Provided in Cooperation with: Ifo Institute Leibniz Institute for Economic Research at the University of Munich

Transcription

1 eonstor Der Open-Aess-Publikationsserver der ZBW Leibniz-Informationszentrum Wirtshaft The Open Aess Publiation Server of the ZBW Leibniz Information Centre for Eonomis Pagano, Maro; Immordino, Giovanni Working Paper Optimal regulation of auditing CESifo working paper, No Provided in Cooperation with: Ifo Institute Leibniz Institute for Eonomi Researh at the University of Munih Suggested Citation: Pagano, Maro; Immordino, Giovanni (2007) : Optimal regulation of auditing, CESifo working paper, No This Version is available at: Nutzungsbedingungen: Die ZBW räumt Ihnen als Nutzerin/Nutzer das unentgeltlihe, räumlih unbeshränkte und zeitlih auf die Dauer des Shutzrehts beshränkte einfahe Reht ein, das ausgewählte Werk im Rahmen der unter nahzulesenden vollständigen Nutzungsbedingungen zu vervielfältigen, mit denen die Nutzerin/der Nutzer sih durh die erste Nutzung einverstanden erklärt. Terms of use: The ZBW grants you, the user, the non-exlusive right to use the seleted work free of harge, territorially unrestrited and within the time limit of the term of the property rights aording to the terms speified at By the first use of the seleted work the user agrees and delares to omply with these terms of use. zbw Leibniz-Informationszentrum Wirtshaft Leibniz Information Centre for Eonomis

2 OPTIMAL REGULATION OF AUDITING MARCO PAGANO GIOVANNI IMMORDINO CESIFO WORKING PAPER NO CATEGORY 2: PUBLIC CHOICE MAY 2007 PRESENTED AT CESIFO AREA CONFERENCE ON APPLIED MICROECONOMICS, MARCH 2007 An eletroni version of the paper may be downloaded from the SSRN website: from the RePE website: from the CESifo website: Twww.CESifo-group.deT

3 CESifo Working Paper No OPTIMAL REGULATION OF AUDITING Abstrat We study regulation of the auditing profession in a model where audit quality is unobservable and enforing regulation is ostly. The optimal audit standard falls short of the first-best audit quality, and is inreasing in the riskiness of firms and in the amount of funding they seek. The model an enompass ollusion between lients and auditors, arising from the joint provision of auditing and onsulting servies: defleting ollusion requires less ambitious standards. Finally, banning the provision of onsulting servies by auditors eliminates ollusion but may not be optimal in the presene of eonomies of sope. JEL Code: G28, K22, M42. Keywords: auditing, regulation, enforement, ollusion. Maro Pagano Faulty of Eonomis University of Naples Federio II Via Cintia Monte S. Angelo Naples Italy mrpagano@tin.it Giovanni Immordino Faulty of Eonomis University of Salerno Fisiano (SA) Italy giimmo@tin.it April 11, 2007 We thank Ron Anderson, Claudio Borio, Mike Burkart, Rajna Gibson, Christian Gollier, Antoine Faure-Grimaud, Denis Gromb, Fahad Khalil, Gianarlo Spagnolo, Oren Sussman, Roberto Tizzano, David Webb and seminar partiipants at Columbia University, University of Bologna, University of Salerno, University of Lausanne, London Shool of Eonomis, Oxford University, University of Zurih, the Bank of International Settlements, the 2004 Stokholm CEPR-SITE Workshop Understanding Finanial Arhiteture: On the Eonomis and Politis of Corporate Governane, the 2005 Toulouse Conferene in Tribute to Jean- Jaques Laffont, the 20th Annual Congress of the European Eonomi Assoiation and the 2007 CESifo Area Conferene on Applied Miroeonomis (keynote leture). Maro Pagano gratefully aknowledges finanial support from the Fondazione IRI and from the Italian Ministry of University and Researh (MIUR). A former version of this paper was irulated under the title Optimal Auditing Standards.

4 1. Introdution The reent orporate sandals involving major ompanies (Enron, Worldom, Qwest, Sunbeam, Parmalat, et.) have highlighted that the regulation of auditing and its enforement are key determinants of the reliability of orporate information. For many ompanies involved in orporate sandals, auditors failed to report any misbehavior or substantive inauray. These audit failures have damaged auditors reputation as independent experts and monitors of aounting information. As a result of this loss of onfidene, there has been a shift from self-regulation and litigationbased enforement of audit rules towards government regulation and publi-driven enforement. 1 In the United States, the Sarbanes-Oxley At of 2002 established the Publi Company Aounting Oversight Board (PCAOB), whih, under the oversight of the Seurities Exhange Commission (SEC), will register publi aounting firms, and establish rules for auditing, quality ontrol, ethis, independene and other standards. Moreover, it will inspet aounting firms, arry out disiplinary proeedings and impose penalties. The Sarbanes-Oxley At has also greatly inreased the finanial resoures devoted to the ativity of the SEC. A similar shift is under way in other ountries. The United Kingdom moved away from self-regulation by widening the sope and powers of the Finanial Reporting Counil, and speifially by reating a new subsidiary board (the Professional Oversight Board for Aountany), entrusted with oversight of the auditing and aountany profession. Also the Italian parliament is drafting new legislation to extend the powers of the national seurities ommission (CONSOB) on regulation and oversight of auditors ativity. Now that the role of the publi regulation of auditing is widely reognized, the natural question arises of what is the optimal design of suh regulation. The job of auditors is to ertify the reliability of aounting information, but in turn the reliability of this ertifiation an vary depending on the quality standards set by regulators. These standards may onern the auditing proedure, the organization of the auditing firm, or both. For example, regulation an affet the auditing proedure by mandating external onfirmation of the audited ompany s redits, and by alibrating the evidene required for suh onfirmation aording to the redit s magnitude. While relatively small redits may be heked by a telephone all to the debtor or a fax, original douments may be required for redits whose existene and terms an affet the solveny of the audited firm. The importane of rules on external onfirmation was highlighted by the reent Parmalat sandal, where massive fraud went undeteted beause of 1 Auditing rules apply to the ondut of auditors: They presribe how audits must be onduted. In ontrast, aounting standards apply to firms: They onern the reporting priniples and proedures that firms an use. 1

5 insuffiient evidene on a major redit of the ompany. 2 The Parmalat sandal undersored also the importane of another proedural rule, whih applies to the auditing of onglomerates by multiple auditors: when an auditor ertifies the aounts of the group s holding ompany and other auditors verify those of its subsidiaries, a problem of moral hazard in teams arises unless the responsibility for the ertifiation of the onsolidated balane sheet lies with a single auditor. 3 But regulation an also affet the organization of auditing firms, for instane by providing guidelines regarding the system of quality ontrol for audits and by setting standards of ompetene, independene and honesty of their employees. 4 In this paper, we show that the optimal design of auditing regulation depends on three main ingredients. First, the ost of enforement, whih inludes both the neessary publi funding (salaries of bureaurats and judges, paperwork, investigations, et.) and the ompliane osts borne by audit firms and their ustomers. Seond, the aountants inentives to ollude with their lients. Third, the eonomies of sope that may be reaped through the joint provision of auditing and onsulting servies to the same firm. We haraterize optimal regulation of auditing under the assumption that the quality of the information ertified by auditors is unobservable, so that in the absene of regulation the equilibrium level of audit quality would be ineffiiently low. To avoid this loss of informational effiieny and the implied misalloation of investment, the regulator an impose a minimum quality standard on auditors, but this hoie must take enforement osts into aount. As a result, the optimal standard will fall short of the first-best audit quality level, and must be lower the less effiient the enforement tehnology. 2 On 6 Marh 2003, the auditing firm Grant Thornton aepted a opy of a fax sent by Bank of Ameria as valid evidene of a 3,6bn redit and 336m ash held by Bonlat (a subsidiary of Parmalat), altogether worth 36 perent of Parmalat s debt and aounting for almost all the liquidity of the onglomerate. On 18 Deember 2003, the fax was revealed to be false. If the auditor had heked diretly the existene of the redit with Bank of Ameria, this fraud would have been revealed. 3 Deloitte Touhe was group auditor and Grant Thornton dealt with subsidiaries, inluding the Bonlat offshore unit in the Cayman Islands that is at the entre of the sandal. Deloitte failed to detet the frauds beause it took at fae value the reports produed by Grant Thornton about Bonlat. If regulation had made Deloitte Touhe the only auditor responsible for the entire group, it would have raised its inentive to hek the subsidiaries aounts diretly. Italy, the US and South Afria are among the few ountries where this rule is absent, although legislation urrently being passed in Italy will eliminate this defiieny. (We thank Roberto Tizzano for bringing this point to our attention.) 4 Suh rules on how to struture and operate auditing firms are not only present in the legislation of several ountries, but also detailed in the standards issued by the International Auditing and Assurane Standards Board: see IAASB (2004). 2

6 This baseline model assumes that the moral hazard problem lies only in the ativity of auditors, while managers always seek a truthful report, sine they are assumed to maximize shareholders value. But the reent orporate sandals suggest that also the behavior of managers may be plagued by moral hazard, insofar as they want to go ahead with investment irrespetive of its profitability, in order to extrat private benefits from the firm s expansion. In this ase, they will attempt to bribe auditors into produing positive reports under all irumstanes. To do so, managers an award profitable onsulting ontrats to auditors, onditioning on reeiving a favorable audit. The potential for ollusion inreases the auditors inentives to misreport, so that more resoures are required to enfore any given audit standard. Thus, when the managers of lient firms seek to orrupt their auditors, the regulator must optimally hoose a less ambitious standard. In this ase, an additional regulatory tool is to sever the link between onsulting and auditing ativity, by forbidding auditors to provide onsulting servies, as indeed presribed by the Sarbanes-Oxley At. If this is the only way in whih lient firms may bribe their auditors, this poliy would appear as a superior option to tampering with auditing standards. Indeed, in our model it would allow the regulator to leave the standard at the seond-best level. However, this option is not neessarily superior one one takes into aount that the profits arising from the joint provision of auditing and onsulting are themselves part of soial welfare. Indeed, these profits have two onfliting effets on soial welfare: a harmful effet, insofar as they raise the potential for ollusion between auditors and their lients; and a benefiial effet, insofar as they reflet effiienies of sope. We find that the regulator will want to allow bundling when the implied profits do not exeed a threshold that depends on the marginal effiieny of enforement. The struture of the paper is as follows. Setion 2 reviews the related literature and plaes the paper in perspetive, by omparing regulation with alternative mehanisms that an improve the informativeness of audits. In Setion 3 we present the model, derive the first-best audit quality, and haraterize the seond-best audit standard to be hosen if audit quality is privately unobservable. In Setion 4, we analyze the optimal regulation when firms ollude with auditors. Setion 5 onsiders how the design of regulation is affeted by eonomies of sope arising from the joint provision of auditing and onsulting servies. Setion 6 onludes. 3

7 2. Related literature Our model is related to the miroeonomi analysis of the auditor-firm relationship proposed by Dye (1993), where auditors an ontribute to the effiient alloation of investment but the quality of their audits is unobservable, leading to a moral hazard problem. In ontrast to Dye s model, however, in our setting this problem is not left to litigation between investors and aountants, but entrusted to regulation and its enforement by publi offiials. The optimal regulatory response must take into aount its enforement ost, as in Immordino and Pagano (2005). The result is a normative analysis of the regulation of the auditing profession, whih takes into aount the possible onflits of interest of auditing firms, as well as the ageny problems between managers and shareholders within lient firms. Therefore, our paper is relevant to the ongoing debate about the appropriate regulatory response to the reent orporate sandals. Our analysis fouses on regulation and publi enforement as the only devie to temper ageny problems in auditing and ollusion between auditors and lient ompanies. Private mehanisms have been suggested in the literature as alternative remedies to these problems: auditors self regulation assisted by litigation-based enforement, reputational mehanisms, ertifiation by intermediaries, finanial statements insurane, whistleblowing, et. But none of these mehanisms is indisputably superior to publi regulation. Indeed, as mentioned in the introdution, the reent orporate sandals highlighted the weaknesses of self-regulation and private litigation as exlusive disipline devies for auditors. The reliability of audit ompanies as self-regulating entities has been tarnished by the inreasing onflit of interest between their auditing role and their onsulting role, as the share of onsulting fees kept inreasing in their revenues. This point is reminisent of the result in DeMarzo, Fishman and Hagerty (2005), that a self-regulatory organization aountable to its members tends to hoose laxer enforement than ustomers would. Of ourse, even in a system of publi regulation, enforement an be entrusted to litigation rather than to regulatory intervention, as postulated in our model. The limitation of litigation-based enforement is that the osts of suing auditors may deter dispersed investors from taking ation against violations, due to olletive ation problems. In ontrast, a well-endowed and highly motivated publi proseutor an be very effetive against finanial fraud, as witnessed by the ativity of New York State Attorney General Eliot Spitzer sine Reputation is another deentralized mehanism that might enhane the reliability of auditors, espeially onsidering the limited number of ative auditing firms, the repeated nature of their 4

8 interations with lient firms and investors, and espeially the large stakes represented by the auditors equity base. Therefore, in priniple this mehanism ould be effetive. However, it appears not to have deterred negligent or fraudulent behavior so extensive as to wipe out established ompanies suh as Arthur Andersen. Even though the reasons why reputation has been ineffetive are still unlear, its limitations suggest that it needs to be omplemented by regulatory intervention. Alternative mehanisms that ould in priniple be used to improve the reliability of auditors are intermediaries that ertify the quality of their information (Lizzeri, 1999) or the reation of a finanial statement insurane (FSI) sheme, by whih ompanies purhase insurane that protets investors against losses due to misrepresentation in finanial reports and the insurer itself appoints and pays auditors (Dontoh, Ronen and Sarath, 2004). Certifiation intermediaries would tend to realign the inentive of auditors towards truth-telling by reating a seond layer of monitoring, while the FSI sheme would eliminate the onflit of interest arising from managers hiring auditors. But neither of these mehanisms is ompletely ollusion-proof: both a ertifiation intermediary and an insurane ompany providing FSI might ollude with the lient firm, unless they are themselves under the purview of a publi enforement mehanism. 5 Thus, rather than dispensing with publi intervention, these mehanisms simply shift the need for suh intervention to a higher layer. The FSI sheme suffers also from another problem: auditors appointed by outsiders may be unable to aess sensitive data plaed under management s ontrol, and this lak of ooperation by management may prevent them from performing effetive audits. The danger of ollusion between auditors and their lients suggests that whistle-blowing mehanisms might be partiularly suited in this ase, being designed to break ollusion. Suh mehanisms ould reward individuals (e.g., employees) who report fraud by auditors and/or their lients, by entitling them to a portion of the penalty paid by the latter. Suh shemes may deter ollusion, as shown by Spagnolo (2004) in the ontext of antitrust regulation. But in our ase whistle-blowing is unlikely to be effetive, sine the whistle-blower would hardly be able to doument the tait exhange of favorable audits against onsulting ontrats. 6 This disussion undersores the importane of publi regulation and enforement as residual mehanisms to disipline the auditing profession. 5 Collusion an hinder the effetiveness of a FSI sheme in two ways. First, the fat that auditors are hosen and paid by the insurane ompany does not rule out that lient firms might try to bribe them. Seond, the lient ompany might bribe the insurane ompany to underreport FSI premiums in exhange for higher premiums on other insurane ontrats, and thereby mishievously appear in the eyes of investors as ompanies with high-quality finanial statements. 6 We thank Gianarlo Spagnolo for pointing this out to us. 5

9 3. The model This setion explains the rationale for regulation of auditing in a setting where auditing has informational value in raising new finane, as in Dye (1993). As a benhmark ase, we first analyze a setting where the auditors ativity is observable and ontratible, and the eonomy ahieves the first-best outome. We then examine what happens if investors annot observe the level of effort that auditors invest in their task. This moral hazard problem in auditing implies that auditors will hoose the minimal level of quality. Under our assumptions, the soial ost of this moral hazard is that investors will alloate their funds less effiiently Informational value of auditing Consider an eonomy with risk neutral agents and a ontinuum of firms. The representative ompany is managed in the interest of the shareholders, so that the manager s objetive is to maximize its urrent value. To ontinue operating, the ompany needs a ash injetion (investment) of size I. Absent suh refinaning, the ompany is liquidated at a value that for simpliity is normalized to zero. The firm also needs to raise ash to pay for any fees F required by its auditors. Assuming that the required rate of return on new apital is standardized to zero, shareholders provide the needed ash infusion in exhange for shares that are worth at least I + F. 7 Eventually, the ompany may turn out to be a suess (state s = H) or a failure (state s = L). State H ours with unonditional probability p, and state L with probability 1 p. If the ompany is suessful, its final value V % is V H ; if not, it is VL < I < VH. Thus, in the bad state it is not worth refinaning the ompany. Sine there is a ontinuum of firms, p is also the fration of suessful firms. The initial shareholders and the manager are supposed to have no private information about the ompany s future value. 8 So, absent any additional information, its market prie P is the unonditional expetation of its final value, V = pv + (1 p) V. We assume that V > I, so that it is worth refinaning the ompany even if no information is gathered via an audit report. H L 7 A smaller stake would violate their partiipation onstraint. It is indifferent whether this ash infusion is ontributed by the initial shareholders or by new shareholders. 8 In many real-world irumstanes, however, managers have private information about the firm s true value. In this ase, auditors also perform the role of monitoring the manager s information. We analyze this alternative setting in a ompanion paper (Immordino and Pagano, 2007). 6

10 However, an audit may still be worthwhile as it allows investors to ondition the refinaning deision on an early signal of the firm s prospets. If the ompany is audited before it raises additional equity, its market prie will reflet also the information ertified by the audit. Auditors have a ostly tehnology to eliit a signal that aids in distinguishing high-value from low-value firms, and relate this signal by filing a report r on the value of the firm. In pratie, auditors assess only the reliability of the historial and prospetive information provided by the ompany s aountants, and deliver this filtered information to investors who use it to evaluate the ompany. 9 As in Dye (1993), we ollapse these two phases (the validation of aounting information and the evaluation made by the market) in a redued-form proess, by viewing the auditor s report as an assessment of the value of the ompany. An auditing firm an hoose the preision of its signal that it observes. The signal is assumed to be perfetly aurate when the state is H, while it may be inaurate if the state is L. The auditor truthfully relates the signal in his report r. 10 Formally the onditional probabilities of the auditor s report being orret are: Pr( r = L s = L, q) = q, Pr( r = H s = H, q) = 1. (1) We assume that the auray q of the auditor s signal is influened by the quality of the audit, as determined by the proedures adopted (e.g., external onfirmation of aounting data) and on the auditor s internal organization (e.g., personnel seletion riteria). But better audit quality omes at a ost. Formally, the auditor hooses the audit quality q [0,1) at a ost C(q), whih is ontinuous, inreasing and onvex in q, with C (0) = 0, lim C'( q) = 0 and q 0 lim C'( q) =. q 1 Using Bayes rule, the probability that the ompany will sueed onditional on a good report is: Pr( s = H r = H) p Pr( s = H r = H) = =. Pr( r = H) p+ (1 p)(1 q) (2a) while the probability that it will sueed onditional on a bad report is zero: 9 Auditors also soliit the prodution of additional information from the ompany s managers and aountants, when they are dissatisfied with its reliability. If met with refusal, they an threaten a dislaimer of opinion. 10 Even though the auditor annot misreport the observed signal, he an hoose to rely on a totally uninformative one. In this ase, the signal will always be favorable to ontinuation of investment, as will be shown below. 7

11 Pr( s = H r = L) Pr( s = H r = L) = = 0. Pr( r = L) (2b) The initial prie of the ompany P equals the expetation of the ompany s value onditional on the auditor s report, whih using the probabilities in (2a) and (2b) an be written as if the report is favorable, or V (1 )(1 ) ( ) H p VL p q EV % + r= H = V p+ (1 p)(1 q) (3a) EV ( % r= L) = VL (3b) if it is unfavorable. If no audit is arried out, the probabilities of the states H and L will be the unonditional ones, p and 1 p, and the market prie of the ompany will be equal to V. The shareholders surplus from ontinuation, before netting out the audit fee F (if an audit is performed), EV ( % r) I, takes different values in eah of these ases. With no audit, the surplus from ontinuation is V I, whih is positive by assumption. So a fortiori the surplus is positive when the audit report is favorable. If instead the report is unfavorable, the surplus is V I < 0. Therefore, the firm is refinaned only if no audit report is filed or if the report is favorable. It is liquidated if the report is unfavorable, so that in this ase the surplus, onditional on the optimal investment deision, is zero. Whenever an audit is ommissioned, its ost F must be deduted from the shareholders surplus. Let us denote by r the shareholders surplus net of the audit fee F and onditional on a report r and on the optimal investment deision. This net surplus is H = E( V% r = H) I F and = F, depending on the report filed by the auditor. The informational value of an audit is the L differene between the expeted value of without an audit, V I: r with an audit and the surplus from ontinuation L Ω ( q) = Pr( r = H) + Pr( r = L) ( V I) = q(1 p)( I V ) F. (4) H L L This expression is inreasing in the quality of auditing q, dereasing in the firms quality p (the worse the pool, the more valuable is information), and inreasing in the losses that would arise from investing in bad firms. The term I V is a measure of the potential misalloation of investment that an be prevented by auditors information. L 8

12 3.2. The unregulated outome If the audit fee F just equals the auditors ost C(q), i.e. if auditors make zero profits, then expression (5) beomes the net soial surplus (on a per-firm basis): W( q) = q(1 p)( I V ) C( q), (5) Indeed, sine auditors earn zero profits, the entire net soial surplus arues to the shareholders. The first-best outome is obtained by maximizing the net soial surplus W(q). Given our assumptions about the auditor s ost funtion, W(q) is onave and has an internal maximum where the marginal value of audit quality equals its marginal ost. This identifies the first-best quality fb value q (0,1) : Sine the ost funtion C(q) is onvex, L fb (1 p)( I V ) = C'( q ). (6) L fb q is dereasing in the quality of the pool and inreasing in the potential misalloation of investment, just as the informational value of auditing. If the audit quality is ontratible, the first-best outome emerges as the ompetitive market equilibrium. Firms managers hoose their demand for audit quality by maximizing the informational value of auditing, Ω ( q). Auditors hoose their supply of audit quality by maximizing their profit per audit, Fq ( ) Cq ( ), and make zero profits. The market-learing prie of fb an audit will then be the fee orresponding to the first-best level of audit quality, Fq ( ). It is easy to show that if quality is observable, the first-best alloation oinides with the Bertrand equilibrium of the model, that is, the Nash equilibrium of an extensive-form game where auditors hoose the quality q of the audit and a fee funtion F(q). The strategy of auditor j is a hoie of quality and fee, whih is the best response to the qualities and fees hosen by ompeting auditors. The situation in whih all firms hoose the first-best quality and prie is a Nash equilibrium, sine no firm an profitably deviate. If instead the audit quality is privately unobservable, then for any positive audit quality expeted by investors, auditors have an inentive to hoose a lower level and save the orresponding ost. As a result, the equilibrium audit quality is zero, the market prie will equal the unonditional 9

13 expetation V, and an unprofitable firm will be more likely to ontinue operating. 11 So in this ase there is a rationale for publi intervention. To this we turn in the next subsetion. 3.3 Regulated auditing The government sets an auditing quality standard q. This implies that auditors must hoose a quality level at least equal to q. If they deviate, they are liable to pay a penalty l. The quality hosen by auditors is observable and verifiable at a ost by a regulator, who hooses also the amount of resoures e devoted to enforement, i.e. to detetion of violators. The penalty is monetary and annot exeed an upper bound, 12 denoted by l. This bound an be thought as the entire wealth of the auditing ompany, whih is taken as exogenous in the ontext of the relationship with a speifi lient. 13 Figure 1 illustrates the sequene of moves. First, nature hooses the state s. Seond, the regulator hooses the audit standard q, the penalty l and the enforement e. Third, auditors hoose the quality level of their audit q harging the fee F(q), and produe the orresponding report r. Fourth, bureaurats enfore the standard by inspetion, deteting non-ompliane with probability f () e. 14 Next, the stok market sets the prie P of the ompany at a level refleting the atual report filed by the auditor and the pereived quality of auditing; depending on the report, shareholders may ontribute equity to finane the ompany. Finally, the ompany s atual value is determined. 11 Here we are assuming that the auditor s fee is not onditional on the ex-post auray of the report. Admittedly, suh a ontrat ould eliit a positive level of effort from the auditor, but it would not generally eliit the first-best auditing quality if auditors have limited liability. Moreover, in the auditing profession ontingent ontrats annot be used as an inentive devie beause auditors ode of professional ethis (rule 302) prohibits them from aepting inentive ontrats for audit-related work (Dye, 1993, p. 888, footnote 5). This ban arises from effiieny reasons: auditors are hired by managers, and inentive ontrats would provide easy ways to bribe auditors. 12 Auditors limited liability implies that the first-best annot be ahieved even via a more sophistiated design of the penalty or of enforement, for instane by making them ontingent on the auray of the auditor s report. (The point is analogous to that made in footnote 10.) 13 Auditors wealth an derive from interation with previous ustomers and from the sale of non-audit servies. In reality it may be impossible to onfisate the entire wealth of the auditor, due to the danger of subversion of justie (Glaeser and Shleifer, 2003). 14 In our setting, the regulator ommits to the probability of detetion f () e, by allotting the level of resoures e to enforement ativity. One an think of e as the salaries paid to the offiials in the authority that oversees the appliation of audit standards: one hired, these detet violations with a probability given by their enforement tehnology. 10

14 Nature hooses the state of the world s. Firm buys audit servie at fee F; auditor hooses audit quality q, and files report r. Market sets prie of ompany. The ompany raises funds I + F. The investment deision is taken. Regulator hooses legal standard q, penalty l and enforement e. Bureaurats detet ompliane with standard q with probability f(e). The ompany s atual value V is determined.. Figure 1. Time line To enfore the audit standard, the regulator must bak it up by a suffiiently high expeted penalty L in ase of non-ompliane. The auditor s profit is: Π = Fq ( ) Cq ( ) L, (7) where the expeted penalty L is the produt of the probability f(e) of deteting a non-omplying auditor and the statutory penalty l. The probability of detetion is inreasing and onave in the regulator s effort: f '( e ) > 0, f "( e ) < 0, with f (0) = 0, lim f '( e) e 0 = and lim f '( e) = 0, that is, e the enforement tehnology has dereasing marginal produtivity. So the expeted penalty is: f () el if q q, L < = 0 otherwise. Being benevolent, the regulator hooses the auditing standard q, the enforement level e and the penalty l so as to maximize the soial surplus from auditing quality net of the assoiated 11

15 enforement ost e, subjet to the inentive-ompatibility onstraint of auditors. Formally, the problem is to maximize the seond-best welfare level lq,, e subjet to the inentive ompatibility onstraint: sb sb W : L max W ( q ) = q (1 p)( I V ) C( q ) e (8) Fq ( ) Cq ( ) Fq ( ) Cq ( ) fel ( ) for any q q, (9) where the auditor s fee F on both sides of the inequality orresponds to the presribed audit quality expeted by investors, while the ost C depends on the quality level atually hosen by the auditor. As in Beker (1968), for any positive enforement level it is optimal to set the penalty at the maximum feasible level: 15 l = l. To obtain the optimal enforement level, we use the inentive ompatibility onstraint (9) with equality, sine the optimal poliy requires this onstraint to be binding. If not, the regulator ould inrease welfare by lowering enforement e, for any given l. Next, notie that, in ase of non-ompliane, the auditor would optimally deviate to a zero quality level, sine this would minimize his ost. Finally, sine the detetion probability f(e) is monotonially inreasing, it an be inverted to yield the optimal enforement: 1 eq ( ) = f ( Cq ( )/ l ). (10) From the properties of the enforement and audit tehnologies, it is immediate that the optimal enforement e is an inreasing and onvex funtion of the audit standard q, and a dereasing funtion of the maximum penalty l. 16 The positive relationship between enforement and audit standards highlights their omplementarity: a more demanding audit standard invites nonompliane by auditors, so that it must be assisted by more intensive monitoring by the regulator. Replaing the optimal enforement (10) into the objetive funtion, the problem of maximizing (8) an be rewritten as: L max q (1 p)( I V ) C( p ) e( q ), (11) q 15 To see why, notie that if the penalty were set at a lower level, inreasing it would enable the regulator to derease enforement e while keeping L onstant. The soial surplus in the objetive funtion would be unhanged but the enforement ost would be lower, so that welfare would be higher. 16 The first derivative of enforement with respet to the standard is e' = f ' 1 ( C( q )/ l ) C'( q )/ l > 0. Its seond derivative is e" = f ' 1 ( C( q )/ l ) C"( q )/ l + f " 1 ( C( q )/ l ) ( C'( q )/ l ) 2 > 0. 12

16 whose first-order ondition is (1 p)( I VL ) = C'( q ) + e'( q ) (12) Under our hypotheses on the limiting behavior of the Cq ( ) and f ( e ) funtions, this optimality ondition identifies an interior solution q > 0. More importantly, it implies that: Proposition 1 (Seond-best audit standard). The optimal audit standard q is smaller than the first-best standard fb q. The proof of this proposition (and subsequent ones) is in the Appendix. The intuition for why the optimal standard is lower than the first-best level is simple: the regulator must take into aount the resoure ost of enforing it. It is interesting to explore how the optimal standard q varies depending on the parameters of the eonomy: Proposition 2 (Comparative statis). The optimal standard is dereasing in the fration of suessful ompanies p and inreasing in the required investment I, in the effiieny of the auditing and in the effiieny of the enforement tehnology. Intuitively, the audit standard should be more demanding if audits allow investors to pik the few winners in a bad pool, and/or if the audit ost is spread over a large investment. These are situations in whih the soial value of a reliable auditor is very high. Moreover, a ountry an afford higher standards if its auditors beome more effiient in their job and/or regulators beome better at monitoring them. 4. Auditors onflit of interest and ollusion with audited firms As disussed in the introdution, one of the alleged soures of the reent orporate sandals has been the ability of managers to buy the aquiesene of auditing firms by exploiting the onflit of interest between their onsulting arm and their auditing arm. Auditing firms an provide advisory servies in the area of tax, aounting or management information systems as well as strategi advie, whih for brevity shall be referred to as advisory servies. If managers behave opportunistially rather than in the best interest of investors, they an use the fees for the purhase 13

17 of onsulting servies as bribes to buy the aquiesene of auditors to their objetives. To apture this point, we shall introdue the following new assumptions. Instead of maximizing the firm s profits as assumed in previous setions, let us onsider the situation where managers maximize the firm s investment, so as to reap a private benefit B stemming from the ontinued operation of the firm. This is a sum of money that he an appropriate at the shareholders expenses without being punished. As a result, the shareholders payoff from the investment in ase of ontinuation is V % I B. Under these assumptions, managers will try to indue auditors to issue positive reports so as to indue the firm s ontinuation as often as possible. We assume that, owing to olletive ation problems, shareholders annot ditate the appointment of auditors, and therefore annot prevent suh opportunisti behavior by managers. However, we rule out diret bribes to the firm s auditor via monetary transfers, on the ground that suh openly illegal bribes would be detetable via whistle-blowing, and therefore punishable by law enforers. Diret monetary bribes are not allowed even when disguised as auditors fees in exess of the ompetitive level: we assume that this would be detetable by inspetion of the ompany s aounts. 17 Companies an purhase two servies from auditors: auditing servies at the fee F(q) = C(q), as before; and advisory servies, whih ompetitors of the inumbent auditor supply at the prie R. Companies are assumed to value advisory servies more than their prie R. The inumbent auditor has a ost advantage over his ompetitors: he produes advisory servies at zero ost, due to his aquaintane with the ompany and to eonomies of sope in the joint provision of auditing and advising. 18 So, if the inumbent auditor gets the onsulting ontrat at the going prie R for the provision of advisory servies to the firm, he earns a rent R. In the hands of an opportunisti manager, this rent ould be used to buy the aquiesene of auditors. That is, the manager may ondition a onsulting ontrat with the inumbent auditor to reeiving a positive report r = H, irrespetive of the true state of nature. Although formally legal, 17 If monetary bribes were allowed, the possibility of orruption would trivially arise even in the absene of the onflit of interest arising from the joint provision of auditing and onsulting servies. However, pratitioners regard this onflit as the main soure of auditors involvement in orporate sandals (see Crokett, Harris, Mishkin and White, 2003). 18 The zero-ost assumption is made purely to save notation and implies no loss of generality: all that is needed is a potential ost advantage of the inumbent auditor over the ompetition. 14

18 this transation obviously plays the same role of a bribe. 19 If he were to aept this bribe, the auditor would optimally invest no resoures in auditing, that is, he would hoose zero quality sine in our setting this will yield a positive report with ertainty: Pr( r = H q= 0) = 1. In this event, the inumbent auditor would earn Cq ( ) + R, that is, the sum of the ost saving in auditing Cq ( ) obtained by hoosing zero quality instead of the statutory quality q and the ost saving R in the prodution of advising. The manager s payoff from bribing the auditor derives from the implied inrease in the probability of grabbing the private benefit: with the bribe, the firm ontinues with ertainty, whereas without the bribe it would ontinue only if the auditor s report were positive, that is, with probability Pr( r = H). Therefore the additional expeted gain to the manager is [1 Pr( r = H)] B. 20 This hanges in two ways the regulator s problem of hoosing the optimal audit standard and the assoiated enforement spending. On one hand, the ost effiieny of the inumbent auditor inreases soial welfare. On the other hand, the implied rent an be used by the firm s manager to bribe the auditor. Formally, the regulator s objetive funtion must be modified by inluding the ost savings R of the inumbent auditor: where ˆ max W( q ) = q (1 p)( I V ) ( ) L + R C q e, (13) lq,, e q denotes the standard hosen by the regulator when there is the danger of ollusion. The new symbol ˆ W indiates that the soial welfare funtion has been modified to inlude also the effiieny gain R. Instead, the presene of private benefits B does not affet the expression for soial welfare, sine they are a pure transfer. The inentive ompatibility onstraint of the auditor must be rewritten to take into aount that managers an use the rent R to orrupt inumbent auditors: 19 Admittedly, the potential for ollusion may arise even within the market for auditing servies, if this market is not perfetly ompetitive or if the inumbent has a ost advantage relative to the ompetition in the provision of auditing servies. Also in this ase, auditors ould be bribed by the threat of losing their rents. But in reality the rents appear to arise mainly from the joint provision of auditing and onsulting servies. 20 Therefore, the total surplus available to the manager and the inumbent auditor is the sum of the auditor s ost savings Cq ( ) + R and the manager s additional private benefits Pr( r = L) B. Sine no monetary transfers are allowed, they annot split their joint surplus in any way that differs from the one onsidered in the text. However, it must be notied that this assumption is not ruial to our results. We ould allow monetary transfers and therefore any possible split of the joint surplus between manager and auditor, as long as the bargaining leaves part of the total surplus to the inumbent auditor, so that the potential for ollusion remains. 15

19 Fq ( ) Cq ( ) Fq ( ) + R fel ( ). (14) Now regulation and enforement must be designed also to deter the auditor from aepting the bribe. Going through the same steps as in the previous setion, one finds that now the optimal enforement level is: ( ) 1 e ( q ) = f ( C( q ) + R)/ l. (15) This expression, whih is inreasing and onvex in q, shows that the enforement neessary to uphold a given audit standard q exeeds the seond-best level identified in equation (10) beause the possibility of obtaining the rent R from onsulting raises the enforement ativity needed to prevent ollusion. The optimal audit standard is orrespondingly lower: intuitively, the potential bribe raises the ost of enforement, induing the regulator to hoose a less ambitious standard. The larger is the rent R from advisory servies, the greater the potential for orruption, and therefore the less ambitious the auditing standard must be: Proposition 3 (Audit standard with onflit of interest). If the rent R from advising is positive, then the optimal standard q ( R) is lower than the seond best R. q for R > 0 and it is dereasing in Note that the onflit of interest may indue the regulator not to intervene: the rent R inreases disretely the inentive to deviate and thereby in ertain ases the maximal value of welfare attainable with intervention would fall below zero: ˆ W < 0 for any q > 0. In this ase, the regulator will rather hoose q = 0, in ontrast with the seond best. One might suggest that an even better poliy would be to eliminate the onflit of interest at its root, by severing the link between auditing and advisory ativity. In this ase, the optimal aounting standard would inrease to the seond best level q. However, this does not neessarily inrease welfare: even though auditing standards would inrease, soiety would forgo the ost savings arising from the joint provision of auditing and advisory servies. The next setion analyzes under whih irumstanes a ban on suh joint provision is justified. 16

20 5. Conflit of interest versus effiieny gains from non-audit servies Now we onsider the ase in whih the regulator, beside setting an auditing standard and an enforement level, may forbid the provision of advisory servies by auditors. Indeed, this is one of the provisions ontained in the Sarbanes-Oxley At (SOX). SOX prohibits all registered publi aounting firms from providing audit lients, ontemporaneously with the audit, ertain non-audit servies inluding internal audit outsouring, finanial-information-system design and implementation servies and expert servies. These sope-of-servie restritions go beyond previous SEC independene regulations. In addition, all other servies, inluding tax servies, are permissible only if pre-approved by the issuer s audit ommittee and all suh pre-approvals must be dislosed in the issuer s periodi reports to the SEC. Similar reforms have reently been approved also in other ountries. For instane, Italy s 2005 Law for the Protetion of Saving and Finanial Market Regulation (artile 18) ontains a prohibition that is very similar to that ontained in SOX. If regulation allows bundling of auditing and advisory servies, soiety reaps the effiieny gain arising from eonomies of sope, but opens the door to the managers attempt to bribe their auditor. If instead the regulator forbids bundling of these servies, it forgoes the effiieny gain R arising from eonomies of sope. If this strategy is hosen, the problem reverts to that of Setion 3, and the resulting auditing standard beomes simply the seond-best level Setion 3. q, derived at the end of Whih of these strategies maximizes soial welfare? The answer depends on two fators: (i) the magnitude of the rents R from onsulting and (ii) the marginal effiieny of enforement. Indeed, an inrease in R has two onfliting effets: an inrease in the potential for ollusion between manager and auditor, and an effiieny gain from auditors eonomies of sope. If the marginal effiieny of enforement is suffiiently high ( e'( q ) > C'( q ) ), the negative effet of an inrease in R is dampened by the law, so that the positive effet from the effiieny gain dominates. But this holds only for relatively low values of R: one the rents from onsulting that an be used to bribe onsultants beome too high, the negative effet dominates anyway. This is illustrated in the top panel of Figure 2. The solid urve represents the welfare level that obtains when the regulator allows the joint provision of auditing and onsulting servies, ˆ W( q ; R ), whereas the flat line indiates the welfare level assoiated with the prohibition of bundling, that is, sb the seond-best welfare level W ( q ). Below the threshold R %, the solid urve lies above the flat 17

21 Panel A W ˆ ( ; ) if '( ) '( > ) W q R e q C q sb W ( q ) ˆ ( ; ) if '( ) '( < ) W q R e q C q 0 R % R q Panel B q q ( R ) 0 R % R Figure 2. Soial welfare W (panel a) and optimal auditing standards q (panel b) as funtions of the rents from onsulting R 18

22 line, so that it is optimal to allow the joint provision of auditing and onsulting servies. Above this threshold, instead, it is optimal to forbid bundling. The bottom panel of Figure 2 shows how the optimal audit standard hanges in response to different values of the rent R. The audit standard must be set at the seond-best level q when R = 0, dereases as R inreases (as shown by Proposition 3) up to the threshold R %, and reverts to the seond-best level q for even larger R. Intuitively, in the region where bundling is allowed, the standards must be gradually mellowed as R inreases in order to aommodate the inreasing severity of the onflit of interest. This proess stops as soon as the eonomy enters the region where it is optimal to forbid bundling, thereby removing the onflit of interest at its root. But also another ase is possible, whih is illustrated by the dashed line in Figure 2: if the marginal effiieny of enforement is suffiiently low ( e'( q ) < C'( q ) ), then the welfare level ˆ ( W q ; R ) that obtains allowing the joint provision of auditing and onsulting (the dashed urve) lies entirely below the seond-best welfare level sb W ( q ) (the flat line). In this ase, bundling should always be forbidden, and the audit standard should always be set at its seond-best level for any R. In this ase, the legal system is so ineffiient that even a very small bribe to the auditor is too ostly to deter. 21 These points are summarized in the following proposition: q Proposition 4 (Interdition of bundling and audit standards). If e'( q ) > C'( q ), then there exists a threshold value R % >0 suh that for R R % it is optimal to allow bundling of auditing and onsulting, and to forbid it for R standard is q ( R ) for R R % and q for R > R %. > R %. The optimal audit If e'( q ) < C'( q ), then it is optimal to forbid bundling of auditing and onsulting and set the audit standard equal to q for any R. One may wonder whether, rather than banning the joint provision of audit and non-audit servies, the regulator may not want to limit its intervention simply to mandating dislosure of information about the identities of the auditor and the onsultant of the firm, trusting that 21 In the proof of Proposition 4 we also show that in the borderline ase where e'( q ) = C'( q ), soial welfare is the same in the two regimes, so that the regulator is indifferent between allowing and forbidding bundling for any R. 19

23 shareholders will intervene to forbid the joint provision of the two types of servie in the regions where this is ineffiient. However, in our setting this type of mandatory dislosure would not prevent ollusion between managers and auditors, due to the assumption of olletive ation problems among shareholders. Indeed starting from 2000 the SEC had already mandated firms to dislose audit and nonaudit fees paid to their auditors, hoping to provide investors with transparent information about the auditor-lient relationship. The approval of SOX in 2002 indiates that this dislosure was not deemed suffiient to ahieve the desired degree of auditors independene. As explained above, the results of Proposition 4 rest on the balane between the two onfliting effets of the rent R earned in the provision of non-audit servies: on the one hand, the easier apture of auditors by managers; on the other, the effiieny gains from eonomies of sope. Various papers have attempted to measure empirially whih of these two effets dominates. Some event studies show that unexpeted inreases in non-audit fees have a negative or insignifiant effet on the abnormal returns of ompanies (Frankel, Johnson and Nelson, 2002, Ashbaugh, Lafond and Mayhew, 2003, and Chaney and Philipih, 2002). Other studies have instead unovered evidene of eonomies of sope. Some of them have done so by showing that audit fees are positively orrelated with non audit-fees (Simuni, 1984; Palmrose, 1986; Parkash and Venable, 1993). The study by Antle, Gordon, Narayanmoorthy and Zhou (2004) shows that this positive orrelation survives even after ontrolling for the joint determination of audit and non-audit fees as endogenous variables. In a reent paper, Cho, Han and Brown (2006) use a different empirial strategy: they investigate how stok market pries respond to aounting information, and test whether the responsiveness of pries hanges depending on the ratio of audit to non-audit fees paid the ompany. They find that the prie relevane of aounting information is greater in ompanies where non-audit fees are omparably larger, whih an be seen as further evidene in favor of eonomies of sope in the ativity of auditors. 6. Conlusion The reent orporate sandals have highlighted the need for tighter regulation of the audit profession. However, one it is reognized that the enforement of suh regulation is ostly, three important lessons an be drawn onerning the optimal standards to be imposed on auditors. First, audit quality standards must be based on a ost-benefit analysis of audit ativity. On the ost side, they must be less ambitious in eonomies that have less effiient enforement. On the 20