Analysis of the Consequences of the Disclosure of Key Audit Matters. in the Audit Report

Transcription

1 Analysis of the Consequences of the Disclosure of Key Audit Matters in the Audit Report Jean Bédard Université Laval, Chair in Corporate Governance Nathalie Gonthier-Besacier University of Grenoble, CERAG, UMR CNRS Alain Schatt HEC Lausanne (UNIL), Chair in Financial Accounting Keywords. Audit Report; Key Audit Matters; France; Market Reaction; Audit Effort. We thank the participants at the 2015 Audit & Assurance (BAFA) Conference at Edinburgh, 2015 Conférence de l Association Francophone de Comptabilité at Toulouse, 2014 ISAR Conference at Maastricht, 2014 Conférence Internationale de Gouvernance at Dijon, 2014 Workshop on Audit Quality at Venice, and workshop participants at the Bordeaux University, for their helpful comments. We acknowledge financial support from Laval University s Corporate Governance Chair. 1

2 Analysis of the Consequences of the Disclosure of Key Audit Matters in the Audit Report This article investigates some economic consequences for investors and the audit of the disclosure of justifications of assessments (JOAs) in French audit reports. These JOAs, which have been mandatory since 2003, are similar to the concept of key audit matters (KAMs) introduced recently by standards setters and regulators to enhance the informative value of audit reports. Based on French audit reports issued from 2003 to 2011, we analyze the impact of JOAs, when they were first introduced, and in subsequent years, on the market reaction (measured by abnormal returns and abnormal trading volume) and on the audit (measured by audit fees, audit report lag, and unexpected accruals). Regarding the market reaction, we find that first time JOAs are associated with reduced information asymmetry for smaller firms, which are characterized by a weaker information environment. Regarding the audit, we find that first time JOAs are negatively associated with unexpected accruals. In subsequent years, however, we do not find any significant association between new JOAs and our five measures. These results suggest that, past the first disclosures, JOAs have no significant information content. 2

3 Analysis of the Consequences of the Disclosure of Key Audit Matters in the Audit Report INTRODUCTION It is generally admitted that the audit report is not effective in communicating important information about the audit and the audit process to the users of financial statements (International Organization of Securities Commissions 2009; Gray et al. 2011; Vanstraelen et al. 2012). In this standardized report, which contains boilerplate language and uses a pass or fail approach, auditors usually conclude with an unqualified opinion. Consequently, the report has come to be interpreted as a single symbol that is no longer read (The Commission on Auditor's Responsibilities 1978) and has little informative value (Church et al. 2008; Mock et al. 2013). In response to stakeholders demands for more informative audit reporting, the International Auditing and Assurance Standards Board (IAASB) has issued ISA 701- Communicating Key Audit Matters in the Independent Auditor s Report (IAASB 2015) and the Public Company Accounting Oversight Board (PCAOB) has proposed for public comment (PCAOB 2013) a new auditing standard requiring auditors to communicate Critical Audit Matters (CAMs). These are matters that, in the auditor s professional judgment, were of most significance in the audit of the financial statements of the current period. The purpose of communicating KAMs is to enhance the communicative value of the audit report (IAASB 2015). 3

4 Since 2003, auditors of French public companies have already disclosed such matters in the audit report. Indeed, French auditors must provide commentaries called Justifications of Assessments (JOAs). As indicated by the French accounting profession, JOAs are a very similar mechanism (CNCC 2013) in that they must enable the user of the report to obtain a better understanding of the reasons behind the statutory auditors opinion on the financial statements (Haut Conseil des Commissaires aux Comptes 2006). In this study, we examine the consequences of JOAs (hereafter also called French KAMs ) for investors and for the audit. For the investors we examine whether, as expected by regulators, KAMs are informative. Previous studies on the audit report show that going concern opinions have informative content when they are not expected, or provide new information to the market (Taffler et al. 2004; Menon and Williams 2010). Moreover, a recent study by Czerney et al. (2014a) finds that explanatory language added to standard audit reports (e.g. adoption of new accounting standards, division of auditor responsibility, previous restatements, and emphasis of a matter paragraphs) are informative to investors, suggesting that the audit report may also have an attention directing effect. To date, however, no academic study has yet dealt with the specific French report 1. For the audit, we examine the potential consequences of KAMs on audit costs, delay in issuing the audit report, and audit quality. Regarding the audit costs, KAMs may result in additional audit effort because of increased accountability (Carcello and Li 2013), in increased litigation risk and/or reputational risk associated to additional disclosures (DeFond and Zhang 2014), and in the auditor s additional time in preparing and reviewing the report. As in previous studies we use audit fees to proxy for audit intensity and effort (e.g., Gul 2006; 1 A study conducted by consultants (Footprint Consultants 2011) for the French accounting professional organization, examines users perception through interviews of 34 persons and finds that French KAMs are generally well perceived by users. However, this survey provides limited information about the economic consequences of the new audit report for investors and the audit. 4

5 Srinidhi and Gul 2007; Rice and Weber 2012). Since audit report lag is significantly longer on engagements where more total audit effort is needed (Knechel and Payne 2001), one may expect that KAMs result in an increase in the audit report delay, if KAMs require additional time and effort, particularly given that part of this additional work is to finalize the auditor s report that occurs near the end of the audit. Finally, KAMs could potentially increase auditor effort and skepticism, induced from greater accountability, which may increase the quality of KAMs related items, thereby producing better quality financial statements (PCAOB 2011b). 2 While standard setters and regulators have high expectations regarding KAMs, it is however possible that KAMs will not have the expected consequences on the investors and the audit. In fact, the information provided in KAMs could already be known or expected by the financial market, for various reasons (e.g., nature of KAMs, lack of auditor s incentives). Also, the use of boilerplate or ambiguous sentences, and technical language, may reduce the information content of KAMs. These ideas are in line with the results of Gutierrez et al. (2015), who report no significant investor reaction to the new audit report introduced in 2012 in the UK. Similarly, less relevant KAMs would not have the same accountability effect on the auditors, and would reduce the expected effect of KAMs on the audit. To test the consequences of French KAMs, we adopt an empirical strategy that exploits the change in audit report using a pre/post French KAMs requirement design as well as the changes in the KAMs reported by auditors of French companies in the years following the introduction of KAMs in August We use a sample varying from 1,750 to 2,407 nonfinancial firm-year observations for the period to examine the effect of KAMs for 2 This effect could be especially present in the French context of weak legal protection where auditors may act as a substitute to curb earnings management (Choi and Wong 2007; Han et al. 2011). In such a context, KAMs give auditors an effective additional lever to influence the quality of financial statements. 5

6 investors and the audit. Additionally, to control for firms time-invariant unobserved variables, we use regression models that include firm fixed effects for all of our tests. Regarding the impact of KAMs on investors, we find no significant effects on both abnormal market returns and abnormal trading volume, either for the first time French KAMs became mandatory (2003) or for subsequent years ( ). In supplementary analyses, we examined whether the information content of KAMs varies depending on the firm s information environment. Using the firms market capitalization as a measure of the strength of that information environment, we find that for smaller firms, KAMs are associated with reduced information asymmetry (i.e. less abnormal trading volume) the first time KAMs were mandated. For larger firms, first time KAMs have no significant effect. This result does not hold in subsequent years for new KAMs. Overall, these results suggest that French KAMs do not have information content except in the first year of application where, in weaker information environments, they are associated with lower information asymmetry. Regarding the effect of KAMs on the audit, we do not find a significant effect on the level of audit fees for French KAMs that are disclosed for the very first time nor for new KAMs in subsequent years. We do, however, find that JOAs are negatively associated with unexpected accruals (measured by the level of unexpected accruals) the first year KAMs were disclosed. We also show a significant increase of 19 percent in the audit report lag during the first year KAMs were disclosed. In subsequent years, audit quality and audit report lag are not associated with the disclosure of new KAMs. Such results are consistent with the idea that in the first year of application of the law, auditors had to adapt and took more time in selecting, drafting and discussing KAMs. 3 3 It is also possible that the effect is caused by other requirements of the Law on Financial Security adopted in France in August 2003, which not only included requirements for KAMs, 6

7 Overall, our results highlight that French KAMs have little effect on the information content of the audit reports, nor on the audit cost, efficiency and quality. Therefore, our paper mainly contributes to the audit report literature and, in particular, the literature on KAMs that reflects a significant change in the format and content of the audit report. It complements previous experimental KAMs studies (e.g. Christensen et al., 2014) by providing archival evidence on the effect of KAMs and shows that it does not lead to a greater informative value of the audit report in France. Further, it does not significantly impact audit effort (proxied by audit fees, audit delay and audit quality). Thus, our results show that this new audit report has merely a symbolic value. Our paper also contributes to the audit market regulation literature. DeFond and Zhang (2014) discuss the economic consequences of SOX in the US. They find that some aspects (for instance, the prohibition of some non-audit services) of this new regulation does not necessarily impact audit quality. Our paper extends this stream of research by examining the effect of the regulatory change regarding the content of the audit report. We show that the disclosure of KAMs in France is not associated with significant economic consequences for investors and for the audit. Finally, our paper provides empirical evidence that could be useful for standards setters and regulators in developing the audit report s standards and evaluating its effect on the information gap and auditors behaviors. The remainder of this paper is organized as follows. In a second section, we present the institutional background relating with the disclosure of KAMs. In the third section, we discuss some economic consequences associated to the introduction of such disclosures in France and develop our hypotheses. The fourth section is devoted to our methodology. The fifth section presents and discusses our results. The final section concludes this paper. 7

8 INSTITUTIONAL BACKGROUND Following the recent financial crisis, many economic and political actors around the world have questioned the role of auditors (e.g.vanstraelen et al. 2012; Simnett and Huggins 2014). In particular, they have criticized the fact that the audit report is too standardized and does not respond to users information needs relating to the audited entity, its financial statements and the audit process (IASB 2013). In response to these pressures and, after a long consultation process, in January 2015 the IAASB issued a new standard requiring that auditors of financial statements of listed entities communicate KAMs in their report (ISA 701). This standard is effective for audits of financial statements for periods ending on or after December 15, 2016 (IAASB 2015). The purpose of KAMs is to increase the communication value of the auditor s report to users. KAMs are those matters that, in the auditor s professional judgment, were of most significance in the audit of the financial statements of the current period, including areas identified as significant risks or involving significant auditor judgment, areas in which the auditor encountered significant difficulty during the audit, and circumstances that required significant modification of the auditor s planned approach to the audit. In the U.S., the PCAOB has issued a similar proposal (PCAOB 2013) that has not yet resulted in a standard. 4 The French justification of Assessments KAMs are not a totally new concept. Indeed, the disclosure of audit matters in the audit report has been mandatory in France since The French Financial Security Act, which was adopted as a response to SOX, requires auditors to justify their assessments. More specifically, the French Code of Commerce states that in its report the auditor must justify the 4 In the US, CFOs and auditors are reluctant to adopt this new report because of the potential increase of litigation risk (e.g. Katz 2014). 8

9 findings made during the audit (Article L ). The justifications of assessments are supposed to help users of the audit report to gain a better understanding of the reasons behind the auditors opinion. The audit standard NEP 705: Justification of assessments (Haut Conseil des Commissaires aux Comptes 2006) clarifies the nature of French KAMs and provides guidance for their application. Thus, the French KAMs are an explanation of the auditor s assessments and substantiation of the audit opinion. They generally relate to important matters for the understanding of the financial statements. These important matters include: options used in the selection of accounting policies or in their implementation, critical accounting estimates (including those involving significant professional judgment in their assessment), and the overall presentation of the financial statements. NEP 705 also emphasizes that KAMs should not disclose information that is the responsibility of the entity's management and cannot replace the need to make a qualified opinion or a disclaimer of opinion. Regarding the presentation of the KAMs in the audit report, NEP 705 specifies that the formulation of the KAMs must be clear and include for each assessment: an identification of the matter and, if possible, reference to the notes, a summary of the procedures performed by the auditor to support his/her assessment, and a conclusion, expressed positively. Further, the formulation of the KAMs should not be perceived as providing specific assurance on items that are referred to. Finally, the KAM must appear in a separate section of the report following the auditor s opinion. Appendix 1 presents a part of the audit report that discusses French KAMs in the 2010 annual report of the company France Telecom. THEORETICAL BACKGROUND AND HYPOTHESES DEVELOPMENT The expected consequences of the disclosures of KAMs are considered along with their dual nature. On the one hand, KAMs may be viewed as supplementary information in the audit 9

10 report. This conception leads to an appreciation of their benefits through their informative content, which can mainly be measured by empirical studies based on market reactions. On the other hand, the disclosure of KAMs can be viewed as a new procedure imposed to auditors. Consequences of French KAMs for investors Investors may react to a KAM if it reduces the information asymmetry between auditors and investors. Indeed, previous research on the audit report shows that the market reacts when explanatory language is added in the unqualified audit report and provides new information to investors. For example, Czerney et al. (2014b) find significantly higher abnormal trading volumes for unqualified audit reports with explanatory language on prior restatements, other consistency matters, or on the emphasis of a matter. Similarly, Menon and Williams (2010) find that the market reacts to going concern opinions only when they provide new information. Recent experimental evidence suggests that KAMs may have an effect on users. Christensen et al. (2014) find that nonprofessional investors react to the disclosure of KAMs related to the fair value in the financial statements. Thus, investors who receive a report with KAMs are more likely to change their investment decision than investors who receive a standard audit report, or investors who receive the same information but in the financial statements disclosures. This result underlies both an information effect and a source credibility effect. In addition, in their eye-track study, Sirois et al. (2014) show that KAMs may have an attention directing effect, increasing users probability of attending the KAMrelated financial statement disclosures and drawing users attention to these disclosures. These studies suggest that KAMs may have an information effect, a source credibility, or an attention directing effect. 10

11 Investors, however, may not react to KAMs for various reasons. First, investors may have difficulty understanding KAMs because of the use of subjective concepts (e.g., Arnedo et al. 2008), the use of technical language (e.g. Asare and Wright 2012), the use of boilerplate KAMs, and cognitive overload (Hirshleifer and Teoh 2003). Second, KAMs could be expected by the financial market. For instance, when some assets are material, it is likely that the auditor discloses a KAM related to these assets, especially when managers have to use a model to estimate their fair value (i.e. the discounted cash flows model for the impairment test of the goodwill). Finally, financial statements are just one small subset of information used by investors, and they generally do not provide a substantial amount of new information to investors (e.g. Ball 2013). It is then possible that KAMs, related to the audit of the financial statements, will not provide new information to users. This is even more likely when auditors are prohibited to disclose information that is the responsibility of the entity's management. The previous arguments lead to the following hypothesis, stated in alternate form: H1: Ceteris paribus, there is an association between KAMs disclosed in the audit report and the investor s reaction to the filing of the annual report. Consequences of KAMs on the Audit While the main objective of KAMs is to reduce the information gap (IAASB 2012), and help users of the audit report to have a better understanding of the reasons behind the auditors opinion, the disclosure of KAMs may also affect the audit. We examine the potential consequences of KAMs on audit fees, delay in issuing the audit report, and audit quality. 11

12 Effect of KAMs on audit fees KAMs should require additional audit effort 5 by the most senior members of the engagement team for the determination, preparation, documentation, and reviewing of the KAM section of the audit report (IAASB 2012; PCAOB 2013). In addition, even if the standard does not require additional audit procedures, auditors may feel more accountable for matters to be reported (e.g., gathering more and better evidence to audit these items). For example, Carcello and Li (2013) find an increase in audit fees following a requirement that partners sign the audit report, a finding that they attribute to increased accountability. Such accountability may also prevail in France, where partners also have to sign the audit report. Finally, since these key matters are more visible to users, their disclosure may increase the potential costs borne by auditors if misstatements or accounting fraud are discovered in the future (e.g., Brasel et al. 2014; Kachelmeier et al. 2014; Gimbar et al. 2014). These costs are related to the increase of reputational risk and litigation risk, which are two key drivers of audit effort (DeFond and Zhang 2014). On the other hand, some stakeholders argue that the disclosure of KAMs is a formal exercise that does not really affect the overall effort made by the auditor. Many respondents to the IAASB invitation to comment believe this because KAMs are related to information already addressed in the audit file and communicated to the audit committee. Thus, the main additional audit effort required is limited to determining, preparing the language for communication, and documenting the KAMs (PCAOB 2013). Consequently, the auditor's incremental efforts and cost to report KAMs would be minimal. Moreover, auditors may not charge these additional reporting hours because of client reluctance to pay for such disclosure. 5 As in previous studies, we use audit fees to proxy for audit intensity and effort (e.g., Gul 2006; Srinidhi and Gul 2007; Rice and Weber 2012). 12

13 Finally, in France, the litigation risk being quite low, French KAMs are not expected to increase auditors fees in response to increased liability. The previous arguments allow the formulation of the following hypothesis, stated in alternate form: H2a: Ceteris paribus, KAMs disclosed in the audit report are associated with an increase in audit fees. Effect of KAMs on the audit report lag As indicated in the previous section, KAMs require effort to determine, prepare the language for communication, and document, KAMs. They may also result in additional audit effort because of accountability and liability effects. Audit report lag being significantly longer for engagements where more audit effort is needed (Knechel and Payne 2001), the inclusion of KAMs may result in a longer audit delay. Indeed, the PCAOB (2011a) indicates that the drafting of KAMs, including the discussion among management, the auditor, and the audit committee regarding their nature and extent, could increase the time to complete the audit and issue the audit report. Given that the information was already communicated to management and the audit committee, however, and that the disclosures could be boilerplate (Mock et al. 2013; Footprint Consultants 2011), French KAMs may not result in much discussion and debate among the auditor, management, and the audit committee. Furthermore, even if there is an increased effort, given that the audit matters are identified before the audit report date, the auditor may execute most of the additional audit effort in advance, thereby reducing the effect on the audit report lag (Knechel et al. 2012). 13

14 The previous arguments allow us to develop the following hypothesis, stated in alternate form: H2b: Ceteris paribus, KAMs disclosed in the audit report are associated with an increase in audit report lag. Effect of KAMs on audit quality Finally, KAMs may also have an impact on audit quality. As indicated before, due to greater accountability, KAM may induce auditors to gather more and better audit evidence regarding the KAMs related items, thereby increasing audit quality (PCAOB, 2013). In addition, such accountability could result in an increase in the auditor s professional skepticism, another driver of audit quality (IAASB 2012). Indeed, as suggested by the Council of Institutional Investors in its response to the IAASB exposure draft, the disclosure of KAMs could increase quality competition among audit firms, particularly in the area of professional skepticism". 6 Ultimately, KAMs may result in increased attention by management and those charged with governance to KAMs related disclosures in the financial statements, thereby improving the quality of financial reporting (IAASB 2012). In the French specific context, KAMs may have a greater impact on audit quality because of the weaker investors legal protection (e.g., Leuz 2010). In such as context, auditors may act as a substitute to curb earnings management (e.g., Choi and Wong 2007; Han et al. 2011). 7 As argued in the previous section, however, the accountability and legal liability effects of KAMs may not be sufficient to induce auditors to gather more and better audit evidence and increase auditor skepticism. Furthermore, because KAMs could be boilerplate, In addition, starting in 2005, French public companies had to apply IFRS, which allow managers more discretion than the French accounting standards. Managers have then more opportunities for error and discretion to manage earnings, thereby increasing the auditors role. In our analysis, we control for IFRS and in supplementary analysis, we further test the effect of KAMs when IFRS were implemented. 14

15 they may not result in increased attention by management and those charged with governance to KAMs related disclosures. Accordingly, KAMs could have no effect on audit quality. These arguments lead us to specify our last hypothesis, stated in alternate form: H2c: Ceteris paribus, KAMs disclosed in the audit report are associated with an increase in audit quality. SAMPLE SELECTION AND EMPIRICAL MODELS Sample To test our hypotheses, we selected a large sample of French public companies across the years This period includes the first year of application of KAMs disclosures in France (2003) and allows pre/post comparisons. To create our sample, we start with the lists of companies composing the SBF 250 index during the period This index is based on the 250 largest market capitalizations in Euronext Paris at the end of each year. We exclude non-french companies because their auditors are not subject to the KAMs requirement, and companies in the financial services sector because their financial statements are not comparable to those of non-financial companies. We require the remaining companies to have their annual financial statements available on the French market authority (Autorité des Marché Financiers) website, or on their own website. In addition, we require firm-year observations to have the necessary financial data on the Worldscope database. We also require valid returns data from the Datastream database, non-missing audit fee data from the annual report (for the audit fee model), as well as the filing date on the French market authority website (for the audit delay model). This procedure leaves us with 1,767 firm-year observations for the market reaction model based on returns, 1,967 firm-observations for the market reaction model based on trading volume, 2,140 firm-year observations for the audit quality model, 2,407 firm-year observations for the audit delay model, and 1,750 firm-year 15

16 observations for the audit cost model. Table 1 shows the distribution of the sample by year, and the last column also highlights the average number of KAMs. In 2003, the French firms in our sample disclosed on average 1.56 KAM and this number increases to 2.33 in The main change occurred in 2005, when IFRS became mandatory. [TABLE 1] Models specification We use five different models to test our hypotheses on the consequences of the KAMs disclosed in France. Two models assess the investor reaction to KAMs (proxied by market reaction using abnormal returns and abnormal trading volume) and three models assess the impact of KAMs on audits (measured with audit fees, audit report lag, and unexpected accruals). Design Since French KAMs were required for all auditors reports on French companies with a year-end after August 1, 2003, all companies with a 2003 year-end prior to August 2, 2003 had their first KAMs communicated in 2004 instead of In our models, these companies act as control group for the companies disclosing their first KAMs in In 2004 it is the contrary, these companies are the treatment group and the others form the control group. These two implementation years allow us to control for other contemporary factors that occurred the first time French KAMs were disclosed and in particular the other measures included in the Law on Financial Security as well as the greater public attention to the audit induced by the adoption of that law in Because of the use of a firm-fixed effect model along with these two implementation dates, the coefficient on KAM_FIRST is the difference-in-difference between the first KAM year and the other years for the treatment and control groups. 16

17 Besides, in each model we include firm fixed effects to control for all firms timeinvariant unobserved variables. As a consequence, we do not include variables that do not change over time, such as industry, because the effects of these variables are included in the firm fixed effects. French KAMs To test the effect of KAMs, we use two main variables. The first variable allows us to test the effect of the introduction of the KAMs requirement using a pre/post KAMs research design. KAM_FIRST is a dummy variable that equals one when the company's auditor report was subjected to the KAM requirement for the first time, and zero otherwise. We expect that the consequences of KAMs will be stronger when KAMs are disclosed for the first time, because that information was unknown in the previous year. Second, to examine the effect of new matters mentioned after the first year, we use NEW_KAM, the percentage of new matters in the KAM section of the audit report of the current year (i.e. number of new matters/total number of matters). If new matters have information content, then the effect should be stronger when a greater proportion of new matters are disclosed in a given year. Appendix 2 summarizes each variable s definition. Investors reaction We measure investors reaction to KAMs using two different models dealing with abnormal returns and abnormal trading volumes. Using abnormal trading volumes allows us to capture changes in expectations of individual investors in addition to the changes in expectations of the market measured by the abnormal returns (Causholli et al. 2010; Bamber et al. 2011). The first model examines the reaction of the financial market by measuring changes in the market value of the French companies when KAMs are disclosed. More precisely, we 17

18 conduct an event study to measure the cumulative abnormal returns (CAR) around the announcement date of the audit report. 9 CAR = β 0 + β 1 KAM_FIRST + β 2 NEW_KAM + β 3 LOG_MARKET_CAP + β 4 MB + β 5 LEVERAGE + β 6 VAR_NI_TA + β 7 ROA + β 8 IFRS + β 9 FILING_DELAY + year fixed effects + firm fixed effects + ε (1) We use the absolute value of cumulative abnormal returns ( CAR ) over two days (the filing day and the following day). If the KAMs disclosures reduce the information asymmetry and are useful for investors, then the coefficient on KAM_FIRST and NEW_KAM will be positive. Since other variables may affect the market reaction, we control for firm size (LOG_MARKET_CAP), market-to-book ratio (MB), leverage (LEVERAGE), variation in net income (VAR_NI_TA), return on assets (ROA), as well as the use of IFRS (IFRS) and the delay in filing the annual report (FILING_DELAY). The second model examines the reaction of the financial market by measuring changes in trading volume. This second analysis is important because investors can trade more stocks, without any effect on the stock price (abnormal return) if the new information disclosed increases the disagreement between investors (e.g. Bamber et al. 2011). Thus, we conduct an event study to measure the abnormal trading volume (AVOL) around the announcement date of the audit report. This abnormal volume is calculated as the difference between announcement period trading minus the expected trading. The calculation of the latter is explained in detail in Appendix 2. In this model, we also control for other factors that may impact the abnormal trading volume. 9 We measure the announcement date of the audit report as the date that the annual report is filed with the French Market Authority. 18

19 AVOL= δ 0 + δ 1 KAM_FIRST + δ 2 NEW_KAM + δ 3 LOG_MARKET_CAP + δ 4 M_B + δ 5 LEVERAGE + δ 6 VAR_NI_TA + δ 7 ROA + δ 8 IFRS +δ 9 FILING_DELAY + δ 10 RETURN + year fixed effects + firm fixed effects + ε (2) Audit fees We measure audit fees as the log of the sum of the audit fees paid to the pair of auditors (LOGAUDFEES), because joint audit is mandatory in France (Gonthier-Besacier and Schatt 2007; Francis et al. 2009; André et al. 2015). We manually collect the audit fees from the companies annual report from 2003 to We start in 2003 because it is the first year of mandatory disclosure of audit fees. We use the following regression to test the association between KAMs and audit fees. LOGAUDFEES = α 0 + α 1 NEW_KAM + α 2 LOG_ASSETS + α 3 LEVERAGE + α 4 ROA + α 5 LOSS + α 6 INV_REC + α 7 IFRS + α 8 TWO_ΒIG4 + α 9 ONE_ΒIG4 + α 10 BUSY_SEASON + firm fixed effects + year fixed effects + ε (3) Given that the publication of the audit fees data started the same year as the KAMs disclosure requirement, we only examine the effect of the KAMs using the new matters mentioned in a given year (NEW_KAM). If the disclosure of KAMs increases audit fees, then the coefficient of the variable will be positive. Following previous audit fee studies (see the meta analysis of Hay et al. 2006), we also control for firm size (LOG_ASSETS), leverage (LEVERAGE), profitability (ROA and LOSS), receivables and inventory intensity (INV_REC), use of IFRS (IFRS), the year end (BUSY_SEASON), and whether the auditors dyad is composed of two Big 4 auditors (TWO_ΒIG4 ) or one Big 4 and a non-big 4 (ONE_BIG4). The two latter variables allow us to assess the differential amounts of fees charged by different pairs of auditors in France (Gonthier-Besacier and Schatt 2007; André et al. 2015). 19

20 Audit report lag We measure the audit report lag (LOGAUDLAG) as the number of calendar days between the date of the audit report and the company's year-end. We use the following model to test the effect of KAMs: LOGAUDLAG = Ω 0 + Ω 1 KAM_FIRST + Ω 2 NEW_KAM + Ω 3 LOG_ASSETS + Ω 4 LEVERAGE + Ω 5 ROA + Ω 6 LOSS + Ω 7 TWO_ΩIG4 + Ω 8 ONE_ΩIG4 + Ω 9 BUSY_SEASON + year fixed effects + firm fixed effects + ε (4) We expect positive coefficients on KAM_FIRST and NEW_KAM if the disclosure of KAMs leads to greater audit report lag. In addition, we control for firm size (LOG_ASSETS), audit risk (LEVERAGE, ROA, LOSS), type of joint audit (TWO_ΒIG4, ONE_ΒIG4), and year end (BUSY_SEASON). Audit quality We measure audit quality with the absolute value of unexpected accruals (DeFond and Zhang 2014) 10, and estimate the following model to examine the relation between audit quality and KAMs. DAC = γ 0 + γ 1 KAM_FIRST + γ 2 NEW_KAM + γ 3 LOG_ASSETS + γ 4 LEVERAGE + γ 5 ROA + γ 6 LOSS + γ 7 IFRS + γ 8 TWOΒIG4 + γ 9 ONEΒIG4 + year fixed effects + firm fixed effects + ε (5) Our dependent variable DAC is the absolute value of unexpected accruals generated by the modified Jones (1991) approach (Dechow et al. 1995). If the KAMs have an effect on audit quality, KAM_FIRST and NEW_KAM will be negatively associated with DAC. 10 DeFond and Zhang (2015) note that abnormal accruals are an imperfect measure of audit quality. We use this measure because there is no other way to measure audit quality in France. Indeed, there are no restatements and no going-concern opinions in France. 20

21 Following prior research (Becker et al. 1998; Frankel et al. 2002; Ashbaugh et al. 2003), we also control for firm size (LOG_ASSETS), leverage (LEVERAGE), profitability (ROA and LOSS), use of IFRS (IFRS), and whether the auditors dyad is composed of two Big 4 auditors (TWO_ΒIG4 ) or one Big 4 and a non-big 4 (ONE_BIG4). These two last variables allow to assess the quality of auditors in France, where companies are forced to use two auditors (Francis et al. 2009). RESULTS Descriptive statistics Table 2 presents descriptive statistics for the variables used in multivariate tests 11. The average absolute value of the cumulative abnormal returns over two days ( CAR ) is 3.2% and the average abnormal trading volume (AVOL) is equal to 2.1%. The average audit fees are equal to 2.8 million. This number is lower than fees reported in previous studies on French companies (Gonthier-Besacier and Schatt 2007; Broye 2009; Ben Ali and Lesage 2013; André et al. 2015). The average audit delay is 99 days, a number slightly larger than the 85.1 days found by Piot (2008) on the French SBF-120 index market. All of those differences are explained by the fact that our sample includes more companies and smaller ones (with lower audit fees and less pressure on audit delay). The average unexpected accruals are equal to 9.1%, which is consistent with Piot (2008) who finds an accrual percentage of 8.1% with a similar model. [TABLE 2] 11 All continuous variables are winsorized at the 1st and 99th percentiles. 21

22 Table 2 shows an average percentage of new KAMs of 16.2 percent per year and a median of zero percent. Accordingly, a majority of the KAMS are repeated from year to year, suggesting that important matters hold over many years. Consequences of KAMs for investors In Table 3, the results of our regressions show that the disclosure of KAMs does not impact the market reaction. Our two variables FIRST_KAM (first year of KAMs disclosure) and NEW_KAM (relative increase of the number of KAM from one year to another) are not significantly associated with abnormal returns (Model 1) or with abnormal trading volumes (Model 2). These two models account for 32.1% and 18.2% of the total variance respectively. It is interesting to note, however, that in controlling for all time-invariant unobserved variables with the firm fixed effects and the year fixed effects, none of the control variables are statistically significant, except IFRS and FILING_DELAY in Model 1, which are marginally significant. These results indicate that the market reaction is lower for French public companies using IFRS, and larger for companies increasing their filing delay. 12 In summary, our results suggest that, in general, French KAMs do not have a significant effect on investors. [TABLE 3] Consequences on the audit Table 4 presents the regression results for audit fees, audit report lag, and abnormal accruals. The three models are significant and have R 2 value of 0.98, 0.75, and 0.26 respectively. These R 2 are higher than usual because of the use of firm fixed effects. [TABLE 4] 12 The negative coefficient for the IFRS result is consistent with the idea that managers have greater discretion under IFRS, which results in less reliable numbers, leading to a lower market reaction. 22

23 Audit Fees The first column of Table 4, reports the results for the regression, examining the effect of New_KAMs on the audit fees (Model 3). As explained before, FIRST_KAM could not be estimated for audit fees because the fees are not available before The coefficient on NEW_KAM is not significant, suggesting that the inclusion of new KAMs in the audit report is not associated with an increase in audit fees. Results for the control variables show that increase in size is associated with higher audit fees. Further, audit fees decrease when profitability (ROA) is lower and when the accounting year-end is not December 31 (BUSY_SEASON). Audit Report Lag Table 4 also presents the results of the regression, examining the association between audit report lag (Model 4) and the disclosure of KAMs. Both the coefficients on FIRST_KAM and NEW_KAM are not significant. These results suggest that, contrary to regulators concerns, the additional auditor effort to prepare, discuss, and review the KAMs in France is not associated with an increased audit report lag both when KAMs are identified and drafted for the first time and in subsequent years when new KAMs are disclosed. With regard to the control variables, in addition to the firm and year fixed effects, only LEVERAGE is significant, with a positive association. Abnormal Accruals Finally, Table 4 reports the results for the regression on abnormal accruals (Model 5). The significant negative coefficient on FIRST_KAM suggests that abnormal accruals are lower when a company discloses KAMs for the first time. However, the non-significant coefficient on NEW_KAM indicates that such an effect does not hold over time. Further, we show that 23

24 more debt (LEVERAGE) and the disclosure of a negative net income (LOSS) are associated with higher unexpected accruals. Summary of the findings To sum-up, our analyses suggest that French KAMs generally do not have significant consequences on the audit. The only exception concerns audit quality in the first year KAMs were disclosed, where it increased. This result is consistent with the greater accountability explanation. The lack of effect in the following years for new KAMs, however, suggests that this effect was temporary and limited to the first year. It is also possible that the effect is caused by auditors reactions to the Law on Financial Security adopted in August 2003, which not only included requirements for KAMs, but also for internal control, and which placed the profession under the oversight of an independent body. 13 These results lead us to further examine the effect of the first disclosure of KAMs, and to provide supplementary analyses. Supplementary Analyses In this section, we analyze more precisely the specific effect of the first disclosure of KAMs, the influence of the nature of the KAMs, the influence of the information environment and the influence of IFRS adoption. First Disclosure of KAMs In the previous section, the models are estimated over the complete period ( ). Accordingly, the effects of the control variables are assumed to be the same over the period (pre- and post-kams) and more importantly, the effects of the first KAMs are assumed to be the same no matter whether they were implemented in 2003 or Compared to SOX 404 which requires management and auditors opinion on the effectiveness of the internal control over financial reporting, the French requirements are more lenient. The Chairman of the Board of Directors presents, in a report on corporate governance, a description the internal control procedures related to all types of risks, not only those related to the financial reporting. Based on their audit of the financial statements, the auditors present their observations on those internal control procedures related to the preparation and processing of accounting and financial information. 24

25 We re-estimate models (1), (2), (4) and (5) 14 using the same control variables and KAM_FIRST for a new First KAMs in 2003 model and KAM_FIRST, NEW_KAM for a new First KAMs in 2004 model. In each model, we use a subset of companies for which we have the data for the KAM year and the previous year (i.e., the year 2002 for 2003 KAMs, and the year 2003 for 2004 KAMs). We use a firm-fixed effect model to obtain the difference between the two years. 15 Accordingly, the coefficient on KAM_FIRST is the difference-indifference between the first year of the KAMs disclosure and the previous year for both groups: the treatment group (disclosure of KAMs) and the control group (no disclosure of KAMs). For instance, in the First KAM in 2004 model, NEW_KAM controls for changes in KAMs for firms that had disclosed their first KAMs in Table 5 presents the results of the regression for each model. For the consequences on investors, we do not find any significant association between CAR and first KAMs in both years (Panel A). There is a significant negative association between abnormal volume and both the disclosure of first KAMs and new KAMs in 2004 (Panel B). Thus, in this second year, new KAMs for first time KAMs disclosers, and new KAMs for disclosers that already disclosed KAMs the previous year, appear to help increase the consensus among investors even if they do not affect the market return. Combined with the lack of significant results for the first year (2003) these results could be an indication of a learning effect (i.e. in 2003 investors did not understand KAMs, or they were not paying attention to them until 2004). Similarly, as in the main analysis over the full period, for the post first KAMs periods, we do not find any significant effect for new KAMs, suggesting that this effect is temporary. [TABLE 5] 14 We do not use the audit fees model (model 3) because we do not have the fees for the year Since we have two observations per firms, coefficients for the fixed effects model are identical to OLS using difference scores for all the time-varying variables for the two years. 25

26 For the consequences on the audit (Panels C and D), the coefficient on FIRST_KAM is only significant for the audit report lag in Thus, compared to companies not yet subjected to KAMs, the change in audit report lag between 2002 and 2003, is 19 percent longer for first time KAM s audit reports. These results are consistent with the idea that in the first year of application of the law, auditors had to adapt to the new law and took more time before issuing the audit report. Given that French firms also had to implement the new reporting regulation on internal control and that the auditor had to report on them, we cannot determine whether this delay is caused by the KAMs or the report on internal control. No matter the source, the absence of effect in 2004 suggests that these difficulties were limited to the initial year of implementation. Nature of KAMs To further explore the effect of KAMs, we classified each matter within the KAM section into two groups, based on the two themes identified in the French auditing standard: accounting principles and accounting estimates. Following Bédard and Gonthier Besacier (2013), we identify 4 categories for the accounting principles group (General, Going Concern, Choice of methods, and Change in methods) and 13 categories, based on the KAM related financial statements item, for the accounting estimates. Column 1 and 2 of Table 6 present the frequency of KAMs in each category for the first year of application (2003). Depreciation or impairments of fixed assets (66.5%), choice of accounting methods (38.1%), provisions for risks and charges (30%), differed income tax (21.8%), intangible assets (19.1%), pension liability (18.3%), are the most frequent categories Because a KAM section includes more than one matter, a KAM could be coded in more than one category and accordingly, the total does not add up to 100 percent. 26

27 [TABLE 6] To analyze the effect of the nature of KAMs, we repeat the 5 analyses reported in Tables 3 and 4, adding 17 indicator variables for each category of KAMs. As FIRST_KAM and NEW_KAM are included in the models, the coefficients for the categories indicate whether the effect of KAMs is different for a given category. Columns 3 to 7 of Table 6 present the results for the coefficients that are significant at the 0.10 level for each model. Of the 85 coefficients (17 x 5 regressions), 9 coefficients are significant at the 0.10 level. For the measures of consequences for investors, Table 6 shows that investors react to some specific KAMs. Thus CAR is higher when a firm discloses a KAM related to investments, while AVOL it is higher when a firm discloses a KAM about the Going Concern assumption, and accounting estimates concerning Depreciation or impairments of fixed assets. Regarding the impact on the audit, the three last columns in Table 6 highlight that the audit fees are lower for firms with a KAM related to Investments. Audit report lag is higher for firms with KAMs related to the Going Concern assumption and to Provisions for risks & charges, suggesting that these KAMs require more time for discussion with management and audit committee. As shown by the significant negative coefficients on Going Concern, Expenses and other income and Revenues, these types of KAMs are associated with even lower DAC. Overall, the results of this exploratory analysis suggest that the impact of KAMs on investors and the audit varies depending on the nature of the KAM, albeit it is only for a small number of them. The significant coefficients on Going Concern for AVOL and LOGAUDLAG and DAC are consistent with previous research on going concern opinion 27

28 (e.g., Carson et al. 2013) and, as such, provide assurance regarding our results, and in particular the general lack of impact of KAMs. Information environment The effect of information disclosure on investors reactions depends on the strength of the firm information environment, this effect decreasing with the strength of the firm information environment (Botosan 1997). We thus expect that KAMs will have a greater impact as the strength of the firm information environment decreases. We use the firm s market capitalization (LOG_MARKET_CAP) as a measure of the strength of its information environment. Indeed, larger firms have higher disclosure levels (Lang and Lundholm 1993) and more analysts following them than smaller firms (Bhushan 1989). To examine this effect, we estimate models (1) and (2) and add an interaction between LOG_MARKET_CAP and our two KAMs measures (FIRST_KAM, NEW_KAM). Table 7 presents the results for CAR and AVOL. In Model 1, the two interaction terms are not significant for CAR and, as in our main analysis of Table 3, the coefficients on FIRST_KAM and NEW_KAM are not significant. Thus, the effect of KAMs on abnormal returns is not associated with the strength of the firm s information environment. For AVOL, the interaction between LOG_MARKET_CAP and FIRST_KAM is positive and marginally significant. Thus, the first time they are disclosed, KAMs are associated with less information asymmetry and the higher the size of the firm, the lower the decrease. We further examine this effect, by estimating Model (2) for small firms (first quartile of LOG_MARKET_CAP) and larger firms (other three quartiles). For small firms, the coefficient on FIRST_KAM (not tabulated) is negative and significant, while it is not significant for larger firms. For both types of firms, the coefficients on NEW_KAM are not significant. Accordingly, for firms with weak information environment, KAMs are associated with less information asymmetry when 28