European Accounting Review Vol. 16, No.

2, 429 454, 2007

External Auditors, Audit Committees and Earnings Management in France

MI JANIN CHARLES PIOT & RE
GSCM Montpellier Business School, CEROM, Montpellier, France and ` s France University, Grenoble Business School, CERAG-CNRS, Pierre Mende Grenoble, France

ABSTRACT We investigate the effect of various audit quality dimensions (i.e. auditor reputation and tenure, audit committee existence and independence) on earnings management in France. We thus contribute to the empirical audit quality literature in a Continental European environment that markedly differs from the USA in terms of auditing and corporate governance. The main ndings are that: (1) the presence of an audit committee (but not the committees independence) curbs upward earnings management; and (2) the presence of a Big Five auditor makes no difference regarding earnings management activities. Implications of these ndings are discussed with regard to the specicities of the French auditing and governance settings. In particular, although the audit committee acts as a device to control the more egregious (i.e. incomeincreasing) forms of earnings management, the monitoring incentive of outside directors may be hampered by the collective board responsibility for nancial reporting quality. Second, the lack of differentiation among Big Five auditors in terms of accounting conservatism is consistent with the lower litigation risk offered by the French Civil Code (vs. the US Common Law system), which is likely to eliminate the deep pockets incentive for investors.

Correspondence Address: Charles Piot, Associate Professor, CEROM Research Centre, Montpellier Business School, 2300, avenue des Moulins, F-34185 Montpellier Cedex 4, France. Tel.: 33 4 67 10 28 02; Fax: 33 4 67 45 13 56; E-mail: cpiot@supco-montpellier.fr 0963-8180 Print/1468-4497 Online/07/02042926 # 2007 European Accounting Association DOI: 10.1080/09638180701391030 Published by Routledge Journals, Taylor & Francis Ltd on behalf of the EAA.

430 1.

C. Piot & R. Janin Introduction

In the context of the many nancial scandals challenging the credibility of the audit function, numerous studies tend to establish a positive relation between audit quality surrogates auditor and audit committee characteristics and the quality of nancial reporting by listed companies. Among these empirical works, a growing number address earnings management,1 which can be dened as the use of managerial discretion to inuence the accounting gures published to the companys stakeholders (Degeorge et al., 1999, p. 2). Virtually all these studies address US rms. In this paper, we extend the earnings management literature to the French context, which proves especially interesting in a Continental European environment that markedly differs from the Anglo-Saxon systems in terms of auditing and corporate governance. First, the strong legal protection of auditor independence in France (i.e. a six-year audit engagement, mandatory joint-auditing) vs. the more self-regulated Anglo-Saxon model suggests a greater ability of auditors to resist managerial pressure and keep earnings management practices in check. Second, however, the lower litigation risk, as compared to the responsive US litigation system, may increase the tolerance of audit rms toward opportunistic accounting practices. Third, the setting up and organisation of audit committees are currently recommendations, thus allowing listed companies important discretion whether to involve corporate directors in audit quality dard matters. Hence, following the US investigations by Klein (2002a) and Be et al. (2004), the French setting offers the opportunity to appraise the role of audit committees with regard to the quality of earnings, and to provide some insight into normative developments regarding the functioning of these monitoring devices. In this paper we perform empirical tests using abnormal accruals estimated by cross-sectional versions of the Jones Model (1991), for the main French listed companies over a three-year period (1999 2001). Abnormal accruals are considered in signed value form (proxying for conservatism), and in absolute value (measuring the overall propensity to earnings management). The main empirical ndings are that: (1) signed abnormal accruals decrease when an audit committee is present, but the audit committees independence has no signicant effect on accruals measurements; and (2) that Big Five-audited companies do not differ from others in terms of absolute and signed abnormal accruals. These ndings contribute to the institutional debates regarding nancial reporting quality, in relation to the French characteristics of auditing and corporate governance. First, audit committees stand as potentially valuable audit quality devices, because they constrain the more egregious (i.e. income-increasing) form of earnings management. However, contrary to the US ndings of Klein dard et al. (2004), the role of independent audit committees (2002a) and Be does not emerge. This could be due to the lack of explicit guidelines about the monitoring duties of outside directors, and to the collective board responsibility

Audit Committees and Earnings Management in France

431

for nancial reporting quality. This French collegial liability system may render sterile any further recommendation or regulation regarding the role of outside directors: reforming the corporate law on this question might then be of interest. Second, the lower litigation risk offered by the French Civil Code (vs. the more responsive US Common Law system) makes the large audit rms less exposed to the deep pockets incentive. Therefore, they do not have to handle this litigation threat by adopting a more conservative attitude with respect to earnings management, as empirically documented in the US market (Becker et al., 1998; Kim et al., 2003). Section 2 reviews the literature on relations between accruals and audit quality attributes, and then species the research hypotheses with reference to the French setting. Section 3 deals with methodological aspects. Section 4 reports and discusses the empirical results, and Section 5 reviews the main contributions of the paper. 2. Literature, Institutional Background and Hypotheses

Conceptually, audit quality is the joint probability that the external auditor detects an anomaly in nancial statements, and then reveals it to the external users (DeAngelo, 1981). The probability of detection is a matter of competence, whereas the probability of revelation depends on independence, that is, the auditors willingness to face the pressure exerted by the preparers of nancial statements. In reaction to nancial scandals, efforts implemented in the last few years to restore investors condence in nancial reporting suggest that audit quality does not solely focus on external auditors, but also includes a monitoring oversight by audit committees. In the USA, this reaction nds expression by challenging the self-regulation of the accounting profession (e.g. the Public Company Accounting Oversight Board), and by more restrictive rules on audit committees (e.g. Sarbanes Oxley Act, Sections 301 and 407). In France, the institutional context is markedly different. Historically, the auditing profession has been strictly state-regulated, especially with regard to independence. However, audit committees are not required, but only recommended by corporate governance not, 1995, 1999; Bouton, 2002). This allows listed companies importreports (Vie ant latitude whether to involve directors in the audit process. 2.1. External Auditors and Earnings Management The quality of the external audit, unobservable to the public, is commonly proxied by auditor size or reputation (DeAngelo, 1981; Klein and Lefer, 1981), or by the length of the auditor client relationship (Knapp, 1991). Auditor reputation and earnings management. In contrast with the scandals that challenge the credibility of well-known audit rms, many academic studies tend to demonstrate that the brand name audit networks (the Big Eight,

432

C. Piot & R. Janin

todays Big Four) are statistically more conservative in their opinions, and are more likely to constrain opportunistic accounting practices. DeFond and Jiambalvo (1993) nd that auditor client conicts relating to income-increasing accounting practices are more likely to occur if the auditor is one of the Big Eight, and conclude that these audit rms are more likely to resist managerial pressure and maintain an independent opinion. Becker et al. (1998) observe a positive relation between abnormal accruals and the presence of a non-Big Six auditor. On the NASDAQ, Francis et al. (1999) also observe a lower level of abnormal accruals among Big Six-audited companies. These ndings are commonly interpreted by the fact that Big N auditors have to maintain their reputation capital, and therefore provide quality-differentiated services which imply a lower tolerance towards earnings management in general. Given that this quality differentiation is perceived to operate homogenously worldwide, we state the following rst hypothesis in the French context: H1a: Big Five-audited companies exhibit less earnings management in general. However, researchers also bring forward auditors asymmetric monitoring of the earnings management problem, that is, less tolerance in the case of income-increasing earnings management, and more leniency toward incomedecreasing accruals. Francis and Krishnan (1999) nd that rms with largely positive abnormal accruals are more likely to receive a modied audit report. Nelson et al. (2002) emphasise that audit adjustments are more frequent in income-increasing (vs. income-decreasing) situations. In the UK, Gore et al. (2001) nd that the Big Five curb threshold-induced, income-increasing abnormal accruals. An explanation for this conservative attitude is generally believed to be the risk of legal action, and more specically the deep pockets incentive commonly attached to brand name auditors. Kim et al. (2003) show that the Big Five are more cautious than other auditors only in situations of incomeincreasing earnings management. They interpret this asymmetric monitoring as a consequence of the pressure the US judicial system exerts on the Big audit rms. The deep pockets argument is less likely in France because of the lower responsiveness of the Civil Law litigation system in protecting investors rights (La Porta et al., 1997, 1998). Also, Coffee (2001) stresses the historical passivity of the French stock market authority in developing a self-regulatory structure and high-quality disclosure standards. The litigation system remains specically more protective; by imposing a heavier formalism than Common Law principles, the Civil Code implicitly limits auditors third-party liability. Notably, it is difcult to establish an auditors professional fault in view of his due diligence (i.e. an obligation of means), and it is frequently impossible to prove a direct causality between this fault and the damages suffered by the plaintiffs. In practice, the reluctance of French judges to convict management and dit Lyonnais auditors stands out in two recent court decisions, namely, the Cre

Audit Committees and Earnings Management in France

433

and Pallas-Stern affairs (Stolowy, 2005). Further, the US judicial system allows class action suits and contingent fees. According to the Securities Act, it is enough for investors to claim that they trusted audited nancial statements when investing in a company, that these statements proved to be deceptive or fraudulent, which caused a drop in stock prices (Menon and Williams, 1994). Class actions and contingent fees are not allowed in France, and the activism of minority shareholders generally requires a formal association of plaintiffs, thus reducing the deep pockets incentive for investors. Consistently, Piot (2005) nds that riskier rms do not necessarily appoint high-prole auditors. Hence, the asymmetric monitoring of earnings management by large auditors should not prevail in France. H1b: Big Five-audited companies do not differ from others with respect to the conservatism of reported earnings. Auditor tenure and earnings management. Standard-setters tend to argue that an overly long auditor client association is a threat to independence. They assert that personal ties and familiarity tend to develop and may lead to less vigilance by the auditor, and even to a tolerant attitude towards the companys management. Aside from this threat to independence, the audit engagement may become routine over time, and the auditor may devote less effort to identifying the weaknesses in internal control and potential risks. Knapp (1991) studied auditor tenure and competence using the perception of US audit committee members, nding that the likelihood that an auditor detects an anomaly increases in the rst years of appointment, and then decreases gradually, reaching its weakest level after 20 years of service. As a whole, it is commonly assumed that audit quality decreases with auditor tenure. In reaction, and without mandating auditor rotation,2 stock market authorities generally impose a rotation of engagement partners. In the USA, this rotation was reduced from seven to ve years by the Sarbanes Oxley Act. In May 2002, the European Commission recommended a seven-year rotation of engagement partners. Most Member States, including the French market authority, have followed this recommendation. However, the theory that auditor tenure is inversely related to audit quality is far from being corroborated empirically. Lys and Watts (1994) report that the likelihood of an auditor being sued is not related to his tenure. Geiger and Raghunandan (2002) nd that the issuance of a going-concern report prior to bankruptcy is a positive function of auditor tenure, suggesting that the quality of auditor reporting improves over time. Furthermore, the magnitude of abnormal accruals is found to be negatively associated with auditor tenure (Frankel et al., 2002; Myers et al., 2003), and positively inuenced by a dummy indicating a short, two- or three-year tenure (Johnson et al., 2002). Finally, Ghosh and Moon (2005) document that the perceived quality of earnings, proxied by earnings response coefcients, increases with auditor tenure. Hence, these US studies refute the hypothesis that audit quality deteriorates over time.

434

C. Piot & R. Janin

In France, auditors are appointed for six scal years. Their mandate and to some extent their tenure thus enjoys a legal protection, initially enforced to mitigate opinion-shopping opportunities.3 But this protection may have adverse effects; by cutting market forces, it accelerates the formation of rents and the , 2000). If so, the capacity to auditors economic dependence on his client (Pige resist managerial pressure is likely to decrease over time. In line with this, the hypothesis of a positive association between earnings management and auditor tenure is proposed. H2: Auditor tenure is associated with higher levels of earnings management. 2.2. Audit Committees and Earnings Management An effective audit committee adds to the quality of the audit process at two levels. First, by supervising major accounting choices, the committee should mitigate earnings management practices. Second, by coordinating the internal and external audits, and by protecting external auditors independence from managerial pressure (McMullen, 1996), the audit committee should maximise the likelihood that irregularities discovered by auditors will be reported at a sufciently high level. US empirical studies tend to highlight a negative relation between the independence and/or expertise of the audit committee and the magnitude of abnormal accruals, although the consensus is unclear due to the diversity of the approaches used. Klein (2002a) nds that board or audit committee independence measured either by the proportion of independent directors or by their majority in both structures reduces abnormal accruals in absolute value. However, she reports insignicant effects for completely independent audit committees. dard et al. (2004) nd that the likelihood of aggressive earnings management Be (i.e. markedly positive or negative abnormal accruals) decreases if the audit committee includes a nancial expert or an expert in corporate governance, and if it is composed entirely of independent directors. Their ndings are thus in line with Sarbanes Oxley regarding fully independent audit committees, contrary to the study by Klein (2002a). However, Xie et al. (2003), studying the S&P500, as did Klein, do not observe any impact of the proportion of independent audit committee members on abnormal working capital accruals. In the UK, Peasnell et al. (2000a) nd that the presence of an audit committee4 has no direct inuence on abnormal working capital accruals, but allows board independence to mitigate income-increasing earnings management, and thus indirectly enhances the monitoring role of independent directors. Audit committees have developed among French listed companies since the not Report in 1995. Although they now tend to be present in the rst Vie largest companies, their creation remains a voluntary decision, and no explicit regulation exists on audit committees (except for nancial institutions). However, advisory panels have dened the role and missions of audit committees

Audit Committees and Earnings Management in France

435

not 1 Report states that the committee similar to the American model. The Vie should: (1) examine the relevance and consistency of accounting methods with a special focus on transactions potentially associated with conicts of interests; (2) appreciate the reliability of internal control and reporting systems; and not 2 (3) meet privately with the CFO, internal and external auditors. The Vie Report (1999) completes this framework insisting on the appreciation of external auditors independence, and on the examination of accounting standards for consolidated nancial statements. Piot (2004) documents that the existence of audit committees in France is positively associated with the extent of shareholder manager and debtholder conicts. Therefore, audit committees seem to respond to a monitoring demand by fund providers, either outside shareholders or debtholders. We then formulate the following third hypothesis: H3: The existence of an audit committee is associated with less earnings management. However, as evidenced by institutional debates and empirical research, the question of independence is essential to effective monitoring by audit committees. The related French guidelines remain well behind the stricter not 2 Report recommends a minimum of only US requirements. The Vie one-third of independent directors on audit committees; the Bouton Report (2002) mentions an objective of two-thirds. This evolution suggests an increasing pressure for more independent audit committees, but still provides listed companies with a lot of discretion on that question. Piot (2004) observes a negative relation between several measures of audit committees independence and inside ownership, supporting the view that more independent committees are a response to monitoring needs. We then test the following fourth hypothesis: H4: The independence of the audit committee is associated with less earnings management.

3.

Data and Methodology

3.1. Sample We investigate the main companies on the French stock market, and more specically the SBF 120 Index companies, of which approximately half had set up an ry-Dubuisson, 2002). These companies audit committee by the end of 1998 (Thie offer relatively precise disclosures about their auditors and their corporate governance organisation. To avoid the limitations of purely cross-sectional analyses, we considered three consecutive nancial years: 1999, 2000 and 2001. In the not French corporate governance context, this timeframe spans the second Vie Report (July 1999) and the Bouton Report (September 2002). The targeted

436

C. Piot & R. Janin

sample comprises 102 non-nancial rms5 which were included in the SBF 120 Index at least once during 1998 2002. Details about the sampling procedure are provided in the Appendix. 3.2. Determination of Abnormal Accruals We assume that the discretionary component of accounting earnings is mostly found in accruals. Total accruals are calculated for each rm i in the year t with the following indirect formula:6 TAit DSTAit DSTLit DepExpit AmortExpit DLTDefExpit DProvit where i 1 102, t 1999, 2000, 2001 and: . DSTA change in Short-Term Assets (STA inventories customers receivables other receivables short-term deferred expenses, all in gross values), . DSTL change in Short-Term Liabilities (STL payables to suppliers social and scal payables other payables deferred revenues), . DepExp Depreciation Expenses relating to Short-Term Assets, . AmortExp Amortisation Expenses concerning xed assets, . DLTDefExp change in Long-Term Deferred Expenses, . DProv change in Provision for risks and charges. We further estimate the abnormal component of total accruals using the Jones (1991) cross-sectional industry model. Industries are classied according to the three-digit NAF code (French codication roughly similar to the US SIC one). Industries that do not contain at least six companies in the Diane database used to estimate regression coefcients are dropped. Formally, the following Jones Model is estimated for each industry j and year t: TA jt a1 b1:GPPE jt g1:DREV jt 1 jt where (all variables including the intercept a1 are scaled by lagged total assets): . TA Total Accruals, . GPPE Gross Property Plant and Equipment Long-Term Deferred Expenses,7 . DREV change in Revenue, . e random error term. The sign of coefcient b1 is expected to be negative due to the increase in amortisation expenses when companies invest. The sign of g1, generally expected to be positive, accounts for the normal increase in working capital

Audit Committees and Earnings Management in France

437

when companies turnover grows.8 For each rm i and year t, the abnormal accruals according to the Jones Model (designated by AbnAcc1) are then calculated by difference between total and expected accruals: AbnAcc1it TAit a1 b1:GPPEit g1:DREVit : Furthermore, Dechow et al. (1995) have shown that the Jones Model is not well specied for companies with extreme cash ows. Hence, given the usually strong correlation between accruals and cash ows (Dechow, 1994), the following CFO Model (Jeter and Shivakumar, 1999) is also estimated for each industry j and year t: TA jt a2 b2:GPPE jt g2:DREV jt d2:CFO jt 1 jt with CFO obtained by difference between net income before goodwill amortisation and total accruals (all scaled by lagged total assets). The sign of coefcient d2 is expected to be negative. Abnormal accruals according to the CFO Model (designated by AbnAcc2) are then calculated as follows: AbnAcc2it TAit a2 b2:GPPEit g2:DREVit d2:CFOit : Descriptive statistics for the 146 industry-year OLS regressions9 to estimate the accruals models are disclosed in the Appendix. As expected, the coefcient of DREV is positive on average and in the median, and the coefcients of GPPE and CFO are negative. When CFO is introduced in the accruals model, the coefcient of GPPE decreases sharply in absolute value (from 20.081 to 20.019), while the coefcient of DREV increases noticeably (from 0.027 to 0.055). Descriptive statistics for total and abnormal accruals are reported in Table 1. Of the 306 rm-years initially targeted, we obtain 255 usable observations. This loss of observations is attributable to: (1) missing data in the Diane database, (2) the minimal requirements of six rms to estimate the accruals models, and (3) the elimination of outliers, when the observed value for AbnAcc1 or AbnAcc2 falls out of a range of three standard deviations around the average. Total accruals are signicantly negative on average ( p , 0.01), reecting the weight of income-decreasing components (e.g. amortisation expenses, provisions for risks and charges). The ANOVA does not reveal any signicant differences for TA over the three years. Abnormal accruals are signicantly positive on average ( p , 0.01) when estimated with the CFO Model, and not statistically different from zero when derived from the Jones Model. Also, the Jones Model produces more dispersed abnormal accruals as evidenced by the extrema (20.439 and 0.478 compared to 20.242 and 0.290 for the CFO Model). These statistics suggest that model specication affects the estimated abnormal component of accruals; it is thus of interest to run empirical tests using both approaches in parallel.

438

C. Piot & R. Janin

Table 1. Descriptive statistics for the pooled sample (1999 2001)

Variable

Meana

Median 20.030 0.003 0.021 0.050 0.035 1 7.553 1 0 0.005 0.715 2,485 0.429 0.083

Std. dev. 0.103 0.113 0.074 0.083 0.053 0.373 7.966 0.485 0.479 0.215 1.181 20,782 0.235 0.121

Min 20.380 20.439 20.242 0.000 0.000 0 0.504 0 0 0.000 0.000 57 0.000 20.397

Max 0.482 0.478 0.290 0.478 0.290 1 46.559 1 1 0.932 14.630 150,738 1.000 0.455

N 255 255 255 255 255 246 221 241 234 247 254 255 239 255

ANOVA f-stat.b 0.482 0.643 2.631 3.250 1.948 0.604 0.081 0.851 1.258 0.716 0.613 0.513 0.172 1.604

Accruals variables TA 20.026 AbnAcc1 0.004 AbnAcc2 0.019 jAbnAcc1j 0.077 jAbnAcc2j 0.055 Audit quality variables Big5 0.833 AudTen 9.647 AC 0.627 ACInd50 0.355 Control variables %Managers 0.139 Lev 0.965 Assets (ME) 10,774 BdInd 0.417 CFO 0.079
a

t-Statistic from the nullity test: TA (2 3.96 ), AbnAcc1 (0.58), AbnAcc2 (4.00 ). Percentage of positive observations: TA (30%), AbnAcc1 (53%), AbnAcc2 (67%). b One-way ANOVA according to the year distribution (1999, 2000 or 2001). , and denote signicance at p , 0.05, 0.01 and 0.001, respectively (two-tailed). Variables denition: TA Total Accruals, AbnAcc1 Abnormal Accruals derived from the Jones Model, AbnAcc2 Abnormal Accruals derived from the CFO Model, jAbnAcc . . .j absolute value of Abnormal Accruals, Big5 1 if one of the statutory auditors (at least) is a Big Five auditor, and 0 otherwise, AudTen time (in years) between the nancial year end and the date of rst nomination of the leading auditor, AC 1 if an audit committee is present, and 0 if not, ACInd50 1 if an audit committee with a majority of independent directors is present, and 0 if not, %Managers proportion of common shares owned by top executives, Lev debt-to-equity ratio, Assets total consolidated assets (in million euros), BdInd proportion of independent directors on the board, CFO (net income before goodwill amortisation total accruals)/lagged total assets. The accruals variables, rm size, leverage and CFO are extracted or calculated from the Diane database. Other exogenous variables have been coded manually from the companies annual reports.

3.3. Audit Quality Variables As for the global audit market, the Big Five dominate the segment of French listed companies during 1999 2001. The legislation requires that consolidated nancial statements be certied by at least two distinct statutory auditors (Commerce Code, article L823-2), who are jointly liable for the opinion delivered.10 This joint-auditorship, which applies to the entire sample of this study, is perceived to promote audit quality in two ways: (1) by offering a reciprocal check of auditors diligences; (2) by diluting managers pressure. However, the

Audit Committees and Earnings Management in France

439

effectiveness of this mechanism is based, to a large extent, on the allocation of equivalent means of control to both auditors. In practice, joint-auditorship is often characterised by a leading auditor, who imposes its quality standards, and a second auditor of a lower calibre. Investigations by Le Maux (2004) show, in a more disquieting manner, that the distribution of fees between the co-auditors and probably the budgeted audit hours is far from being equal. In this context, the quality of the external audit can be proxied by the characteristics of the dominant auditor. Auditor reputation is captured with a dummy variable (Big5), coded 1 if the rm has at least one Big Five among its auditors, and 0 if not. Auditor tenure refers to the leading auditor, based on the market structure. Piot (2001) identies four categories of auditors according to size and reputation criteria: (1) the Big Five, (2) the two largest rard, Salustro Reydel), (3) the ve other national networks (Mazars & Gue Majors, and (4) the smaller audit rms.11 When both co-auditors pertain to the same category, the tenure retained is that of the oldest, assumed to be the leader. Hence, AudTen designates the time elapsed (in years) between the nancial year end and the date when the leading auditor rst entered into function. The presence of an audit committee is captured by a dummy variable (AC). Independent audit committees are qualied according to the majority rule (Klein, 2002a), which is overall consistent with the two-thirds of independent members recommended by the Bouton Report given that most audit committees comprise three members. The dummy variable ACInd50 is then coded 1 if the rm has an audit committee composed of more than 50% of independent directors, and 0 if not.12 The concept of an independent director in France is comparable to the Anglo-Saxon views: a director is independent if he has no relationship whatsoever with the company, its group or its managers, which could compromise his freedom of judgement (Bouton, 2002, p. 9, translation). Practically, directors qualify as independent if they conform to the following criteria: . They are not part of the top management nor have the same family lineage as one of the top managers (presumption of family ties). . They hold neither executive functions in a company of the group nor in the parent company, if any. . They are not themselves, or do not represent, a signicant shareholder of the company.13 3.4. Control Variables We control for contracting motivations regarding earnings management, and for the monitoring constraint offered by the whole board of directors. Contracting motivations. Agency conicts are likely to provide incentives and opportunities for accounting manipulations. If the separation between ownership

440

C. Piot & R. Janin

and control is important, managers may be tempted to manipulate earnings to maximise their bonus, improve their performance and reputation, or protect themselves from hostile takeovers (Salamon and Smith, 1979). Thus, the weaker the managerial ownership, the more likely a strategy of income-increasing earnings management exists. The direct and indirect ownership fraction of top managers (%Managers) is used to control for the extent of shareholder manager agency conicts (Jensen and Meckling, 1976). The intensity of shareholder debtholder conicts increases with leverage. The higher the leverage ratio, the greater the risk that some debt covenants will be breached (Smith and Warner, 1979), and the higher the cost of debt nancing. As debt increases, companies may therefore resort to income-increasing accounting practices in order to present a more favourable nancial position when negotiating with lenders. A positive relation should then be observed between abnormal accruals and the debt-to-equity ratio (Watts and Zimmerman, 1986). This ratio is denoted by variable Lev in our model. Finally, large companies may engage in income-decreasing earnings management in order to mitigate political pressure (Watts and Zimmerman, 1986). This political costs hypothesis suggests a negative relation between abnormal accruals and rm size, measured with the natural logarithm of total assets (LnAssets). Monitoring constraint offered by the board of directors. The board of directors is ultimately responsible for the quality of nancial reporting. Vigilant monitoring by the board itself may keep opportunistic earnings management in check. Given that effective monitoring requires directors who are sufciently independent from those proposing the accounting policy, an independent board is a priori more inclined to cast a critical eye over accounting choices made by managers. Empirical studies tend to document a negative relation between abnormal accruals and board independence (Peasnell et al., 2000a; Klein, 2002a; Xie et al., 2003). We control for this effect with the proportion of independent directors on the board (BdInd), dened according to the independence criteria mentioned above. Table 1 provides descriptive statistics. It can be seen that 83% of nancial statements are certied by at least one Big Five, and that the average tenure of the leading auditor is 1.5 times the six-year legal mandate (9.6 years). In addition, 63% of observations mention the existence of an audit committee; but when present, the committee comprises a majority of independent directors in only 55% of the cases. Independent directors account for 42% of the board on average; the recommended minimum threshold increasing over the period from not 2 Report, 1999) to 50% for companies with diffused ownership 33% (Vie (Bouton Report, 2002). The correlation matrix in Table 2 conrms the strong negative correlation between abnormal accruals drawn from the Jones Model and the CFO variable (20.613, p , 0.0001), whereas this correlation is virtually not different from zero (20.104, p 0.096) if the CFO Model is used to estimate abnormal accruals.

Audit Committees and Earnings Management in France

Table 2. Spearman rank correlation matrix AbnAcc1 AbnAcc2 jAbnAcc1j jAbnAcc2j AbnAcc2 jAbnAcc1j jAbnAcc2j Big5 AudTen AC ACInd50 %Managers Lev LnAssets BdInd CFO

Big5

AudTen

AC

ACInd50 %Managers

Lev

LnAssets BdInd

0.585 0.014 0.256 20.004 0.006 20.144 20.028 20.008 20.025 20.061 20.072 20.613

0.085 0.381 20.013 20.020 20.190 20.069 0.060 20.164 20.077 20.062 20.104

0.385 20.061 20.008 20.066 20.021 0.072 0.025 20.082 20.078 20.021

0.032 20.157 20.146 20.186 0.313 20.052 20.165 0.168 20.028 0.586 20.012 20.248 20.055 20.407 20.227 20.250 0.175 0.087 0.178 0.127 20.130 20.255 0.192 0.073 0.488 0.350 20.471 0.490 20.072 0.282 0.075 0.159 0.438 20.169 0.104 0.285 20.096 20.057 0.026 20.004 20.111 0.115 20.220 20.040 20.043

and denote signicance at p , 0.05, 0.01 and 0.001, respectively (two-tailed).

441

442 4.

C. Piot & R. Janin Empirical Results

4.1. Univariate Tests Table 3 reports univariate tests on abnormal accruals after having partitioned the sample according to external auditors characteristics (reputation, tenure). In absolute values, Big Five-audited companies exhibit slightly inferior abnormal accruals according to the Jones Model, but similar ones according to the CFO Model. In both cases, the comparison proves insignicant, thus rejecting H1a. In signed values, although abnormal accruals are lower for Big Five-audited companies on average, insignicant comparison tests suggest that the Big Five are not more conservative than other audit rms, consistent with the absence of deep pockets incentive in France (H1b). Concerning auditor tenure, abnormal accruals are statistically equivalent in signed value, and appreciably more important in absolute value for low tenure observations when derived from the CFO Model ( p , 0.10). These ndings are then contrary to the hypothesis that audit quality decreases over time (H2).14 Table 4 groups the observations into three sub-samples: (A) no audit committee, (B) existence of an audit committee not having a majority of independent directors, and (C) existence of an audit committee that is majority independent. The rms with no audit committee display the highest levels

Table 3. Univariate tests according to the external auditor characteristics The leading auditor is a Big Five (H1) No (N 41) Yes (N 205)
b

Audit tenure of the leading auditor . annual median valuea (H2) No (N 111) 2 0.003 0.014 0.071 0.058 Yes (N 110) 0.008 0.017 0.079 0.046 t-Stat. 2 0.770 2 0.311 2 0.755 1.791

t-Stat. 0.953 0.447 0.802 0.063

Abnormal accruals in signed value AbnAcc1 0.021 0.002 AbnAcc2 0.023 0.017 Abnormal accruals in absolute value b jAbnAcc1j 0.088 0.077 jAbnAcc2j 0.055 0.055

Reported are the mean values for measures of abnormal accruals, and Student t-statistics (adjusted for heterogeneity of variances when necessary) for the comparison of independent NoYes subsamples. Non-parametric MannWhitney U-tests (untabulated) provide similar results. , and denote signicance at p , 0.10, 0.05 and 0.01, respectively (two-tailed). a Median values (in years) for the tenure of the leading auditor are 8.00 (N 71), 6.90 (N 78) and 7.54 (N 72) for 1999, 2000 and 2001, respectively. b AbnAcc1 Abnormal Accruals derived from the Jones Model; AbnAcc2 Abnormal Accruals derived from the CFO Model.

Audit Committees and Earnings Management in France

Table 4. Univariate tests according to audit committee presence and independence No audit committee (A) (N 90) Non-independent ACa (B) (N 61) Independent ACa (C) (N 83) 0.000 0.015 0.074 0.045 (A) vs. (B) t-stat. 2.635 1.959 0.773 1.582 (A) vs. (C) t-stat. 1.387 1.516 0.791 2.557 (B) vs. (C) t-stat. 2 1.486 2 0.750 2 0.090 0.762 (A) vs. (B C) t-stat. 2.145 2.066 0.810 2.731 (A B) vs. (C) t-stat. 0.231 0.676 0.485 2.053

Abnormal accruals in signed value b AbnAcc1 0.024 2 0.026 AbnAcc2 0.032 0.006 Abnormal accruals in absolute value b jAbnAcc1j 0.084 0.073 jAbnAcc2j 0.066 0.052

Reported are the mean values for measures of abnormal accruals, and Student t-statistics (adjusted for heterogeneity of variances when necessary) for the comparison of independent sub-samples. Non-parametric MannWhitney U-tests (untabulated) provide consistent results. , and denote signicance at p , 0.10, 0.05 and 0.01, respectively (two-tailed). a An audit committee is classied as independent if it is composed of more than 50% of independent directors (i.e. non-executive directors with no apparent connection with top executives or signicant shareholders). b AbnAcc1 Abnormal Accruals derived from the Jones Model; AbnAcc2 Abnormal Accruals derived from the CFO Model.

443

444

C. Piot & R. Janin

of signed abnormal accruals, whereas those with a non-independent committee display the lowest. The difference in mean between groups (A) and (B) is 5% of total assets using the Jones Model, and 2.6% using the CFO Model. Both are statistically signicant, whereas they remain generally insignicant in comparisons of (A) vs. (C), (B) vs. (C) and (A B) vs. (C). Hence, two comments arise. First, consistent with H3, the presence of an audit committee may prevent income-increasing earnings management. Second, however, a greater independence of the audit committee does not seem to be associated with more conservative accounting earnings. Nonetheless, the tests on absolute values conrm that an independent audit committee may constrain the overall magnitude of earnings management, but only for abnormal accruals derived from the CFO Model, thus making the support for H4 quite weak.

4.2. Multivariate Analysis Table 5 reports regression results of the following model, estimated to appraise the marginal effect of each predictor on earnings management (see the table notes for the denitions of variables): Abnormal Accruals a b1:Big5 b2:AudTen b3:AC b3:ACInd5015 b4:%Managers b5:Lev b6:LnAssets b7:BdInd b8:CFO16 b9:Y1999 b10:Y2000 1: Regarding the audit quality variables, the presence of a Big Five auditor does not affect abnormal accruals, whatever their form. Two other variables have a signicant effect on abnormal accruals derived from the CFO Model. On the one hand, the presence of an audit committee drives down the level of signed abnormal accruals, offering a partial corroboration of H3 ( p , 0.05 onetailed). On the other hand, contrary to H2, auditor tenure slightly mitigates abnormal accruals in absolute value. These ndings are open to three main interpretations. First of all, involving directors in the audit process through audit committees seems, at the very least, to curb income-increasing earnings management. Audit committees are associated with a greater conservatism of reported earnings (approximately 2% of total assets according to the coefcient of AC). However, untabulated regressions substituting ACInd50 for AC show no signicant impact of audit committees with an independent majority on dard earnings management,17 in contrast with US ndings (Klein, 2002a; Be et al., 2004), and questions the role of independent directors in the French model of corporate governance. Legal exposure considerations may explain this point. Specically, the strong collegiality principle reafrmed by the

Audit Committees and Earnings Management in France

445

Table 5. Time-series adjusted regressions for the pooled sample (1999 2001)a Abnormal Accruals a b1.Big5 b2.AudTen b3.AC b4.%Managers b5.Lev b6.LnAssets b7.BdInd [ b8.CFO] b9.Y1999 b10.Y2000 e
Abnormal accruals in signed value AbnAcc1 Intercept 0.035 (0.49) Big5 ?/ 20.023 (21.11) AudTen 0.000 (0.68) AC 20.018 (21.21) %Managers 20.027 (20.59) Lev 20.016 (23.80) LnAssets 0.004 (0.88) BdInd 20.033 (21.08) CFO 20.620 (26.33) Y1999 ? 0.037 (2.57) Y2000 ? 0.015 (1.53) R2 0.427 F-Stat. 14.37 N 216
a

Abnormal accruals in absolute value jAbnAcc1j jAbnAcc2j 0.192 (4.84) 0.006 (0.57) 20.001 (23.03) 20.007 (20.80) 20.019 (20.77) 0.007 (2.54) 20.009 (23.72) 20.011 (20.63) Omitted 0.005 (0.80) 0.018 (2.49) 0.188 5.09 216

AbnAcc2

20.062 (20.81) 0.078 (1.24) 20.007 (20.47) 20.005 (20.37) 20.001 (20.98) 0.000 (20.49) 20.021 (21.74) 0.007 (0.56) 20.001 (20.03) 0.052 (1.98) 20.014 (24.09) 0.005 (1.13) 0.007 (1.62) 20.001 (20.32) 20.012 (20.42) 20.030 (21.16) Omitted 0.002 (0.03) 0.010 (0.89) 0.020 (1.53) 0.039 (3.19) 0.021 (2.27) 0.106 0.077 6.14 1.92 216 216

Cluster Regression procedure from the STATAw Software, allowing for the potential time-series dependence of observations relating to the same rm. The initial pooled sample provides N 216 observations with non-missing data, representing 84 rms or clusters. The White estimator is used to compute standard errors and t-statistics. Reported items are regression coefcient and t-statistic between parentheses. , and denote signicance at p , 0.10, 0.05 and 0.01, respectively (two-tailed). Variables denition: AbnAcc1 Abnormal Accruals derived from the Jones Model, AbnAcc2 Abnormal Accruals derived from the CFO Model, Big5 1 if one of the statutory auditor (at least) is a Big Five, and 0 otherwise, AudTen time (in years) between the nancial year end and the date of rst nomination of the leading auditor, AC 1 if an audit committee is present, and 0 if not, %Managers proportion of common shares owned by top executives, Lev debt-to-equity ratio, LnAssets natural logarithm of total assets, BdInd proportion of independent directors on the board, CFO (net income before goodwill amortisation total accruals)/lagged total assets, Y1999 (Y2000) year dummies to control for xed time effects over the 19992001 period.

Bouton Report (2002, p. 6) in the sense that board sub-committees have a consultative role, not being able to substitute for board responsibilities renders the board of directors collectively responsible for nancial reporting quality, which is likely to hamper the individual monitoring incentives of independent directors. Second, Big Five auditors do not mitigate earnings management in France, contrary to what has been documented in the USA. Weaker enforcement mechanisms, resulting in a less risky litigation environment, may explain the absence of asymmetric monitoring by these large audit rms (H1b). Third, auditor tenure does not seem to threaten the quality of reported earnings; even the contrary is supported, although the effect is weak in magnitude. Regarding French audit regulation, one could then argue that the time protection of the auditors mandate is likely to promote audit quality, or that the reciprocal monitoring engendered by joint-auditorship preserves the respective scrutiny of co-auditors over time.

446

C. Piot & R. Janin

Regarding control variables, our results are in contradiction with the positive accounting hypotheses of earnings management. It seems that managers do not have strong incentives in income-increasing practices. As explained by Coffee (2005), this is consistent with the ownership concentration that characterises European countries and France in particular. The frequent presence of a dominant (or even controlling) shareholder provides a direct monitoring of managers and reduces the use of bonuses and stock options in remuneration packages. The shareholder manager agency problem then turns into a dominant minority shareholder conict, in which the use of expropriation mechanisms is more likely than earnings management strategies. Else, if large rms resort to less earnings management in absolute value (using the CFO Model), the political cost hypothesis cannot be retained given the insignicant effect of rm size on signed abnormal accruals. Most notably, the negative relation between leverage and signed abnormal accruals has contracting and legal interpretations that are consistent with the French setting of debtholders protection. First, it is in contradiction with the positive accounting view that managers resort to income-increasing earnings management to avoid covenant breaches. The lesser use of covenants in France probably eliminates the income-increasing earnings management incentive. Second, this nding is consistent with a conservative accounting attitude that responds to debtholders concerns in assessing potential loans, or in monitoring borrowers ability to pay back existing loans (Watts, 2003, p. 212). The strong emphasis placed on prudence in the French accounting framework (e.g. limited re-evaluation possibilities, extensive use of provisions for risks and charges) primarily addresses these concerns. As debt increases, managers may report more conservatively, by recording assets at their orderly liquidation value (e.g. impairments, write-offs), in order to maintain the debt contracting efciency and to minimise potential bankruptcy costs. The latter point is supported by legal provisions that managers can be held personally liable for a bankrupt companys debt on several grounds, notably imprudence or negligence, once a judicial liquidation phase has been pronounced (Fried Frank Harris Shriver & Jacobson LLP, 2005). 4.3. Simultaneity Issues The relation between audit committees and abnormal accruals, as previously estimated, may be biased by a joint-determination, or a simultaneity problem. Empirically, the setting-up of an audit committee depends on agency and governance variables, which may also intervene as explanatory factors of earnings management. Therefore, to some extent, the existence of an audit committee might be endogenous to the accounting strategy. We control for this issue using a two-stage least squares method (e.g. Klein, 2002b). The rst stage consists of predicting the existence of an audit committee with a group of exogenous instruments. Consistent with the literature, we estimate an audit committee

Audit Committees and Earnings Management in France

447

Table 6. Time-series adjusted two-stage least squares (2SLS) regressions with AC instrumented for the pooled sample (1999 2001)a Abnormal Accruals a b1.Big5 b2.AudTen b3.ACinstrumented b4.% Managers b5.Lev b6.LnAssets b7.BdInd [ b8.CFO] b9.Y1999 b10.Y 2000 e Second stage Dep. Var. AbnAcc1 20.058 (20.62) 20.008 (20.45) 0.000 (0.61) 20.016 (22.49) 20.052 (21.13) 20.016 (23.96) 0.009 (1.64) 20.033 (21.09) 20.619 (26.40) 0.031 (2.25) 0.012 (1.30) Second stage Dep. Var. AbnAcc2 20.117 (21.30) 0.001 (0.06) 20.001 (21.01) 20.011 (22.03) 20.013 (20.40) 20.014 (24.22) 0.010 (1.87) 20.011 (20.40) Omitted 0.007 (0.60) 0.020 (2.17)

First stage Probit [AC] Intercept Big5 AudTen AC %Managers Lev LnAssets BdInd CFO Y1999 Y2000 Control BdSize Dual UsList Complex R2 Chi2/f-stat. N
a

25.480 (23.54) 1.527 (3.78) 20.010 (20.59) 21.135 (21.39) 20.066 (20.87) 0.384 (3.76) 0.494 (0.64) 20.648 (23.05) 20.292 (21.72) 20.151 (20.43) 0.229 (4.87) 20.640 (21.40) 0.620 (1.52) 21.548 (21.48) 0.456 65.11 216

0.438 13.13 216

0.108 5.87 216

Cluster 2SLS Regression procedure from the STATAw Software, allowing for the potential timeseries dependence of observations relating to the same rm. The initial pooled sample provides N 216 observations with non-missing data, representing 84 rms or clusters. The White estimator is used to compute standard errors and t-statistics. Reported items are regression coefcient, and z- or t-statistic between parentheses. , and denote signicance at p , 0.10, 0.05 and 0.01, respectively (two-tailed). Variables denition: Big5 1 if one of the statutory auditor (at least) is a Big Five auditor, and 0 otherwise, AudTen time (in years) between the nancial year end and the date of rst nomination of the leading auditor, AC 1 if an audit committee is present, and 0 if not (AC is the predicted value of AC), %Managers proportion of common shares owned by top executives, Lev debtto-equity ratio, LnAssets natural logarithm of total assets, BdInd proportion of independent directors on the board, CFO (net income before goodwill amortisation total accruals)/lagged total assets, Y1999 (Y2000) year dummies to control for xed time effects over the 19992001 period. Additional instruments for AC: Control 1 if one shareholder controls more than 50% of the companies voting rights, and 0 if not, BdSize residuals from the univariate OLS regression of the number of directors on LnAssets, that is, marginal effect of board size over rm size (procedure used due to the high correlation between board and rm sizes), Dual 1 if the positions of CEO and Chairman of the board are separated (two-tiered board), and 0 if not, UsList 1 if the rm is listed on a US stock market, and 0 if not, Complex (proxy for audit risk/complexity) (gross inventories receivables)/total assets.

448

C. Piot & R. Janin

Probit model including the predictors of the accruals model and other exogenous variables recognised for their impact on the voluntary formation of an audit committee.18 The second step is the estimation of the accruals model itself, using the predicted value of AC derived from the rst stage. The results are reported in Table 6 for signed abnormal accruals as dependent variables. The rst stage regression underlines the main determinants of the existence of an audit committee found in the literature, that is, auditor reputation, rm size and board size. The second stage regressions conrm a negative relation between signed abnormal accruals and the instrumented AC variable ( p , 0.05). Although the magnitude of coefcients is slightly lower relative to their OLS equivalent, their signicance is globally stronger. This suggests that endogeneity may overestimate the real magnitude of the effect of audit committees on signed abnormal accruals, while disturbing the power of this relation. If anything, the 2SLS results are consistent with the univariate tests on that point.19 Finally, regarding abnormal accruals in absolute value, the 2SLS analyses (not reported for clarity) provide qualitatively similar regression coefcients as the one disclosed in Table 5.

5.

Conclusion

This paper investigates the effect of various audit quality dimensions on earnings management in France. We extend the earnings management literature to an auditing environment that differs from the Anglo-Saxon systems in at least three ways: (1) the legal requirement for joint-auditorship and a guaranteed mandate of six scal years for auditors; (2) the relatively lower litigation risk for audit rms as compared to Common Law countries; (3) the non-required formation of audit committees, as well as the less stringent guidelines on audit committees independence (vs. the corresponding US rules). In this context, we adopt a composite view of audit quality, based on external auditors characteristics and on the potential contribution of audit committees to, notably, the probability that irregularities are revealed. We hypothesise that brand name auditors curb earnings management as a whole, but, that there is no differentiation between auditors in terms of accounting conservatism; that auditor tenure is positively associated with earnings management; and that the presence of an audit committee, and its independence, mitigates earnings management. Empirical tests address the main listed companies from 1999 to 2001. Abnormal accruals are considered in signed value as a proxy for earnings conservatism, and in absolute value as a proxy for the overall extent of earnings management. The main ndings are that signed abnormal accruals decrease when an audit committee exists, but that the audit committees independence has no signicant effect on accruals measurements. Also, Big Five-audited companies do not differ from others in terms of absolute and signed abnormal accruals. Finally, we nd no evidence that earnings management increases

Audit Committees and Earnings Management in France

449

with auditor tenure. In fact, the opposite relation is partially supported, depending on the accruals measurement. These ndings contribute to the debates on nancial reporting quality regarding the role of audit committees and the status of external auditors in France. Audit committees act as potentially valuable audit quality devices in the sense that they control the most egregious (i.e. income-increasing) form of earnings management. However, our ndings challenge the role of independent directors in French corporate governance. Are they competent to appreciate earnings quality? Do they have real monitoring incentives? The question of competence has been addressed in very general terms by the Bouton Report (2002). The individual monitoring incentives might be hampered by the liability regime of directors, as the full board remains collectively responsible for nancial reporting quality matters. Factual evidence supporting this view has recently been reported in the Rhodia affair, when Thierry Breton the former President of Rhodias audit committee tried to dilute his responsibility among the other directors by claiming that the audit committee has no judicial existence.20 Hence, future recommendations or regulations about audit committees independence are likely to be sterile if the duties and responsibilities of outside directors are not clearly specied. Regarding external auditors, Big Five audit quality differentiation does not operate in France with respect to accounting earnings, contrary to the US ndings. Specically, the absence of asymmetric monitoring regarding abnormal accruals is consistent with the lower litigation risk incurred by audit rms in France, compared to the US Common Law environment, where investors benet from easier lawsuit opportunities (e.g. class actions, contingent fees). Thus, large audit rms, less exposed to the deep pockets incentive, would not have to deal with this threat by adopting more conservative attitudes with respect to earnings management. In summary, our ndings have implications for policy-makers regarding audit quality and legal matters in France. In line with prior studies, we nd no evidence that earnings quality decreases with auditor tenure, suggesting that the legal time protection of audit engagements is not a threat to audit quality. Further, reforming the litigation system may stimulate the monitoring exerted by large audit rms, and the vigilance of independent directors who sit on audit committees. Future research on earnings management in France may investigate two important dard et al., 2004); issues: (1) the competence of audit committee members (Be and (2) the relation (if any) between earnings management and the position in time of nancial statements with respect to the renewal date of the auditors mandate. Acknowledgements We thank the Research Alliance on the New Economy, funded by the Social Sciences and Humanities Research Council of Canada, as well as the Chair in

450

C. Piot & R. Janin

Governance and Forensic Accounting at HEC Montreal, for their nancial support. We are also grateful to workshop participants at the 28th EAA Congress in Gothenburg, the 25th Congress of the Francophone Accounting Association in ans, the Third International Conference on Governance and Forensic Orle al and the CERAG-Finance Group for their comAccounting at HEC Montre lin Gura u for ments. Special thanks to C. Richard Baker, Claude Laurin and Ca fruitful suggestions on previous drafts. Two anonymous reviewers have also contributed to the quality of this paper. Notes
See, for instance, DeFond and Jiambalvo (1993), Dechow et al. (1996), Becker et al. (1998), Francis et al. (1999), Peasnell et al. (2000a), Frankel et al. (2002), Klein (2002a), Kim et al. dard et al. (2004). (2003), Xie et al. (2003) and Be 2 The effectiveness of such regulation, for example, in Italy, has not been demonstrated. In practice, it is more likely to result in a game of musical chairs than in a strengthening of auditors independence. 3 It is not possible for a manager to dismiss an auditor during the course of his mandate, other than by a Court decision. Judges may revoke the auditor if it can be proved that the latter has committed a fault which caused damages to the audited company (this is an extremely uncommon situation). 4 The formation of audit committees is not mandatory in the UK. 5 The methodology used for estimating abnormal accruals does not apply to nancial companies. 6 Some authors (Peasnell et al., 2000b, Xie et al., 2003) posit that only the short-term component of accruals can actually be manipulated, and as such keep only these in their model. We prefer considering also the long-term component of accruals, because of the importance placed on provisions for risks and charges in the French accounting system. The indirect formula, based on balance sheet and income statement items, is preferred given that cash ow statements are not systematically supplied in the French Diane database at the time of our study. The items used in this formula are the ones prescribed by the French accounting format, replicated in Diane. English translations of these items are recommended by the authors so that the reader may be able to appreciate the equivalent in an Anglo-Saxon accounting system. 7 Long-term deferred expenses constitute amortisable entries in French nancial statements; as such, they are added to the amortisable xed assets. 8 If this coefcient is expected to be positive for industries where companies have a structural need for working capital, it should be negative for industries in which companies post a surplus in working capital. 9 The initial 102 listed companies were classied into 54 industries. Of the resulting 162 industryyears, 16 did not meet the minimum data requirement of six observations to estimate the accruals models. 10 The professional standard states that the two co-auditors agree on the audit opinion. In a case of deep divergence, it remains possible to include both opinions in the audit report (extremely rare case). 11 Early 1998, three groups of auditors are identied on the French market of listed companies: (1) rard, Salustro Reydel, the Big Six; (2) seven national networks called the Majors (Mazars & Gue Amyot Exco, Fidulor, Calan Ramolino, Constantin, BDO Gendrot); and (3) the other Local rard and Salustro Reydel) are disaudit rms. Among the Majors, two networks (Mazars & Gue tinguishable by their revenues which are fairly close to those of the Big Six (Piot, 2001, p. 492). 12 We have also considered the fact that audit committees may be composed of 100% independent members. But this alternative concerns only 11% of available observations, which severely limits the power of statistical tests.
1

Audit Committees and Earnings Management in France

13

451

not Reports of 1995 The question of the shareholder linkage is not addressed in detail in the Vie and 1999. The Bouton Report (2002, p. 10) stipulates that Beyond a threshold of 10% in capital or in voting rights, it is suitable that the board, based on the report of the nomination committee, should systematically inquire into the independent qualication, taking into account the capital structure of the company and the potential for conicts of interest. By default, all the directors who hold more than 10% of the capital or voting rights are not considered to be independent; the same applies to the representatives of other companies that pass one of these thresholds. 14 The results reported are potentially biased by the arbitrary dichotomy used. It is also possible that the relation between a high tenure and a loss in audit quality is not linear. Furthermore, considering only the tenure of the leading auditor inevitably causes a loss of information. In order to integrate the joint-auditorship, these tests were replicated with the average tenure of co-auditors, and the results were insignicant. 15 Given the strong contingency between the presence of an audit committee (AC 1) and the presence of a committee independent in majority (ACInd50 1), hypotheses H3 and H4 are tested separately. 16 This control variable is included only when abnormal accruals are derived from the Jones Model. The control for cash ows is endogenous when the CFO Model is used. 17 Regressions with the ACInd50 variable are not reported for clarity. They are qualitatively similar to those of Table 5, excepted that ACInd50 is never signicant. 18 See Piot (2004) for an empirical analysis in France, and for a review of studies on the determinants of the existence of audit committees. 19 The univariate effect of the presence of an audit committee has also been tested by regressing signed abnormal accruals on AC. The results are as follows (both regressions exhibit statistically signicant coefcients for AC at p , 0.05 one-tailed): AbnAcc1 0:033:AC 0:024, AbnAcc2 0:020:AC 0:032: As a whole, regression diagnostics (VIF) show that multicolinearity problems are not likely to affect the multiple regressions, either traditional or 2SLS analyses. 20 T. Breton is the current Minister for the Economy. His words are translated from the business press (Les Echos, No. 19446, 30 June 2005, p. 5).

References
Becker, C., DeFond, M., Jiambalvo, J. and Subramanyam, K. R. (1998) The effect of audit quality on earnings management, Contemporary Accounting Research, 15, pp. 1 24. dard, J., Chtourou, S. M. and Courteau, L. (2004) The effect of audit committee expertise, indepenBe dence, and activity on aggressive earnings management, Auditing: A Journal of Practice and Theory, 23(2), pp. 1335. Blue Ribbon Committee (1999) Report and recommendations of the Blue Ribbon Committee on improving the effectiveness of corporate audit committees, New York Stock Exchange and National Association of Securities Dealers, Stamford, CT, NY. es Rapport du groupe de Bouton, D. (2002) Pour un meilleur gouvernement des entreprises cote side par Daniel Bouton, Mouvement des Entreprises de France (MEDEF) & Associtravail pre es (AFEP-AGREF), September, Paris. ation Franc aise des Entreprises Prive Coffee, J. C. (2001) The rise of dispersed ownership: the role of law in the separation of ownership and control, Working Paper No. 182, The Center for Law and Economic Studies, Columbia University Law School. Coffee, J. C. (2005) A theory of corporate scandals: why the U.S. and Europe differ, Working Paper No. 274, The Center for Law and Economic Studies, Columbia University Law School.

452

C. Piot & R. Janin

DeAngelo, L. E. (1981) Auditor size and audit quality, Journal of Accounting and Economics, 3(2), pp. 183 199. Dechow, P. M. (1994) Accounting earnings and cash ows as measures of rm performance: the role of accounting accruals, Journal of Accounting and Economics, 18, pp. 342. Dechow, P. M., Sloan, R. G. and Sweeney, A. P. (1995) Detecting earnings management, Accounting Review, 70(2), pp. 193225. Dechow, P. M., Sloan, R. G. and Sweeney, A. P. (1996) Causes and consequences of earnings manipulation: an analysis of rms subject to enforcement actions by the SEC, Contemporary Accounting Research, 13(1), pp. 1 36. DeFond, M. L. and Jiambalvo, J. (1993) Factors related to auditorclient disagreements over incomeincreasing accounting methods, Contemporary Accounting Research, 9, pp. 415 431. Degeorge, F., Patel, J. and Zeckhauser, R. (1999) Earnings management to exceed thresholds, Journal of Business, 27, pp. 1 33. European Commission (2002) Statutory auditors independence in the UE: a set of fundamental principles, Recommendation from the EC, May, Brussels. Francis, J. R. and Krishnan, J. (1999) Accounting accruals and auditor reporting conservatism, Contemporary Accounting Research, 16, pp. 135 166. Francis, J. R., Maydew, E. L. and Sparks, H. C. (1999) The role of Big 6 auditors in the credible reporting of accruals, Auditing: A Journal of Practice and Theory, 18(2), pp. 1734. Frankel, R. M., Johnson, M. F. and Nelson, K. K. (2002) The relation between auditors fees for nonaudit services and earnings management, Accounting Review, 77(Suppl.), pp. 71 105. Fried Frank Harris Shriver & Jacobson LLP (2005) Reform of the French Bankruptcy Law Adoption of the Business Safeguard Act, Memorandum, 17 November. Available at: http:// www.ffhsj.com/cmemos/051117_reform_french_bankruptcy_law.pdf (accessed 30 January 2006). Geiger, M. A. and Raghunandan, K. (2002) Auditor tenure and audit reporting failures, Auditing: A Journal of Practice and Theory, 21(1), pp. 67 78. Ghosh, A. and Moon, D. (2005) Auditor tenure and perceptions of audit quality, Accounting Review, 80(2), pp. 585 612. Gore, P., Pope, P. F. and Singh, A. K. (2001) Non-audit services, auditor independence and earnings management, Working Paper, Lancaster University. Jensen, M. C. and Meckling, W. (1976) Theory of the rm: managerial behavior, agency costs and ownership structure, Journal of Financial Economics, 3(4), pp. 305 360. Jeter, D. C. and Shivakumar, L. (1999) Cross-sectional estimation of abnormal accruals using quarterly and annual data: effectiveness in detecting event-specic earnings management, Accounting and Business Research, 29(4), pp. 299319. Johnson, V. E., Khurana, I. K. and Reynolds, J. K. (2002) Audit-rm tenure and the quality of nancial reports, Contemporary Accounting Research, 19(4), pp. 637660. Jones, J. J. (1991) Earnings management during import relief investigations, Journal of Accounting Research, 29(2), pp. 193 228. Kim, J.-B., Chung, R. and Firth, M. (2003) Auditor conservatism, asymmetric monitoring and earnings management, Contemporary Accounting Research, 20(2), pp. 323 359. Klein, A. (2002a) Audit committee, board of director characteristics, and earnings management, Journal of Accounting and Economics, 33, pp. 375 400. Klein, A. (2002b) Economic determinants of audit committee independence, Accounting Review, 77(2), pp. 435 452. Klein, B. R. and Lefer, K. (1981) The role of market forces in assuring contractual performance, Journal of Political Economy, 89(4), pp. 615 641. Knapp, M. C. (1991) Factors that audit committee members use as surrogates for audit quality, Auditing: A Journal of Practice and Theory, 10(1), pp. 35 52. La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1997) Legal determinants of external nance, Journal of Finance, 52(3), pp. 11311150.

Audit Committees and Earnings Management in France

453

La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998) Law and nance, Journal of Political Economy, 106(6), pp. 11131155. mune ration pour quel contro le? La situation des commissaires aux Le Maux, J. (2004) Quelle re comptes, Proceedings of the Third International Conference on Governance and Forensic Accounting: Issues and Challenges, HEC Montreal, 3 and 4 June, CD-ROM. Lys, T. and Watts, R. L. (1994) Lawsuits against auditors, Journal of Accounting Research, 32(Suppl.), pp. 65 93. McMullen, D. A. (1996) Audit committee performance: an investigation of the consequences associated with audit committees, Auditing: A Journal of Practice and Theory, 15(1), pp. 87103. Menon, K. and Williams, D. D. (1994) The insurance hypothesis and market prices, Accounting Review, 69(2), pp. 327 342. Myers, J. N., Myers, L. A. and Omer, T. C. (2003) Exploring the term of the auditor client relationship and the quality of earnings: a case for mandatory auditor rotation?, Accounting Review, 78(3), pp. 779 799. Nelson, M. W., Elliott, J. A. and Tarpley, R. L. (2002) Evidence from auditors about managers, and auditors earnings management decisions, Accounting Review, 77(Suppl.), pp. 175 202. Peasnell, K. V., Pope, P. F. and Young, S. (2000a) Board monitoring and earnings management: do outside directors inuence abnormal accruals?, Working Paper, Lancaster University. Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id 249557 Peasnell, K. V., Pope, P. F. and Young, S. (2000b) Detecting earnings management using crosssectional abnormal accruals models, Accounting and Business Research, 30(4), pp. 313 326. , B. (2000) Qualite de laudit et gouvernement dentreprise: le ro le et les limites de la concurrence Pige de laudit, Comptabilite Contro le Audit, 6(2), pp. 133 151. sur le marche Piot, C. (2001) Agency costs and audit quality: evidence from France, European Accounting Review, 10(3), pp. 461 499. Piot, C. (2004) The existence and independence of audit committees in France, Accounting and Business Research, 34(3), pp. 223 246. Piot, C. (2005) Auditor reputation and model of governance: a comparison of France, Germany, and Canada, International Journal of Auditing, 9(1), pp. 2144. Salamon, G. L. and Smith, E. D. (1979) Corporate control and managerial misrepresentation of rm performance, Bell Journal of Economics, Spring, pp. 319328. Smith, E. D. and Warner, J. (1979) On nancial contracting: an analysis of bond covenants, Journal of Financial Economics, 5, pp. 117161. Stolowy, H. (2005) Nothing like the Enron affair could happen in France (!), European Accounting Review, 14(2), pp. 405 415. Teoh, S. H., Welch, L. and Wong, T. J. (1998) Earnings management and the underperformance of initial public offerings, Journal of Finance, 53, pp. 1935 1974. ry-Dubuisson, S. (2002) Exigences actionnariales et re seaux dadministrateurs: a ` quoi re pond Thie s daudit en France?, Comptabilite Contro le Audit, 8(1), la mise en place des comite pp. 129150. not, M. (1995) Le conseil dadministration des socie te s cote es, Association Franc Vie aise des Entre es (AFEP) & Conseil National du Patronat Franc prises Prive ais (CNPF), July, Paris. not, M. (1999) Rapport du comite sur le gouvernement dentreprise pre side par M. Marc Vie not, Vie es (AFEP) & Mouvement des Entreprises de France Association Franc aise des Entreprises Prive (MEDEF), July, Paris. Watts, R. L. (2003) Conservatism in accounting. Part I: explanations and implications, Accounting Horizons, 17(3), pp. 207 221. Watts, R. L. and Zimmerman, J. L. (1986) Positive Accounting Theory (Englewood Cliffs, NJ: Prentice Hall). Xie, B., Davidson, W. N., III and DaDalt, P. J. (2003) Earnings management and corporate governance: the role of the board and the audit committee, Journal of Corporate Finance, 9, pp. 295316.

454

C. Piot & R. Janin

Appendix
Table A1. Sampling procedure Steps Companies that compose the SBF 120 Index on 17 December 2002, according to the French Market Authority (www.cob.fr) Financial, insurance and investment companies 2 Real estate companies Companies (except those in the above industries) that appeared at least once on the index between 1998 and 2002, and absent from the beginning lista 2 Companies recently incorporated or introduced on the market,b or for which at least one annual report for nancial years 1999 2001 could not be consulted 2 Companies unreferenced or that do not publish consolidated nancial statements in the Diane database 2 Companies for which the industry cannot be clearly identied Number of rms remaining
a b

N 116 213 23 23 25 212 24 102

Taken from the Index InOut Movement tables published on www.cob.fr Some research suggests that companies specically manage their earnings upward just before an IPO (e.g. Teoh et al., 1998).

Table A2. Accruals models. Descriptive statistics for estimated models parameters based on a cross-sectional approach of 146 industry-years OLS regressionsa Mean Median Std. dev. Min Max % Positive

Jones Model: TAjt a1 b1.GPPEjt g1.DREVjt e jt a1 125.7 2.8 2,546.8 211,706.3 15,472.4 b1 20.081 20.068 0.099 20.594 0.266 g1 0.027 0.012 0.151 20.283 0.843 CFO Model: TAjt a2 b2.GPPEjt g2.DREVjt d2.CFOjt e jt a2 419.0 187.1 1,680.7 26,181.2 7,605.8 b2 20.019 20.007 0.073 20.369 0.237 g2 0.055 0.042 0.105 20.284 0.577 d2 20.666 20.667 0.253 21.528 0.524 b N 35 37 22 6 216

14 55

44 75 1

TA Total Accruals; GPPE Gross Property Plant and Equipment plus Long-Term Deferred Expenses; DREV the change in net sales; CFO Cash Flow from Operations. a Corresponding to the total number of industry-years regressions that could be run (i.e. 49 for 1999 and 2000, 48 for 2001), given the minimum data requirement of six observations. b N designates the number of observations included in each regression.