular exercise. This study is not without limitations. First, our ﬁndings are based on selfassessments; hence, determining the reliability of responses with any precision is difﬁcult. Additionally, as with any proposed model, other signiﬁcant factors that were not included may exist; for instance, some past studies have indicated that other variables might attenuate the effect of the theory of planned behavior dimensions on actual behavior, such as prior behavior, intention certainty, and self-efﬁcacy (Arau´jo, McIntyre, & Sniehotta, 2009; Hagger et al., 2002). Finally, two CPA ﬁrms were surveyed; thus, the extent to which the ﬁndings are generalizable to other CPA ﬁrms is unknown. Rising health care costs are a concern for all organizations. This study provides information that could be useful for companies attempting to improve employees’ health and may lead to lower medical-related expenses. Healthier employees may contribute to a ﬁrm’s proﬁt in other ways as well. For instance, future research could examine the effect of exercise on work productivity and employee satisfaction. Other research could focus on alternate methods of increasing exercise behavior, such as incentive programs. Rising health care costs are becoming considerably more burdensome for employers and employees, not to mention the human suffering that accompanies unhealthiness; in this light, researchers can add

163

Exercise in the Audit Profession

value to society by investigating various ways to improve the healthiness of people in all professions.

NOTES 1. The dataset indicated a violation of multivariate normality (Martia’s coefﬁcient 68.94, critical ratio 20.77, po.01). Thus, the asymptotically distribu tion free method of estimation was used to calculate the structural equation model. 2. The dataset indicated a violation of multivariate normality (Martia’s coefﬁcient 80.98, critical ratio 29.21, po.01). Thus, the asymptotically distribu tion free method of estimation was used to calculate the structural equation model. 3. Parameter estimates are compared in multigroup analysis because standardized coefﬁcients can differ merely based on different variances. 4. The dataset indicated a violation of multivariate normality (Martia’s coefﬁcient 72.45, critical ratio 27.91, po.01). Thus, the asymptotically distribu tion free method of estimation was used to calculate the structural equation model.

REFERENCES Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In: J. Kuhl & J. Beckman (Eds), Action control: From cognition to behavior. New York: Springer Verlag. Ajzen, I. (2006). Constructing a TbP questionnaire: Conceptual and methodological considerations. Available at http://people.umass.edu/aizen/pdf/tpb.measurement.pdf Aon Corporation. (2009). Most employers making changes to their 2010 medical program expect to increase employee contributions: Aon Consulting Survey. Available at http://ir.aon.com/ phoenix.zhtml?c¼ 105697&p¼ irol newsArticle_print&ID¼ 1339658&highlight Arau´jo, S., McIntyre, & Sniehotta. (2009). Predicting changes in physical activity among adolescents: The role of self efﬁcacy, intention, action planning and coping planning. Health Education Research, 24(1), 128. Bostic, T. J., Rubio, D. M., & Hood, M. (2000). A validation of the subjective vitality scale using structural equation modeling. Social Indicators Research, 52(3), 313 324. CCH Inc. (2002). 2002 CCH unscheduled absence survey. Available at http://www.cch.com/ press/news/2002/20021016h.asp CDC. (1996). Physical activity and health: A report of the surgeon general. Atlanta, GA: US Department of Health and Human Services. Chartered Institute of Personnel & Development. (2008). Annual report 2008: Absence management. Available at http://www.cipd.co.uk/NR/rdonlyres/6D0CC654 1622 4445 8178 4A5E071B63EF/0/absencemanagementsurveyreport2008.pdf Chatzisarantis, N. L. D., & Biddle, S. J. H. (1998). Functional signiﬁcance of psychological variables that are included in the theory of planned behaviour: A self determination

164

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

theory approach to the study of attitudes, subjective norms, perceptions of control and intentions. European Journal of Social Psychology, 28(3), 303 322. Courneya, K. S., & Friedenreich, C. M. (1999). Utility of the theory of planned behavior for understanding exercise during breast cancer treatment. Psycho Oncology, 8(2), 112 122. Courneya, K. S., Friedenreich, C. M., Arthur, K., & Bobick, T. M. (1999). Understanding exercise motivation in colorectal cancer patients: A prospective study using the theory of planned behavior. Rehabilitation Psychology, 44(1), 68 84. Courneya, K. S., & McAuley, E. (1994). Are there different determinants of the frequency, intensity, and duration of physical activity. Behavioral Medicine, 20(2), 84 90. Dolan, P. L. (2007). Fewer back higher premiums for unhealthy living. American Medical Association. Available at http://www.ama assn.org/amednews/2007/11/26/bisd1126.htm Eccles, J. (1983). Expectancies, values, and academic behaviors. In: J. T. Spence (Ed.), Achievement and achievement motives: Psychological and sociological approaches (pp. 75 146). San Francisco: Freeman. Elley, C. R., Kerse, N., Arroll, B., & Robinson, E. (2003). Effectiveness of counselling patients on physical activity in general practice: Cluster randomised controlled trial. British Medical Journal, 326(7393), 793 796. Godin, G., & Gionet, N. J. (1991). Determinants of an intention to exercise of an electric power commissions employees. Ergonomics, 34(9), 1221 1230. Godin, G., & Kok, G. (1996). The theory of planned behavior: A review of its applications to health related behaviors. American Journal of Health Promotion, 11, 87 89. Godin, G., Valois, P., Jobin, J., & Ross, A. (1991). Prediction of intention to exercise of individuals who have suffered from coronary heart disease. Journal of Clinical Psychology, 47(6), 762 772. Godin, G., Valois, P., Shephard, R. J., & Desharnais, R. (1987). Prediction of leisure time exercise behavior: A path analysis (Lisrel V) model. Journal of Behavioral Medicine, 10(2), 145 158. Hagger, M. S., Chatzisarantis, N. L. D., & Biddle, S. J. H. (2002). A meta analytic review of the theories of reasoned action and planned behavior in physical activity: Predictive validity and the contribution of additional variables. Journal of Sport & Exercise Psychology, 24(1), 3 32. Haskell, W. L., Lee, I. M., Pate, R. R., Powell, K. E., Blair, S. N., Franklin, B. A., Macera, C. A., Heath, G. W., Thompson, P. D., & Bauman, A. (2007). Physical activity and public health: Updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Medicine and Science in Sports and Exercise, 39(8), 1423 1434. Hausenblas, H. A., Carron, A. V., & Mack, D. E. (1997). Application of the theories of reasoned action and planned behavior to exercise behavior: A meta analysis. Journal of Sport & Exercise Psychology, 19(1), 36 51. Hill, R. K., Thompson, J. W., Shaw, J. L., Pinidiya, S. D., & Card Higginson, P. (2009). Self reported health risks linked to health plan cost and age group. American Journal of Preventive Medicine, 36(6), 459 562. Jones, A., III., Norman, C. S., & Wier, B. (2010). Healthy lifestyle as a coping mechanism for role stress in public accounting. Behavioral Research in Accounting, 22(1), 21 41. Kerner, M. S., & Grossman, A. H. (1998). Attitudinal, social, and practical correlates to ﬁtness behavior: A test of the theory of planned behavior. Perceptual and Motor Skills, 87(3), 1139 1154.

Exercise in the Audit Profession

165

Kerse, N., Elley, C. R., Robinson, E., & Arroll, B. (2005). Is physical activity counseling effective for older people? A cluster randomized, controlled trial in primary care. Journal of the American Geriatrics Society, 53(11), 1951 1956. Kimiecik, J. (1992). Predicting vigorous physical activity of corporate employees: Comparing the theories of reasoned action and planned behavior. Journal of Sport & Exercise Psychology, 14, 192 206. Kwan, M. Y. W., Bray, S. R., & Ginis, K. A. M. (2009). Predicting physical activity of ﬁrst year university students: An application of the theory of planned behavior. Journal of American College Health, 58(1), 45 52. Luepker, R. V., Johnson, S. B., Breslow, L., Chobanian, A. V., Davis, C. E., Duling, B. R., Kumanyika, S., Lauer, R. M., Lawson, P., McBride, P. E., Oparil, S., Prineas, R. J., & Washington, R. L. (1996). Physical activity and cardiovascular health. Journal of the American Medical Association, 276(3), 241 246. Nichols, M. (2008). How can employers limit the cost of absenteeism? An international look. Citizen Economist. Available at http://www.citizeneconomists.com/blogs/2008/10/30/ how can employers limit the cost of absenteeism an international look/ Norman, P., Conner, M., & Bell, R. (2000). The theory of planned behaviour and exercise: Evidence for the moderating role of past behaviour. British Journal of Health Psychology, 5, 249 261. Pate, R. R., Pratt, M., Blair, S. N., Haskell, W. L., Macera, C. A., Bouchard, C., Buchner, D., Ettinger, W., Heath, G. W., King, A. C., Kriska, A., Leon, A. S., Marcus, B. H., Morris, J., Paffenbarger, R. S., Patrick, K., Pollock, M. L., Rippe, J. M., Sallis, J., & Wilmore, J. H. (1995). Physical activity and public health: A recommendation from the centers for disease control and prevention and the American College of sports medicine. Journal of the American Medical Association, 273(5), 402 407. PricewaterhouseCoopers (PwC). (2009). Behind the numbers: Medical cost trends for 2010. Available at www.pwc.com/medicalcosts2010 Ryan, R. M., & Frederick, C. (1997). On energy, personality, and health: Subjective vitality as a dynamic reﬂection of well being. Journal of Personality, 65(3), 529 565. Schifter, D. E., & Ajzen, I. (1985). Intention, perceived control, and weight loss: An application of the theory of planned behavior. Journal of Personality and Social Psychology, 49(3), 843 851. Sisko, A., Truffer, C., Smith, S., Keehan, S., Cylus, J., Poisal, J. A., Clemens, M. K., & Lizonitz, J. (2009). Health spending projections through 2018: Recession effects add uncertainty to the outlook. Health Affairs, 28(2), W346 W357. Stewart, K. J., Turner, K. L., Bacher, A. C., DeRegis, J. R., Sung, J., Tayback, M., & Ouyang, P. (2003). Are ﬁtness, activity, and fatness associated with health related quality of life and mood in older persons? Journal of Cardiopulmonary Rehabilitation, 23(2), 115 121. Tessier, S., Vuillemin, A., Bertrais, S., Boini, S., Le Bihan, E., Oppert, J. M., Hercberg, S., Guillemin, F., & Briancon, S. (2007). Association between leisure time physical activity and health related quality of life changes over time. Preventive Medicine, 44(3), 202 208. Theodorakis, Y. (1994). Planned behavior, attitude strength, role identity, and the prediction of exercise behavior. Sport Psychologist, 8(2), 149 165. VanRyn, M., Lytle, L. A., & Kirscht, J. P. (1996). A test of the theory of planned behavior for two health related practices. Journal of Applied Social Psychology, 26(10), 871 883.

166

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

Vuillemin, A., Boini, S., Bertrais, S., Tessier, S., Oppert, J. M., Hercberg, S., Guillemin, F., & Briancon, S. (2005). Leisure time physical activity and health related quality of life. Preventive Medicine, 41(2), 562 569. Warburton, D. E. R., Nicol, C. W., & Bredin, S. S. D. (2006). Health beneﬁts of physical activity: The evidence. Canadian Medical Association Journal, 174(6), 801 809. Wendel Vos, G. C. W., Schuit, A. J., Tijhuis, M. A. R., & Krombout, D. (2004). Leisure time physical activity and health related quality of life: Cross sectional and longitudinal associations. Quality of Life Research, 13(3), 667 677.

APPENDIX Survey participants answered the following items. The items are organized by construct in this appendix, as indicated by the subheadings, which did not appear on the survey. On the survey, all of the items (up to the demographic items) were randomized, and there were three randomized versions. Subjective vitality scale (Ryan & Frederick, 1997) 1. I feel alive and vital. Very Untrue _____ _____ _____ _____ _____ 2. I don’t feel very energetic. a Very Untrue _____ _____ _____ _____ _____ 3. Sometimes I am so alive that I just want to burst. Very Untrue _____ _____ _____ _____ _____ 4. I have energy and spirit. Very Untrue _____ _____ _____ _____ _____ 5. I look forward to each new day. Very Untrue _____ _____ _____ _____ _____ 6. I nearly always feel awake and alert. Very Untrue _____ _____ _____ _____ _____ 7. I feel energized. Very Untrue _____ _____ _____ _____ _____

_____ _______ Very true _____ _______ Very true _____ _______ Very true _____ _______ Very true _____ _______ Very true _____ _______ Very true _____ _______ Very true

Theory of Planned Behavior (Ajzen, 1985) Exercise Behavior 8. In the course of the past month, how often have you exercised for at least 30 contiguous minutes? ___ Never, ___ A few times ___ A number of times, but less than half ___ About half of the days ___ Most days ___ Almost every day ___ Every day

Exercise in the Audit Profession

167

9. In the course of the three months, how often have you exercised for at least 30 contiguous minutes? ___ Never, ___ A few times ___ A number of times, but less than half ___ About half of the days ___ Most days ___ Almost every day ___ Every day 10. In the course of the six months, how often have you exercised for at least 30 contiguous minutes? ___ Never, ___ A few times ___ A number of times, but less than half ___ About half of the days ___ Most days ___ Almost every day ___ Every day 11. In the course of the 12 months, how often have you exercised for at least 30 contiguous minutes? ___ Never, ___ A few times ___ A number of times, but less than half ___ About half of the days ___ Most days ___ Almost every day ___ Every day Attitude toward exercising for at least 30 contiguous minutes per day: For me to exercise for at least 30 contiguous minutes each day is: (place a check mark at one place on the scale) 12. Harmful _____ _____ _____ _____ _____ _____ _____ Beneﬁcial 13. Pleasanta _____ _____ _____ _____ _____ _____ _____ Unpleasant 14. Gooda _____ _____ _____ _____ _____ _____ _____ Bad 15. Worthless _____ _____ _____ _____ _____ _____ _____ Valuable 16. Enjoyablea _____ _____ _____ _____ _____ _____ _____ Unenjoyable Subjective norm toward exercising for at least 30 contiguous minutes each day of the month: 17. Most people who are important to me think thata I should _____ _____ _____ _____ _____ _____ _____ I should not exercise for at least 30 contiguous minutes each day of the month 18. Most people in my life whose opinions I value think that exercising for at least 30 contiguous minutes is: Harmful _____ _____ _____ _____ _____ _____ _____ Beneﬁcial

168

D. KIP HOLDERNESS JR. AND JAMES E. HUNTON

19. Most people about whom I care would want me to exercise for at least 30 contiguous minutes each day of the month: Disagree _____ _____ _____ _____ _____ _____ _____ Agree Totally Not sure Totally Perceived behavioral control over exercise 20. For me to exercise for at least 30 contiguous minutes each day of the month would be: Impossible _____ _____ _____ _____ _____ _____ _____ Possible 21. If I wanted to, I could exercise for at least 30 contiguous minutes each day of the month. Disagree _____ _____ _____ _____ _____ _____ _____ Agree Totally Not sure Totally 22. It is mostly up to me whether I exercise for at least 30 contiguous minutes each day of the month. Disagree _____ _____ _____ _____ _____ _____ _____ Agree Totally Not sure Totally Health assessment 23. I would consider myself to beb: Very _____ _____ _____ _____ _____ _____ _____ Very Underweight Average Overweight 24. How would you assess your overall medical health relative to other people in your age range? Much _____ _____ _____ _____ _____ _____ _____ Much Worse Average Better 25. About how many days of work did you miss during the past calendar year due to sickness or illness?c Please give your best estimate in whole days _____ Demographics: 26. What is your gender? 27. What is your age? 28. What is your position Level? a

_____ _____

Female _____ Staff _____

Male

Senior Manager _____ _____

Partner _____

The scale is reverse coded. For data analysis purposes, ‘‘average’’ was coded as 0, and the overweight responses were coded as 1, 2, and 3 (very overweight) to be consistent with the notion that more overweight is less healthy. One could also argue that being underweight is less healthy, to a certain extent. Only eight participants indicated that they were underweight, all of whom checked the ﬁrst blank line to the left of ‘‘average.’’ To be conservative, these responses were coded as 1 (the same as we coded the ﬁrst blank line to the right of ‘‘average’’). c For data analysis purposes, responses were reverse coded so that more days reﬂect less healthiness. b

AN EXAMINATION OF PERCEPTIONS OF AUDITOR INDEPENDENCE AND FINANCIAL REPORTING QUALITY WHEN FORMER AUDITORS ARE HIRED Brian Daugherty and Denise Dickins ABSTRACT This study examines perceptions of auditor independence (AI) and ﬁnancial reporting quality (FRQ) when former auditors are hired by public companies into accounting oversight positions under differing strengths of corporate governance. Although the Sarbanes–Oxley (SOX) mandate of a one-year cooling-off period for the hiring of former audit engagement team members into accounting oversight positions (e.g., chief ﬁnancial ofﬁcer) may enhance perceptions of AI, it potentially sacriﬁces FRQ by restricting the hiring of candidates most familiar with a particular company’s industry, risks, and controls. The results of this experiment suggest when a company (i) has strong corporate governance and (ii) hires an audit engagement team member without a one-year cooling-off period, stakeholders perceive ﬁnancial statement quality to be highest as compared to all other experimental conditions. Interestingly, we also ﬁnd hiring a former auditor who has not cooled-off one-year results in roughly the same perception of AI as hiring an auditor observing the one-year Advances in Accounting Behavioral Research, Volume 13, 169 194 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013011

169

170

BRIAN DAUGHERTY AND DENISE DICKINS

cooling-off requirement. Collectively, results suggest stakeholders may not perceive a beneﬁt from the cooling-off requirement as independence is not viewed as enhanced and FRQ is viewed as diminished. Requiring disclosure of auditor alumnus hires, in lieu of a mandated cooling-off period, coupled with external measures of companies’ strength of corporate governance may be sufﬁcient to protect AI and FRQ.

INTRODUCTION The stated intention of the Sarbanes–Oxley Act of 2002 (SOX) is ‘‘to protect investors by improving the accuracy and reliability of corporate disclosures made pursuant to the securities laws, and for other purposes’’ (U.S. House of Representatives, 2002, p. 1). This is to be accomplished, in part, by strengthening corporate governance, enhancing audit committee and auditor independence (AI), and increasing ﬁnancial reporting quality (FRQ) through a mixture of SOX regulation and disclosure solutions. The relative costs and beneﬁts of SOX have been extensively debated, with some academics and professionals questioning whether SOX was necessary and whether market forces could have efﬁciently addressed the underlying issues (Aggarwal & Williamson, 2007). DeFond and Francis (2005) contend many of the intended solutions embedded in SOX are unlikely to solve problems in the auditing profession and may, instead, lead to serious unintended negative consequences. Certain SOX regulations, imposed in lieu of prior disclosure requirements, may have diminished audit quality for some registrants (Abbott, Parker, & Peters, 2010); and ﬁnancial reporting may have become more conservatively biased, not less biased, since SOX (Lobo & Zhou, 2006). Aggarwal and Williamson (2007) report no strengthening of the relationship between strong corporate governance and market values, pre- and post-SOX, and posit SOX compliance costs, especially for smaller entities, may outweigh the intended beneﬁts. The possibility exists that stakeholders may not prefer ‘‘one size-ﬁts all’’ regulation over the alternative of company-speciﬁc disclosures to address corporate governance, AI, and FRQ issues. Disclosure allows entities increased ﬂexibility in addressing issues related to AI (e.g., auditor provided consulting services, magnitude of audit fees, or hiring a former auditor), while considering other characteristics of the entity such as size or corporate governance structure. Disclosure solutions allow stakeholders to adjust their

Perceptions of Hiring Former Auditors

171

reactions to companies’ actions, dependent on other company-speciﬁc characteristics. Stakeholders may signal their perceptions of the value relevance of disclosure through proxy statement votes (shareholders), buyhold-sell ratings (equity analysts), modiﬁcation of contractual terms (creditors/lenders), and so forth. Section 206 of SOX states, ‘‘It shall be unlawful for a registered public accounting ﬁrm to perform for an issuer any audit service required by (SOX) if a chief executive ofﬁcer, controller, chief ﬁnancial ofﬁcer, chief accounting ofﬁcer, or any person serving in an equivalent position for the issuer, was employed by that registered independent public accounting ﬁrm and participated in any capacity in the audit of that issuer during the 1-year period preceding the date of the initiation of the audit’’ (U.S. House of Representatives, 2002). Thus, a public company can immediately hire any employee of their audit ﬁrm into an accounting oversight position if they did not participate in the current or prior year’s audit, or a member of a prior engagement team as long as they did not participate in any audit services during the year before the initiation of the current audit. The SOX mandate applies to any member of the external audit engagement team without respect to position within the audit ﬁrm or tenure on the audit engagement. Before SOX, there was no pre-employment ‘‘cooling-off’’ requirement. The results of prior research provide some evidence that AI may be impaired when a company’s chief ﬁnancial ofﬁcer (CFO) is an alumnus of its auditing ﬁrm. Companies with auditor alumni in accounting oversight positions are more likely to receive unqualiﬁed audit opinions (Lennox, 2005).1 Menon and Williams (2004) and Dowdell and Krishnan (2004) use different methodologies to compare earnings management of companies with former audit partner-CFOs to other companies. Both ﬁnd former audit partner-CFOs are associated with an increased likelihood of earnings management and interpret these results as being due to auditor alumnusCFOs exerting inﬂuence on their former colleagues in a manner that reduces audit quality. Although these studies do not isolate their analyses to former auditors working on the company’s prior audit engagement(s), Dowdell and Krishnan (2004) provide some evidence that when the time period between leaving the audit ﬁrm and joining the client company is short, earnings management may be more pronounced. Imhoff’s (1978) results conﬁrm that stakeholders question AI when supervising auditors (partners and managers) are hired by clients within six months of working on the audit engagement, and further ﬁnd stakeholders’ independence concerns are only mitigated after supervising auditors have cooled-off for at least 18 months, a period longer than the current one year SOX mandate.

172

BRIAN DAUGHERTY AND DENISE DICKINS

On the contrary, many companies ﬁnd their auditing ﬁrm to be a source of quality candidates for accounting oversight positions. Approximately 20 percent of auditors leaving CPA ﬁrms accepted positions with a client (Imhoff, 1978) and one-third of CFOs of Fortune 1000 companies had prior work experience with their current auditing ﬁrm (Behn, Carcello, & Hermanson, 1999). Experienced former auditors hired by clients into accounting oversight positions bring with them a unique set of qualiﬁcations and experiences. These individuals have extensive knowledge of the client’s speciﬁc industry, inherent risks, and the design and effectiveness of internal controls over ﬁnancial reporting; and have an existing and established professional relationship with key personnel, including those charged with corporate governance (senior executives, board of directors, audit committees, etc.). Other research suggests there are no adverse effects on FRQ when former auditors are hired. In stark contrast to the ﬁndings of Menon and Williams (2004) and Dowdell and Krishnan (2004), Geiger, North, and O’Connell (2005) found no signiﬁcant differences in earnings management for companies that hired an individual from their auditing ﬁrm into accounting oversight positions, as compared to companies that hired from industry sources, other auditing ﬁrms, or retaining incumbent executives. Furthermore, the pre-SOX market valued the auditor-to-client direct hiring practice more positively than other appointments (unafﬁliated auditor and industry hires) to accounting oversight positions (Geiger, Lennox, & North, 2008). Thus, hiring former auditors directly into accounting oversight roles, without cooling-off, may potentially improve the credibility and reliability of ﬁnancial statements. This apparent difference in ﬁndings may, in part, be due to an unexamined variable, companies’ strength of corporate governance. When companies have strong (weak) corporate governance, directly hiring audit team alumni may increase (reduce) perceptions of AI and FRQ. If true, disclosure solutions may be an appropriate and less costly alternative to the current regulation. In the case of former auditor hiring practices, although the cooling-off mandate may enhance perceptions of AI, it may also prevent companies from making optimal hiring decisions that maximize FRQ. Disclosure would enable stakeholders of companies with varying strengths of corporate governance to evaluate the relative beneﬁts and costs associated with the hiring of former audit engagement team members into accounting oversight positions. In addition to contributing evidence to the ongoing debate over the optimal balance of regulation versus disclosure, the current study

Perceptions of Hiring Former Auditors

173

incrementally contributes to the prior literature on the impact of former auditor hires on perceptions of AI and FRQ by considering two other factors. First, the inﬂuence of strength of corporate governance and, second, the inﬂuence that the timing and experience of former auditor hires have on such perceptions. Since data regarding whether a former auditor hire was also a member of the company’s audit engagement team (including recency of former auditor involvement) are not currently available in the United States, an experimental setting is used to investigate the hypotheses. Consistent with prior literature, results of this study suggest concerns about AI impairment are greatest when companies hire alumni of their audit engagement team into the position of CFO, as compared to hiring auditors from an unafﬁliated ﬁrm (unafﬁliated auditor). However, stakeholders do not appear to perceive a difference in AI when comparing an auditor alumnus immediately hired (current auditor) to an auditor alumnus that has cooled-off for a year (former auditor). The experimental design manipulates a potential CFO candidate as a current auditor, a former auditor, or an unafﬁliated auditor; fully crossed with strong and weak corporate governance environments. As cooling-off does not appear to mitigate perceived impairment of independence (a stated SOX intention), disclosure of current or former auditor hires into accounting oversight positions may be a reasonable substitute for mandatory cooling-off periods.2 Results also suggest stakeholders perceive companies hiring a current auditor to have the highest FRQ. Stakeholders appear to perceive cooling-off as mitigating ﬁnancial reporting beneﬁts accruing from the former team members’ lack of recent familiarity with a company’s industry, risks, and controls. Strong corporate governance appears to enhance this relationship. Companies hiring their current auditor and having strong corporate governance have the highest perceived level of FRQ as compared to all other experimental settings. Companies hiring a current auditor have a signiﬁcantly smaller perceived decrease in AI when corporate governance is strong, relative to when governance is weak. Strength of corporate governance did not inﬂuence perceived AI for companies hiring a former auditor. Collectively, when hiring auditor alumni, the one-year cooling-off requirement may not be optimal. Perceived FRQ is highest when companies hire current auditors, whereas perceived AI is similar for current and former auditors. From a stakeholder perspective, the SOX cooling-off mandate does not appear to achieve its goal of enhancing AI or FRQ, particularly when corporate governance is strong. When stakeholders view a company as being both credible and transparent, disclosure may be an optimal solution to simultaneously maximize perceptions of AI and FRQ.

174

BRIAN DAUGHERTY AND DENISE DICKINS

The remainder of the chapter is organized as follows. The Background and Hypotheses section provides background and discusses the hypotheses. The Methodology section describes the methodology. Results are presented in the Results section, and conclusions and implications are discussed in the Conclusions section.

BACKGROUND AND HYPOTHESES The ﬁrst four title sections of SOX deal extensively with the improvement of the quality of the independent auditor’s work. Certain SOX provisions employ regulation to address these issues (e.g., the ban on the provision of most consulting services by auditors to attest clients), whereas other SOX provisions rely on less costly disclosure solutions [e.g., audit committee disclosure of pre-approved non-audit services (NAS) performed by the independent auditor]. SOX requirements apply to all issuers of registered securities on U.S. stock exchanges. There has been extensive discussion that SOX implementation and compliance have been especially onerous for smaller registrants.3 In 2007, commissioners of the Securities and Exchange Commission (SEC) voted unanimously to ease SOX rules, stating the changes were especially urgent for smaller registrants, and noting the SEC sought to strike a balance between investor protection and arduous record keeping requirements (Gordon, 2007). The Public Company Accounting Oversight Board (PCAOB) adopted a revised standard for audits of internal control over ﬁnancial reporting, noting the revised standard is more scalable to ﬁt the varying size and complexity of different companies (PCAOB, 2007). These actions support the notion that ‘‘one size-ﬁts all’’ regulations may not always be the best solution in matters of enhancing AI, FRQ, and corporate governance.

Regulation Versus Disclosure Legislation banning certain behaviors or policies, as opposed to adopting disclosure strategies, implies that the dissemination of information is not fully efﬁcient. Results of prior studies generally support the notion that the market is not fully or uniformly efﬁcient (e.g., Ball & Brown, 1968; Bhattacharya, Black, Christensen, & Mergenthaler, 2007). Other studies conﬁrm disclosures are indeed value relevant. For example, both Shu (2000) and Whisenant, Sankaraguruswamy, and Raghunandan (2003) found

Perceptions of Hiring Former Auditors

175

signiﬁcant market reactions for disclosures related to auditor changes. In most instances, SOX mandates prescribe regulation over disclosure. However, Abbott et al. (2010) found evidence supporting the efﬁcacy of the market-based disclosure approach related to auditor fees and suggest the SOX regulation banning most NAS may have harmed certain registrants who previously experienced increased audit quality through the receipt of NAS (fees required to be disclosed prior to the SOX ban) by their independent auditor. Currently, public companies are permitted to engage their independent auditors to perform NAS not speciﬁcally prohibited by SOX, subject to preapproval by the company’s audit committee. Such approvals must be disclosed to stakeholders in periodic reports ﬁled with the SEC (U.S. House of Representatives, 2002). Thus the audit committee, in protecting the interests of stakeholders, is charged with determining the likely impact (positive, neutral, or negative) of engaging their independent auditor to perform NAS. In this circumstance, SOX relies on corporate governance oversight and disclosure to increase the likelihood that optimal decisions are made and that such decisions are transparent to the investor.

Hiring Former Auditors: Auditor Independence, Financial Reporting Quality, and Corporate Governance Considerations Auditor Independence and Financial Reporting Quality As previously discussed, one of the regulatory mandates enacted by SOX was a one-year cooling-off between the time that a current auditor works on a company’s audit engagement and the time that auditor may be hired by the client into an accounting oversight position. The intent of this rule is to enhance independence between client management and their auditing ﬁrm. Beasley, Carcello, and Hermanson (2000) suggest audit quality may suffer due to misplaced conﬁdence in the former auditor’s ability or integrity. Similarly, Parlin and Bartlett (1994) suggest the audit team may over rely on the former auditor and fail to exercise sufﬁcient professional skepticism. Auditors’ reputation and litigation risks may not override the potential for these occurrences, and the results of certain studies (previously discussed) support regulators’ concerns that AI may be impaired when a company’s CFO is an auditing ﬁrm alumnus. Social identity theory (Tajfel & Turner, 1985) suggests alumni are likely to identify with their former ﬁrms. If identiﬁcation is strong, the effect of such identiﬁcation may not be in the best interest of stakeholders (Iyer, Bamber, & Bareﬁeld, 1997). Consistent

176

BRIAN DAUGHERTY AND DENISE DICKINS

with social identity theory, together with the results of prior literature, stakeholders are likely concerned about AI when companies hire current and former auditors, into accounting oversight positions. Formally stated: H1. When companies hire current or former auditors, potential stakeholders will consider the audit ﬁrm to be less independent than when companies hire an unafﬁliated auditor. Despite the potential for reduced AI, many publicly traded companies have found their independent audit ﬁrm to be a source of quality candidates for accounting oversight positions due to their familiarity with the company’s speciﬁc industry, inherent risks, and the design and effectiveness of internal controls over ﬁnancial reporting. Prior research has found that former auditors may contribute to enhanced FRQ (Geiger et al., 2005). Although AI is generally proposed as an antecedent to audit quality and FRQ, in the case of current and former auditor hires, tension exists in the optimal balance between enhancing AI and FRQ. Taylor, DeZoort, Munn, and Thomas (2003) advocate a framework where ﬁnancial statement reliability is emphasized over AI. They characterize reliability as resulting in ﬁnancial statements that are credible, dependable, and provide an accurate representation of the ﬁnancial position. Taylor et al. (2003) also demonstrate that independence is an antecedent to ﬁnancial statement reliability; however, they believe other antecedents such as expertise should not be sacriﬁced to enhance independence if reliability will ultimately be reduced. This suggests the SOX mandated cooling-off period may have negative ramiﬁcations. For example, former auditors will not be as familiar with a particular company’s current inherent and control risks. Current practices and intervening personnel changes may also diminish the legacy institutional knowledge held by the former auditor. Additionally, the opportunity cost of waiting one year may make hiring such quality candidates difﬁcult, and a company may be required to employ a less qualiﬁed candidate in the interim if they choose to wait (or do not have the luxury of waiting) one year to hire a former auditor.4 Consistent with prior arguments, such a result would be counter to the fundamental goal of engaging a public auditor in the ﬁrst place (i.e., reliable ﬁnancial statements). On the basis of the earlier discussion, stakeholders are expected to perceive FRQ is highest when companies hire current and former auditors. Formally stated: H2a. When companies hire current and former auditors, potential stakeholders will consider ﬁnancial reporting quality to be higher than when companies hire an unafﬁliated auditor.

Perceptions of Hiring Former Auditors

177

Given the loss of recent institutional knowledge resulting from the oneyear cooling-off period, stakeholders are expected to perceive FRQ is higher when companies hire current auditors than when companies hire former auditors. Formally stated: H2b. When companies hire a current auditor, potential stakeholders will consider ﬁnancial reporting quality to be higher than when companies hire a former auditor. Corporate Governance Recall another common complaint regarding the SOX legislation is, for the most part, it adopts a ‘‘one size-ﬁts all’’ approach. However, companies differ in many aspects. One characteristic crucial to the necessity and effectiveness of SOX mandates is the strength of an entity’s corporate governance structure. Companies signal the strength of corporate governance in various ways including having independent board members, separating the chief executive ofﬁcer and chairman of the board roles, and by being transparent; because doing so reduces agency and contracting costs. Prior studies conﬁrm these actions have a positive effect on companies’ AI, FRQ, and market values. Earnings management is constrained by independent boards (Klein, 2002) and board member ﬁnancial sophistication (Xie, Davidson, & DaDalt, 2003). Fully independent audit committees are associated with a lower cost of debt (Anderson, Mansi, & Reeb, 2004), and earnings are more informative when audit committees and boards are more independent (Chang & Sun, 2009). Companies are less likely to reappoint a current or former auditor’s ﬁrm if audit committees are more independent, suggesting independent audit committees may strengthen AI by deterring inappropriate afﬁliations between auditing ﬁrms and auditor alumnus-CFOs (Lennox & Park, 2007). These ﬁndings support the notion that the potential for reduced AI may be mitigated if certain governance structures are in place. As suggested by the Independence Standards Board (ISB, 2000), the appearance of AI may be enhanced if the audit ﬁrm reports to an independent audit committee (as currently mandated by SOX). Although the variation among public companies in terms of corporate governance has lessened, due in part to the imposition of SOX mandates requiring all members of the audit committee to be independent and ﬁnancially literate (with at least one to also be a ﬁnancial expert), requiring senior management to opine on the effectiveness of their internal control over ﬁnancial reporting, and restricting the provision of NAS, differences still exist. These differences, evidenced by corporate governance ‘‘scores’’

178

BRIAN DAUGHERTY AND DENISE DICKINS

issued by independent agencies, likely affect stakeholders’ perceptions of corporate actions.5 Jennings, Pany, and Reckers (2006) found perceptions of AI were incrementally improved by strengthening corporate governance beyond the SOX mandated minimum levels. Thus, strong corporate governance structures are expected to mitigate stakeholders’ concerns about AI and FRQ. Formally stated: H3a. For companies hiring a current auditor, strong corporate governance will result in a smaller perceived threat to independence and ﬁnancial reporting quality compared to similar companies with weak corporate governance. FRQ is expected to be perceived as greatest when corporate governance is strong and the auditor alumnus-hire has not cooled-off. Such a ﬁnding would provide further evidence that disclosure may be an appropriate substitute for regulation. Recall that a motivation for the current study is an interest in studying previously unexamined variables in the extant literature, namely the inﬂuence that strength of corporate governance may have on AI and FRQ. Formally stated: H3b. Companies hiring a current auditor, coupled with strong corporate governance, will result in the highest perceived ﬁnancial reporting quality as compared to all other conditions (former and unafﬁliated auditors with strong corporate governance, and current, former and unafﬁliated auditors with weak corporate governance).

METHODOLOGY Participants The experimental instrument was administered to 139 participants composed of two sets: (i) a group of 33 practicing ﬁnancial analysts, all with undergraduate degrees in ﬁnance or accounting, and (ii) 106 ﬁnance and accounting major students enrolled in senior-level undergraduate ﬁnance and accounting classes at two different universities.6 Practicing ﬁnancial analysts were solicited through industry sources known to the authors. Neither set of participants received compensation (or credit) for participating in the study. The experimental instruments were administered in single sessions (at the beginning of a class period for students, supervised by the researchers) or distributed to participating analysts (with responses

179

Perceptions of Hiring Former Auditors

Table 1. CFO Candidate Unafﬁliated auditor Current auditor Former auditor

Design of Analyses and Cell Size.

Weak Corporate Governance

Strong Corporate Governance

Cell 1 (n ¼ 23) Cell 3 (n ¼ 22) Cell 5 (n ¼ 25)

Cell 2 (n ¼ 27) Cell 4 (n ¼ 23) Cell 6 (n ¼ 19)

mailed directly to the researchers). Comparing degree of study (ﬁnance or accounting), there were no signiﬁcant differences between the two sets of participants (pW0.10). Furthermore, t-tests of difference of means (not tabulated) suggest no differences between the groups in terms of the study’s variables of interest discussed later (all p-values W0.10). Participants were generally aged 20–29 and 61 percent were male. Participants were randomly assigned to one of the six treatment combinations representing strength of corporate governance (CG-manipulated at two levels), fully crossed with CFO candidate source (manipulated at three levels). The Appendix provides a description of the experimental manipulations. Table 1 presents the six-cell design. The hypotheses are tested using pre-planned comparisons of these cells.

Experimental Instrument Participants were informed that the hypothetical company, a large publicly traded U.S. manufacturer, had been audited by an external auditing ﬁrm for 22 years, consistent with the average auditor tenure of Fortune 1000 companies following the implementation of SOX (Arel, Brody, & Pany, 2005). The manufacturing industry was chosen as it is a common industry that should be familiar to both sets of participants. Participants were provided information on the impending retirement of the current CFO, as well as background information on a potential replacement CFO candidate, and were asked to evaluate the external audit ﬁrm’s independence (AI) and the company’s FRQ assuming the candidate is hired. The position of CFO was used in the experiment as it is a position that SOX speciﬁcally mandates a one-year cooling-off period to avoid impairment of the audit ﬁrm’s independence (U.S. House of Representatives, 2002) and is a position frequently ﬁlled by public registrants with audit ﬁrm alumni (Behn et al., 1999). Audit practitioners, members of the academy, and practicing

180

BRIAN DAUGHERTY AND DENISE DICKINS

ﬁnancial analysts were consulted and provided guidance in developing the instrument.7 The instrument was pre-tested using undergraduate and graduate accounting students as well as executive MBA students at a major state university. CG Manipulations As further described in the Appendix participants were provided background information about Corporate Governance Quotients (CGQs) assigned to public companies by independent rating agencies and were told that CGQ scores are based on various CG characteristics including the composition and independence of the company’s Board of Directors (including the Audit Committee), the company’s relationship with its auditors, and the company’s executive and board member compensation. They were informed CGQs range from 0 (very poor) to 20 (very good), and that the average CGQ score for all publicly traded companies is 15.8 Participants assigned to the weak (strong) setting were informed the company’s CGQ was 5 (19). Those in the weak setting were informed the company’s CG structure just met the minimum requirements of the U.S. stock exchanges and SOX, whereas those in the strong setting were told the company’s CG structure far exceeded such minimums.

Source of CFO Candidate Manipulations As further described in the Appendix CFO candidate source is manipulated in three ways. For all three CFO candidates, participants are informed the potential candidate is a Senior Audit Manager with a registered public accounting ﬁrm in good standing with the PCAOB. The CFO candidate is further described in all scenarios as being a certiﬁed public accountant, well respected by peers and colleagues, having relevant industry experience, and having 10 years of work experience. In the unafﬁliated auditor manipulation, the CFO candidate is described as being the employee of an auditing ﬁrm that has no previous afﬁliation or interaction with the hypothetical public company (allowed pre- and postSOX). In the current auditor manipulation, the CFO candidate is described as being a ﬁve-year member of the current audit engagement team, including the most recently completed audit (allowed pre-SOX, prohibited post-SOX). In the former auditor manipulation, the CFO candidate is described as being a ﬁve-year member of the prior audit engagement team, including the

Perceptions of Hiring Former Auditors

181

audit completed one year ago (allowed pre-SOX and meeting post-SOX cooling-off criteria). Dependent Variables To measure perceptions related to AI, participants are asked to evaluate their view of the independence of the auditor if the CFO candidate is hired. Responses are measured on an 11-point Likert scale, ranging from 5 (‘‘Low Audit Firm Independence’’) to þ5 (‘‘High Audit Firm Independence’’), with a midpoint of 0 (‘‘Moderate Audit Firm Independence’’). To measure perceptions related to FRQ, participants are asked to evaluate their view of the company’s FRQ if the CFO candidate is hired using a similar scale, ranging from 5 (‘‘Low Financial Reporting Quality’’) to þ5 (‘‘High Financial Reporting Quality’’), with a midpoint of 0 (‘‘Moderate Financial Reporting Quality’’). To determine whether participants attended to the experimental materials, two manipulation checks were included. The ﬁrst asked whether the CFO candidate was currently employed by the hypothetical company’s current public accounting ﬁrm or another public accounting ﬁrm. The second asked whether the CFO candidate had (i) never provided services to the company (unafﬁliated auditor), (ii) had provided services in the most recent year audited (current auditor), or (iii) had provided services in the prior year audited (former auditor). Participants failing either of these two manipulation checks (one analyst and eight students) were not included in the analyses that follow.9

RESULTS The impact of source of candidate and strength of CG on AI and FRQ are presented in Table 2. Pre-planned contrasts presented in Table 3 and Table 4 serve as the primary tests of the study’s hypotheses. Data in Table 2, panel A, show stakeholder perceptions of AI vary signiﬁcantly (po0.001) depending on whether the candidate is an unafﬁliated auditor (mean of 2.42) or a current auditor (mean of 1.86); and also vary signiﬁcantly (po0.001) comparing unafﬁliated auditors to former auditors (mean of 1.49). There is no difference in AI (pW0.10) when comparing current auditors and former auditors. Data in Table 2, panel A, show perceptions of FRQ vary dependent on the candidate. Comparing an unafﬁliated auditor (mean of 1.54) and current

182

BRIAN DAUGHERTY AND DENISE DICKINS

Table 2. Impact of Source of Candidate and Corporate Governance on Auditor Independence and Financial Reporting Quality. CFO Candidate

n

Auditor Independencea

t test of Difference

Financial Reporting Qualityb

t test of Difference

Means Panel A: Impact of source of candidate on auditor independence and ﬁnancial reporting quality 2.431 1. Unafﬁliated 50 2.42 8.591 (1 vs. 2) 1.54 (1 vs. 2) auditor 2. Current auditor 44 1.86 0.622 0.27 2.893 (2 vs. 3) (2 vs. 3) 1.82 0.660 3. Former auditor 45 1.49 6.977 (1 vs. 3) (1 vs. 3) Overall 139 0.20 1.23 Corporate governance

n

Auditor independencea

t test of difference

Financial reporting qualityb

t test of difference

Means Panel B: Impact of strength of corporate governance on auditor independence and ﬁnancial reporting quality 0.86 1.819 1. Weak 70 0.97 2.839 (1 vs. 2) (1 vs. 2) 2. Strong 69 0.58 1.61 Overall 139 0.20 1.23 Signiﬁcant at po0.10. Signiﬁcant at po0.05. Signiﬁcant at po0.01. Signiﬁcant at po0.001. a

For auditor independence, participants were asked to evaluate their view of the independence of the auditor if the CFO candidate is hired. Responses were measured on a scale ranging from 5 (‘‘Low Audit Firm Independence’’) to þ5 (‘‘High Audit Firm Independence’’) with a midpoint of 0 (‘‘Moderate Audit Firm Independence’’). b For ﬁnancial reporting quality, participants were asked to evaluate their view of the company’s FRQ if the CFO candidate is hired. Responses were measured on a scale, ranging from 5 (‘‘Low Financial Reporting Quality’’) to þ5 (‘‘High Financial Reporting Quality’’) with a midpoint of 0 (‘‘Moderate Financial Reporting Quality’’).

auditor (mean of 0.27), stakeholder perceptions of FRQ vary signiﬁcantly (po0.05); and comparing the current auditor and former auditor (mean of 1.82), stakeholder perceptions of FRQ vary signiﬁcantly (po0.05). Stakeholders report no difference (pW0.10) comparing the unafﬁliated and former auditors.

183

Perceptions of Hiring Former Auditors

Table 3. CFO Candidate

Contrast Tests for Perceived Auditor Independence. Weak Corporate Governance

Panel A: Mean (standard deviation) Unafﬁliated auditor

Strong Corporate Governance

a

Current auditor Former auditor

1.52 (2.63) 2.36 (2.46) 2.04 (2.42)

Contrast coefﬁcients Panel B: Preplanned contrasts H1: 2 2 1 1 1 1 H3a: 0 0 1 1 0 0

3.19 (1.71) 0.65 (3.49) 1.63 (2.69)

F

p valueb

76.4 4.89

o0.0001 0.014

a Auditor independence was measured on a scale ranging from 5 (‘‘Low Audit Firm Independence’’) to þ5 (‘‘High Audit Firm Independence’’) with a midpoint of 0 (‘‘Moderate Audit Firm Independence’’). b p values are based on directional predictions (with the exception of the test of H1b).

Table 4.

Contrast Tests for Perceived Financial Statement Quality.

CFO Candidate

Weak Corporate Governance

Panel A: Mean (standard deviation)a Unafﬁliated auditor Current auditor Former auditor

Contrast coefﬁcients Panel B: Preplanned contrasts H2a: 2 2 1 1 1 1 H2b: 0 0 1 1 1 1 H3b: 1 1 1 5 1 1

1.43 (2.09) 1.23 (2.20) 0.00 (2.77)

Strong Corporate Governance

1.63 (2.15) 2.39 (1.78) 0.63 (3.17)

F

p valueb

1.24 8.69 6.70

0.14 0.004 0.005

a Financial statement quality was measured on a scale, ranging from 5 (‘‘Low Financial Reporting Quality’’) to þ5 (‘‘High Financial Reporting Quality’’) with a midpoint of 0 (‘‘Moderate Financial Reporting Quality’’). b p values are based on directional predictions.

184

BRIAN DAUGHERTY AND DENISE DICKINS

Table 2, panel B, reports stakeholders’ perceptions of AI and FRQ under varying strengths of CG. As depicted, AI is judged to be signiﬁcantly higher (po0.01) when comparing strong CG (mean of 0.58) to weak CG (mean of 0.97). CG also marginally enhances perceived FRQ (po0.10) when comparing strong CG (mean of 1.61) and weak CG (mean of 0.86). Recall H1 predicts companies hiring current and former auditors as CFO will result in perceptions that the audit ﬁrms are less independent than when an unafﬁliated auditor is hired as CFO. Referring to Table 1, this relationship suggests that cells 1 and 2, on average, are greater than the average independence score in cells 3 through 6. The perceived AI means are provided in Table 3, panel A. AI is perceived as being highest when a newly hired CFO is an unafﬁliated auditor, and CG is strong (mean of 3.19); and AI is perceived as being most impaired when the new hire is a current auditor, and CG is weak (mean of 2.36). Planned contrasts provided in Table 3 Panel B conﬁrm H1. AI is highest when former auditors hired into accounting oversight positions are unafﬁliated auditors (F ¼ 76.4; directional p-value o0.001). This ﬁnding suggests perceived AI may be maximized when former audit engagement team members are not hired, although such a regulatory solution is not practical, may violate various state and federal labor regulations, and would ostensibly limit companies’ ability to hire the most competent personnel into accounting oversight positions. H2a and H2b consider how FRQ perceptions are inﬂuenced by the hiring of current, former, and unafﬁliated auditors. Recall H2a predicts stakeholders will perceive FRQ to be higher when companies hire current and former auditors (versus an unafﬁliated auditor). Referring back to Table 1, these expectations suggest that cells 1 and 2 are, on average, smaller than the average of cells 3 through 6. As reported in Table 4, panel B, this contrast is not signiﬁcant (F ¼ 1.24; p-value ¼ 0.14), and H2a is rejected. An examination of the cell means suggests the primary reason is that companies hiring former auditors are perceived as having lower FRQ. Focusing on the FRQ difference between companies hiring former auditors compared to hiring current auditors (H2b), we expect cells 3þ4 W 5þ6. Table 4, panel B, provides the results of this contrast, which is signiﬁcant (F ¼ 8.69; p-value ¼ 0.004), in support of H2b. These results suggest that when companies hire a current auditor, FRQ is perceived to be stronger than for companies hiring a former auditor. H3a and H3b consider how a company’s CG structure inﬂuences perceptions of AI and FRQ. Recall, perceived independence risk is reduced with strong CG (Jennings et al., 2006). Thus, for companies hiring a current

Perceptions of Hiring Former Auditors

185

auditor, strong CG is expected to enhance perceived AI relative to similar companies with weak CG. This expectation suggests that cell 4W3. This contrast is reported in Table 3, panel B (F ¼ 4.89; p-value ¼ 0.014), suggesting strong CG offsets the negative impact on AI of hiring a current auditor, and conﬁrming H3a. These ﬁndings are particularly relevant in light of the ﬁnal prediction, H3b, that companies with strong CG hiring a current auditor provide the highest FRQ. This contrast suggests that cell 4 is greater than the other ﬁve cells [cell 4W(1 þ 2 þ 3 þ 5 þ 6)/5]. As depicted in panel B of Table 4, this contrast is signiﬁcant (F ¼ 6.70; p-value ¼ 0.005), providing support for H3b. Cell 4 is also signiﬁcantly greater than all other cells when compared individually (all p-values r0.05, not tabulated).

Additional Analyses A number of additional post-experimental questions are included in the experimental materials. First, participants were asked, Indicate your degree of familiarity with the one year cooling off period requirement when publicly traded companies hire members of their external audit engagement team into accounting oversight positions.

This question was included to determine whether participants’ responses were inﬂuenced based on their depth of knowledge of the SOX requirement. Responses were measured on an 11-point Likert scale ranging from 5 (‘‘Not at all Familiar’’) to þ5 (‘‘Very Familiar’’), with a midpoint of 0 (‘‘Somewhat Familiar’’). Responses ranged from 5 to þ5 and averaged 0.68, a result signiﬁcantly different from 0 (t ¼ 2.30, p ¼ 0.02), suggesting a general lack of familiarity with the current cooling-off mandate. Using participants’ responses to this question in an ANCOVA (not tabulated) does not alter the study’s results. Next, participants were asked to explicitly report their perceptions about the inﬂuence of the strength of CG on AI and FRQ. They were asked two separate questions as follows: Indicate your belief as to the degree of inﬂuence that the strength of corporate governance has on the independence of a publicly traded company’s external auditor. Indicate your belief as to the degree of inﬂuence that the strength of corporate governance has on the ﬁnancial reporting quality of a publicly traded company.

186

BRIAN DAUGHERTY AND DENISE DICKINS

Responses were measured on an 11-point Likert scale ranging from 5 (‘‘Not at All Inﬂuential’’) to þ5 (‘‘Very Inﬂuential’’), with a midpoint of 0 (‘‘Somewhat Inﬂuential’’). Responses ranged from 5 to þ5 and averaged 1.73 for AI and 2.15 for FRQ, results signiﬁcantly different from 0 (t ¼ 9.74, po0.001, and t ¼ 11.62, po0.001, respectively), suggesting participants generally believe CG strength inﬂuences both AI and FRQ. Participants were also explicitly asked to report their opinions about optimal cooling-off periods. They were asked, Currently, publicly traded companies in the U.S. are prohibited from hiring members of their independent audit engagement team into accounting oversight positions if the member participated in any capacity on the audit during the preceding one year period. Please indicate your preference related to the duration, if any, of the waiting period for hiring members of the independent audit engagement team into accounting oversight positions.

Responses were measured on an 11-point Likert scale ranging from 5 (‘‘No waiting period is preferable’’) to þ5 (‘‘A longer waiting period is preferred’’), with a midpoint of 0 labeled as (‘‘One year waiting period is preferable’’). Responses ranged from 5 to þ5 and averaged 1.14, results signiﬁcantly different from 0 (t ¼ 5.92, po0.001), suggesting that a coolingoff period of longer than one year would be preferred by potential stakeholders. Participants were also asked their views on whether companies’ hiring of former audit engagement team members should be required to be disclosed as follows: Companies that hire former engagement team members who have observed the one year waiting period are not required to disclose the hiring or the former afﬁliation with the independent audit ﬁrm. Please indicate your preference related to disclosure of hiring members of the independent audit engagement team into accounting oversight positions.

Responses were measured on an 11-point Likert scale ranging from 5 (‘‘Disclosure of hiring members of an audit engagement team should not be required’’) to þ5 (‘‘Disclosure of hiring of members of an audit engagement team should be required’’), with a midpoint of 0 labeled as (‘‘Indifferent’’). Responses ranged from 5 to þ5 and averaged 2.82, results signiﬁcantly different from 0 (t ¼ 14.78, po0.001), suggesting disclosure of the hiring of a company’s former audit engagement team member is considered useful to potential stakeholders.

Perceptions of Hiring Former Auditors

187

CONCLUSIONS Enhancing AI was the rationale for instituting the SOX-imposed one-year cooling-off period when companies hire members of their current auditing ﬁrm. However, the experimental results suggest when a company hires a current auditor into an accounting oversight position, the presence of strong corporate governance signiﬁcantly reduces any of the negative effects on perceived AI associated with the current auditor’s prior afﬁliation with the company’s external auditor. Furthermore, results suggest perceptions of FRQ are maximized by hiring a current auditor (versus an unafﬁliated auditor or former auditor), presumably due to the current auditor’s recent institutional knowledge of the client. These results suggest the SOX-mandated cooling-off requirement may unintentionally reduce FRQ by potentially eliminating the most qualiﬁed candidates for accounting oversight positions. AI is perceived as being higher when unafﬁliated auditors are hired into accounting oversight positions, as compared to hiring current auditors or former auditors, but we observe no difference in perceptions of AI when comparing the hire of a current auditor or former auditor. This ﬁnding suggests a requirement more stringent than the current one-year cooling-off period may be perceived as having a greater impact on stakeholder perceptions of AI; however, such a requirement would be arduous, perhaps untenable, and in light of our results regarding the relationship between CFO candidate and FRQ, is likely a sub-optimal solution. Future research could consider the effect of increases or decreases in the cooling-off period, including whether potential stakeholders perceive value in differing coolingoff periods dependent on the auditor’s tenure and level within the public accounting ﬁrm. This research, in part, strives to examine potential stakeholder views on regulation and disclosure related to hiring former auditors into accounting oversight positions at publicly traded companies. We do this by describing the potential CFO candidate’s background in the experimental instrument. If disclosure of the hiring of former auditors into accounting oversight positions were to be adopted, a practical method to accomplish this would be Form 8-K ﬁlings by registrants with the SEC.10 The experimental instrument did not include a hypothetical 8-K ﬁling in an attempt to avoid confounds resulting from possible perceived signaling resulting from such an overt disclosure. As discussed earlier, participants indicated a preference for both a cooling-off period longer than the current one-year requirement (regulatory

188

BRIAN DAUGHERTY AND DENISE DICKINS

solution) and a preference that the hiring of former auditors be publicly disclosed (disclosure solution). Although our experimental instrument does not directly manipulate the effects of disclosure to avoid possible confounds, these reported preferences suggest further research about the costs and beneﬁts of regulatory versus disclosure solutions may be informative. Relatively costless disclosure may be an efﬁcient and effective alternative to the current regulatory mandate. Alternatively, disclosure may be an enhancement to the current cooling-off mandate. This study makes several contributions to the discussion regarding the appropriateness of existing regulations, potential modiﬁcations of such regulations, and potential disclosure solutions in auditing. First, it provides indications for the debate regarding the beneﬁts of regulation versus disclosure, although additional investigation is warranted. In this context, knowledge of the CFO candidate’s background appears to achieve similar stakeholder perceptions of AI as does the current, and potentially more costly, regulatory requirement. Second, this study provides preliminary evidence that such requirements may need to be sensitive to other characteristics of the company (i.e., ‘‘one size-ﬁts all’’ may not be the best solution). Companies with strong CG may be better served through increased transparency. Strength of CG has generally not been considered in prior research examining the impact of AI on FRQ and may, in part, provide an explanation of why some prior studies have drawn inconsistent conclusions. Third, the ﬁndings are relevant to practitioners, standard-setters, and regulators (including the international community) contemplating changes to audit professionals’ acceptance of employment with publicly traded clients. Finally, this study is consistent with the views of some that standard setters and regulators should be careful not to sacriﬁce ﬁnancial statement reliability when striving to maximize AI (e.g., Taylor et al., 2003). The results suggest that, with regard to the SOX-imposed one-year cooling-off requirement, FRQ may be sacriﬁced in an attempt to maximize perceptions of AI.

NOTES 1. This ﬁnding could be interpreted as indicating independence impairment of the audit ﬁrm. Conversely, it may be that auditor alumni represent the most highly qualiﬁed personnel for employment into accounting oversight positions and, by extension, better able to produce ﬁnancial statements conforming to generally accepted accounting principles.

Perceptions of Hiring Former Auditors

189

2. AU 220.03 states, in part, ‘‘Independent auditors should not only be independent in fact; they should avoid situations that may lead outsiders to doubt their independence’’ (PCAOB, 2009). Thus, auditor independence has two components: in fact and in appearance (i.e., perceived independence). Although SOX addresses auditor independence in fact (a non measurable construct), the experiment employed in this study measures market participants’ perceptions related to the appearance of independence. 3. The most vocal discussion surrounds the Section 404 internal control compliance requirements. Accelerated ﬁlers (the largest of publicly traded companies) were required to comply with the new internal control disclosures for ﬁscal years ending on or after November 15, 2004, with others required to comply shortly thereafter (PCAOB, 2004). The SEC granted extension relief a number of times to accelerated ﬁlers with less than $700 million in market capitalization and to non accelerated ﬁlers (SEC, 2005). In November 2009, the House Financial Services Committee narrowly approved the Garrett Adler amendment to the proposed Investor Protection Act to provide non accelerated ﬁlers with permanent relief from Section 404 compliance. As of March 2010, opposition to the amendment continues. 4. Anecdotal evidence suggests former auditors may continue to be a viable source for ﬁlling accounting oversight positions in the United States. Consider the hiring policies of Robert Half International Inc. regarding the hiring of former employees of the company’s outside auditors (after satisfying minimum SOX and company imposed cooling off periods). ‘‘No former outside auditor professional y may be hired in an Assistant Controller or higher ﬁnancial reporting oversight role without the advance approval of the Audit Committee’’ and ‘‘No former outside auditor professional y may be hired in any other accounting role or ﬁnancial reporting oversight role without the advance approval of the Chief Financial Ofﬁcer’’ (Robert Half International, 2009). A number of similar hiring policies for public companies were noted during an internet search. 5. One of the more common corporate governance ratings is published by RiskMetrics Group (riskmetrics.com/cgq). As advertised, ‘‘RiskMetrics Group’s Corporate Governance Quotient (CGQs) measures the strengths, deﬁciencies and overall quality of a company’s corporate governance practices and board of directors. CGQ is a reliable tool to help analysts, portfolio managers, research directors and others understand a company’s corporate governance structure and practices.’’ CGQs are developed based on corporate attributes in four key areas: Board Structure, Anti Takeover Defenses, Audit, and Compensation/Ownership. See http://www.riskmetrics.com/cgq. 6. The career services center at one of the participating universities conﬁrmed that employers seeking entry level ﬁnancial analysts all speciﬁed an undergraduate degree in accounting or ﬁnance (or dual major) as a prerequisite. Anecdotally, a number of the student participants reported having accepted positions as ﬁnancial analysts. 7. Consulted analysts did not participate in the experiment. 8. Different independent rating agencies may use different scales in evaluating the relative strength of an entity’s corporate governance structure. We use 0 to 20 in our hypothetical range of CGQs. 9. Results presented in the Results section are unchanged when these nine observations are included.

190

BRIAN DAUGHERTY AND DENISE DICKINS

10. Similar to the current Form 8 K ﬁling requirements related to auditor dismissals and resignations, the audit ﬁrm would be required to indicate its concurrence or disagreement with the client’s characterization in the 8 K of the newly hired employee’s background, ﬁrm tenure, engagement tenure, etc.

ACKNOWLEDGMENTS This manuscript has beneﬁted from the comments of Patricia Arnold, Leslie Kren, and John Lere (University of Wisconsin – Milwaukee), Bill Hillison (Florida State University), John Reisch (East Carolina University), and Wayne Tervo (Murray State University). We are especially grateful to Rick Hatﬁeld (University of Alabama), for his substantial commentary and input on design of the analyses. We also thank Dan Neely (University of Wisconsin – Milwaukee) and Steve Platau (University of Tampa) for assistance in administering the research instrument.

REFERENCES Abbott, L. J., Parker, S., & Peters, G. F. (2010). Agency based demand for audit quality: The impact of disclosure imposed agency costs on non audit services. Working paper. University of Wisconsin Milwaukee, Milwaukee, WI. Aggarwal, R., & Williamson, R. (2007). Did new regulations target the relevant corporate governance attributes? Working paper. Georgetown University. Available at http:// papers.ssrn.com/sol3/papers. cfm?abstract_id ¼ 891411 Anderson, R. C., Mansi, S. A., & Reeb, D. M. (2004). Board characteristics, accounting report integrity, and the cost of debt. Journal of Accounting and Economics, 37(3), 315 342. Arel, B., Brody, R. G., & Pany, K. (2005). Audit ﬁrm rotation and audit quality. The CPA Journal, 75(1), 36 39. Ball, R., & Brown, P. (1968). An empirical evaluation of accounting income numbers. Journal of Accounting Research, 6, 159 178. Beasley, M. S., Carcello, J. V., & Hermanson, D. (2000). Should you offer your job to an external auditor? The Journal of Corporate Accounting and Finance, 11(4), 35 42. Behn, B. K., Carcello, J. V., & Hermanson, D. R. (1999). Client satisfaction and big 6 audit fees. Contemporary Accounting Research, 16(4), 587 608. Bhattacharya, N., Black, E. L., Christensen, T. E., & Mergenthaler, R. D. (2007). Who trades on pro forma earnings information? The Accounting Review, 82(3), 581 619. Chang, J. C., & Sun, H. L. (2009). Crossed listed foreign ﬁrms’ earnings informativeness, earnings management and disclosures of corporate governance information under SOX. The International Journal of Accounting, 44(1), 1 32. Defond, M. L., & Francis, J. R. (2005). Audit research after Sarbanes Oxley. Auditing: A Journal of Practice and Theory, 24(Suppl.), 5 30.

Perceptions of Hiring Former Auditors

191

Dowdell, T. D., & Krishnan, J. (2004). Former audit ﬁrm personnel as CFOs: Effect on earnings management. Canadian Accounting Perspectives, 3(1), 117 142. Geiger, M. A., North, D. S., & O’Connell, B. T. (2005). The auditor to client revolving door and earnings management. Journal of Accounting, Auditing and Finance, 20(1), 1 26. Geiger, M. A., Lennox, C. S., & North, D. S. (2008). The hiring of accounting and ﬁnance ofﬁcers from audit ﬁrms: How did the market react? Review of Accounting Studies, 13(1), 55 86. Gordon, M. (2007). SEC approves framework that would ease SOX rules. Associated Press, April 5. Imhoff, E. A., Jr. (1978). Employment effects on auditor independence. The Accounting Review, 53(4), 869 881. Independence Standards Board (ISB). (2000). Statement of independence concepts: A conceptual framework for auditor independence. Exposure Draft. New York: Independence Standards Board. Iyer, V. M., Bamber, E. M., & Bareﬁeld, R. M. (1997). Identiﬁcation of accounting ﬁrm alumni with their former ﬁrm: Antecedents and outcomes. Accounting, Organizations and Society, 22(3 4), 315 336. Jennings, M. M., Pany, K. J., & Reckers, P. M. J. (2006). Strong corporate governance and audit ﬁrm rotation: Effects on judges’ independence perceptions and litigation judgments. Accounting Horizons, 20(3), 253 270. Klein, A. (2002). Audit committee, board of director characteristics, and earnings management. Journal of Accounting and Economics, 33(3), 375 400. Lennox, C. S. (2005). Audit quality and executive ofﬁcers’ afﬁliations with CPA ﬁrms. Journal of Accounting and Economics, 39(2), 201 231. Lennox, C. S., & Park, C. W. (2007). Audit ﬁrm appointments, audit ﬁrm alumni, and audit committee independence. Contemporary Accounting Research, 24(1), 235 258. Lobo, G. J., & Zhou, J. (2006). Did conservatism in ﬁnancial reporting increase after the Sarbanes Oxley act? initial evidence. Accounting Horizons, 20(1), 57 73. Menon, K., & Williams, D. D. (2004). Former audit partners and abnormal accruals former audit partners and abnormal accruals. The Accounting Review, 79(4), 1095 1118. Parlin, J. C., & Bartlett, R. W. (1994). Prior employment effects and independence in fact. Business and Professional Ethics Journal, 13(1 2), 185 202. Public Company Accounting Oversight Board (PCAOB). (2004). Auditing Standard No. 2. An Audit of internal control over ﬁnancial reporting performed in conjunction with an audit of ﬁnancial statements. Available at http://www.pcaobus.org (retrieved on March 9). Public Company Accounting Oversight Board (PCAOB). (2007). Board approves new audit standard for internal control over ﬁnancial reporting. Press Release, May 24. Public Company Accounting Oversight Board (PCAOB). (2009). PCAOB interim auditing standards (AU). Available at http://www.pcaobus.org/Standards/Interim_Standards/ Auditing_Standards/index.aspx Robert Half International Inc. (2009). Hiring policies regarding outside auditors. Available at http://www.rhi.com/External_Sites/content/RHI/rhi us/Shared/downloads/Hiring AuditorEmployees.pdf Securities and Exchange Commission (SEC). (2005). Final rule: Release No. 33 8545. Management’s Report on Internal Control over Financial Reporting and Certiﬁcation of Disclosure in Exchange Act Periodic Reports of Non Accelerated Filers and Foreign Private Issuers. March 2nd.

192

BRIAN DAUGHERTY AND DENISE DICKINS

Shu, S. Z. (2000). Auditor resignations: Clientele effects and legal liability. Journal of Accounting and Economics, 29(2), 173 205. Tajfel, H., & Turner, J. C. (1985). The social identity theory of intergroup behavior. In: S. Worchel & W. G. Austin (Eds), Psychology of Intergroup Relations (l, pp. 7 24). Chicago, IL: Nelson Hall. Taylor, M. H., DeZoort, F. T., Munn, E., & Thomas, M. W. (2003). A proposed framework emphasizing auditor reliability over auditor independence. Accounting Horizons, 17(3), 257 266. U.S. House of Representatives, Committee on Financial Services. (2002). Sarbanes Oxley act of 2002. Public law no (pp. 107 204). Washington, DC: Government Printing Ofﬁce. Whisenant, S., Sankaraguruswamy, S., & Raghunandan, K. (2003). Market reactions to the disclosure of reportable events. Auditing: A Journal of Practice and Theory, 22(1), 181 194. Xie, B., Davidson,, W. N.,, III., & DaDalt, P. J. (2003). Earnings management and corporate governance: The role of the board and the audit committee. Journal of Corporate Finance, 9(3), 295 316.

APPENDIX. CFO CANDIDATE MANIPULATIONS Current Auditor This candidate is a Senior Audit Manager with Cooke, Poole, and Adams (ABC’s auditor). This candidate has been the audit manager on the ABC audit for the past ﬁve years, including the just completed audit. Cooke, Poole, and Adams is a registered public accounting ﬁrm in good standing with the PCAOB. This candidate is a certiﬁed public accountant, is well respected by peers and colleagues, and has 10 years of work experience. The candidate has relevant manufacturing experience and has current experience working with ABC’s complicated accounting, internal control, and ﬁnancial reporting systems for the past ﬁve years (including the just completed audit). Thus, the candidate has a current understanding of the unique aspects of these environments. The candidate has current experience with ABC’s personnel and those charged with corporate governance oversight, including members of the Board of Directors and the Audit Committee. Former Auditor This candidate is a Senior Audit Manager with Cooke, Poole, and Adams (ABC’s auditor). This candidate had been the audit manager on the ABC audit for ﬁve years, including the audit completed one year ago. The

Perceptions of Hiring Former Auditors

193

candidate was not involved in the just completed audit. Cooke, Poole, and Adams is a registered public accounting ﬁrm in good standing with the PCAOB. This candidate is a certiﬁed public accountant, is well respected by peers and colleagues, and has 10 years of work experience. The candidate has relevant manufacturing experience and has experience from just over one year ago working with ABC Company’s complicated accounting, internal control, and ﬁnancial reporting systems for ﬁve years (including the audit completed one year ago). Thus, the candidate has a fairly recent understanding of the unique aspects of these environments. The candidate has experience from just over one year ago with ABC’s personnel and those charged with corporate governance oversight, including members of the Board of Directors and the Audit Committee.

Unafﬁliated Auditor This candidate is a Senior Audit Manager from another public accounting ﬁrm. The Senior Audit Manager’s public accounting ﬁrm has never provided any services to ABC Company. This ﬁrm is a registered public accounting ﬁrm in good standing with the PCAOB. This candidate is a certiﬁed public accountant, is well respected by peers and colleagues, and has 10 years of work experience. The candidate has relevant manufacturing experience but does not have any speciﬁc experience with ABC Company’s complicated accounting, internal control, or ﬁnancial reporting systems. Thus, the candidate does not have a current understanding of the unique aspects of these environments. The candidate does not have any speciﬁc experience with ABC’s personnel or those charged with corporate governance oversight, including members of the Board of Directors and the Audit Committee.

CORPORATE GOVERNANCE MANIPULATIONS Common to Both Manipulations Public companies are assigned CGQ by independent rating agencies. CGQs are based on a number of governance factors, including the composition and independence of the company’s Board of Directors (including the

194

BRIAN DAUGHERTY AND DENISE DICKINS

Audit Committee), the company’s relationship with its auditors, and the company’s executive and board member compensation. Strong Corporate Governance CGQs may range from 0 (very poor) to 20 (very good). The average company CGQ for all publicly traded companies is 15. ABC Company’s CGQ is 19, suggesting that ABC Company’s corporate governance structure far exceeds the minimum requirements mandated by the U.S. stock exchanges and the SOX. Weak Corporate Governance CGQs may range from 0 (very poor) to 20 (very good). The average company CGQ for all publicly traded companies is 15. ABC Company’s CGQ is 5, suggesting that ABC Company’s corporate governance structure just meets the minimum requirements mandated by the U.S. stock exchanges and the SOX.

SELF-PROMOTE OR NOT? AN EXAMINATION OF THE EFFECT OF MANAGEMENT’S SELF-PROMOTION IN MANAGEMENT DISCLOSURES Wei Li ABSTRACT This study provides experimental evidence on whether and how management’s use of self-promotion, as a type of proactive impression management strategies in its disclosures, inﬂuences nonprofessional investors’ judgments and decisions. The results show that management’s use of self-promotion inﬂuences nonprofessional investors so that investors (1) expect management’s future performance to be better and (2) are likely to invest more in the company. These positive effects are more prominent when management’s credibility is perceived to be high than when it is low. The ﬁndings of this study provide implications for both practice and research.

Advances in Accounting Behavioral Research, Volume 13, 195 218 Copyright r 2010 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 1475 1488/doi:10.1108/S1475 1488(2010)0000013012

195

196

WEI LI

INTRODUCTION Imagine that a nonprofessional investor is reading a company’s annual report. After analyzing its past operation results and predicting its future performance, the company concludes, ‘‘we have the people and the skills in place to widen our strong competitive advantage, and we think it would be very difﬁcult, if not impossible, for any other company to replicate our overall capabilities in the foreseeable future y We are extremely well positioned to enhance our capabilities and build momentum as the leading player in the competitive markets through the world, and we look forward to delivering continued strong results to our shareholders.’’ On the basis of these statements, will the investor form an impression that the company’s management is very conﬁdent about its capabilities and thus predict that management will manage the company very well in the near future? Or will the investor suspect that management is just bragging and thus ignore these statements? Will the investor’s impression and inferences inﬂuence his/her decisions to invest in this company? These statements come from Enron’s 1998 annual report and illustrate one impression management strategy that the management of public companies may use in their disclosures: self-promotion. To directly maintain and enhance its image of competence, management can strategically use these self-promotion statements to explicitly self-describe its prior achievements and honors, advertise its competence or other strengths, avoid weaknesses, and exude conﬁdence about itself (Ogden & Clark, 2005; Jones, 1964; Godfrey, Jones, & Lord, 1986). Whereas prior research (e.g., Staw, McKechnie, & Puffer, 1983; Barton & Mercer, 2005; Kaplan, Pourciau, & Reckers, 1990; Sanders & Coelho, 2001) has provided evidence that investors respond to reactive impression management strategies that management uses in its disclosures to account for its past performance/ events, none of them has examined whether and how self-promotion, as a proactive impression management strategy (i.e., a strategy initiated by management to proactively establish a particular identity instead of being merely reactive to past performance), is effective. Thus, with the above unanswered questions, this study attempts to examine whether and how management’s use of self-promotion in its disclosures inﬂuences nonprofessional investors’ judgments and decisions. On the one hand, archival accounting research suggests, but does not directly test, that positive statements in management disclosures are irrelevant because they are just sugar-coated without real information value (Abrahamson & Amir, 1996), or that investors tend to discount transparent

Self Promote or Not?

197

self-promotional behavior in management disclosures because they recognize that management sometimes exaggerates its achievements and capabilities (Aerts, 2005). Thus, management’s self promotion in its disclosures may not inﬂuence investors’ judgments and decisions. On the other hand, psychology research and experimental accounting research show that the mere presence of information, regardless of its relevance to people’s judgments or decisions at hand, can have unintentional effects on the information processing of people who lack expertise, like nonprofessional investors (e.g., Frederickson & Miller, 2004; Tversky & Kahneman, 1974; Nisbett, Zukier, & Lemley, 1984). Drawing on this literature, the mere presence of self-promotion statements in management disclosures may possibly affect nonprofessional investors’ judgments and decisions. Speciﬁcally, based on prior research in psychology and management (e.g., Ferris & Judge, 1991; Kristof-Brown, Barrick, & Franke, 2002; Ellis, West, Ryan, & DeShon, 2002), this study posits that management’s use of selfpromotion in its disclosures will lead investors to make favorable judgments and decisions, and such a favorable effect is more prominent when management’s credibility is perceived to be high than when it is low. Nonprofessional investors are examined in this study because they are a signiﬁcant group in the current market, with 33.8 million individuals investing directly in the stock market (New York Stock Exchange [NYSE], 2002) and owning nearly 34% of all shares outstanding (Bogle, 2005). However, these investors lack well-developed valuation models or investment expertise to properly acquire and integrate ﬁnancial and nonﬁnancial information into their judgments and decisions (Frederickson & Miller, 2004). In addition, nonprofessional investors may not have access to investment-related information through other sources than management or analysts (Clarkson, Kao, & Richardson, 1994). Thus, information revealed in management disclosures is likely to inﬂuence these investors’ decisions. As a result, examining whether and how impression management used in management disclosures inﬂuences nonprofessional investors’ judgments and decisions is important. To test the predictions, this study uses a 2 2 between-subject experiment that varies (1) the use of self-promotion in management disclosures (used vs. not used) and (2) perceived credibility of management (high vs. low). An experimental approach is used for at least two reasons. First, while management disclosures usually contain multiple components (Barton & Mercer, 2005), using an experimental method can help separate these components and provide better control over extraneous variables (Krische, 2005), so that the effect of self-promotion statements can be more clearly

198

WEI LI

studied. Second, because virtually all ﬁrms use self-promotion statements in their disclosures, using an experimental method can help examine the effect of self-promotion in ‘‘what if’’ scenarios (Maines, 1994). That is, what is the effect if management uses or does not use self-promotion in its disclosures? In sum, an experimental method enables examination of issues that are difﬁcult to examine when archival data are not available (Libby, Bloomﬁeld, & Nelson, 2002). The results of the experiment show that management’s self-promotion in its disclosures inﬂuences nonprofessional investors’ judgments and decisions. Speciﬁcally, when management uses self-promotion statements in its disclosure, investors (1) expect management’s future performance to be better and (2) are likely to invest more in the company. These favorable effects are more prominent when management is perceived to have high credibility than when it is perceived to have low credibility. This study contributes to both practice and research. At the practice level, the results provide guidance to management about whether and how nonprofessional investors respond to its self-promotion statements. The ﬁnding indicates that if management is perceived to be credible and uses self-promotion statements, investors will make more favorable judgments and decisions relative to the company. This ﬁnding can be beneﬁcial for management in deciding whether, when, and how to promote itself in its disclosures to obtain the desired image, given that management disclosures (i.e., annual reports) have evolved from ﬁnancially driven documents to the ones used to construct a corporate image (Lee, 1994). In addition, the study answers the call from the Securities and Exchange Commission (SEC) to improve the usefulness of management disclosures. Speciﬁcally, it provides insights into how the usefulness of management disclosures can be improved for nonprofessional investors. This study also makes several contributions to research. First, the results shed some light on the research on narratives in management disclosures. Prior research on management narratives (e.g., Abrahamson & Amir, 1996; Aerts, 2005) suggests that it is not worthwhile to study the effect of positive narratives in management disclosures because positive narratives are meaningless and irrelevant. However, this study, together with the concurrent archival work by Davis, Piger, and Sedor (2007), provides direct evidence that sugar-coated narratives can affect investors’ judgments and decisions, even though such information seems irrelevant. Such evidence may (1) help explain why management, who is always criticized for providing sugar-coated information, continues to provide such ‘‘meaningless’’ positive narratives in its disclosures and (2) indicate a need for researchers to consider

Self Promote or Not?

199

the validity of their assumption about sugar-coated information in future studies. Second, the current study complements research on impression management in management disclosures. Whereas prior studies (e.g., Staw et al., 1983; Barton & Mercer, 2005; Kaplan et al., 1990; Sanders & Coelho, 2001) have provided evidence on whether impression management strategies in management disclosures are effective, these studies have largely focused on reactive strategies that management uses to account for its past performance outcomes. This study, along with Siegel and Brockner (2005), examines the effects of impression management in a more proactive manner and indicates that investors also respond to a more proactive effort of management to manage its image. Third, the results extend research on self-promotion in two ways. While prior research on self-promotion has largely focused on the effect of this strategy at the individual level within an organization (i.e., mostly on human resource decisions), this study examines self-promotion at the organizational level and shows that self-promotion can also have signiﬁcant organizational-level consequences. In addition, while prior research suggests that a self-promoter’s perceived credibility is important in the effectiveness of his/her self-promotion (Aerts, 2005; Gardner & Cleavenger, 1998; Godfrey et al., 1986; Jones & Pittman, 1982), none of them have empirically examined this issue. This study provides empirical evidence on the importance of management credibility in the effectiveness of management’s self-promotion tactic. This evidence conﬁrms that more credible managers are better able to communicate information to the capital markets (e.g., Hirst, Koonce, & Miller, 1999; Hirst, Koonce, & Venkataraman, 2007; Hodge, Hopkins, & Pratt, 2006; Mercer, 2004, 2005; Williams, 1996). The remainder of this study proceeds as follows. The next section reviews the relevant research and develops the study’s hypotheses. Then the section following describes the research method and the next section reports the results of the study. The ﬁnal section summarizes and concludes.

BACKGROUND AND HYPOTHESES DEVELOPMENT Impression Management in Management Disclosures To compete for investor support, management seeks to maintain or enhance its image of in-control or competence (e.g., Kaplan et al., 1990; Salancik & Meindl, 1984). As a result, management is motivated to use impression management to manage its image to investors. Impression management refers to strategies that people use to create desired social images or

200

WEI LI

identities (Tetlock & Manstead, 1985), and prior research has provided evidence that management uses these strategies in its disclosures such as letters to shareholders, Management Discussion and Analysis (MD&A), or earnings announcements. For example, Bettman and Weitz (1983), Salancik and Meindl (1984), Tsang (2002), and Clatworthy and Jones (2003) ﬁnd that when explaining past years’ performance in the letters to shareholders, management tends to use two impression management strategies: excuse (i.e., attribute bad performance to external factors such as ﬁerce competition) and entitlement (i.e., attribute good performance to internal factors such as a successful strategy). More recently, Aerts (2005) ﬁnds that when explaining its past years’ accounting outcomes in the annual reports, management also uses causality denials (i.e., dissociate management from performance outcomes, therefore reducing its responsibility for negative outcomes), justiﬁcation (i.e., implicitly accept responsibility of an outcome, but simultaneously reduce its negative repercussions by pointing to its intermediary character as a step to achieve higher goals), and enhancements (i.e., portray positive outcomes within the contexts of negative external inﬂuences). A few other studies have provided evidence that impression management in management disclosures inﬂuences investors’ judgments and decisions. For example, Staw et al. (1983) ﬁnd that management’s justiﬁcations for its past performance in the letters to shareholders are associated with subsequent improvements in stock price. Barton and Mercer (2005) ﬁnd that plausible (implausible) excuses provided by management in its MD&A lead analysts to provide higher (lower) earnings forecasts and stock valuations than if excuses had not been provided. Kaplan et al. (1990) ﬁnd that justiﬁcation and change strategies (i.e., avoid excuse or justiﬁcation for poor performance but merely focus on future prospects) in the letters to shareholders are effective in inﬂuencing investors’ judgments of future corporate proﬁt as well as their decisions of tendering a proxy to current management and buying/holding stocks. Sanders and Coelho (2001) ﬁnd that impression management in the form of management commentary in its earnings announcements (i.e., excuse and entitlement) signiﬁcantly shifts investors’ stock price expectations when past ﬁnancial information is negative.

Self-Promotion The abovementioned studies have largely focused on the impression management strategies that management uses in response to past organizational performance. These strategies are classiﬁed as reactive strategies

Self Promote or Not?

201

because they are reactive to situational demands such as when management is faced with a predicament (Tedeschi & Melburg, 1984). However, management does more than simply react to its environments; it sometimes tries to take the more proactive stance of doing what it believes will be organizationally (or personally) beneﬁcial (Siegel & Brockner, 2005). That is, management may use proactive impression management, which includes strategies initiated by management to proactively establish a particular identity instead of being merely reactive to situational demands (Tedeschi & Melburg, 1984). For example, in anticipation of organizationally threatening events, management uses self-handicapping strategies in its disclosures (i.e., claim obstacles before performance) (Siegel & Brockner, 2005). Similarly, to more directly maintain or enhance its image of competence, management is motivated to use self-promotion in its disclosures, similar to the strategies used by Enron. Self-promotion involves explicitly selfdescribing management’s or its company’s prior achievements and honors, advertising their competence or other strengths, avoiding weaknesses, and exuding conﬁdence about themselves (Ogden & Clark, 2005; Jones, 1964; Jones & Pittman, 1982; Godfrey et al., 1986). Such self-promotion statements are proactive (Godfrey et al., 1986) because they are not used to explain the good past performance. Nor are they directly linked to a company’s projected future performance or used to explain a company’s earnings forecast in details. They are just like those used by job applicants in job interviews, by ‘‘asserting’’ their competence, whether with reference to their general ability level or to a speciﬁc skill (Jones & Pittman, 1982; Tedeschi & Melburg, 1984). Until now, no research has provided empirical evidence on whether and how management’s self-promotion in its disclosures inﬂuences nonprofessional investors’ judgments and decisions.1 Some archival accounting research suggests that positive statements in management disclosures are irrelevant and meaningless because they are just sugar-coated without real information value (e.g., Abrahamson & Amir, 1996) or investors tend to discount transparent self-promotional behavior in management disclosures (Aerts, 2005). Thus, this line of research implies that management’s self-promotion may not signiﬁcantly inﬂuence investors’ judgments and decisions. However, experimental research in accounting, as well as research in psychology and management, gives reasons to expect the contradictory. Speciﬁcally, research in psychology and experimental accounting (e.g., Frederickson & Miller, 2004; Tversky & Kahneman, 1974; Nisbett et al., 1984) ﬁnds that the mere presence of information, regardless of its relevance to people’s judgments or decisions at hand, can have unintentional effects on the information

202

WEI LI

processing of people who lack expertise, such as nonprofessional investors. Thus, the presence of self-promotion statements in management disclosures, regardless of their direct relevance to a company’s past or future performance, may have effects on nonprofessional investors. Moreover, research on self-promotion from both psychology and management has found that an individual’s use of self-promotion will lead decision makers to make favorable judgments because the individual is perceived to be more competent at his/her job than others who do not use self-promotion. For example, at the theoretical level, Brickman and Seligman (1974), Schlenker and Leary (1982), and Vonk (1999) demonstrate that before a self-promoter’s actual performance is known, self-promotion can succeed in raising perceivers’ estimates of the self-promoter’s competence/intelligence. At the application level, Ellis et al. (2002), Ferris and Judge (1991), Kacmar, Delery, and Ferris (1992), Kacmar and Carlson (1999), Kristof-Brown et al. (2002), and Stevens and Kristof (1995) show that compared to those who do not use self-promotion tactic, job applicants who use self-promotion tactic in their interviews are perceived to be more competent and thus receive more positive interview outcomes. In sum, based on the above research, compared to management that does not use selfpromotion tactic in its disclosures, management that uses self-promotion tactics should be perceived to be more competent at its job, leading investors to make more favorable judgments/decisions. Such a prediction leads to H1: H1. Management’s use of self-promotion in its disclosures will lead investors to make more positive judgments and decisions.

Management Credibility and its Effect on the Effectiveness of Management’s Self-Promotion Prior research has documented that investors’ judgments and decisions are inﬂuenced not only by the information contained in management disclosures (e.g., self-promotion statements) but also by the credibility of the disclosures (Jennings, 1987; Mercer, 2004). One signiﬁcant factor that inﬂuences disclosure credibility is management credibility, which refers to investors’ perceptions of management’s competence and trustworthiness in ﬁnancial disclosures (Hovland, Janis, & Kelley, 1953; Mercer, 2004, 2005). If management is perceived to have higher credibility, investors are more willing to rely on management’s subsequent disclosures (Mercer, 2005), are more likely to agree with management’s decision to classify a hybrid security

203

Self Promote or Not?

in the equity section of the balance sheet (Hodge et al., 2006), are more likely to predict that the company’s future earnings will be higher, and are more likely to feel conﬁdent in their predictions (Hirst et al., 1999). Taken together, these studies suggest that management credibility positively inﬂuences investors’ judgments and decisions. This leads to H2: H2. High management credibility will lead investors to make more positive judgments and decisions. Management’s credibility is especially important in inﬂuencing investors’ reactions to its self-promotion because there is a phenomenon referred to as the ‘‘self-promoter’s paradox’’: As a self-promoter increases claims of competence and positive achievements, perceivers become less convinced (Jones and Pittman, 1982). That is why Aerts (2005, p. 498) speciﬁcally states that credibility concerns undermine the effectiveness of self-promotion in management disclosures. To be perceived as competent, self-promoters must express their competence claims in a credible way (Godfrey et al., 1986, p. 106). Thus, for self-promotion in management disclosures to be more effective, these selfpromotion statements should be credible or management that uses selfpromotion should appear credible. As the credibility of management’s statements depends on the credibility of management, it is expected that management’s self-promotion statements will be more credible and thus more effective at leading investors to make positive judgments and decisions when management is perceived to have high credibility than when management is perceived to have low credibility. This expectation leads to H3: H3. The positive effect of management’s self-promotion on investors’ judgments and decisions will be more prominent when management is perceived to have high credibility than when it is perceived to have low credibility. Fig. 1 (Panel A) summarizes the hypotheses at the theoretical level.

RESEARCH METHOD To test the hypotheses, this study conducted a 2 2 between-subject experiment that varies (1) the use of self-promotion in management disclosures (used vs. not used) and (2) perceived credibility of management (high vs. low).

204

WEI LI

Panel A: Theoretical Level Management Credibility

Management’s Use of Self-Promotion

Investors’ Judgments and Decisions

Panel B: Operational Level Explanations in the MD&A • Self-Promotion Statements in the MD&A

•

Fig. 1.

Investors’ Expectations for Management’s Future Performance; Investors’ investment decisions

Proposed Model.

Experimental Procedure Each participant ﬁrst received a letter of introduction outlining the study’s purpose, procedure, sequence, and expected duration. Then, each participant was instructed to assume the role of an investor who planned to invest a $10,000 year-end bonus in the medical supply industry and read preliminary information of two hypothetical companies in this industry: Amax and Blueco. Preliminary information was designed to convey that the two ﬁctitious companies had almost identical business backgrounds, historical and current ﬁnancial performances, and analyst recommendations. For example, both companies had a slight increase in net income in the current year and both predict an increase in sales for next year. Thus, the superiority of either company was not evident from that information alone, and participants were then advised to read both companies’ MD&A.2 Next, participants were provided with short excerpts from both companies’ annual reports, including excerpts from ﬁnancial statements and the MD&A. The MD&A contains management’s discussion of current year’s

205

Self Promote or Not?

results and analysis of the future year’s performance. After reading about the two companies, participants responded to a series of questions regarding the dependent variables (i.e., participants’ expectations for the companies’ future performance and their investment decisions), manipulation check, and demographic information. The experiment took an average of 30 minutes for participants to complete.

Participants Similar to previous experimental research (e.g., Hirst et al., 1999; Frederickson & Miller, 2004; Hodge et al., 2006; Mercer, 2005), this study used graduate business students as surrogates for nonprofessional investors. Speciﬁcally, 73 MBA students, enrolled in ﬁnancial accounting courses at three universities, voluntarily completed the experiment in controlled class room settings.3 Of the participants, 54.8 percent and 45.2 percent were male and female respectively. The mean work experience was 8.9 years. Participants on average had completed four accounting courses and two ﬁnance courses. Forty-four percent of the participants reported having had experience buying and selling individual companies’ stock. Of the participants who were not investing at the time of the experiment, 81 percent planned to invest within the next ﬁve years.4 In sum, these participants were appropriate for this study because their business experience and coursework have provided them with a basis for understanding management disclosures, and most of them invest or are interested in investing (Elliott, Hodge, Kennedy, & Pronk, 2007).

Manipulated Variables Following the approach used by Hodge, Kennedy, and Maines (2004), this study manipulated information in Blueco’s MD&A, whereas the information in Amax’s MD&A was used as a benchmark against which Blueco’s MD&A could be compared. Thus, in all experimental conditions, the information about Amax was identical and presented ﬁrst. This was intended to increase the effectiveness of the manipulated variables (see below and Fig. 1 Panel B).

206

WEI LI

Self-Promotion In the self-promotion used condition, Blueco’s MD&A was designed to include one paragraph of self-promotion statements immediately following management’s forecasts in its Forward-Looking Statements section. The statements were adapted from those in Enron’s 1998 annual reports and included the following: ‘‘Based on our prior experience, we feel very conﬁdent about our unparalleled ability to adjust to dynamic market conditions and our exceptional skills at managing the company. Such abilities and skills will be very difﬁcult, if not impossible, for any other company to replicate in the foreseeable future. As a result, we are conﬁdent that our sales of both products will increase in the next year by at least as much as our forecast noted above.’’ These statements were not explanations for the current year’s results nor linked to speciﬁc performance in the future (such as explanations for sales/expense forecasts). In the self-promotion not used condition, such a paragraph was not included in Blueco’s MD&A. In all experiment conditions, Amax’s MD&A did not include self-promotion statements. Management Credibility Similar to prior studies (e.g., Hirst et al., 1999; Mercer, 2005), this study manipulated perceived management credibility through management’s disclosure behaviors. Speciﬁcally, this study manipulated perceived management credibility through the explanations provided in the MD&A. The SEC requires management of public companies to provide clear explanations for its current and future performance in its MD&A (SEC, 1989). However, the SEC notes that many companies’ MD&A deviate from those requirements by simply identifying only changes that are evident from the face of ﬁnancial statements without clear reasons for their performance (SEC, 2003a). According to the Correspondent Inference Theory (Jones & Davis, 1965), such a deviation can be attributed to management’s dispositions, speciﬁcally management’s inability to disclose or management’s unwillingness to disclose clear explanations, which are two dimensions of management credibility. Thus, management that issues an unclear MD&A is expected to have lower credibility than management that issues a clear MD&A. Drawing on the above, this study provided participants in the high credibility condition a clear MD&A in which Blueco’s management followed the SEC’s guidance (SEC, 2003b) by explaining the reasons for the changes in current year’s sales and cost of goods sold as well as the reasons that would cause the next year’s changes. In contrast, participants in the low credibility condition received a MD&A in which Blueco’s management simply repeated the sales and other

207

Self Promote or Not?

numbers that were already included in its income statements without any explanations about the reasons that had caused or would cause the changes in sales or cost of goods sold. Based on a pilot test, Amax’s MD&A was set at a moderate amount of explanations.5 Speciﬁcally, Amax’s management indicated that changes in sales and cost of goods sold were caused and will continue to be caused by its sales of products. However, Amax’s management did not provide further insights into the underlying reasons behind the changes in product sales. This information about Amax was identical in all experimental conditions.

Dependent Variables Following Kaplan et al. (1990), this study examined two categories of judgments and decisions. First, participants were asked to make a judgment regarding their expectations for management’s future performance. Speciﬁcally, participants were asked to indicate the extent to which they agreed that the operation results of Blueco would be better than that of Amax in the next year (expectation for future performance). Responses to this question were recorded on a 7-point Likert-type scale with the ends labeled ‘‘strongly disagree’’ (1) and ‘‘strongly agree’’ (7). Second, participants were asked about the investment decisions they would make in regard to the company. Speciﬁcally, similar to those in Hodge et al. (2004), participants in this study were asked to indicate what percentage of $10,000 they would invest in Blueco (investment decision). Their responses were recorded on an 11-point scale with the endpoints labeled from 0% to 100%.

RESULTS Manipulation Check To verify that management’s credibility can be inferred from whether Blueco’s management provided clear explanations in its MD&A, two sets of manipulation questions were asked. One set of questions asked participants to indicate the extent to which they agreed that Blueco’s management appeared trustworthy and honest on 7-point Likert scales from 1 (strongly disagree) to 7 (strongly agree). Two one-way Analyses of Variance (ANOVAs) were conducted to test participants’ responses. The results show that participants in the high credibility condition perceived management to

208

WEI LI

appear more trustworthy (mean ¼ 4.80, SD ¼ 1.36) than did participants in the low credibility condition (mean ¼ 2.44, SD ¼ 1.23) (F ¼ 59.95, po0.01). Similarly, participants in the high credibility condition perceived management to appear more honest (mean ¼ 4.83, SD ¼ 1.39) than did participants in the low credibility condition (mean ¼ 3.77, SD ¼ 1.66) (F ¼ 8.75, po0.01). The other set of questions asked participants to indicate the extent to which they agreed that Blueco’s management provided clear explanations for its performance in its MD&A on 7-point Likert scales with the endpoints labeled from 1 (strongly disagree) to 7 (strongly agree). The results from oneway ANOVAs show that participants in the high credibility condition agreed that management provided much clearer explanations for its current year performance in the MD&A (mean ¼ 5.33, SD ¼ 1.51) than did participants in the low credibility condition (mean ¼ 2.34, SD ¼ 1.60) (F ¼ 67.05, po0.01). Similarly, participants in the high credibility condition (mean ¼ 5.28, SD ¼ 1.55) agreed that management provided much clearer explanations for its forecasts of next year performance than did participants in the low credibility condition (mean ¼ 2.60, SD ¼ 1.56) (F ¼ 54.26, po0.01). Together, these results indicate that the manipulation for management credibility was successful.6

Tests of Hypotheses As the two dependent variables were found to be correlated (correlation ¼ 0.67, p o0.01), a multivariate analysis of variance (MANOVA) was ﬁrst conducted with self-promotion (used vs. not used) and management credibility (high vs. low) as the independent variables and expectation for future performance and investment decision as the dependent variables.7 The results of MANOVA, as reported in Table 1, indicate that self-promotion has a signiﬁcant main effect on participants’ judgments and decisions (F ¼ 2.78, p ¼ 0.03), as does management credibility (F ¼ 7.00, po0.01). Moreover, there is a signiﬁcant interaction effect of self-promotion and management credibility on participants’ judgments and decisions (F ¼ 3.29, p ¼ 0.02). Thus, the overall MANOVA results provide supports for H1–H3. Then, two separate ANOVAs were conducted to examine the effects of self-promotion and management credibility on expectation for future performance and investment decision respectively. Table 2, Panel A, provides the descriptive statistics for participants’ expectations for future performance, whereas Panel B provides the results of the ANOVA. Panel B shows a signiﬁcant main effect of self-promotion on participants’

209

Self Promote or Not?

Table 1.

Multivariate Analysis of Variance.

Factor Self promotion Management credibility Self promotion management credibility

df

F value

p value (one tailed)

2, 68 2, 68 2, 68

2.78 7.00 3.29

0.03 o0.01 0.02

Dependent Variables: Expectation for Future Performancea and Investment Decisionb Signiﬁcant at 0.05 level. Signiﬁcant at 0.01 level. a

Participants’ expectations for future performance were measured by the extent to which they agreed that the operation results of Blueco would be much better than that of Amax in the next year on a 7 point Likert type scale from 1 (strongly disagree) to 7 (strongly agree). b Participants’ investment decisions were measured by the percentage of $10,000 they chose to invest in Blueco on an 11 point Likert scale labeled from 0 (0%) to 10 (100%).

expectations for Blueco’s next year operation results (F ¼ 4.12, p ¼ 0.02). Speciﬁcally, participants expected Blueco’s operation results in the next year to be much better if management used self-promotion statements in its MD&A (mean ¼ 4.07, SD ¼ 1.36) than if management did not use self-promotion (mean ¼ 3.43, SD ¼ 1.34) as shown in Panel A of Table 2. Table 3, Panel A, provides the descriptive statistics for participants’ investment decisions, whereas Panel B provides the results of the ANOVA. Similarly, Panel B shows a signiﬁcant main effect of self-promotion on participants’ investment decisions (F ¼ 5.04, p ¼ 0.01). Speciﬁcally, participants were likely to invest more in Blueco if management used self-promotion statements in its MD&A (mean ¼ 6.14, SD ¼ 2.05) than if management did not use self-promotion (mean ¼ 5.07, SD ¼ 2.40) as shown in Panel A of Table 3. Taken together, the ANOVA results provide support for H1 that management’s use of self-promotion in its disclosures will lead investors to make more positive judgments and decisions. Table 2, Panel B, also shows a signiﬁcant main effect of management credibility on participants’ expectations for future performance (F ¼ 3.28, p ¼ 0.04). Speciﬁcally, participants expected Blueco’s operation results in the next year to be much better if management was perceived to have high credibility (mean ¼ 3.99, SD ¼ 1.44) than if management was perceived to have low credibility (mean ¼ 3.47, SD ¼ 1.27) as shown in Panel A of Table 2. Similarly, Panel B of Table 3 shows a signiﬁcant main effect of management credibility on participants’ investment decisions (F ¼ 13.69, po0.01). Speciﬁcally, participants were likely to invest more in Blueco if management was perceived to have high credibility (mean ¼ 6.43,

210

WEI LI

Table 2.

Effect of Self-Promotion and Management Credibility on Participants’ Expectations for Future Performance.

Panel A: Mean (Standard Deviation) for Participants’ Expectations for Future Performancea Credibility level Self promotion not used Self promotion used Overall Low credibility

High credibility

Overall

3.50 (1.25) n ¼ 18 3.38 (1.44) n ¼ 20 3.43 (1.34) n ¼ 38

Panel B: Analysis of Variance Factor

3.44 (1.32) n ¼ 17 4.67 (1.14) n ¼ 18 4.07 (1.36) n ¼ 35

3.47 (1.27) n ¼ 35 3.99 (1.44) n ¼ 38

df Sum of squares F value p value (one tailed)

Self promotion 1 management Credibility 1 self promotion management credibility 1 Error 69

6.91 5.51 8.29

4.12 3.28 4.94

0.02 0.04 0.02

Panel C: Simple Effects

Effect of self promotion in the low credibility condition Effect of self promotion in the high credibility condition

df

Sum of squares

F value

p value (one tailed)

1

0.03

0.02

0.45

1

15.81

9.26

o0.01

Signiﬁcant at 0.05 level. Signiﬁcant at 0.01 level. a

Participants’ expectations for future performance were measured by the extent to which they agreed that the operation results of Blueco would be much better than that of Amax in the next year on a 7 point Likert type scale from 1 (strongly disagree) to 7 (strongly agree).

SD ¼ 2.14) than if management was perceived to have low credibility (mean ¼ 4.66, SD ¼ 2.10) as shown in Panel A of Table 3. Taken together, the ANOVA results provide support for H2 that high management credibility will lead investors to make more positive judgments and decisions. Moreover, Table 2, Panel B, shows a signiﬁcant interaction effect of selfpromotion and management credibility on participants’ expectations for Blueco’s next year operation results (F ¼ 4.94, p ¼ 0.02). Speciﬁcally, when management was perceived to have high reporting credibility, participants

211

Self Promote or Not?

Table 3.

Effect of Self-Promotion and Management Credibility on Participants’ Investment Decisions.

Panel A: Mean (Standard Deviation) for Participants’ Investment Decisionsa Credibility level Self promotion not used Self promotion used Low credibility

High credibility

Overall

4.22 (2.16) n ¼ 18 5.83 (2.40) n ¼ 20 5.07 (2.40) n ¼ 38

Panel B: Analysis of Variance Factor

Overall

5.12 (2.00) n ¼ 17 7.11 (1.60) n ¼ 18 6.14 (2.05) n ¼ 35

4.66 (2.10) n ¼ 35 6.43 (2.14) n ¼ 38

df Sum of squares F value p value (one tailed)

Self promotion 1 Management credibility 1 Self promotion management credibility 1 Error 69

21.64 58.80 0.69

5.04 13.69 0.16

0.01 o0.01 0.34

Panel C: Simple Effects

Effect of self promotion in the low credibility condition Effect of self promotion in the high credibility condition

df

Sum of squares

F value

p value (one tailed)

1

7.01

1.62

0.11

1

15.67

3.68

0.03

Signiﬁcant at 0.05 level. Signiﬁcant at 0.01 level. a

Participants’ investment decisions were measured by the percentage of $10,000 they chose to invest in Blueco on an 11 point Likert scale labeled from 0 (0%) to 10 (100%).

expected Blueco’s operation results in the next year to be much better if it used self-promotion statements in the MD&A (mean ¼ 4.67, SD ¼ 1.14) than if it did not use self-promotion (mean ¼ 3.38, SD ¼ 1.44) as shown in Panel A of Table 2. Such a difference is signiﬁcant (F ¼ 9.26, po0.01) (Panel C of Table 2). In contrast, when management was perceived to have low reporting credibility, participants’ expectations for Blueco’s next year operation results were not signiﬁcantly different between the self-promotion used condition (mean ¼ 3.44, SD ¼ 1.32) and the self-promotion not used condition (mean ¼ 3.50, SD ¼ 1.25) (F ¼ 0.02, p ¼ 0.45).

212

WEI LI

Regarding to participants’ investment decisions, Panel A of Table 3 indicates that when management was perceived to have high reporting credibility, participants were likely to invest much more in Blueco if it used self-promotion statements in the MD&A (mean ¼ 7.11, SD ¼ 1.60) than if it did not use self-promotion (mean ¼ 5.83, SD ¼ 2.40). Such a difference is signiﬁcant (F ¼ 3.68, p ¼ 0.03) (Panel C of Table 3). In contrast, when management was perceived to have low reporting credibility, participants’ investments in Blueco were not signiﬁcantly different between selfpromotion used condition (mean ¼ 5.12, SD ¼ 2.00) and self-promotion not used condition (mean ¼ 4.22, SD ¼ 2.16) (F ¼ 1.62, p ¼ 0.11). Taken together, the ANOVA results provide support for H3 that the positive effect of management’s self-promotion on investors’ judgments and decisions will be more prominent when management is perceived to have high credibility than when it is perceived to have low credibility.

CONCLUSIONS This study examines whether and how self-promotion, as a type of proactive impression management strategies in management’s disclosures, inﬂuences nonprofessional investors’ judgments and decisions. The results of the experiment show that management’s use of self-promotion will inﬂuence nonprofessional investors so that investors will (1) expect management to perform better in the future and (2) invest more in that company. These positive effects are more prominent when management is perceived to have high credibility than when it is perceived to have low credibility. The results provide implications for management and regulators. As management has opportunities to choose what narratives it discloses (SEC, 1989), these results could help management to decide what and how narratives should be disclosed to maintain and enhance its competent image. Speciﬁcally, the ﬁndings suggest that management should use this strategy thoughtfully because investors will respond positively only when management is perceived to have high credibility. Also, the ﬁndings may have implications for how to improve the usefulness of management disclosures for nonprofessional investors, consistent with the call from the SEC for more quality disclosures (Meiers, 2006). The ﬁndings of this study have implications for researchers as well. First, although archival studies (e.g., Abrahamson & Amir, 1996; Aerts, 2005) suggest that positive narratives are meaningless and irrelevant, this study, based on the research in psychology, experimental accounting, and

Self Promote or Not?

213

management, provides direct evidence to the contrary. Such evidence highlights the importance of positive narratives that have been relatively ignored by prior archival accounting research. Second, these ﬁndings, combined with those from prior studies (e.g., Barton & Mercer, 2005; Kaplan et al., 1990; Sanders & Coelho, 2001; Siegel & Brockner, 2005), show that investors respond to both reactive and proactive impression management strategies in management’s disclosures. Third, this study indicates that the competence assessments applied to an individual selfpromoter are applicable to a management or company that uses selfpromotion. That is, management’s self-promotion can also have signiﬁcant organizational-level consequences. Furthermore, this study provides empirical evidence that investors’ responses to management’s self-promotion depend critically on management credibility. This evidence adds to those from prior studies (e.g., Hirst et al., 1999; Hirst et al., 2007; Hodge et al., 2006; Mercer, 2004, 2005) to underscore the importance of management credibility in inﬂuencing the effectiveness of its disclosures. Speciﬁcally, this study shows that whether management provides clear explanations in its MD&A inﬂuences investors’ perceptions of management credibility, which will inﬂuence investors’ reactions to management’s self-promotion statements as well as investors’ ﬁnal judgments and decisions. This study has several limitations that may lead to future research. First, this study focuses only on nonprofessional investors’ reactions to management’s self-promotion, and half of the participants did not have experience investing in individual stocks. Thus, whether the ﬁndings can be generalized to more experienced investors, such as professional investors, is unknown. On the one hand, if self-promotion statements are deemed to be irrelevant for investment decisions, professional investors, who have expertise, are not expected to be inﬂuenced by these irrelevant statements in management disclosures. On the other hand, the ﬁnding in Davis et al. (2007) that optimistic tones in earnings press releases can inﬂuence the market at whole suggests that professional investors, as signiﬁcant market participants, may be inﬂuenced by management’s self-promotion. Thus, whether and how management’s self-promotion inﬂuences professional investors differently can be explored in future research. Second, this study manipulated management credibility by varying whether management provided clear explanations in its MD&A, while management credibility can be inferred from other disclosure behaviors of management. Thus, future research may examine how other disclosure behaviors of management inﬂuence investors’ perceptions of management credibility and then further inﬂuence the effectiveness of management’s self-promotion. Third, to draw participants’

214

WEI LI

attentions to management’s narratives in the MD&A, participants were provided with information about the SEC’s guidelines for the MD&A. Future studies may examine whether investors will react to the information contained in the MD&A in the same way if they are not informed about the SEC’s guidelines. Finally, this study examines how investors respond to management’s self-promotion when management’s actual future performance is unknown. Future research may examine whether and how investors’ judgments and decisions will change as management’s actual future performance becomes known.

NOTES 1. Some prior studies, such as Lang and Lundholm (2000), investigate whether optimistic tones in management disclosures affect the market. However, these prior studies do not clearly distinguish whether optimistic tones in management disclosures come from optimism in content (i.e., positive earnings or negative earnings) or optimism in linguistic style (i.e., describing information in an optimistic way). Concurrent work by Davis et al. (2007) distinguishes disclosure’s content and disclosure’s linguistic style and ﬁnds that after controlling the content in earnings press releases, the use of optimistic words in these releases is positively related to market responses. However, Davis et al. (2007) are different from this study in at least three ways. First, Davis et al. (2007) assume that management behaves truthfully by using different linguistic styles to convey credible information to the market. In contrast, this study is based on the assumption that management behaves strategically by using languages to manage its impression to investors. Second, Davis et al. (2007) focus on providing archival evidence to show the association between management’s use of optimistic tones in earnings press releases and the market at whole. In contrast, this study takes advantages of an experimental design to show when and how management’s use of languages inﬂuences market participants. Third, Davis et al. (2007) do not provide theories to explain why and how optimistic tones in earnings press releases inﬂuence market responses. Due to the lack of a theoretical basis, Davis et al. (2007) state that whether or not investors will respond to the levels of optimistic tones in earnings press releases is uncertain. In contrast, this study develops hypotheses based on research in psychology, experimental accounting, and management. 2. To draw participants’ attention to the MD&A and help them understand whether management follows the SEC’s guidelines for the MD&A, participants were provided with a section describing these guidelines. This might have possibly biased the study in favor of obtaining the predicted effects. In particular, participants might have paid more attention to the MD&A and are thus more likely to be inﬂuenced by the information within. Not including this section may diminish the effects, but the directions of the effects should not be altered. 3. Instructors of each course helped recruit participants in this study. They ﬁrst informed students enrolled in their courses about the study. Then, students who voluntarily chose to participate in the study came to the class for the experiment.

215

Self Promote or Not?

In contrast, students who chose not to participate did not attend the portion of class in which the experiment was conducted. As a result, participants were self motivated and interested in this study, although there was no monetary compensation. University afﬁliation was not signiﬁcantly related to any dependent variables. 4. There were no signiﬁcant differences in any of the dependent measures between current investors and noninvestors. Also, the reported results did not change when any of other demographic information was included as covariates in the tests. 5. A pilot test was conducted among nonprofessional investors from two local investment clubs. Speciﬁcally, 11 investors read the case materials and answered the questions used in the experiment. All of them perceived the case materials to be realistic and engaging. Despite the small group, the direction of their responses showed that these investors perceived the differences in the amount of explanations for high vs. low management credibility conditions. Moreover, their feedback provided more insights about what should be set as a moderate amount of explanations included in Amax’s MD&A. 6. A MANOVA was also conducted to test participants’ responses to the four manipulation check questions. The results from the MANOVA were consistent with those from the ANOVAs. 7. Bivariate correlation matrix showed that participants’ perceptions of Blueco management’s credibility were not signiﬁcantly related to Blueco’s use of self promotion in its MD&A (Pearson correlation 0.03, one tailed p 0.42). In other words, the two independent variables were not correlated in the data analysis.

ACKNOWLEDGMENTS This chapter is based, in part, on my dissertation completed at Washington State University. I especially thank Bernard Wong-On-Wing (my dissertation chair) for his advice, support, and encouragement. I also appreciate comments provided by Tom Nunamaker, Steven Thornburg (committee members), Lan Guo, Ruby Lee, and participants at the 2009 American Accounting Association Ohio Region Meeting. In addition, my thanks also extend to the two anonymous reviewers and Vicky Arnold (editor).

REFERENCES Abrahamson, E., & Amir, E. (1996). The information content of the president’s letter to shareholders. Journal of Business Finance and Accounting, 23(8), 1157 1182. Aerts, W. (2005). Picking up the pieces: Impression management in the retrospective attributional framing of accounting outcomes. Accounting, Organizations and Society, 19(4/5), 337 353. Barton, J., & Mercer, M. (2005). To blame or not to blame: Analysts’ reactions to external explanations for poor ﬁnancial performance. Journal of Accounting and Economics, 39(3), 509 533.

216

WEI LI

Bettman, J. R., & Weitz, B. A. (1983). Attributions in the boardroom: Causal reasoning in corporate annual reports. Administrative Science Quarterly, 28(2), 163 183. Bogle, J. (2005). The ownership of corporate America rights and responsibilities. Remarks by John C. Bogle, the Vanguard Group, 20th Anniversary Meeting of the Council of Institutional Investors. Available at http://www.vangaurd.com/bogle_site/ sp20050411.htm Brickman, P., & Seligman, C. (1974). Effects of public and private expectancies on attributions of competence and interpersonal attraction. Journal of Personality, 42, 558 568. Clarkson, P. M., Kao, J. L., & Richardson, G. (1994). The voluntary inclusion of forecasts in the MD&A section of annual reports. Contemporary Accounting Research, 11(1), 423 450. Clatworthy, M., & Jones, M. J. (2003). Financial reporting of goods news and bad news: Evidence from accounting narratives. Accounting and Business Research, 33(3), 171 185. Davis, A. K., Piger, J. M., & Sedor, L. M. (2007). Beyond the numbers: Managers’ use of optimistic and pessimistic tone in earnings press releases. Working Paper. University of Oregon and University of Washington. Elliott, W. B., Hodge, F. D., Kennedy, J. J., & Pronk, M. (2007). Are M.B.A. students a good proxy for nonprofessional investors? The Accounting Review, 82(1), 139 168. Ellis, A. P. J., West, B. J., Ryan, A. M., & DeShon, R. R. (2002). The use of impression management tactics in structured interviews: A function of question type? Journal of Applied Psychology, 87(6), 1200 1208. Ferris, G. R., & Judge, T. A. (1991). Personnel/human resources management: A political inﬂuence perspective. Journal of Management, 17(2), 447 488. Frederickson, J. R., & Miller, J. S. (2004). The effects of pro forma earnings disclosures on analysts’ and nonprofessional investors’ equity valuation judgment. The Accounting Review, 79(3), 667 686. Gardner, W. L., & Cleavenger, D. (1998). The impression management strategies associated with transformational leadership at the world class level. Management Communication Quarterly, 12(1), 3 41. Godfrey, D. K., Jones, E. E., & Lord, C. G. (1986). Self promotion is not ingratiating. Journal of Personality and Social Psychology, 50(1), 106 115. Hirst, D. E., Koonce, L., & Miller, J. (1999). The joint effect of management’s prior forecast accuracy and the form of its ﬁnancial forecasts on investor judgment. Journal of Accounting Research, 37(Suppl.), 101 124. Hirst, D. E., Koonce, L., & Venkataraman, S. (2007). How disaggregation enhances the credibility of management earnings forecasts. Journal of Accounting Research, 45, 811 837. Hodge, F. D., Hopkins, P. E., & Pratt, J. (2006). Management reporting incentives and classiﬁcation credibility: The effects of reporting discretion and reputation. Accounting, Organizations and Society, 31(7), 623 634. Hodge, F. D., Kennedy, J. J., & Maines, L. A. (2004). Does search facilitating technology improve the transparency of ﬁnancial reporting? The Accounting Review, 79(3), 687 703. Hovland, C., Janis, I., & Kelley, H. (1953). Communication and persuasion. New Haven, CT: Yale University Press. Jennings, R. (1987). Unsystematic security price movements, management earnings forecasts, and revisions in consensus analyst earnings forecasts. Journal of Accounting Research, 25(1), 90 110.

Self Promote or Not?

217

Jones, E. E. (1964). Ingratiation. New York, NY: Appleton Century Crofts. Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in person perception. In: L. Berkowitz (Ed.), Advances in experimental social psychology. New York, NY: Academic Press. Jones, E. E., & Pittman, T. S. (1982). Toward a theory of strategic self presentation. In: J. Suls (Ed.), Psychological perspectives on the self (pp. 231 262). Hillsdale, NJ: Erlbaum. Kacmar, K. M., & Carlson, D. S. (1999). Effectiveness of impression management tactics across human resource situations. Journal of Applied Social Psychology, 29(6), 1293 1315. Kacmar, K. M., Delery, J. E., & Ferris, G. R. (1992). Differential effectiveness of applicant impression management tactics on employment interview decisions. Journal of Applied Social Psychology, 22(16), 1250 1272. Kaplan, E. S., Pourciau, S., & Reckers, P. M. J. (1990). An examination of the effect of the president’s letter and stock advisory service information on ﬁnancial decisions. Behavioral Research in Accounting, 2, 63 92. Krische, S. D. (2005). Investors’ evaluations of strategic prior period benchmark disclosures in earnings announcements. The Accounting Review, 80(1), 243 268. Kristof Brown, A., Barrick, M. R., & Franke, M. (2002). Applicant impression management: Dispositional inﬂuences and consequences for recruiter perceptions of ﬁt and similarity. Journal of Management, 28(1), 27 46. Lang, M. H., & Lundholm, R. J. (2000). Voluntary disclosure and equity offerings: Reducing information asymmetry or hyping the stock? Contemporary Accounting Research, 17(4), 623 662. Lee, T. A. (1994). The changing form of the corporate annual report. The Accounting Historians Journal, 21(1), 215 232. Libby, R., Bloomﬁeld, R., & Nelson, M. (2002). Experimental research in ﬁnancial accounting. Accounting, Organizations and Society, 27(8), 775 810. Maines, L. A. (1994). The role of behavioral accounting research in ﬁnancial accounting standard setting. Behavioral Research in Accounting, 6(Suppl.), 204 212. Meiers, D. H. (2006). The MD&A challenge. Journal of Accountancy, 201(1), 59 66. Mercer, M. (2004). How do investors assess the credibility of management disclosures? Accounting Horizon, 18(3), 185 196. Mercer, M. (2005). The ﬂeeting effects of disclosure forthcomingness on management’s reporting credibility. The Accounting Review, 80(2), 723 744. New York Stock Exchange. (2002). Share ownership 2000 highlights. Available at http:// www.nysedata.com/nysedata/asp/factbook/viewer_edition.asp?mode ¼ text&key ¼ 51 &category ¼ 11 Nisbett, R. E., Zukier, H., & Lemley, R. E. (1984). The dilution effect: Nondiagnositc information weakens the implications of diagnostic information. Cognitive Psychology, 13(April), 248 277. Ogden, S., & Clark, J. (2005). Customer disclosures, impression management and the construction of legitimacy: Corporate reports in the UK privatized water industry. Accounting, Auditing and Accountability Journal, 18(3), 313 345. Salancik, G. R., & Meindl, J. R. (1984). Corporate attributions as strategic illustrations of management control. Administrative Science Quarterly, 29, 238 254. Sanders, D. E., & Coelho, J. L. (2001). The effect of impression management strategies on earnings release announcements. Journal of Business Strategies, 18(1), 25 45.

218

WEI LI

Schlenker, B. R., & Leary, M. R. (1982). Audiences’ reactions to self enhancing, self denigrating, and accurate self presentations. Journal of Experimental Social Psychology, 18, 89 104. Securities and Exchange Commission (SEC). (1989). Management’s discussion and analysis of ﬁnancial condition and results of operations; Certain investment ﬁrm disclosures. Release No. 33 6835 and 34 26831. Available at http://www.sec.gov/rules/interp/ 33 6835.htm Securities and Exchange Commission (SEC). (2003a). Summary by the division of corporation ﬁnance of signiﬁcant issues addressed in the review of the periodic reports of the Fortune 500 companies. Available at http://www.sec.gov/divisions/corpﬁn/fortune500rep.htm Securities and Exchange Commission (SEC). (2003b). Commission guidance regarding management’s discussion and analysis of ﬁnancial condition and results of operations. Release Nos. 33 8350 and 34 48960. Available at http://www.sec.gov/rules/interp/ 33 8350.htm Siegel, P. A., & Brockner, J. (2005). Individual and organizational consequences of CEO claimed handicapping: What is good for the CEO may not be so good for the ﬁrm. Organizational Behavior and Human Decision Processes, 96, 1 22. Staw, B., McKechnie, P., & Puffer, S. (1983). The justiﬁcation of organizational performance. Administrative Science Quarterly, 28, 290 312. Stevens, C. K., & Kristof, A. L. (1995). Making the right impression: A ﬁeld study of applicant impression management during job interviews. Journal of Applied Psychology, 80(5), 587 606. Tedeschi, J. I., & Melburg, V. (1984). Impression management and inﬂuence in the organization. In: S. B. Bacharach & E. J. Lawler (Eds), Research in the sociology of organizations (Vol. 3, pp. 31 58). Greenwich, CT: JAI Press. Tetlock, P. E., & Manstead, A. S. R. (1985). Impression management versus inter psychic explanations in social psychology: A useful dichotomy? Psychological Review, 92(1), 59 77. Tsang, E. W. K. (2002). Self serving attributions in corporate annual reports: A replicated study. Journal of Management Studies, 39(1), 51 65. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185, 1124 1131. Vonk, R. (1999). Impression formation and impression management: Motives, traits, and likeability inferred from self promoting and self deprecating behavior. Social Cognition, 17(4), 390 412. Williams, P. A. (1996). The relation between a prior earnings forecast by management and analyst response to a current management forecast. The Accounting Review, 71(1), 103 113.

Advances in Accounting in Behavioural Research (Advances in Accounting Behavioral Research, Vol. 13)

Advances in Accounting Behavioral Research ~ Volume 12

Advances in Accounting Behavioral Research, Volume 10

Advances in Accounting Behavioral Research, Volume 5 (Advances in Accounting Behavioral Research)

Advances in Accounting Behavioral Research, Volume 6 (Advances in Accounting Behavioral Research)

Advances in Accounting Behavioral Research, Volume 9

Advances in Accounting Behavioral Research, Volume 11

Advances in Accounting Behavioral Research, Volume 9

Advances in Accounting Behavioral Research, Volume 14

Advances in Accounting Behavioral Research, Volume 8 (Advances in Accounting Behhavioral Research)

Advances in Management Accounting, Volume 13 (Advances in Management Accounting)

Research in Accounting Regulation, Vol. 16 (Research in Accounting Regulation) (Research in Accounting Regulation)

Advances in Accounting, Volume 18 (Advances in Accounting) (Advances in Accounting)

Advances in International Accounting, Volume 18 (Advances in International Accounting)

Advances in Management Accounting, Volume 12 (Advances in Management Accounting)

Advances in International Accounting, Volume 17 (Advances in International Accounting)

Advances in Management Accounting, Volume 11 (Advances in Management Accounting)

Advances in Management Accounting

Advances in Accounting, Volume 19 (Advances in Accounting)

Advances in International Accounting

Advances in International Accounting

Advances in International Accounting, Volume 18 (Advances in International Accounting)

Advances in International Accounting, Volume 16 (Advances in International Accounting)

Advances in Management Accounting, Volume 9 (Advances in Management Accounting)

Research in Accounting Regulation, Volume 18 (Research in Accounting Regulation)

Advances in International Accounting, Volume 17 (Advances in International Accounting)

Advances in International Accounting, Volume 16 (Advances in International Accounting)

Research methods in accounting

Research Methods in Accounting

Research in Accounting Regulation, Volume 14 (Research in Accounting Regulation)

Research in Accounting Regulation, Volume 15 (Research in Accounting Regulation)

Advances in Accounting in Behavioural Research (Advances in Accounting Behavioral Research, Vol. 13)

Advances in Accounting Behavioral Research ~ Volume 12

Advances in Accounting Behavioral Research, Volume 10

Advances in Accounting Behavioral Research, Volume 5 (Advances in Accounting Behavioral Research)

Advances in Accounting Behavioral Research, Volume 6 (Advances in Accounting Behavioral Research)

Advances in Accounting Behavioral Research, Volume 9

Advances in Accounting Behavioral Research, Volume 11

Advances in Accounting Behavioral Research, Volume 9

Advances in Accounting Behavioral Research, Volume 14

Advances in Accounting Behavioral Research, Volume 8 (Advances in Accounting Behhavioral Research)

Advances in Management Accounting, Volume 13 (Advances in Management Accounting)

Research in Accounting Regulation, Vol. 16 (Research in Accounting Regulation) (Research in Accounting Regulation)

Advances in Accounting, Volume 18 (Advances in Accounting) (Advances in Accounting)

Advances in International Accounting, Volume 18 (Advances in International Accounting)

Advances in Management Accounting, Volume 12 (Advances in Management Accounting)

Advances in International Accounting, Volume 17 (Advances in International Accounting)

Advances in Management Accounting, Volume 11 (Advances in Management Accounting)

Advances in Management Accounting

Advances in Accounting, Volume 19 (Advances in Accounting)

Advances in International Accounting

Advances in International Accounting

Advances in International Accounting, Volume 18 (Advances in International Accounting)

Advances in International Accounting, Volume 16 (Advances in International Accounting)

Advances in Management Accounting, Volume 9 (Advances in Management Accounting)

Research in Accounting Regulation, Volume 18 (Research in Accounting Regulation)

Advances in International Accounting, Volume 17 (Advances in International Accounting)

Advances in International Accounting, Volume 16 (Advances in International Accounting)

Research methods in accounting

Research Methods in Accounting

Research in Accounting Regulation, Volume 14 (Research in Accounting Regulation)

Research in Accounting Regulation, Volume 15 (Research in Accounting Regulation)

Recommend Documents