Journal of Accounting and Public Policy 23 (2004) 415440 www.elsevier.

com/locate/jaccpubpol

The auditors going concern decision and Types I and II errors: The Coase Theorem, transaction costs, bargaining power and attempts to mislead
Paul Barnes
*
Nottingham Business School, Burton Street, Nottingham NG1 4BU, UK

Abstract It is shown that, in the absence of transaction costs and in line with the Coase Theorem, the going concern decision is ecient in the sense that bias arising from either Type I or II errors is not expected. However, when transaction costs in the form of legal costs, are introduced, bias is expected. The direction of the error depends upon the auditors relative bargaining power. It is also shown that its relative bargaining power provides an incentive for the client company to mislead. Finally, certain empirical observations pertinent to this analysis are discussed together with the regulatory implications. 2004 Elsevier Inc. All rights reserved.
Keywords: Audit; Going concern; Bargaining power; Misinformation

Tel.: +44 115 848 2824; fax: +44 115 848 6175. E-mail address: paul.barnes@ntu.ac.uk

0278-4254/$ - see front matter 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.jaccpubpol.2004.10.003

416

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

1. Introduction An investor needs to know whether a company in which he is interested is in immanent danger of failure. The easiest way to do this is to examine the audit report and whether in the auditors opinion there is substantial doubt about the companys ability to continue as a going concern for a reasonable period of time (SAS # 59). As the auditor is a trained professional with inside information into the commercial future of his client, the investor should be condent in the inferences he makes from the auditors report. Normally, the report would be unqualied but contain a paragraph expressing such uncertainty. Also, if the auditor concludes that the nancial statements inadequately indicate the companys inability to continue, the auditor should express either a qualied or adverse opinion in his report. A similar requirement and set of procedures exist in the UK. See ICAEW (1994). It will probably matter little to the investor as to what form such a conclusion may take in the auditors report and for the remainder of this paper the term adverse report will be used to cover them all. The going concern decision has been the focus of a considerable amount of academic research over many years, primarily because it is an excellent example in which its independence may be tested. See Antle (1984), Asare (1990), Bartlett (1997), Chow and Rice (1982), Johnson et al. (1989), Jones (1996), Kida (1980), Knapp (1985), Krishnan and Stevens (1995), Lavin (1976), Lee and Stone (1995) and Pearson (1987) which are just a few paper to appear in academic journals. More recently, the decision may be seen to involve even greater stakes, given the concerns, particularly in the USA, about limited liability and the provision by audit rms of non-audit consultancy services (Davis et al., 1993). Early research on the going concern decision focussed on audit quality involving (a) the possibility of incompetence (due to a lack of practical appreciation and understanding of the industry in which the client company operates) and (b) lack of independence (due to economic considerations, such as audit switching, aecting the audit rm that may arise from an adverse report). Research strongly supports the hypothesis that auditors are competent at making the going concern decision (the competence hypothesis). However sometimes, they do not issue an adverse report when they should, perhaps because of the fear of loss of the audit and the nancial consequences to the audit rm (the independence hypothesis). For evidence to support both the competence and independence hypotheses see Mutchler (1984, 1985), Campisi and Trotman (1985), Menon and Schwartz (1987), Barnes and Huan (1993), Krishnan and Krishnan (1996), Matsumura et al. (1997) and Lennox (1999a) but for evidence to reject the independence hypothesis, see Louwers (1998). The relationship between the size of the auditing rm and independence has also been raised as well as the increased diculty the auditor faces in maintaining his objectivity in the face of the potential loss of a large client paying substantial audit and consultancy fees, i.e. there is greater economic dependence (DeAngello, 1981c; McKeown et al.,

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

417

1991; Carcello et al., 2000; Lennox, 1999b). 1 More recent studies have recognised the importance to the audit rm of its reputation and the importance of auditor independence as a means of protecting it. See DeFond et al. (2002) who also cite a large body of research that shows it is in the auditors interest to remain independent. 2 Further, Reynolds and Francis (2000) show how very large audit rms may act conservatively for larger clients (for example, by making an adverse report), suggesting that reputationprotection and fear of litigation dominate audit considerations. But, of course, all these studies were prior to Enron. The Reynolds and Francis (2000) ndings also contrast with other empirical evidence, particularly in the UK, which indicate that (A) In a number of major rm collapses (for example the Robert Maxwell Group and Mirror Group Newspapers) the auditors had not made an adverse report (Type II errors), suggesting that for, perhaps economic reasons other than reputation eects, they were reluctant to do so. 3 Of course, all these examples are surpassed by the Enron case. 4 (B) In many cases where an adverse report was made, the client companies did not fail (Type I errors). See Peel (1989), Citron and Taer (1992), Barnes and Huan (1993) and Lennox (1999a) for UK evidence. What is also remarkable is that there are no cases of the surviving rms suing their auditors for damages arising from the adverse report. 5
1 The comment by DeAngello (1981c), although old, is typical: consumers view auditors with established reputations as having more to lose from misrepresentation. Therefore, we expect auditor reputation to be positively associated with the size of the audit rm. Empirically, this has been shown to aect its going concern decision (Krishnan and Krishnan, 1996). 2 The reader is also referred to Chaney and Philipich (2002) who show by means of an event study how the Enron scandal aected the credibility of the annual reports of other Andersen clients, providing an interesting case study of the impact and fragility of auditor reputation. 3 For a discussion of these and some evidence concerning audit quality in which large audit rms are perceived to have a higher reputation for better quality audits see Lennox (1999b). Some of the audit rms involved in these company failures subsequently suered considerably. 4 Although the principal auditing and accounting issues in the Enron case related to departures from GAAP, there was no adverse report. Probably the most useful early summary of the issues involved is Benston and Hartgraves (2002). 5 The best evidence for this is an M.A. dissertation supervised by the author (Kandasamy, 1998). This shows that after a considerable search of UK accountancy and law magazines and reports, discussions with practitioners etc., not a single case could be found. There are many possible reasons for this: as these companies survived, there may have been little demonstrable damage suered; the low priority of suing for negligence when the company is continuing, e.g. the opportunity costs involved; and, of course, the possibility of the company still being in danger of failing. It should also be remembered that the companies suering are also, almost by denition, short of funds to mount a legal action and are, of course, relatively small. Nevertheless, there is a likelihood of such an action which has a real cost to the auditor which needs to be taken into account. In the US, the UK and many other countries, often legal actions do not come to Court but are settled outside. Given the reasons for the diculties of bringing the case to Court mentioned here, it is very likely that they would be settled in this way.

418

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

It is the purpose of this paper to examine the inter-relationships of these and their eects. It is arranged as follows. First, these costs will be examined in the context of a simple rational economic decision-making model excluding transaction costs. It is shown how the Coase Theorem applies if eciency is extended to include the eects of unbiased information. That is, in the absence of transaction costs, the going concern decision is unlikely to be biased. Transaction costs are then introduced in the form of legal costs. It is shown that either a Type I or II error may occur, simply depending upon the bargaining power of the auditor/client. The paper then moves on to consider the situation where there is information asymmetry oering scope for the client company to mislead and the auditors skills in handling this are a factor. This is followed by a short discussion of the conclusions and their regulatory implications. The approach of this paper is along similar lines to that originally developed by Antle (1984) who examined auditors incentives and possible actions using a rational economic model. It also follows (although it does not focus on auditor eort levels and extended procedures) Antle and Nalebu (1991), Krishnan and Krishnan (1996), Boritz and Zhang (1999) and Lennox (1999a) by examining the risk of Types I and II errors and the economic trade-os by placing the auditors decision in a bargaining model. Its origins are a paper by Zhang (1999) who used a bargaining model to examine the eects of auditor and client incentives. He showed that an auditors independence will be preserved if the rm-specic quasi-rents from an audit (the value arising from the dierence between expected audit fees and costs in future engagements with the client) are zero. It is compromised if they are positive. Unfortunately, Zhang (1999) did not examine the possibility of the client company providing misleading information and the eect that this may have. For this we may use the insights provided by Coase (1960). Coase showed that, in the absence of transaction costs, the parties to an economic transaction will continue to negotiate until there is a Pareto ecient outcome as there is no incentive to bargain further. He showed that ineciency can arise either through explicit transaction costs or imperfect information about the gains from bargaining. The critical assumption, therefore, is that bargaining costs are zero and information is perfect. 6 Saraydar (1983) has shown that where information is not perfect and there is asymmetry of information, then there is an incentive for participants to provide misleading information (or as Saraydar calls it dissimulation).
6 Coase stresses the importance of transaction costs for explaining real world phenomena. In the presence of transaction costs, agents will cease bargaining if these costs exceed eciency gains: Such a rearrangement of rights will only be undertaken when the increase in the value of production consequent upon the rearrangement is greater than the costs which would be involved in bringing it about . . .In these conditions the initial delimitation of legal rights does not have an eect on the eciency with which the economic system works.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

419

Bargaining power has been examined in the context of private information (for a review see Kennan and Wilson, 1993; Aumann and Heifetz, 2002) although the theoretical relationship between bargaining power and dissimulation has not been extended. 7 The Saraydar proposition has been applied in the study of business mergers and the determination of the terms of a bid. (The information disclosed at the time of a bid is likely to aect negotiations between the directors of the companies involved, the determination of the nal oer price and its outcome, Barnes et al., 1990). 8 It has also been examined empirically as an explanation of the hubris hypothesis and why bidding companies typically overpay (Barnes, 1998). In any game model, the outcome is aected by the procedures involved (Kennan and Wilson, 1993, emphasize this point). Here the bargaining model is dened so as to reect the institutional setting of auditing and follows Zhang (1999). The bargaining process is as follows. During negotiations, both the auditor and the client can propose the content of the nancial statements. If they agree, the auditor will issue a clean report. If they do not, either the client can report what the auditor requires, in which case there will be a clean audit report, or the client reports what he wants, in which case, there will be an adverse audit report. 9 The only other institutional assumptions to be made are that if the audit rm and client company disagree to such an extent as to threaten an adverse report, they will attempt to negotiate with one another, either implicitly or explicitly, in an attempt to resolve the matter. It is also assumed that, whilst there may be no immediate limit to the length of the negotiating process, the period is nite and within the constraints imposed by law on the period within which the nancial statements and an auditors report must be
Research on bargaining with private information usually involves a price of a good or service and not, as here, a yesno decision. Emphasis has not been on dissimulation or misinformation but on signaling and how best (or how most advantageously) to communicate private information. The typical context (certainly the area of most empirical research) is in the area of wage negotiation where, for example, an employer knows more about the value of labors product than do the workers and its acceptance of a strike may be seen as a way in which this may be communicated (Kennan, 2001). A model applying roughly similar principles and arguments to that of Saraydar (1983) is Farmer and Pecorino (2003). A more prescriptive and applied approach to procedures, strategies and tactics has emerged and is known as negotiation analysis. It de-emphasizes the game-theoretic approach and incorporates the ndings of behavioral scientists and economists. For a useful survey as it relates to management see Sebenius (1992). 8 This is an area ripe for the application of game theory and some of the most famous scholars of economic game theory have applied their skills to it; for instance, Grossman and Hart (1980). 9 It is interesting that Zhang notes after a disagreement, the client may impose a penalty on the auditor by changing the auditors working conditions or terms of engagement or dismissing the auditor. He cites the evidence provided by Chow and Rice (1982) who identify a positive association between a rms propensity to switch auditors and an adverse opinion in the year before the switch.
7

420

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

presented to shareholders. This is in line with Saraydar who also adds a condition to those of Coase: that of a nite period for the negotiation process. The other assumptions used by Coase are negligible: simply, that the parties are rational, self-interested, and, more crucially, they may negotiate freely. It is assumed here that, in the same way that other actors negotiate when it is in their interests to do so, it is axiomatic that the audit rm and the client company will attempt to negotiate the outcome; especially, if they disagree to such an extent as to threaten an adverse audit report. 10 In fact, it has been recognised that negotiation plays an important part of auditing practice. 11 This paper makes certain other assumptions. Given their competence to make such a decision and the same data and other information being available to them (including managements plan for addressing the problem), a dierent auditing rm, if required to act as the client companys auditor, would come to a similar conclusion about the probability of the company failing. It is also assumed for simplicity that the costs and benets relating to the audit, in particular those relating to an action for negligence, would be similar for the same audit across other audit rms. It should be noted, however, that when the concept of bargaining power is introduced later in the paper, this does not mean that all audit rms will make a similar decision about whether to issue an adverse report. The assumption relating to the homogeneity of auditors when interpreting data and deciding on the probability of failure is relaxed towards the end of the paper. There the possibility of the client intentionally misleading the auditor is introduced, together with the concept of audit quality which here relates to the ability of the auditor to see through these attempts.

2. Economic model excluding transaction costs Although it is not reported in probabilistic form, the auditors going concern judgement eectively relates to the probability of the rm continuing and the investor is correct in attempting to make probabilistic inferences. In only the most extreme cases can the auditor be absolutely certain. 12 In all other cases
Zhang (1999) also does this. Negotiation is assumed to play a part in the choice of accounting method. Watts and Zimmerman (1990) suggest that accounting methods may be classied as an acceptable set and an unacceptable set. The role of the auditor is to prevent the client company from adopting the unacceptable set. For research on the negotiating process in the auditorclient relationship see for example Demski and Frimor (1999), Gibbins et al. (2001) in Canada who conclude that negotiation is normal in auditorclient relations, and in the UK, Beattie et al. (2001). 12 The professional literature provides guidance on what information to use, e.g. management reports and plans, but it does not say how. This is left to the auditors professional judgment based on his experience concerning the client company, the industry and the economy generally.
11 10

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

421

there must be a cut-o or threshold to represent reasonable belief. That is, if there is an adverse report, the investor may infer that p(D) > p(P)* where p(D) represents the probability of failure in the judgement of the auditor and p(P)* represents the probability level threshold above which an auditor is professionally expected to issue an adverse report. Similarly, the investor may infer p(D) < p(P)*, if there is no adverse report. Unfortunately for the investor, p(P)* is not expressed in probabilistic terms. If he were to consult the auditing standard he would nd it vague and open to dierent interpretations. Certainly, he would be unable to infer a probability value from it. On the other hand, the investor should feel satised that the situation is not completely unregulated. If a rm fails and there had been no adverse report, the investor could sue the auditor for damages arising from negligence. 13 The basis of the judgement would be the best practice of the auditors peers as determined by the Court, i.e. p(P)*. Why should the auditor decide otherwise? There may be factors other than the fear of being sued for negligence. 14 These include: lost revenue arising from the loss of the audit (an audit switch) in the event of an adverse report but no resulting failure by the client company, 15 and lost revenue arising from lost reputation in the event of no adverse report but failure by the client company. An alternative point to p(P)* may be identied, therefore: p(Z)*, where the costs or lost revenues arising from issuing an adverse report or not are minimized. The distinction between p(Z)* and p(P)* which are both thresholds (hence the superscript *) should be noted. p(P)* is the threshold above which the auditor should issue an adverse report according to externally-determined professional standards, whereas p(Z)* is the threshold above which the auditor should issue an adverse report determined by his own costs and benets and self-interested objectives. We may examine the alternative outcomes in the following way. An adverse report will occur if: pP < pZ < pD;

Krishnan and Krishnan (1996) argue that the magnitude of the auditors payment is likely to be related to the size of damages incurred by investors and, therefore, the size of their investment. 14 Carcello and Palmrose (1994) report that bankruptcy is one of the most frequent sources of litigation against auditors and that the issuance of an adverse report reduces the likelihood of litigation. 15 For empirical evidence to support this see Chow and Rice (1982), McKeown et al. (1991) and Krishnan and Stevens (1995). Krishnan and Krishnan (1996) identify the expected loss of losing a client as the loss of quasi-rents associated with that client. The relative importance of the client in the audit rms portfolio determines the ease with which it can replace the quasi-rents. Second, assuming that audit fees are proportional to client size, the client companys importance depends on its size relative to the audit rms other clients. This approach forms the basis of the paper by Zhang (1999).

13

422

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

pZ < pP < pD: Similarly, there will be no adverse report if pD < pZ < pP ; pD < pP < pZ :

2 3 4

Two situations remain, the outcomes of which depend on the objective function of the audit rm. These are pP < pD < pZ and pZ < pD < pP : 6 5

In case (5), the auditor will not issue an adverse report as p(D) < p(Z)* if rational economic decisions over-rule professional considerations, i.e. p(P)* is irrelevant. This causes a Type II error, as p(P)* < p(D) if, as expected, the client company fails. On the other hand, if professional considerations overrule the prot-maximizing objective, p(Z)* will be irrelevant and the auditor will issue an adverse report as p(P)* < p(D). In case (6), the auditor will issue an adverse report as p(Z)* < p(D) if rational economic decisions over-rule professional considerations, i.e. p(P)* is irrelevant. 16 This causes a Type I error as p(D) < p(P)* if, as expected, the client company does not fail. On the other hand, if professional considerations over-rule the prot-maximizing objective, p(Z)* will be irrelevant and the auditor will not issue an adverse report as p(D) < p(P)*. We may now dene the nature of the economic costs and benets. An adverse report will occur if: CNQ > CQ where CNQ = [CA + CB] and CQ = [CC + CD] and CQ and CNQ are the costs of issuing an adverse report and not issuing one, respectively, CA is the expected cost of a lawsuit from a shareholder, or third party, acting on the belief that, because there was no adverse report, the rm was not going to fail. This will be non-zero if there is no adverse report and the rm fails. CB is the expected cost of loss of reputation from not issuing an adverse report in the event of the rm going bankrupt, CC is the expected cost of a lawsuit for negligence in the event of an adverse report and the rm not failing, CD is the expected cost of the loss of the client from audit switching arising from issuing an adverse report and the rm not failing.
16 It is unlikely that p(Z) < p(P)*. This is discussed in the section on transaction costs and bargaining power.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

423

The expected costs are comprised of: a CA CM pD pP ; 1 pP

where CA > 0 if [p(D) p(P)*] > 0, otherwise CA = 0, and CM is the cost of litigation in terms of damages payable if the auditor did not issue an adverse report but was certain [p(D) = 1] that the rm would fail. For simplicity, we assume a simple linear relationship in which CA increases relative to the dierence between p(D) and p(P)*. The nature of the slope is not crucial to the analysis other than CA > 0 and increases to the right of p(P)*. b C B C R pD pP ; 1 pP

where CB > 0 if [p(D) p(P)*] > 0, otherwise CB = 0, and CR is the cost of loss of reputation from not issuing an adverse report and was certain [p(D) = 1] that the rm would fail. CB will be zero below p(P)* because the auditor is not expected to issue one at such low probability of failure levels. For simplicity, we assume a simple linear relationship 17 in which CB increases relative to the difference between p(D) and p(P)*. Again, the nature of the slope is not crucial to the analysis other than CB > 0 and increases to the right of p(P)*. c CC CN pP pD ; pP

where CC > 0 if [p(P)* p(D)] > 0, otherwise CC = 0, and CN is the cost of the litigation in terms of damages payable in a successful action where there was an adverse report yet the auditor was certain that the rm would not fail [p(D) = 0]. Once more, it is assumed that there is a linear relationship between costs, CC, and the dierence between the auditors assessment of the probability of failure [p(D)] and the level above which professional opinion would require an adverse report [p(P)*]. Again, the nature of the slope is not crucial to the analysis other than CC > 0 and increases, this time to the left of p(P)*. d C D where pD < pP and CD represents the cost of the audit switch in terms of lost prots by the audit rm. The threat of an audit switch occurs where the client company seeks to obtain a clean audit report from another audit rm because it is unable to obtain one from its present auditing rm. Assuming that similar interpretations
17 For all costs, CA to CD, a linear relationship is assumed. This is because there is no reason to assume otherwise and their relationship to p(D) clearly suggests constant proportionality. The direction of the slope (i.e. upwards or downwards) and the point at which these costs become positive are what are critical to the argument and conclusions. These would be the same if the relationships were curvilinear.

424

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

would be made by other audit rms and as p(P)* is the same for them all, the cost of an audit switch is zero because other rms would also not issue a clean report where p(D) < p(P)*. Hence, for present purposes, CD is assumed to be zero. When transaction costs are taken into account (Section 3), CD is found to be non-zero and inuential to the decision. A non-zero CD also has important eects in Sections 4 and 5 where bargaining power and the possibility of the client company attempting to mislead the audit rm are introduced. We may also assume that CN and CM are equal as they represent gross damages where p(D) = 1 or 0, if there was and was not an adverse report, respectively. Damages in respect of CN would be based on the dierence between the value of the rm without the adverse report, VNQ, and with the adverse report, VQ. That is, investors were led to believe that it was worth VNQ whereas it was actually worth VQ. In the event of the failure of the rm, they may sue for their losses, being the dierence between what they thought the rm was worth on the basis of no adverse report and what they would have thought it was worth had there been an adverse report, VQ. Thus, CN = VQ VNQ. Damages in respect of CM are based on a similar principle: (VQ VNQ) as the plaintis would claim that the adverse report had harmed the business by causing other rms to be reluctant to do business with the rm, making the acquisition of loans more dicult and so on. The rms value would have been reduced to VQ when it should have been VNQ; that is (VNQ VQ). Therefore, CN = CM and their expected values are zero when p(D) = p(P)*. 18 It will be seen from (a) to (d) above that the costs CA, CB, CC and CD are zero if p(D) = p(P)*. See Fig. 1. This shows costs becoming non-zero as p(D) departs from p(P)*, hence p(Z)* = p(P)*. As p(Z)* = p(P)*, there is no expectation of bias arising from Type I or II errors. This analysis suggests that external eects are brought into the calculations of the responsible parties in line with the Coase Theorem to achieve Pareto eciency assuming rational economic decision making by auditors (Coase, 1960). As long as p(P)* is eectively enforced by auditors in their actions by realising the need to take into account the legal implications of their decisions, the recognition of the costs involved will lead them to a Pareto ecient outcomein this case, non-biased going concern decisions.

3. Transaction costs It should not be thought that there are no transaction costs. These are in the form of legal costs associated with CM and CN. Regarding CA, FA,A represents
18 Empirically, of course, this may not be so. They may not be symmetrical. For instance, the damages re CM may be more clearly measured than CN. However, the directions and origins of the slopes are merely necessary for the argument.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

425

Fig. 1. The auditors costs associated with the decision of whether or not to issue an adverse report and their relationship with the probability of failure by the client company, p(D), where there are no legal costs. Line CA represents the expected value of legal damages from a successful action against the auditor if he did not issue an adverse report and the client company failed (this increases with the probability of failure). Line CB represents the cost of the loss of reputation to the auditor if he did not issue an adverse report and the client company failed. Line CC represents the expected value of legal damages from a successful action against the auditor if he issued an adverse report and the client company did not fail (this decreases with the probability of failure). Point p(Z)* represents probability of failure threshold at which the costs or lost revenues arising from issuing either an adverse report or not are minimized. Point p(P)* represents probability of failure threshold above which the auditor should issue an adverse report according to externallydetermined professional standards.

the legal costs incurred by the audit rm and FA,C represents those incurred by the client company, its shareholders or a third party. In which case C A C A;L F A;A F A;C and CA P 0. CA = 0 where CA,L = 0 as the client company, its shareholders or a third party will not sue unless CA,L > 0. Here C A;L C M pD pP F A;C ; 1 pP

where CA,L P 0. 19 Similarly for CC, where FC,A represents the legal costs incurred by the audit rm and FC,C represents those incurred by the client company, or its shareholders. In which case C C C C;L F C;A F C;C and CC P 0. CC = 0 where CC,L = 0 as the client company, or its shareholders will not sue unless CC,L > 0. Here
19

FA,C is added back in CA as it is deducted in arriving at CA,L.

426

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

Fig. 2. The auditors costs associated with the decision of whether or not to issue an adverse report and their relationship with the probability of failure by the client company, p(D), where there are legal costs and the auditors bargaining power is zero. For denitions of CA, CB, CC, p(Z)* and p(P)* see Fig. 1. Line CD represents the cost of an audit switch arising from the auditor issuing an adverse report and the client company did not fail. Point p(B) represents the probability level of client company failure above which, taking into account legal fees, there is a positive expected value of legal damages and costs from a successful action against the auditor if he did not issue an adverse report and the client company failed. Point p(C) represents the probability level of client company failure below which, taking into account legal fees, there is a positive expected value of legal damages and costs from a successful action against the auditor if he issued an adverse report and the client company did not fail.

C C;L C N

pP pD F C;C ; pP

where CC,L P 0. Here, the threshold p(Z)* rises from p(P)* to p(B), that point above which CA ceases to be zero. That is: CA, CB, CC and CD are zero if p(D) = p(B) where p(B) is dened as where C A;L C M pB pP F A;C 0: 1 pP

See Fig. 2. 20 The variation of assumptions has an eect on CB. In the initial model, CB became non-zero and positive to the right of p(P)* because of the increasing possibility of adverse reputational eects. Under these new assumptions, CA determines the point above which the costs are non-zero. Below p(B) (i.e. to
20 p(B) will be the auditors adverse report threshold unless he is able to negotiate it with the client company without loss of audit at a lower probability of failure level, i.e. his bargaining ability is zero. The concept of bargaining ability is introduced and discussed in the next section.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

427

the left) there are no reputational costs as, because of legal costs, there is no likelihood of legal action. Hence CB is redened as CR pD pB ; 1 pB

where CB > 0 if [p(D) p(B)] > 0, otherwise CB = 0. These changes have an effect on the possibility of an audit switch. At expected failure probability levels above p(B) all audit rms would issue an adverse report, but at points between p(B) and p(C) there may be audit rms which would be prepared to issue a clean report as shown in Fig. 2. Consider one such competing audit rm. Assuming homogeneity of costs as before, it will have similar cost functions and schedules CA and CC. It will therefore be prepared to oer a clean audit report for probability levels up to p(B). 21 The existing auditor therefore faces the risk of a switch if he issues an adverse report. As there is now the possibility of an audit switch below p(B), CD > 0 for values of p less than p(B). Where CD > 0, it is assumed to be constant as there is no change in the probability of the available alternative auditor. 22 In either event, the slope of CD is not critical to the subsequent analysis in this paper, merely the point above which it becomes non-zero below p(B). Hence, where there are transaction costs, self-interested auditors will use p(B) as their criterion for an adverse report and not p(P)*. That is, p(Z)* = p(B). As p(B) > p(P)*, bias arising from Type II errors should, therefore, be expected to occur. This result is also interesting as it relates to reputational costs. The shifting of the origin for CB from p(P)* to p(B) has shifted the point above which there will be an adverse report from p(P)* to p(B). Reputational costs and considerations between these two points have been driven out as a result of competitive pressures (the threat of a switch) which has been made possible by the existence of legal costs.

4. Bargaining power and audit switching Bargaining power relates to the strength of a negotiator to obtain advantageous terms or, more precisely, the ability to secure an opponents agreement on ones own terms. Here it relates to the ability of the auditor to withstand the pressures placed on him by his client, notably the threat of audit switch, and issue an adverse report nevertheless. At one extreme, the auditor is able
21 If the assumption of homogeneity of costs and interpretation by auditors is relaxed, the area in which there is threat from an audit switch [between p(C) and p(B)] may be even wider. It is possible that there may be signicant dierences across audit rms, especially between the incumbent audit rm and its competitors, regarding fees charged (low balling, Jevons Lee and Gu, 1998) and set-up and continuing costs (Zhang, 1999). 22 If it is assumed that the probability of an available auditor does increase below p(B), then the line CD is upward sloping to the left of p(B) where it is zero.

428

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

to withstand all pressures, perhaps because he has such a high reputation and is not prepared to risk this being threatened. Also, an audit switch may be seen to be a signal to the outside world that the client company is unable to stand up to such high standards of scrutiny. That is: there is no threat of an audit switch as it would have a negative signal to the outside world about the client company. Here the auditors bargaining power is high. At its extreme, Ba = 1, where B indicates bargaining power and the subscript indicates the auditor. At the other extreme is a situation in which the auditor is unable to withstand such pressures and the threat of audit switch is very high. That is, Ba = 0. In which case CD > 0 where p(D) < p(B) and p(Z)* = p(B). That is, CD > 0 for values of p(D) below p(B). In which case, he will issue an adverse report only where p(D) > p(B) as in Fig. 2. 23 In terms of the foregoing analysis, if Ba = 1, the auditor should be able to negotiate an adverse report without loss of the audit for a probability of failure as low as p(C) if he is concerned about protecting his reputation, even though there are no explicit expected costs for CB below p(B). Hence p(Z)* = p(C) as in Fig. 3. That is: CA, CB, CC and CD are zero if p(D) = p(C) where p(C) is dened as where: CC CN pP pC F C 0: pP

The important point is that the high value for Ba pushes back the point below which CD > 0 from p(B) to p(C). In the case of auditors who perceive their bargaining power as falling between these extremities of one and zero, then CD > 0 from the point lying between p(B) and p(C), where Ba[p(B) p(C)]. Hence, bias may occur in either direction. If Ba > B0a where B0a pP pC, it pBpC will be of the Type I form and if Ba < B0a , it will be of the Type II form. The point should also be made that whilst it is true that Ba + Bf = 1, where Bf represents the bargaining power of the client company, it does not follow that each knows each others bargaining strength. For instance, the auditor might think Ba = 0.75, hence Bf = 0.25, yet the client company might believe that they are opposite in value. 24 Hence, the addition of transaction costs has introduced the opportunity for bargaining by the auditor and the client company (in order to avoid an audit switch) between the points p(C) and p(B). Whilst the Coase Theorem recognises the eects of bargaining power, here there is no scope for it as p(P)* is the only available point in the absence of transaction costs. With the addition of transaction costs, this is now possible; the precise value for p(Z)*, the thresh23 Note that CD > CD at p(B). If it is less, then p(Z) will be determined by the intersection of CB and CD. 24 How this is decided is outside purpose of this paper. However, it has been shown by Nash (1950) that it may be decided by the two parties depending upon their marginal utilities.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

429

Fig. 3. The auditors costs associated with the decision of whether or not to issue an adverse report and their relationship with the probability of failure by the client company, p(D), where there are legal costs and the auditors bargaining power is at a maximum (unity). For denitions of CA, CB, CC, p(Z)* and p(P)* see Fig. 1. For denitions of CD, p(B) and p(C) see Fig. 2.

old for issuing an adverse report by a self-interested auditor, is then determined by the two parties according to their subjective relative bargaining strengths.

5. Asymmetry of information, misleading information and audit quality It has often been asserted that there is a positive relationship between auditor size and audit quality. 25 Probably the earliest was DeAngello (1981c) who argued that the valuable reputations of large auditors act as an incentive to provide accurate reports. More recent is the empirical evidence of Krishnan and Schaur (2000) who found that organizational compliance with GAAPs reporting requirements (a measure for quality) was related to the size of the auditing rm. One reason for this is their deeper pockets (see Dye, 1993). Insights into the relative bargaining power of participants leading to incentives to mislead have been developed by Saraydar (1983). Saraydar shows how the relative bargaining power of the participants is manifested in a desire to provide misleading information resulting in a Pareto inecient outcome. We may use these insights to extend the analysis of the going concern decision
25 Although this is an unclear concept and a largely unobservable phenomenon in which dierent aspects may relate to dierent parties (for a discussion see Schroeder et al., 1986), here an important aspect of audit quality relates to the skills and ability of the auditor to correctly assess the commercial future of the client company when it may wish to mislead him.

430

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

where there is considerable information asymmetry. When making his decision, the auditor is dependent upon the client company to a greater or less extent. He may not be an expert in the commercial workings of the industry in which the client company operates or even its nancial records (Zimbelman and Waller, 1999). 26 His decision will involve judgement based on qualitative and subjective information provided, to some extent informally, by the client. The client will have the power to emphasize certain aspects and understate, or even hide, other aspects, e.g. managements plans and assessments. It is only with the acquisition of experience and some considerable skill that the auditor may overcome these shortcomings. We can identify three possible probability estimates of client failure that the auditor may make depending upon the quality of the audit. These are: p(D1), where there is no attempt to mislead by the client company, p(D2), where there is an attempt by the client company to mislead and this is not discovered because the quality of the audit is low, p(D3), where there is an attempt to mislead by the client company but this is discovered because the quality of the audit is high. In each case p(D1) is the correct probability of failure that should be inferred from the accounting data. However, the quality of audit may vary in its ability to identify and adjust for the misleading information. The auditor may be able to fully identify and adjust for it. In which case, p(D1) = p(D3). However, perhaps more likely, p(D3) will fall somewhere between p(D2) and p(D1) depending upon the quality of the audit, the magnitude of the misinformation and the skill of the client company in its deception. Here p(D2) < p(D3) < p(D1). It should not be presumed that p(D2) < p(D1) will always be the case. The client company may decide to increase the adverse information so that p(D1) < p(D2) if in its view this may be done without threatening an adverse report; that is p(D1) < p(D2) < p(Z)*. 27 However, as the client company does not desire an adverse report, misinformation which changes an audit decision only occurs where p(D2) 6 p(D3) < p(Z)* < p(D1). Misleading information and the impact of relative bargaining power may act in the following way:

26 This contrasts with, for instance, the auditors ability to assess the client companys internal controls. Here, the auditor may be more expert at knowing whether they are satisfactory than the company itself. For a discussion of the two parties capabilities see Gibbins et al. (2001). 27 There are many reasons why this may be done; for instance, to exaggerate the companys turnaround during the next year.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

431

1. The client company will produce accounting data designed to suggest p(D2) [based on the client companys subjective view of its relative bargaining power], although they should indicate p(D1), the correct probability. The auditor will decide p(D3) which falls somewhere between the two depending upon the quality of the audit. 2. The auditor will then determine p(Z)* which lies between p(C) and p(B) according to his perceived bargaining power relative to that of the client company. 3. p(D3) will then be compared with p(Z)*. If p(D3) < p(Z)* then the auditor will issue a clean audit report but if p(D3 ) > p(Z)* he will issue an adverse one. As a result of this process, either a Type I or II error may occur. Consider the situation where a bank has received information that some of its borrowers may be unable to repay their loans. Its management think that the auditors are likely to conclude that it is no longer a going concern [p(D1)]. Rather than make the necessary provisions in full and provide the auditors with this information, management decide to provide them with information suggesting that the loans are safe [p(D2)]. When reviewing these loans, the auditors may conclude that (i) they need providing for [p(D1)] in full, (ii) they do not need providing for [p(D2)], (iii) only some of the loans need providing for [p(D3)], or (iv) all the suspect loans need providing for, but not in full [p(D3)]. To use a simple illustration, assume that there is a relationship between the total net value of the loans (L, their face value less provision for bad debts) and the probability of failure, p(D), which is determined by the function p(D) = 1.47 0.67L. For instance, if the reported loans are $1.6 bn and there are no provisions required, the probability of failure, p(D1), is 0.4; but if there were a provision of $0.6 bn (L = $1.0 bn) then p(D1) = 0.8. That is: 0.4 = 1.47 0.67(1.6) and 0.8 = 1.47 0.67(1.0). See Case 1 in Fig. 4. Say the correct value of loans should be $1.254 bn. In which case there should be a provision of $0.346 bn [i.e. $1.6 bn $1.254 bn] and the probability of failure, p(D1), should be 0.63 [i.e. 1.47 0.67(1.254)]. However, say the client company makes a provision of $0.175 in order to imply p(D2) = 0.52 [i.e. 1.47 0.67(1.425)] but the audit rm thinks there should be a provision of $0.206 bn which implies p(D3) = 0.54. Also say p(P)* is 0.6 and p(Z)* is 0.62 based on the audit rms subjective estimate of its bargaining power of 0.4 where p(C) and p(B) are 0.5 and 0.7, respectively [i.e. 0.4 = (0.7 0.62)/(0.7 0.5)]. Here there is a Type II error: there should have been an adverse report [because p(D1) > p(P)*] but there was not [because p(D3) < p(Z)*] and Ba < B0a [0.4 < 0.5, where B0a 0:5, i.e. (0.6 0.5)/ (0.7 0.5)]. Next, consider Case 2 in Fig. 4. Here, the correct value of loans should be $1.334 bn. In which case there should be a provision of 0.266 [$1.6 bn $1.334 bn] and the probability of failure, p(D1), should be 0.58 [i.e. 1.47 0.67(1.334)]. However, the client company produces data designed to imply p(D2), i.e. makes provisions of $0.175 bn. However, the auditor decides

432

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

Fig. 4. Diagram to show how Types I and II errors may occur by reference to the probability levels of the client companys failure which determine the auditors decision to issue an adverse report (not to scale). (The numbers in brackets and the data relating to net loans and provisions refer to the numerical illustration in the text.) For denitions of p(B), p(C), p(P)* and P(Z)* see Figs. 1 and 2. p(D1) represents the probability of failure where there is no attempt to mislead by the client company; p(D2) represents the probability of failure where there is an attempt by the client company to mislead and this is not discovered because the quality of the audit is low; p(D3) represents the probability of failure where there is an attempt to mislead by the client company but this is (partly or otherwise) discovered because the quality of the audit is high. The values for p(Z)* and p(P)* are assumed for illustration purposes. The values for p(D) are also derived from the illustration where the relationship between it and reported net loans (L) is p(D) = 1.47 0.67(L) and gross loans are $1.6 bn.

that there should be provisions of $0.236 bn, p(D3), which implies a probability of failure of 0.56 [i.e. 1.47 0.67(1.364)]. Assume that p(P)*, p(C) and p(B) are the same as before but Ba is 0.75, making p(Z)* = 0.55 [i.e. (0.7 0.55)/ (0.7 0.5)]. Here there is a Type I error. There should not have been an adverse report [because p(D1) < p(P)*] but there was [because p(D3) > p(Z)*] and Ba > B0a [i.e. 0.75 > 0.5]. The important point is that increased bargaining power on the part of the auditor also increases the incentive for, and the size of, misinformation. Say Ba = 0, then p(Z)* = p(B). If p(D1) > p(Z)* then the auditor will issue an adverse report unless the client company provides sucient misleading information so that [p(D2) mis0] 6 p(B), where mis0 represents the required eect of the misinformation when Ba = 0. Now, say Ba = 1, then p(Z)* < p(B) and the client company will need to provide sucient misleading information to the ex-

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

433

Fig. 5. Diagram to show how the amount of required misinformation increases from mis0 to mis1 when the auditors bargaining power (Ba) changes from zero to unity. For a denition of p(Z)* see Fig. 1. For denitions of p(B) and p(C) see Fig. 2. For a denition of p(D2) see Fig. 4. mis0 represents the required eect of the misinformation when Ba = 0; mis1 represents the required eect of the misinformation when Ba = 1.

tent mis1, where mis1 represents the required eect of the misinformation when Ba = 1, and [p(D2) mis1] 6 p(Z)*. Here mis1 > mis0 and (mis1 mis0) increases with [p(B) p(C)]. See Fig. 5. Thus, other things being equal, the greater the bargaining power of the auditor, the greater the size of the required misinformation. This is illustrated in Table 1. There the two extreme right hand columns show in terms of the bank illustration how the amount of misleading information (here a deliberate under-provision for bad debts) directly increases with the auditors bargaining power. To put it another way, stronger auditing actually encourages the client rm to attempt to misinform (e.g. deliberately misleading or fraudulent nancial statements). It is only if Ba is directly related to the quality of audit to combat it, that this tendency may be arrested. However, unless there are reasons to the contrary (and there may be empirically) then these two variables (an auditors bargaining power and the quality of audit required to handle dissimulation) should be regarded as independent. This nding has implications for the regulation of the accountancy profession. For instance, it suggests that, contrary to intuition, increasing the power of the auditor may not improve the ecient workings of the capital markets. Further, the case for increased concentration of the auditing industry (or state control) does not lie in improved nancial reporting.

6. Inferences made by the investor What may an investor infer from whether the audit report is adverse or not? Little, unless he makes certain necessary assumptions concerning the audit

434

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

Table 1 A numerical illustration from the text and Fig. 4 showing how the amount of required misinformation (in terms of under-provision for bad debts) increases as the auditors bargaining power, Ba, decreases from one to zero Ba p(Z)*a Reported net loans, Lb ($bn) 1.45 1.42 1.39 1.36 1.33 1.3 1.27 1.24 1.21 1.18 1.15 If p(D1) = 0.63 under-provision requiredc ($bn) 0.2 0.17 0.14 0.11 0.08 0.05 0.02 None None None None If p(D1) = 0.58 under-provision requiredd ($bn) 0.12 0.09 0.06 0.03 None None None None None None None

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0.5 0.52 0.54 0.56 0.58 0.6 0.62 0.64 0.66 0.68 0.7

a For a denition of p(Z)* see Fig. 1. The values for p(Z)* in this table are derived from the proportional relationship between p(Z)* and Ba, where Ba ranges from one to zero, and p(Z)* ranges from 0.5 to 0.7, the assumed values for values for p(C) and p(B), respectively in the illustration. b An increase in p(Z)* reduces the threshold of the minimum net loans, L, for a clean audit report. This column computes L for a given value p(Z)* in column 2. L is determined by rearranging the function D = 1.47 0.67L to L = (1.47 D)/0.67 and substituting p(Z)* for D. For instance, where p(Z)* = 0.5, L = (1.47 0.5)/0.67 = 1.45. c For a denition of D1 see Fig. 4. If p(D1) = 0.63 (as in Case 1 in Fig. 4), loans net of provisions are $1.25 bn [i.e. (1.47 0.63)/0.67]. Hence, if p(Z)* is 0.5, the under-provision is $1.45 bn $1.25 bn = $0.2 bn. d If p(D1) = 0.58 (as in Case 2 in Fig. 4), loans net of provisions are $1.33 bn [i.e. (1.47 0.58)/ 0.67]. Hence, if p(Z) is 0.5, the under-provision is $1.45 bn $1.33 bn = $0.12 bn.

rms motivation and whether an opinion is determined by the audit rms attempts to maximize prots when these conict with independent professional opinions, its bargaining power, and audit quality; matters which have been discussed here. Whether the rm is prot-maximizing or not determines whether p(P)* or p(Z)* is used as the criterion. This was discussed in the Simple Rational Economic Model section concerning cases (5) and (6). We may now also note that the auditor will issue an adverse report in (5) either if for some reason, he is non-prot-maximizing and Ba < 1 or if he is prot-maximizing and Ba = 1. If he is prot-maximizing and Ba = 0, he will not issue one. Regarding (6), as noted in the previous section, this will only occur if Ba has a particularly high value, that is Ba > B0a 0 in which case the auditor will issue an adverse report, irrespective of whether or not he is prot-maximizing. It should also be noted that the investor will have little idea of the cost functions involved and, therefore, unable to estimate p(Z)*. He may nd it easier to make inferences

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

435

based on the auditors reputation. For instance, because the audit rm is one of the largest in the world with a good reputation, its bargaining power is high, perhaps greater than B0a . The investor may decide that although the audit rm is probably prot-maximizing, eectively p(Z)* = p(P)*. However, as noted earlier, if Ba > B0a then p(Z)* < p(P)*. The situation in which p(Z)* < p(D) < p(P)*, a Type I error, may, therefore, arise. Whilst at rst glance the investor may be happy for the possibility of Type I errors arising (because he has been warned about the possibility of a failure which may not occur) on closer inspection he may not. The eect of the adverse report would be to reduce the share price and hence the market value of his investment. The astute investor may even look for situations in which a Type I error may have occurred, in order to make a gain. It should be noted that the market may only be ecient if there are no expectations of bias arising from Type II errors. Of course, the crucial question for the investor is whether the client company has attempted to mislead the auditor and, if so, whether it has been successful. If the investor assumes that there is no misinformation, then his problems are restricted to what have so far been examined. However, if he thinks that the auditor may be misled, this adds further complications. For instance, if there is no adverse report, although the investor may infer that p(D2) < p(Z)*, he does not know whether p(D1) < p(P)*. Here the important point is that brought out in the previous section. That is, the magnitude of misleading information is likely to increase with the auditors bargaining power, even though it would not be zero where Ba = 0.

7. Conclusions It has been shown that, in the absence of transaction costs, the going concern decision is ecient in the sense that bias arising from either Type I or II errors is not expected. This is in line with the Coase Theorem, although the situation here is not precisely the same. When transaction costs, in the form of legal costs arising from a lawsuit against the auditor are considered, Types I and II errors are expected. (Again, this is in line with the relaxation of the standard assumptions of the Coase Theorem.) Here, the auditors relative bargaining power determines the direction of these errors: a Type I error where the auditors bargaining power is relatively high (where Ba > B ) and a Type II era ror when it is relatively low. An important aspect which has not been considered in the bargaining literature in auditing is the eect that this may have on the incentive for misinformation. It is shown here how the relative bargaining powers of the auditor and his client company produce an incentive for the latter to mislead and that this increases with the formers perceived bargaining power. This analysis helps to explain the empirical evidence in the UK (and also, probably, the USA) where (a) there are many cases of lawsuits against

436

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

auditors who had not issued an adverse report prior to the failure of their client company, and (b) a surprisingly large number of cases of companies having received an adverse report have survived. As yet, there is no new empirical evidence to support the principal proposition in the later sections of this paper, that increased bargaining power for the auditor will encourage the client company to provide misleading information. Nevertheless, it does raise important regulatory issues and begs for empirical evidence. For instance, the impact of mergers of large accountancy rms on their relative bargaining power and the eects of developments in GAAP, auditing processes and pricing on audit quality as they aect the degree and nature of misleading information. Taken on its own, this analysis suggests that increased auditor size per se may not be in the interests of investors generally and the ecient working of the capital markets 28 and may lead directly to earnings manipulation and the rest. It should be noted that there has been a considerable increase in concentration of the accountancy profession amongst the very large rms over recent years both in the USA (Wolk et al., 2001) and the UK (Pong, 1999). Although there is no direct evidence for this, the general impression is that it has coincided with an increase in the manipulation of accounting information by large corporations. This paper suggests a mechanism by which these two phenomena may be related. The implication is that it matters little in marginal cases whether the auditor is weak or strong if the client company is determined to avoid an adverse audit report. Regulatory eorts to increase the bargaining power of the auditor (e.g. by protecting and enhancing audit rms independence or encouraging mergers between them in order to increase their size) may be self-defeating, if this merely forces companies into greater eorts to mislead their auditors when they cannot negotiate a favourable decision. Instead (and this is being done), regulatory eorts should focus on audit quality and the ability of the auditor to see through the attempts to mislead him. Similarly, the development and strengthening of mechanisms, such as corporate governance procedures and audit committees, will help to provide additional internal constraints on managements undue inuence on its external nancial information. At rst glance, the Enron/Arthur Andersen case refutes the conclusions here. (After all, much of the analysis in this paper was done before it occurred!) How could the largest rm of auditors in the worldin terms of this paper, therefore possessing huge bargaining powerbe so damaged by apparently a few weak decisions on one audit client company undermining its perceived independence and the quality of its audits elsewhere? (Chaney and Philipich, 2002). It should also be remembered that the bargaining power of the top four
28 As noted by Boritz and Zhang (1999) in a slightly dierent situation, biases arising from Types I and II errors cannot be diversied away and therefore remain within equilibrium stock market prices.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

437

audit rms is boosted where the client company is so large that it is eectively required to be audited by one of them. On the other hand, the impact of the loss of one client company to a large audit rm will, typically, be much greater than for a small or medium-sized rm because the client company will be much larger. Also, both the likelihood of occurrence and the reputational costs and eects for the audit rm of a negligence claim are much greater from a large client company than for a small one. The relevant measure of auditor size as a proxy for bargaining power may not necessarily be the total size of the rm but, especially if decisions are decentralized and taken at local level, the size of the individual practice oce as Reynolds and Francis (2000) suggest. However, as Chaney and Philipich (2002) illustrate in the Enron case, reputational eects and the need for its protection extend, of course, to beyond the local branch level. In other words, some of the costs and benets in this analysis may occur at the local branch level, but others, notably the reputational eects, are likely to be at the rm level. The rational choice model which is developed here may be useful for the basis for subsequent empirical work when identifying the costs and benets involved in independence-related decisions and their incidence. The model here specically relates to the going concern decision but it may be widened considerably to cover all similar types of decisions by the auditor where there is a conict between his interests and those of the client company.

8. Uncited references American Institute of Certied Public Accountants (1988), Bierman and Fernandez (1993), DeAngello (1981a), DeAngello (1981b), Hermanson et al. (2003), Illing (1992)

Acknowledgments The author would like to thank Den Hooi, Edward Barnes, Jill Lambell, Shanti Chakravarty and Dhylan Kanddasamy for their help in various stages of this research. He would also like to thank the two referees for their patience and some very useful comments that have substantially improved the paper.

References
American Institute of Certied Public Accountants (AICPA), 1988. The auditors consideration of an entitys ability to continue as a going-concern. Statement on Auditing Standards no. 59, American Institute of Certied Public Accountants, New York.

438

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

Antle, R., 1984. Auditor independence. Journal of Accounting Research 22 (3), 304330. Antle, R., Nalebu, B., 1991. Conservatism and auditorclient negotiations. Journal of Accounting Research 29 (Supplement), 3154. Asare, S., 1990. The auditors going-concern decision: A review and implications for future research. Journal of Accounting Literature 9, 3964. Aumann, R.J., Heifetz, A., 2002. Incomplete information. In: Aumann, R.J., Hart, S. (Eds.), Handbook of Game Theory with Economic Applications, vol. 3. Elsevier, New York, pp. 1665 1686. Barnes, P., 1998. Why do bidders do badly out of mergers?. Journal of Business Finance and Accounting 25 (56), 571594. Barnes, P., Chakravarty, S., Haslam, J., 1990. Bargaining power, dissimulation, and takeovers in a rational market with asymmetric information. Journal of Business Finance and Accounting 17 (4), 529539. Barnes, P., Huan, H.D., 1993. The auditors going concern decision: Some UK evidence concerning independence and competence. Journal of Business Finance and Accounting 20 (2), 213228. Bartlett, R., 1997. Auditor independence: Five scenarios involving potential conict of interest. Research in Accounting Ethics 1, 245277. Beattie, V., Fearnley, S., Brandt, R., 2001. Behind Closed Doors: What Company Audit is Really About. Palgrave, Basingstoke. Benston, G., Hartgraves, A.L., 2002. Enron: What happened and what we can learn from it. Journal of Accounting and Public Policy 21 (2), 105127. Bierman, S., Fernandez, L., 1993. Game Theory with Economic Applications. Addison-Wesley, New York. Boritz, J.E., Zhang, P., 1999. The auditors objectivity under a negligence liability system. Auditing: A Journal of Practice and Theory 18 (Supplement), 147164. Campisi, S., Trotman, K.T., 1985. Auditor consensus in going concern judgements. Accounting and Business Research 15 (Autumn), 303310. Carcello, J.V., Palmrose, Z., 1994. Auditor litigation and modied reporting on bankrupt clients. Journal of Accounting Research 32 (Supplement), 130. Carcello, J.V., Hermanson, D.R., Huss, H.F., 2000. Going-concern opinions: The eects of partner compensation plans and client size. Auditing: A Journal of Practice and Theory 19 (1), 6777. Chaney, P.K., Philipich, K.L., 2002. Shredded reputation: The cost of audit failure. Journal of Accounting Research 40, 12211246. Chow, C.W., Rice, S.J., 1982. Qualied audit opinions and auditor switching. The Accounting Review 57 (2), 326335. Citron, D.B., Taer, R.J., 1992. The audit report under going concern uncertainties: An empirical analysis. Accounting and Business Research 22 (3), 337345. Coase, R.H., 1960. The problem of social cost. Journal of Law and Economics 3, 144. DeAngello, L.E., 1981a. The AuditorClient Contractual RelationshipAn Empirical Analysis. UMI Research Press. DeAngello, L.E., 1981b. Auditor independence, low balling and disclosure. Journal of Accounting and Economics 3 (August), 113127. DeAngello, L.E., 1981c. Auditor size and audit quality. Journal of Accounting and Economics 3 (December), 183200. Davis, R., Ricchiute, N., Trompeter, G., 1993. Audit eort, audit fees and the provision of nonaudit services to audit clients. The Accounting Review 68 (1), 135150. DeFond, K., Raghunandan, K., Subramanyam, K.R., 2002. Do non-audit service fees impair auditor independence? Evidence from going concern audit opinions. Journal of Accounting Research 40, 12471274. Demski, J., Frimor, H., 1999. Performance measure garbling under renegotiation in multi-period agencies. Journal of Accounting Research 37 (Supplement), 187214.

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

439

Dye, R., 1993. Auditing standards, legal liability and auditor wealth. Journal of Political Economy 101, 887914. Farmer, A., Pecorino, P., 2003. Bargaining with voluntary transmission of private information: Does the use of nal oer arbitration impede settlement?. Journal of Law, Economics and Organization 19 (1), 6482. Gibbins, M., Salterio, S., Webb, A., 2001. Evidence about auditorclient management negotiation concerning clients nancial reporting. Journal of Accounting Research 39 (3), 535563. Grossman, S., Hart, O.D., 1980. Takeover bids, the free-rider problem, and the theory of the corporation. Bell Journal of Economics 11, 4264. Hermanson, H.M., Schulzke, K.S., Shell, G.R., 2003. The role of bargaining style in public company audits. Journal of Forensic Accounting 4 (2), 233248. Illing, G., 1992. Private information as transaction costs: The Coase Theorem revisited. Journal of Institutional and Theoretical Economics 148, 558576. Institute of Chartered Accountants of England and Wales (ICAEW), 1994. Statement of Auditing Standards: 130 The Going Concern Basis in Financial Statements. Institute of Chartered Accountants of England and Wales, London. Jevons Lee, C.-W., Gu, Z., 1998. Low balling, legal liability and auditor independence. The Accounting Review 73 (4), 533555. Johnson, P.E., Jamal, K., Berryman, G.R., 1989. Audit judgment research. Accounting Organisations and Society 14 (12), 8299. Jones, F., 1996. The information content of the auditors going concern evaluation. Journal of Accounting and Public Policy 15 (1), 127. Kandasamy, D., 1998. An Economic Analysis of Auditor Decision Making with Respect to the Going Concern Decision. University of Nottingham, Nottingham, UK, MA. Kennan, J., 2001. Repeated bargaining and persistent private information. Review of Economic Studies 68, 719755. Kennan, J., Wilson, R., 1993. Bargaining with private information. Journal of Economic Literature 31, 45104. Kida, T., 1980. An investigation into auditors continuity and related qualication judgements. Journal of Accounting Research 18 (3l), 506523. Knapp, M.C., 1985. Audit conict: An empirical study of the perceived ability of auditors to resist management pressure. The Accounting Review 60 (1), 202211. Krishnan, J., Krishnan, J., 1996. The role of economic trade-os in the audit opinion decision: An empirical analysis. Journal of Accounting, Auditing and Finance 11 (4), 565586. Krishnan, J., Schaur, P.C., 2000. The dierentiation of quality amongst auditors: Evidence from the not-for-prot sector. Auditing: A Journal of Practice and Theory 19 (2), 925. Krishnan, J., Stevens, R.G., 1995. Evidence of opinion shopping from audit opinion conservatism. Journal of Accounting and Public Policy 14 (3), 179201. Lavin, D., 1976. Perceptions of Independence of the Auditor. The Accounting Review 51 (1), 41 50. Lee, T., Stone, M., 1995. Competence and independence: The congenital twins of auditing?. Journal of Business Finance and Accounting 22 (4), 11691177. Lennox, C.S., 1999a. The accuracy and incremental information content of audit reports in predicting bankruptcy. Journal of Business Finance and Accounting 26 (56), 757778. Lennox, C.S., 1999b. Audit quality and auditor size: An evaluation of reputation and deep pocket hypotheses. Journal of Business Finance and Accounting 26 (78), 779803. Louwers, T.J., 1998. The relation between going-concern opinions and the auditors loss function. Journal of Accounting Research 36 (1), 143153. Matsumura, E.M., Subramanyam, K.R., Tucker, R.R., 1997. Strategic Auditor behaviour and going concern decisions. Journal of Business Finance and Accounting 24 (6), 728751.

440

P. Barnes / Journal of Accounting and Public Policy 23 (2004) 415440

McKeown, J.C., Mutchler, J.F., Hopwood, W., 1991. Toward an explanation of auditor failure to modify the audit opinions of bankrupt companies. Auditing: A Journal of Practice and Theory 10 (Supplement), 113. Menon, K., Schwartz, K.B., 1987. An empirical investigation of audit qualication decisions in the presence of going-concern uncertainties. Contemporary Accounting Research 3 (2), 302315. Mutchler, J.F., 1984. Auditors perceptions of the going concern decision. Auditing: A Journal of Practice and Theory 3 (Spring), 1730. Mutchler, J.F., 1985. A multivariate analysis of the auditors going concern opinion decision. Journal of Accounting Research 23 (2), 668682. Nash, J., 1950. The bargaining problem. Econometrica 18, 155162. Pearson, M.A., 1987. Auditor independence deciencies and alleged audit failures. Journal of Business Ethics 6, 281287. Peel, M., 1989. The going concern qualication debate: Some UK evidence. British Accounting Review 21 (4), 329350. Pong, C.K.M., 1999. Auditor concentration: A replication and extension for the UK audit market. Journal of Business Finance and Accounting 26 (34), 450475. Reynolds, J.K., Francis, J.R., 2000. Does size matter? The inuence of large clients on oce-level auditor reporting decisions. Journal of Accounting and Economics 30, 375400. Saraydar, E., 1983. Bargaining power, dissimulation and the Coase Theorem. Journal of Institutional and Theoretical Economics 31, 599611. Schroeder, M.S., Solomon, I., Vickey, D., 1986. Audit quality: The perceptions of audit-committee chairpersons and audit partners. Auditing: A Journal of Practice & Theory 5 (1), 3552. Sebenius, J.K., 1992. Negotiation analysis: A characterization and review. Management Science 38 (1), 1838. Watts, R.L., Zimmerman, J.L., 1990. Positive Accounting theory: A ten year perspective. The Accounting Review 60 (1), 131156. Wolk, C.M., Michelson, S.E., Wootton, C.W., 2001. Auditor concentration and market shares in the US: 19881999. British Accounting Review 33 (2), 157174. Zhang, P., 1999. A bargaining model of auditor reporting. Contemporary Accounting Research 16 (1), 167184. Zimbelman, M., Waller, W., 1999. An experimental investigation of auditorauditee interaction under ambiguity. Journal of Accounting Research 37 (Supplement), 135155.