wang - 2003 - the impact of mandatory audit firm rotation on auditor-client negotiations

It is hypothesized that the imposition of mandatory rotation will reduce the auditor's incentive to compromise throughout the entire term of the audit engagement with a client and will a

THE IMPACT OF MANDATORY AUDIT FIRM ROTATION ON

AUDITOR-CLIENT NEGOTIATIONS

by

Jianqiang Wang

Bachelor of Arts Fudan University, 1982

Master of Arts Fudan University, 1986

Master of Accountancy University of South Carolina, 1996

Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in

Moore School of Business University of South Carolina

UMI Number: 3098716

UMI

UMI Microform 3098716 Copyright 2003 by ProQuest Information and Learning Company All rights reserved This microform edition is protected against unauthorized copying under Title 17, United States Code.

ProQuest Information and Learning Company

300 North Zeeb Road P.O Box 1346 Ann Arbor, Ml 48106-1346

I dedicate this dissertation to my beautiful wife, Min, who has amazingly made a philosopher an accounting professor, and to our wonderful son, Diexia, who has easily made a lousy accountant a great piano teacher by being one of the best young pianists in America Thank you for your support and making my life so enjoyable

I would like to offer my sincere thanks to my dissertation chair, Dr Brad Tuttle, for being incredibly generous with his talent and time in helping me develop, conduct, and finish this study Without his remarkable and tireless effort, this work would not have been possible I also want to extend my special thanks to Dr Maribeth Coller for giving

me so much help in so many ways since I took my first accounting course from her, to

Dr Scott Jackson for always providing insightful and challenging feedback and comments on my work, and to Dr Lisa Rutstrom for her expert advice on experimental economics and many other contributions to the completion of this dissertation I was very fortunate to have these wonderful scholars on my dissertation committee and I am deeply grateful for their invaluable guidance and support throughout the whole process of writing this dissertation

I also would like to thank the School of Accounting faculty, especially Dr Rich White, Dr Adrian Harrell, Dr Scott Vandervelde, Dr Tim Doupnik, Dr Gene

Chewning, and Dr A1 Leitch, for helping me in various stages of this work and in many years of my study in this great school My gratitude and appreciation also go to Dr Mark Taylor and Dr Todd DeZoort for their advice and encouragement before and after they left this school

It is my great pleasure to acknowledge the friendship and support of my fellow Ph.D students: Young-Won Her, Jennifer Kahle, George Tsakumis, and Danny Wadden

It has been four unforgettable years and we had so much fun

I also want to acknowledge financial support for this study from the Moore School of Business

Mandatory rotation of audit firms has become an increasingly attractive regulatory option to lawmakers in light of a series of recent disastrous corporate accounting scandals The Sarbanes-Oxley Act of 2002 requires a study of potential effects of such a regulation

This study investigates how the mandatory rotation of audit firms will affect auditor-client negotiations when there is a disagreement on accounting values to report in financial statements It is hypothesized that the imposition of mandatory rotation will

reduce the auditor's incentive to compromise throughout the entire term of the audit engagement with a client and will also reduce the client's incentive to compromise in the

final engagement year As a result of such changes in the auditor’s and the client’s incentives to compromise, any regulation that requires mandatory rotation of audit firms may have unanticipated effects on the outcome of auditor-client negotiations, i.e., the financial statements issued to the public

The study was conducted through experiments in a laboratory setting in which participants assumed the role of a manager (client) or of a verifier (auditor) Using a computerized negotiation system, participants negotiated an asset value to report under one of two manipulated conditions—limited auditor term (i.e., mandatory audit firm rotation) or unlimited auditor term The effects of mandatory rotation are measured by the rate of agreement between auditors and clients and the agreed-upon asset values It is found that mandatory rotation results in a lower rate of agreement between auditors and clients and lower agreed-upon asset values

TABLE OF CONTENTS Chapter 1: INTRODUCTION

Chapter 2: THEORY AND HYPOTHESES

2.1 Auditor-Client Disagreements2.2 Incentives Facing the Client2.3 Incentives Facing the Auditor2.4 Hypotheses

Chapter 3: METHOD

3.1 Experimental Task3.2 Experiment Design3.3 Independent Variables3.4 Dependent Variables3.5 Negotiation Procedure and Rules3.6 Payoffs

Chapter 5: CONCLUSION REFERENCES

APPENDIX

1 6 6

8

1114

19

19

22 22

24242731

33

33344152

54 59 66

LIST OF TABLES

Table 4.1: Analysis of Variance of Continuous Demographic Data 34Table 4.2: Descriptive Statistics: Mandatory Rotation Condition 35-36

Table 4.3: Descriptive Statistics: Mandatory Rotation Condition

Table 4.7: Rate of Agreement: Comparison between Mandatory Rotation

Condition (Non-Final periods) and No Mandatory Rotation

Table 4.8: Rate of Agreement: Comparison between Mandatory Rotation

Condition (Final Periods) and No Mandatory Rotation Condition 43Table 4.9: Rate of Agreement: Comparison between Non-Final Periods and

Table 4.10: Correlation between Auditor Tenure and Asset Values within No

Mandatory Rotation Condition: All Negotiation Resulting in Agreement 45Table 4.11: Correlation between Auditor Tenure and Asset Values within No

Mandatory Rotation: Negotiations lasting Longer than Eight Periods 46

Table 4.12: Asset Values: Comparison between Earlier Periods and Later Periods

Table 4.13: Asset Values by Negotiating Pair: Comparison between Mandatory

Rotation Condition and No Mandatory Rotation Condition 48Table 4.14: Asset Values by Verifier or Manager: Comparisons between Mandatory

Table 4.15: Asset Values by Negotiating Pair: Comparison between Mandatory

Rotation Condition (Non-Final Periods) and N o Mandatory Rotation

Table 4.16: Asset Values by Negotiating Pair: Comparison between Mandatory

Rotation Condition (Final Periods) and No Mandatory Rotation

Table 4.17: Asset Values by Negotiating Pair: Comparison between Non-Final

Periods and Final Periods within Mandatory Rotation Condition 51

LIST OF FIGURES

“The Comptroller General of the United States shall conduct a study and review of the potential effects of requiring the mandatory rotation of registered public accounting firms.”

—Sarbanes-Oxley Act o f 2002, H R 3763-31

“All of the technical aspects [of auditing] are important but they can’t substitute for the primary skill, which is the art of negotiation.”

—Michael Buxbaum The CPA journal, 72(5), p.80, May 2002

1 INTRODUCTION

Auditors and clients may not always agree on accounting principles or practices, financial disclosure, or auditing scope or procedures Auditor-client disagreements often occur because the auditor and client have different preferences representing legitimate but divergent beliefs regarding the appropriate application of the current accounting/auditing standards (Magee and Tseng 1990; Dye 1991; Antle and Nalebuff 1991; DeAngelo et al 1994; Defond and Subramanyam 1998) In these circumstances, auditors and clients often attempt to resolve their disagreements through a process that resembles negotiation (Murnighan and Bazerman 1990; Antle and Nalebuff 1991; Dye 1991; Zhang 1999; Gibbins et al 2001)

Prior research suggests that costs associated with an auditor switch should influence auditor-client negotiations (Magee and Tseng 1990; Dye 1991; Antle and

Nalebuff, 1991; Teoh 1992; Zhang 1999) From the auditor’s perspective, there are at

least two types of switching costs: (1) inefficiencies arising from learning to audit a new client and (2) the loss of potential revenue before the lost client is replaced The second

type of switching costs to the auditor relates to auditor tenure more directly than does the first type of switching costs to the auditor It is easy to see that the magnitude of switching costs to the auditor resulting from losing the client is determined by, among other factors, the size of the client revenue, the duration of the engagement relationship,

and the auditor’s ability to find a replacement client Like the auditor, the client also

bears at least two types of switching costs: (1) inefficiencies arising from having to deal with a new auditor and (2) possible negative reactions to the auditor switch from outsiders who have unfavorable perceptions as to the cause of the switch The second type of switching costs to the client also relates to auditor tenure more directly than does the first type of switching costs to the client It is easy to see that a short-lived

relationship appears to be more problematic than a relationship that lasts for as long as

most other relationships An example of this type of switching costs to the client is that outsiders may perceive an auditor switch to be motivated by opining shopping and respond negatively to the client That is, when outsiders have such a negative perception

of the auditor switch, the client will likely suffer substantial costs imposed by investors, lenders, and regulators These costs are, in effect, “political” costs The purpose of this study is to investigate the potential effects of mandatory audit firm rotation (hereafter mandatory rotation) on auditor-client negotiations through its impact on the two particular types of switching costs: switching costs to the auditor resulting from lost revenue until the client is replaced and switching costs to the client resulting from outsiders’ negative perception of an auditor switch, or the political switching costs

Mandatory rotation has been proposed over decades as a regulatory mechanism to reduce the influence that clients are assumed to exert on their auditors (e.g., U.S Senate

1976, 1977; AICPA 1978; Berton 1991; SEC 1994a; Wolf et al 1999; Benson, 2002; Imhoff 2003) In response to the recent audit failures with some of the largest U.S corporations, several bills have been proposed in the House and Senate that contain provisions limiting the term of the auditor-client relationship as part of an effort to improve financial reporting and protect investors.1 The Sarbanes-Oxley Act of 2002 that recently has been signed into law calls for a study and review of the potential effects of mandatory rotation, due July 2003, in contemplation of possible Congressional action on this measure (Sec 207 (a), H.R 3763-31).

Mandatory audit-firm rotation may accomplish some of its intended goals, but there may also be other consequences that have not been sufficiently explored Prior studies argue that because auditor switching will occur more frequently, mandatory rotation will dramatically increase switching costs borne by both the auditor and the client and result in a decline in audit quality (Elitzur and Falk 1996; Arrunada and Paz- Ares 1997; Comunale and Sexton 2002) Prior studies also argue that because each client can be retained only for a limited number of years, mandatory rotation will dramatically reduce the incumbency value of each client to the auditor and therefore enhance auditor independence (Imhoff 2003; Gietmann and Sen 2002; Dopuch et al 2001; Brody and Moscove 1998; Petty and Cuganesan 1996; Wallman 1996; Copley and Doucet 1993) All of these studies, while focusing on different aspects of an audit engagement and making different cost/benefit analyses, share the same view that switching costs will go

up with the imposition of mandatory rotation No study has addressed the possibility that

The “Integrity in Auditing Act o f 2002” bill suggests that auditors should not be considered independent

if they have audited a firm for more than seven years The “Comprehensive Investor Protection Act o f 2002” suggests that non-independence is a problem when the auditor has consecutively audited a firm’s financial statements for more than four years The “Truth and Accountability in Accounting Act o f 2002”

because periodic switching is required, mandatory rotation will eliminate, for these

required switches, the switching costs associated with outsiders’ possible negative perceptions of an auditor switch, or the political switching costs Furthermore, no study has addressed the possibility that because the ability of the auditor to replace a lost client will increase dramatically as a result of more clients looking for new auditors, mandatory

rotation will reduce the switching costs associated with the auditor’s loss of the client

revenue Because mandatory rotation will likely alter switching costs to both the client and the auditor, it is important to investigate how such changes may influence auditor- client negotiations

The research question of this study is investigated using laboratory negotiation methods This type of experimental method, which has been used in other areas of accounting research such as transfer pricing, budgeting, and labor-management negotiations, has the potential to provide unique insights into the process of financial statement preparation that is otherwise unobservable Experiment participants assumed the role of a manager (client) or of a verifier (auditor) in a negotiation where they tried to resolve their disagreements over an asset value The client had an incentive to report a higher asset value, while the auditor had an incentive to report a lower asset value Half

of the negotiating pairs negotiated without restriction on auditor tenure whereas the other half had auditor term limits imposed The outcomes of the negotiations were measured by the rate of agreement and the agreed-upon asset values

The remainder of this dissertation is organized as follows Chapter 2 provides a review of the related literature and develops the hypotheses Chapter 3 describes the

research method Chapter 4 presents proposed analyses Chapter 5 concludes the paper with a discussion of the implications and limitations of the study.

2 THEORY AND HYPOTHESES

This chapter first defines the term “auditor-client disagreement” as used in this study It then discusses the differing incentives facing the client and the auditor that drive auditor-client negotiations and develops hypotheses comparing the outcomes of auditor- client negotiations in the presence and absence of mandatory rotation

2.1 Auditor-Client Disagreements

The term “auditor-client disagreement” as used in this study refers to any situation

in which the auditor and the client have different preferences with respect to the financial statements The term thus has a broader meaning in this study than in the frequently cited SEC definition that focuses only on reportable disagreements—disagreements that, if not resolved to the auditor’s satisfaction, would have caused reference in the auditor’s report (SEC 1974b) Reportable disagreements occur when “an authorized representative of the client persists in a position and the audit partner in charge of the engagement reaches an unequivocal contrary position as to a material matter” (SEC 1974b) Clearly, the auditor can reach an unequivocal position only when the professional standards provide unequivocal solutions The SEC’s definition does not address disagreements where the professional standards are ambiguous and both the auditor and the client’s positions are defensible within these standards (Kay 1976) For example, the auditor may disagree with

“managers’ subjective estimates that cannot, by definition, be objectively verified by auditors prior to occurrence” (Francis and Krishnan 1999, 135)

Magee and Tseng (1990) argue that a dismissal threat from the client, when theauditor disagrees with the client’s position on a reporting issue, is credible only ifauditors in the market disagree among themselves over the same issue They furtherargue that auditors in the market differ in their opinions on an accounting or auditingissue only if the current professional standards allow them to do so These argumentsexplain why auditor-client disagreements that arise in the area where the professionalstandards are ambiguous and vague are particularly worth investigating As the SEC(2001, 16) states in its recent ruling on auditor independence:

It is rarely the black-and-white issues that an auditor faces The danger lies

in the gray area—where the pressure to bend to client interest is subtle, but

no less deleterious

From an economic perspective, auditor-client disagreements occur because each party faces different incentives That is, the possibility for disagreements to exist within the current professional standards is not a sufficient condition for disagreements to occur Rather, the current professional standards are used to legitimize disagreements that are driven by conflicting incentives If both parties have similar incentives, one would expect them to agree Only when the auditor and client face different incentives does one expect them to disagree Taking such an economic perspective, the SEC (2001) suggests that when deliberating the need for accounting regulations, legislators are not supposed to expect anyone, including both auditors and clients, to behave in one way when there is an incentive for him or her to behave in another way

2.2 Incentives Facing the Client

Executives and managers of public companies face pressure to report earnings that meet financial analysts’ forecasts, that do not violate requirements of debt covenants, and that maximize the value of management compensation plans or meet some other needs that are tied with their personal welfare These incentives produce a desire for higher earnings and/or higher asset values in many cases (Beneish 1999; Coakley and Loebbeck 1985; Hylas and Ashton 1982), although in some cases lower earnings could

be preferred (Barta and McKinnon 2003; DeAngelo et al 1994; Antle and Nalebuff 1990) Because o f auditors’ risk structure and the fact that higher earnings or asset values increase auditor legal liability (Lys and Watts 1994), it is assumed that auditors are more likely to disagree with a client who proposes higher rather than lower earnings or asset values Client incentives in this study are to increase asset values and are consistent with several prior studies (e.g., Tuttle et al 2002; Bell et al 1998; Kinney and Martin, 1994)

Switching costs become part of the client’s incentive to negotiate with the auditor when auditor-client disagreements occur If the disagreement is not resolved to the auditor’s satisfaction, the auditor may choose to issue a qualified opinion To respond, the client may consider switching auditors A significant body of research suggests that auditor-client disagreements and qualified opinions can result in auditor switches (e.g., Fried and Schiff 1981; McConnell 1984; Sarhan et al 1991; Dhaliwal et al.1993; Whisenant and Sankaragurusamy 2000; Chow and Rice 1982; Craswell 1988; Citron and Tafflet 1992; Krishnan 1994; Krishnan and Stephens 1995; Krishan et al 1996)

Prior studies attempt to explore, analytically, how switching costs affect the client’s position in auditor-client negotiation (e.g., Magee and Tseng 1990; Dye 1991;

Teoh 1992; Antle and Nalebuff 1991) It has been argued that the client’s decision on whether to switch auditors is made after weighing the potential benefits against the potential costs of an auditor switch The client knows that auditors differ in their approaches to risk and that potential successor auditors may be willing to adopt a less conservative stance than is the current auditor willing to This possibility, however, must

be balanced against likely switching costs to the client If the client is convinced that it is worth taking the risk to switch to a new auditor, opinion-shopping may occur 2 On the other hand, if the client believes that it is not economically wise to switch auditors at this time, the client will be more willing to cooperate with the auditor in order to avoid an auditor switch

There are at least two types of switching costs facing the client One is efficiency- related transaction costs related to finding and working with a new auditor The other is costs associated with outsiders’ possible negative perception of the cause of an auditor switch, i.e., the political switching costs The latter type of switching costs exists because auditor switches can signal auditor-client disagreements (Schwartz and Menon 1985; Chow and Rice 1982; Fried and Schiff 1981; Dhaliwal et al 1993) Even though only a relatively small portion of auditor switches were actually caused by auditor-client disagreements (Whisenant and Sankaraguruswamy 2002), voluntary auditor switches invite outsiders’ suspicions that are likely to be unfavorable to the client Because outsiders may not be able to verify the true cause of an auditor switch, suspicions exist regardless of what actually have caused the auditor switch As long as an auditor switch

2 The SEC’s disclosure requirements for auditor switch may not be an effective deterrent to opinion-

is made voluntarily, i.e., not required by law, it is always possible that outsiders will perceive an auditor switch as motivated by opinion-shopping.

The political switching costs addressed in this study should not be confused with

the concept of political cost discussed in Watts and Zimmerman’s (1986) Positive

Accounting Theory Although large public companies tend to incur high political

switching costs, firm size is not the primary determinant of the political switching costs Small companies, like large companies, face possible political switching costs imposed

by their creditors and others who often closely monitor the organizations in which they have interests Creditors of small companies, in particular, are expected to be very aware

of auditor-client relationship because small companies tend to be financially less stable

The political switching costs are important because they create a constraint on client management behavior That is, the existence of substantial political switching costs can exert a great pressure on client management to accept their auditors’ position The constraint of the political switching costs on client management’s behavior can be significant not only because such costs can be great, but also because they can be imposed by the auditor When audits are based on single-period contracts, as modern-day audits typically are, not only can the client choose to dismiss an auditor, but also the auditor can choose to resign from or not to renew the engagement and thus impose switching costs on the client (Krishnan and Krishnan 1997; MacDonald 1997) DeFond et

al (1992) and Wells and Loudder (1997) investigate the market reaction to auditor resignation announcements and document negative stock price reactions to such announcements, providing evidence that auditor resignation can effectively impose a political cost on uncooperative clients The auditor’s ability to impose the political

switching costs on the client thus becomes an important part of the client’s incentive to negotiate with the auditor, particularly when such costs are significantly greater than the potential benefits of an auditor switch.

It is hypothesized that mandatory rotation will change the client’s incentive to negotiate with the auditor in different engagement years in different ways The political switching costs will continue to exist when an auditor-client disagreement occurs at the final stage of the audit in years prior to the final engagement year Auditor switches before the term is up are still voluntary and potentially signal negative information about the auditor-client relationship As a result, the client will continue to have an incentive to cooperate with the auditor However, the political switching costs to the client will completely disappear if a disagreement regarding audited financial statements occurs in the final engagement year Because the client is mandated to switch to a new auditor in the following year, the auditor switch that follows the current engagement cannot be a negative signal about the client Consequently, the client’s incentive to compromise will

be substantially lower when a disagreement occurs in the final engagement year than in any previous years

2.3 Incentives Facing the Auditor

The auditor also incurs switching costs, including efficiency-related costs of learning to audit a new client and costs associated with acquiring a new client, as well as the loss of revenue until the lost client is replaced Like the client, in determining whether

to compromise in the event of an auditor-client disagreement, the auditor weighs potential costs of an auditor switch against the potential benefits such as eliminating the

litigation risk (Moore and Scott 1989; Magee and Tseng 1990; Teoh 1992; Dye 1993; Krishnan and Krishnan 1997; Lee and Gu 1998) Evidence shows that auditors are less willing to accommodate income-increasing accounting methods used by clients experiencing poor or deteriorating financial performance because the litigation risk with such clients is high (Dhaliwal et al 1993; Pourciau 1993; Defond and Jiambalvo 1993; Defond and Subramnayam 1998; Lys and Watts 1994; Francis and Krishnan 1999) However, when the estimated cost of losing the client is greater than the potential benefit

of being uncompromising, auditors have an incentive to compromise (Magee and Tseng 1990; Antle and Nalebuff 1991; Raghunathan et al 1994; Lee and Gu 1998, Zhang 1999; Schatzberg and Sevcik 1994; Schatzberg et al 1996; Calegari et al.1998)

The costs of losing a client in terms of lost revenue are determined by, among other factors, the expected length of the relationship with the client and the auditor’s ability to replace the client Under the current institution, where auditor tenure is unlimited and auditor-client relationships tend to last for many years, the pool of replacement clients is quite small Beck (1988) et al show that the average auditor tenure with publicly-traded companies in their sample is about 15 years with 74% of the sampled auditor tenures being 17 years or more Auditor-client relationships are even longer with larger companies (Monti-Belkaoui and Belkaoui 1991) If the average auditor tenure is 15-17 years, then only about 6% of all clients are in the market for an auditor in any given year Hence, when a client switches auditors, the previous auditor may be unable to replace the lost client for a lengthy period of time Particularly at the office level of large audit firms where auditor-client negotiations take place, the probability of immediately replacing a lost client is low The partner doing the negotiations with the

client faces an immediate threat of losing revenue, even though the overall portfolio of clients at firm level may be not substantially affected.

It is hypothesized that mandatory rotation, if implemented, will force a significant portion of clients into the market for a new auditor every year and thus substantially increase the auditor’s prospects of replacing a client who voluntarily switches auditors That is, by changing certain structure of the audit market, mandatory rotation will substantially reduce the potential cost of losing a client and thereby reduce the auditor’s incentive to compromise when negotiating with the client For example, adopting a four- year mandatory rotation policy would put at least 25% of all clients in the market for an auditor every year Thus, even at the office level of audit firms, the probability of replacing a client would increase relative to current levels Mandatory rotation is thus predicted to reduce the auditor’s incentive to compromise and consequently result in relatively fewer agreements reached in auditor-client negotiations

In the final engagement year, lost revenues from the client do not differ whether the auditor chooses to compromise or not Differential cost analysis holds that costs that

do not differ between two alternatives are irrelevant to decisions (Garrison and Noreen 2003; Jiambalvo 2001) Thus, loss of potential revenues is irrelevant to the auditor’s decision during the negotiation with the client in the final engagement year since the client is no longer legally retainable regardless of the auditor’s actions Consequently, the auditor has even less incentive to compromise in the final engagement year relative to prior years under the condition of mandatory rotation Because the auditor’s incentive to compromise is further reduced in the final engagement year, mandatory rotation will

likely produce fewer agreements, i.e., a lower rate of agreement, in the final engagement year relative to years leading up to the final year.3

2.4 Hypotheses

To summarize the preceding discussions, mandatory rotation will change both theauditor’s and the client’s incentives to compromise in the following ways First, in years

before the final engagement year, mandatory rotation will not reduce the client’s

incentive to compromise because the political switching costs continue to exist On the

other hand, mandatory rotation will reduce the auditor’s incentive to compromise in these

years because the auditor is more likely to be able to replace a lost client Hence, a lowerrate of agreement can be expected in these years compared to the rate of agreement underthe current institution without mandatory rotation Second, in the final engagement year,auditors’ incentive to compromise is further reduced because lost revenues are no longer

a consideration and at the same time, the client’s incentive to compromise is alsosubstantially reduced because the political switching costs are eliminated Hence, it ispredicted that the rate of agreement will be lower in the final engagement year than inprior years within the condition of mandatory rotation These arguments lead to H I :

H I: Mandatory rotation will result in a lower rate of agreement between auditors and clients

If an agreement is reached in the negotiation, the outcome of the agreement, i.e., whether the disagreement is resolved in the client’s or in the auditor’s favor, is another

3 Some researchers argue that trying to establish a desirable reputation could either “harden” the auditor’s position during the final engagement if independence is concerned (Chi et al 2002), or “soften” the auditor’s position if a perception o f hostility to the client is concerned (Comunale and Sexton 2002) Others argue that mandatory auditor rotation does not allow the auditor to develop a reputation (Gietzmann and Sen 2002) Examining the effects o f these factors is beyond the scope o f this study.

important issue that needs to be investigated separately The quality of financialreporting, or financial statement values reported to the public, has been the primaryconcern leading to the increasing volume of calls for mandatory rotation One of theassumptions held by many proponents of mandatory rotation is that there is a tendencyfor auditors, over time, to gradually align with the wishes of their clients For example,the Metcalf Committee report (U.S Senate 1976, 21) asserts that:

Long association between a corporation and an accounting firm may lead

to such close identification of the accounting firm with the interests of its client’s management that truly independent action by the accounting firm becomes difficult One alternative is mandatory change of accountants after a given period of years

However, research has found little evidence suggesting a negative correlation between auditor independence and auditor tenure To the contrary, evidence suggests that the auditor is more willing to compromise in earlier years than in later years (Geiger and Raghunandan 2002; Myers et al 2002) These studies raise questions about whether mandatory rotation will be effective in making auditors less willing to go along with their clients Because the effects of auditor tenure on auditor independence have not been investigated using experimental methods, i.e., in a setting where the complex confounding effects that archival studies found difficult to deal with can be effectively controlled for, these findings may not be conclusive Assuming auditors are conservative

so that they prefer reporting lower earnings and lower asset values in audited financial statements, disagreements occur when clients prefer reporting higher earnings and higher asset values If the extended auditor tenure indeed gradually erodes auditor independence, then the final reported agreed-upon asset values should gradually increase as the auditor

tenure extends To test this lay notion of the relationship between auditor tenure and auditor independence, the following hypothesis is proposed:

H2: In the absence of mandatory rotation, agreed-upon asset values will be

higher in later engagement years than in early engagement years

The existing negotiation literature generally does not provide a satisfactory theory that can be used to predict agreed-upon asset values resulting from auditor-client negotiations Auditor-client negotiations are rather unstructured in nature and the information setting of auditor-client negotiations, such as the payoffs to the auditor and the client, are rather unique (Zhang 1999) Generally, it is difficult to generate complete equilibrium predictions about unstructured negotiation outcomes using the tools of standard noncooperative game theory because “the rich message-space and the real-time nature of unstructured negotiation create an unmanageably large set of strategies” (Davis and Holt 1993, 244) The alternative cooperative game theory, which introduces a social perspective into economic analysis, deals with issues that are outside the scope of this study Thus, predictions about differences in agreed-upon asset values are based on the same type of reasoning as predictions about the likelihood of agreement used to support

H I That is, these predictions are based on an economic analysis of qualitative differences in the auditor’s and the client’s incentives to compromise under the two specific conditions, rather than on any existing game-theoretic model

As suggested in discussions related to H I, mandatory rotation changes more than the length of auditor-client relationship It changes the nature of this relationship That is, imposing limits on auditor tenure will result not only in shortened auditor-client relationships, but also in fundamental changes in both the auditor’s and the client’s incentives to compromise Specifically, auditors will have less incentive to compromise

in any engagement year while clients have less incentive to compromise primarily in the final engagement year Such changes in incentives to compromise also serve as the theoretical foundation for predicting differences in agreed-upon asset values The

hypothesized impact of mandatory rotation relates to the relative differences in agreed-

upon asset values and not how much they differ in absolute terms Using Roth’s (1995, 260) terminology, hypotheses regarding agreed upon asset values are “qualitative predictions” not “point predictions.”

It has been argued in developing HI that the expected lost revenues the auditor would bear for not compromising are considered during negotiations Under the current institution, it is difficult for the auditor to replace a lost client As such, the current audit market can be characterized as a buyer’s market in which it is easier for a client to replace an auditor than for an auditor to replace a client (Monti-Belkaoui and Belkaoui 1991; Gietmann and Sen 2002) In order to reduce the risk of losing the client, the auditor may be willing to accept a higher asset value proposed by the client If mandatory rotation is implemented, however, the probability of replacing a lost client will increase significantly, changing the audit market to be more like a seller (auditor)’s market, which

in turn results in a substantial reduction in the auditor’s incentive to compromise These insights suggest that the agreed-upon asset values will be lower with mandatory rotation than without mandatory rotation, which leads to H3:

H3: Mandatory rotation will result in lower agreed-upon asset values.

This effect of mandatory rotation on the audit market is expected to exist regardless of whether it is the final engagement year or the prior years However, the client’s incentives to compromise are also reduced in the final engagement year so that

the dampening effect of mandatory rotation on agreed-upon asset values might be somewhat lessened in the final year The analysis will explore this possibility.

3 METHOD

3.1 Experimental Task

Participants assume either the role of a manager (client) or an outside verifier (auditor) who together must negotiate a value to be reported for an asset In all negotiations, the manager and verifier share the same set of information indicating the range and probability of the actual value of the asset, as shown in Figure 3.1, but have conflicting incentives to choose a value within the range to report The manager has an incentive to report a higher value in order to maximize compensation, while the verifier has an incentive to report a lower value in order to avoid or reduce the risk of paying a penalty for reporting too high The actual value of the asset for each negotiation period was predetermined using a computer-generated random process Both managers and verifiers learn the actual value of the asset only after the current negotiation period is

4

over

Figure 3.1: Asset Values Distribution

a judgment about whether the client’s estimate falls into a reasonable range This range can easily be quite large so that a difference between two numbers within the range can

be material The true value of the inventory can be any number within the range according to certain probabilities attached to each number The client may in fact agree with the auditor’s judgment, but prefers to report a higher value in the range while the

auditor prefers to report a lower value in the range Since choosing any value to report

within that range is defensible ex ante, the disagreement between the auditor and client

are thus driven by differing incentives, which provides room for negotiation

Negotiations were conducted via a computerized negotiation system (Wolfe and Murthy 2003; Lim and Benbasat 1993) Media richness theory indicates that many of the cues available in face-to-face communication are absent in text-based communications via a computerized negotiation system (Daft and Lengel 1986) Prior evidence suggests that these cues can lead to increased pressure tactics and that fewer impasses result when the communication medium does not permit these cues (Lewis and Fry 1977) On the other hand, the lack of visual and social cues can also lead to a reduction of social pressure to cooperate, which in turn can result in a greater number of impasses (Roth 1995) Using a transfer-pricing task, Kachelmeier and Towry (2002) conclude that computerized negotiation results in fewer fairness-based price concessions than in face- to-face negotiations In their study, however, the computer did not allow the participants

to communicate with each other except to propose transfer prices and make counter offers As such, the influence of social factors remains an open question in computer mediated negotiations Importantly, all participants in this study used the same computerized negotiation media, therefore controlling for possible effects of media on the negotiation At the same time, the computer permits greater experimental control and makes it possible to easily track the messages sent between negotiators

Yes Political switching costs No Political switching costs

Mandatory Rotation

The between-subject variable is the presence/absence of mandatory audit-firm rotation (MAR/NoMAR) The within-subject variable is the presence of the political switching costs in the first two periods and absence of such costs in the third (final) period in the MAR condition with the same auditor/client pair

3.3 Independent Variables

The presence/absence of mandatory rotation was manipulated by limiting the number of negotiation periods of the same manager/verifier pair in the MAR condition while allowing manager/verifier pairs to negotiate as long as the verifier continues to accept the manager’s proposed values in the NoMAR condition Participants in the MAR

condition were told that “each manager/verifier pair continues to negotiate with each

other for up to three periods as long as the verifier continues to accept the manager’s

submitted values.” Participants in the NoMAR condition were told that “Each manager

and verifier pair continues to negotiate with each other for an unlimited number of

periods as long as the verifier continues to accept the manager’s submitted values.”

All negotiations took place in a laboratory in groups of nine participants consisting of four managers and five verifiers The addition of one more verifier than manager created a buyer (client)’s market such that if a negotiation failed, the manager was guaranteed to be reassigned to a new verifier, while the verifier might or might not

be reassigned to a new manager, in the next period In the NoMAR condition, if a negotiation failed, the verifier had a chance to be reassigned to a new manager only when another manager/verifier pair failed to reach an agreement In the MAR condition, if a negotiation failed, the verifier had the same chance to be reassigned to a new manager as

in the NoMAR condition plus the chance to be reassigned to any manager that had reached the three-period term limit with his or her verifier

The presence/absence of political switching costs was manipulated by imposing a penalty on managers who failed to reach an agreement with the verifier in either the MAR or the NoMAR condition, except in the third (final) period with the same verifier in the MAR condition The penalty, which was 40% of the manager’s earnings for any period in which the penalty was imposed, was predetermined using a computer simulation such that the expected earnings for a manager in each period would be in the middle range of the given probability distribution of the actual asset value This penalty

was assessed against the manager when the period ended with no agreement, i.e., no value proposed by the manager had been accepted by the verifier in that period.

of only those negotiations that produce an agreement

3.5 Negotiation Procedure and Rules

The order of the conditions (MAR or NoMAR) conducted was predetermined by flipping a coin Participants knew only from the recruiting advertisement that they signed

up for participating in a “business game experiment” with the average cash payment of

$10 per hour Before each session began, participants were randomly assigned to the role

of either a manager or a verifier that they would have to play for the entire session and given time to read written instructions (See Appendix) Each participant was given an ID number according to the role played, such as “Manager 1” or “Verifier 5,” and used only one ID number throughout the entire session Each negotiating pair knew each other during the experiment only by ID number, which showed on both negotiators’ computer screen in the beginning of each period before they started to negotiate

All written instructions provided the same set of information regarding the range and the associated probabilities for the possible actual values of the asset as shown in

Figure 3.1 They differed only in regards to term limits Both managers and verifiers received the same asset value distribution for use in all periods of all sessions The written instructions also provided negotiation rules and explanations of payoffs The experiment began with two practice periods in each session to allow participants to have hands-on experience with the entire sequence of events and learn the payoff scheme without financial risk Each negotiating pair was assigned together for only one practice period, i.e., pairs switched after the first practice period, and pairs switched again after the second practice period to insure anonymity between each pair to begin the next period Participants were given the opportunity to ask questions after each practice period.

Each period of negotiation was programmed to last two and one-half minutes.5

To begin each period of negotiation, the manager first submitted to the verifier a value for the asset to be reported within the range of the possible actual values of the asset as presented in Figure 3.1.6 At this point, the verifier could accept the asset value or initiate

a negotiation Once the manager had submitted an asset value, the verifier and the manager were free to negotiate by submitting typed messages back and forth and by the manager re-submitting new asset values on the computer screen (see Appendix for sample screens for both managers and verifiers) The verifier could accept the value submitted by the manager at any time before the current period ended Once the verifier accepted a value submitted by the manager, the current period of negotiation ended and

5 Two or three minutes are commonly used negotiation times in the existing literature (see Roth 1995) A pretest suggests two and one-half minutes is an adequate time for the participants to complete each period

o f negotiations.

6 Research indicates that the outcome o f a negotiation is affected by which party moves first (Ochs and Roth 1989, Bolton 1991) For auditor-client negotiations over a reporting issue, it is reasonable to assume

the actual value of the asset and earnings for that period were shown on the computer screen (each party saw only his or her own earnings on the computer) The actual value

o f the asset was determined for each period prior to the experiment using a computer­generated random process in accordance with the range and probability distribution of the possible actual values of the asset

If the verifier did not accept the value submitted by the manager before the negotiation time ran out, the period ended without an agreement If the negotiation ended without an agreement, the manager was paired with a new verifier in the following period The verifier, however, was reassigned only if another manager was available Verifiers were not reassigned to the same manager in consecutive periods In the MAR condition, each manager/verifier pair continued to negotiate for up to three periods as long as the verifier continued to accept the manager’s submitted values In the NoMAR condition, each manager/verifier pair continued to negotiate for an unlimited number of periods as long as the verifier continued to accept the manager’s submitted values Participants were not told how many total periods the experiment would last; the pre­printed record sheet they were given to record their earnings indicated that the experiment might last for as many as 22 periods Given that the average auditor tenure under the current institution is about 15-17 years, a total of 20 periods of negotiations were conducted in each session Data from the two practice periods (periods 1 and 2) as well as the last period (period 20) were excluded from hypothesis testing.7

7 Period 20 was excluded from hypothesis testing because participants in the later session might have correctly guessed that the experiment was going to last 20 periods However, the same tests were also performed using the data including the outcomes o f period 20 and the results were unchanged.

3.6 Payoffs

Participants received the payoff trees in Figure 4.3 as part of their instructions A manager’s earnings in each period depended on whether the verifier accepted a submitted value and, in the MAR condition, whether it was the third (final) period with the same verifier If the verifier accepted a submitted value, the manager earned that value regardless of the asset’s actual value If the verifier did not accept a submitted value and

it was not the third (final) period with the same verifier in the MAR condition, the manager earned the asset’s actual value minus a 40% penalty (representing political switching costs) for not reaching an agreement with the verifier There was no penalty to the manager for not reaching an agreement in the third (final) period with the same verifier in the MAR condition For example, if a manager proposed $350 and this was accepted by the verifier, the manager earned $350 regardless of the actual asset value If the verifier did not accept the manager’s proposed values and the negotiation ended without agreement, the manager earned the actual asset value minus any applicable penalty Thus, managers had an incentive to propose higher values

Figure 3.3: Payoffs

1 For participants in no mandatory rotation group:

Managers’ Payoff:

Accepted valueYes

YesAccepted value > Actual value

No

$100

Figure 3.3: Payoffs (cont.)

1 For participants in mandatory rotation group:

60% Actual value (in any first and second

periods with the same verifier), or 100%

Actual value (in any third period with the same verifier)

Verifiers’ Payoff:

Yes

$100 - 50% (Accepted Value - Actual value)

Accepted value > Actual value ^

No

$100

In general, verifiers earned a flat fee of $100 for each period they were paired with a manager Verifiers earned $0 if they were not paired with a manager A verifier’s earnings in each period depended on whether a value submitted by the manager was accepted and whether the accepted value was greater than the actual value of the asset If the accepted value was greater than the actual value, the verifier earned a flat service fee

asset’s actual value If the accepted value was equal to or less than the actual value, the verifier earned the flat fee The between-subjects condition (MAR or NoMAR) did not affect a verifier’s earnings in any period when he or she was assigned to a manager, but did affect a verifier’s likelihood of remaining unassigned, and thus the cumulative earnings For example, if a verifier accepted a manager’s proposed asset value of $350 and the actual asset value was $350 or greater, the verifier earned his/her fee of $100 If the actual asset value was less than $350, for example 250, then the verifier earned the fee less one-half of the difference, or $50 ($100 - ($350 - $250) / 2) If the verifier rejected all of the manager’s proposed values, he/she earned the $100 fee but lost this client after this period Thus, verifiers had an incentive to accept lower values.

Because expected earnings were different (asymmetrical) for managers and verifiers, it was necessary to reward mangers and verifiers separately The final cash payment for each participant of the experiment was determined by the rank of his or her cumulative earnings compared to the cumulative earnings of others who played the same role (manager or verifier) in that session Specifically, each participant in each session was paid according to the following schedule:

This type of payoff scheme has been used in prior experimental negotiation studies and proven effective in making economic incentives salient while reducing the effect of non-economic factors such as consideration of fairness (Bolton 1991) Paying participants by the ranks of their total earnings provides incentives for every participant

to perform at his/her best in contrast to the negative effects sometime experienced in a winner-take-all tournament on the lower performing participants.8

3.7 Participants

The objective of this study is to investigate how auditors and clients negotiate when they are facing different economic incentives Given the impossibility of arranging for a group of auditors to meet their respective client-managers in the field, an alternative setting and sample were used The participants were recruited via an online recruiting system among graduate students in the business school at a large state university Fifty- four graduate business students were selected to participate in 6 sessions of the experiment Table 3.1 summarizes the demographics of these participants

About 44 percent of the participants were women The typical participant was26.2 years old with 3.7 years of work experience In the post-experiment survey, the participants were asked to answer the questions “Have you ever bargained a deal with someone (e.g., purchasing a car, a used book, etc.)?” and “Do you consider yourself a good bargainer?” Participants gave their responses by circling a corresponding number

on each 7-point Likert-type scale with “ 1” representing “Never” and “7” representing

“Very often.” About 80% of the participants circled a number at the middle or upper end

of the scale (4-7) for the first question and about 65% of the participants circled a number

at the middle or upper end of the scale (4-7) for the second question

8 One extra session was conducted using an alternative payoff method to reward each participant based on his or her cumulative earnings as a proportion o f the sum o f the cumulative earnings o f all participants in the same role (manager or verifier) The results o f this session were very similar, both in the number o f agreements reached and the average agreed-upon values, to the results o f those sessions under the same condition but using the “ranking” payoff method The “ranking” method was chosen over the

“proportional” method because o f ethical considerations that the “proportional” method potentially could