Acknowledgements: I thank Wietse de Heer for the helpful discussions and feedback during the writing of this thesis. Additionally, I thank Joost Klaver, Myrthe van der Klei, Albert Voortman and Alain Voortman for helpful comments on earlier drafts of this thesis. I also thank the interviewees: Karin van Hulzen and Veronique Stoffels from the NBA; Geert Koster from the VEB; David Tailleur from Rabobank; Wietse Koster and Philip Wallage from KPMG as well as several other colleagues from KPMG, who all provided useful comments, input and insights for my scoring mechanism. I further thank two colleague students in the M.Sc. A&C program: Henk van den Top and Marcel Visser, for helping with the assessment of my scoring mechanism.
Evidence from the Netherlands
Master Thesis, M.Sc. Accountancy & Controlling, profile Accountancy,University of Groningen, Faculty of Economics and Business
DATE 26-6-2015
JUSTIN VOORTMAN Student number: 2169797
Ruychrocklaan 151 2597EM Den Haag tel. +31 6 11 74 39 66
e-mail: g.j.voortman.1@student.rug.nl
Supervisor University P.G. de Heer
A New Auditor’s Report, Higher Audit Quality?
Evidence from the Netherlands
ABSTRACT: This paper examines whether the new auditor’s report, as issued by the Dutch professional body for accountants: “Nederlandse Beroepsorganisatie van Accountants” (NBA), is associated with audit quality. Using both performance adjusted discretionary accruals and unexpected audit fees as proxies for audit quality, I document a positive, marginally significant relationship between the introduction of the new auditor’s report and audit quality. Further investigating whether the differences between the new auditor’s reports have an impact on this relationship, I employed a scoring mechanism to record differences in quality between the new auditor’s reports issued. I find that the relationship between the new auditor’s report and audit quality is stronger and marginally more significant for higher quality new auditor’s reports. In general my results give support (though only weakly significant) for the hypothesis that publishing previously unavailable audit information to the public is correlated with higher audit quality.
Keywords: Auditor’s report; audit information disclosure; audit quality; discretionary accruals; unexpected audit fees.
Data Availability: Data used in this study is available from the public sources identified in the text.
I. INTRODUCTION
On December 12th 2014 the Dutch professional body for accountants: “Nederlandse Beroepsorganisatie van Accountants” (hereafter NBA) issued a revision of the auditing standard regarding the independent auditor’s report. This revision requires those auditors that audit Organizations of Public Interest1, to explain more about their work above and beyond the pass/fail statement that was the previous auditor’s report. This new standard is applicable for audits on financial statement periods ending on or after December 15th 2014. The NBA (2014) requires according to this standard, among other things, that auditors explain more about the key risks they identified in their audit and how they applied materiality2 in their work.
With this new standard, the Netherlands is one of the frontrunners in the world of auditing, preceded only by the United Kingdom which issued a similar standard for financial years commencing in or after October 2012 (Financial Reporting Council, 2013) and preceding the more limited requirements by the International Auditing and Assurance Standards Board (hereafter IAASB), effective as of December 15, 2016. The issue of the IAASB only requires reporting of the key audit matters and how these are addressed in the audit. For instance, a specification of the materiality used is not a requirement. In this paper I explicate my study into the impact of the new auditor’s report, as issued by the NBA in the Netherlands (hereafter the new auditor’s report), on the quality of the audits for which these new auditor’s reports are being issued. Specifically, I examine the impact of the introduction of the new auditor’s report as a whole on audit quality. Furthermore, because substantial differences between new auditor’s reports have been observed in the United Kingdom and in the Dutch pilot3 (Deans & Fisher, 2014; Eumedion, 2014; Financial Reporting Counsil, 2015), I research the impact of the differences in quality of the new auditor’s reports on audit quality.
The debate about the impact of the auditor’s report has been ongoing for several decades. Back in the previous century Bailey et al. (1983) already researched if wording changes in the auditor’s report impacted the perceived message of readers. More recently, Mock et al. (2013, p. 345) state: “users desire more information about the auditor, audit, financial statements and aspects of the entity that are related to financial statements.” This statement is supported by results from the research from Simnett & Huggins (2014). Moreover, Vanstraelen et al. (2012) point out that despite the fact that the debate about audit reports has been ongoing for some time, it is still a long way from being settled. They use recent publications by the Public Company Accounting Oversight Board (hereafter PCAOB) and the IAASB that propose changes to the auditor’s reporting model to substantiate this claim (IAASB, 2011; PCAOB, 2011).
1An organization is of public interest if it is based in the Netherlands and 1* established to Dutch law and listed on a regulated market according to the wet financieel toezicht (Wft) or; 2* a credit institution according to the Wft or; 3* a central credit institution according to the Wft or; 4* a life- or indemnity insurer according the Wft (Autoriteit Financiële Markten, 2015).
2 “Information is material if omitting it or misstating it could influence decisions that users make on the basis of financial information about a specific reporting entity. In other words, materiality is an entity-specific aspect of relevance based on the nature or magnitude, or both, of the items to which the information relates in the context of an individual entity’s financial report” (International Accounting Standards Board, 2010, p. 17).
3 In the Netherlands a number of auditors decided together with their clients to issue a voluntary extended audit report over financial year 2013 (Eumedion, 2014).
The questions with regard to the auditor’s report are concerned with whether the standard auditor’s report conveys the appropriate information and whether or not its current form and content stimulate audit quality (International Organization of Securities Commissions, 2009). The IAASB (2009) recognized in discussions on the topic of audit quality that auditor communications, of which the auditor’s report is one of the most important forms, is closely linked with audit quality. If the auditor’s report has an impact on quality and the new auditor’s report improves the stimulation of quality, than I expect the auditing quality to go up with the introduction of the new auditor’s report. However, if audit quality is not linked to the auditor’s report or if the new auditor’s report does not enhance the stimulation of audit quality, I expect the introduction of the new auditor’s report not to have an impact on audit quality. Moreover I expect this effect to be enhanced by the quality of the new auditor’s report. That is, a higher quality new auditor’s report is associated with a higher quality audit.
My analysis is based on data collected from 112 firms over the fiscal years 2010 to 2014. Next to the dummy variable which measures whether or not a version (pilot or first year legislative) of the new auditor’s report is submitted with the annual report, I develop a score variable that is used to measure the quality of the new auditor’s report. The score is based on the standard as established by the NBA (2014), statements publicized by users and input from interviews with users and preparers of the new auditor’s report. I use two complementary methods to measure audit quality. The first method I employ is the performance adjusted modified Jones model as employed in prior studies (Ashbaugh, et al., 2003; Lim, et al., 2008; Lim, et al., 2012). The second method I utilize uses unexpected audit fees as a measure of audit quality as suggested by Li et al. (2009).
I find a negative, though insignificant, association between the new auditor’s report and discretionary accruals. Between new auditor’s report quality and discretionary accruals I find a negative and marginally significant relation. For the unexpected audit fee tests I find a positive association between unexpected audit fees, the new auditors report and new auditor’s report quality. However, only the results for new auditor’s report quality is significant; the effects of the new auditor’s report variable are marginally insignificant. These findings indicate that a high quality audit is correlated with publishing a new auditor’s report, where the relation is stronger for higher quality new auditor’s reports.
This paper adds to different strands of the auditing literature. First of all, this study contributes to the literature on the auditor’s report. A significant amount of research has been done with regards to the auditor’s report. Much of this research is qualitative in nature using experiments or other explorative methods. Experimental research on this subject showed evidence of improvements in market functioning when for instance materiality is disclosed in the auditor’s reports (e.g., Davis, 2007; Fisher, 1990; van Buuren, et al., 2013). On the other hand, Gold et al. (2012) found that the revised ISA 700 disclosure does not result in a decreased expectation gap in Germany as compared to an audit-opinion-only report and they suggest that an audit opinion by itself signals sufficient relevant information to users.
Explorative research does point yet again in the direction of a positive result of additional disclosure in the auditor’s report. For instance Gray et al. (2011) state that interviewed auditors speculated that lower materiality levels will be set than the disclosed materiality, which results in more procedures to increase confidence that the disclosed materiality is actually achieved. This entails an increase in auditing quality following the new regulations with regard to the auditor’s
report. Further explorative evidence is given by Vanstraelen et al. (2012) who conclude that both users and auditors think that including for instance key areas of risk and critical accounting estimates enhances informational value of the auditor’s report. These studies use either questionnaires or interviews to find out if additions to the auditor’s report are desired and experimental settings to see what might be the results of some of the proposed changes. Moreover the evidence of the proposed changes found thus far is still inconclusive.
In contrast, this study applies a quantitative method using the newly obtainable data available from the regulatory change in the Netherlands to compare the introduction of the new auditor’s report to several established measures of audit quality. In addition, this paper looks at the effect on audit quality where prior literature focusses on market functionality. Furthermore, I investigate the difference between the new auditor’s reports to see whether or not this has an additional impact on audit quality. With this study I, therefore, respond to Coram et al. (2011), who call for further research into how different auditor report formats and content affect audit quality. Similarly Mock et al. (2013) call for research into the changes to the auditor’s report considered at that time, to lead to improvement of the value relevance of auditor services, disclosures and assurances. The results are generally consistent with the hypothesis that a more informative, extended auditor’s report is linked with higher audit quality as proxied by discretionary accruals and unexpected audit fees. This conclusion is limited by the fact that audit quality is inherently unobservable and proxies, like the ones used, are not perfectly interchangeable with audit quality.
By looking at the effect that the new auditor’s report has on audit quality, this paper also contributes to the broad strand of literature on audit quality. Considerable research has been done on factors influencing auditing quality among which: audit firm size, non-audit services and auditor tenure, finding mixed results. Regarding size, studies find that larger audit firms supply higher audit quality (e.g., Colbert & Murray, 1998; DeAngelo, 1981). With regard to non-audit services some studies find that providing these services negatively impacts audit quality (e.g., Frankel, et al., 2002; Wines, 1994). Ruddock et al. (2006) on the other hand find no association between the provision of non-audit services and audit quality. Auditor tenure has generally been found to have a positive effect on audit quality (e.g., Ghosh & Moon, 2005; Johnson, et al., 2002; Myers, et al., 2003). In contrast Davis et al. (2009) find evidence that longer tenure diminishes audit quality, although their results indicate that this relation disappeared post-Sarbanes-Oxley. In a recent article in Accounting Today however, it is stated that mandatory rotation of auditors might have benefits that surmount the costs that are so oftentimes mentioned (James, 2015). Thus far users see the auditor’s report as binary code and only check if it is a pass or fail, with the reputation of the audit firm, dependent on the variables described above, vouching for the quality (Vanstraelen, et al., 2012). This research aims to find the effects on audit quality associated with the new auditor’s report. According to the best of my knowledge this approach has not been attempted before. My evidence suggests that an increase in audit reporting usefulness, like a high quality new auditor’s report, can be used as an indicated driver for audit quality as Knechel et al. (2013) propose.
The final contribution of this research is of a practical nature. Legislators around the world are wondering if a new auditor’s report should be required and how this new report should be shaped. Outcomes from this study suggest that a (high quality) new auditor’s report is linked with
higher audit quality. Therefore, this paper supports the application of a new auditor’s report as implemented in the Netherlands. However, this study does not look at the costs of the new auditor’s report. Consequently it cannot by itself be used as a full substantiation for mandating a new auditor’s report without investigating the cost side of the equation.
The remainder of this paper is organized as follows. Section II discusses the theory and develops hypotheses. Section III describes the research method. In section IV I present the empirical results. Finally, section V deliberates on conclusions and limitations of this paper.
II. BACKGROUND, THEORY AND HYPOTHESIS DEVELOPMENT
This investigation has a dual purpose: to investigate the effect of the new auditor’s report on audit quality in the Netherlands; and secondly, to examine whether differences in the new auditor’s reports have an impact on this relationship. I use DeAngelo (1981) to define audit quality as: “the quality of audit services is defined to be the market-assessed joint probability that a given auditor will both (a) discover a breach in the client's accounting system, and (b) report the breach.” Agency Theory, Roots of the Audit and the Old Auditor’s Report
Humphrey et al. (2009) cite Mautz and Sharaf (1961) stating that the purpose of the audit is to provide parties supplying property with a confirmation of the reliability of information offered by those entrusted with this property. With this statement they point toward the problem agency theory identifies as information asymmetry (Jensen & Meckling, 1976): parties supplying property (principals) require confirmation of the information they get from parties entrusted with this property (agents). This means that the reason for auditing can be explained with agency theory: because of agency problems, principals require attestation of agent-supplied information by an external auditor to monitor whether the agents’ objectives are truly aligned with the principals’. This verification must be communicated to principals and according to the IAASB (2009) the auditor’s report is the main communicative devise the auditor has at his disposal. Whereas to financial statement users (principals), the auditor’s report represents the auditors (primary) means of communicating (Eid, 2014). The purpose and value of this report then also lies in the constraining of agency problems through the independent authentication of information.
Evidence of the assertion that principals trust audited information more is plentiful (Abdel-khalik, 1993; Chow, 1982; Dopuch, et al., 2003; Watts & Zimmerman, 1983). Furthermore, Deumes et al. (2010) confirm the value of the auditor’s report means of communicating in their study, by finding that the auditor’s report is valuable to financial analysts and increases these analysts’ reliance on financial statements. Moreover, Coram et al. (2011) find that the auditor’s report is of importance to users of financial statements, because it signals a level of reliability that unaudited financial statements lack. Nevertheless, they conclude that is the sole communicative value that the auditor’s report seems to have. Humphrey et al. (2009) concur with this view and state that the standard external audit report is not helpful in giving insight into the specific work carried out and findings acquired by auditors. Mock et al. (2013) similarly argue that there is a gap between what financial statement users want to be communicated and what the auditor’s report is actually communicating. Financial statement users not only seem to want confirmation of
information provided by the agents. They expect an extra information contribution of the independent third party auditor to restrain agency problems further.
To tackle this state of relative trivialness of the old auditor’s report, many researchers suggest possible supplements to the auditor’s report to increase its value as an informative communication devise. Humphrey et al. (2009) suggest the addition of a section on key audit findings, inclusion of materiality and operational risk, significant errors detected and also the disclosure of scale, nature and results of the tests done to verify the information in the annual report. According to them this will increase the substance and quality of the audit work demonstrated to the users. Mock et al. (2013) find that additional information desired, consists of information related to audit judgments, auditor independence, the audit process, materiality and the level of assurance provided.
The new auditor’s report responds to these suggestions by requiring, at a minimum, the addition of (NBA, 2014):
1) A description of the key audit risks identified in the audit that had an impact on the audit strategy and the allocation of resources of the audit;
2) An explanation of how the auditor applied the concept of materiality, at a minimum specifying the threshold used by the auditor as materiality for the financial statements as a whole;
3) And the provision of an overview of the scope of the audit, including an explanation of how this scope both addressed the assessed risks of (1) and was influenced by the materiality of (2).
Communication Theory, the New Auditor’s Report and Audit Quality
According to communication theory, research into the auditor’s report is being used to address multiple types of “gaps.” These gaps are: the expectation gap; the information gap; and the communication gap, in addition these gaps are interrelated (e.g., Hronsky, 1998; Maijoor, et al., 2002; Mock, et al., 2013). The expectation gap is the difference between what users expect from the auditor and the financial statement audit, and the reality of what an audit is (IAASB, 2011, p. 7).“Users of corporate financial information point to the existence of a gap between the information they believe is needed to make informed investment and fiduciary decisions and what is available to them through the entities audited financial statements or other publicly available information” (IAASB, 2011, p. 8), this is known as the information gap.“The communication gap reflects differences between what users desire and understand and what is communicated by the assurance provider” (Mock, et al., 2013, p. 327).
The new auditor’s report tries to address the information gap by inserting information wanted by the users (materiality, assessment of risks of material misstatement, etc.). This extra information also aims to close the communication gap, according to the process school of communication theory. This school states that information is the transmission of messages and ignores the contextual factors which might affect interpretations. Furthermore, the new auditor’s report seeks to address the expectation gap by giving an overview of the scope of the audit. This can give readers a more thorough understanding of what actual work the auditor has undertaken and thereby why his responsibilities are as he already states in the original auditor’s report: “to
audit and express an opinion” and not: “to prepare the financial statements giving a true and fair view,” which is the responsibility of the directors of the entity. Wright and Wright (2014) assert that while auditors face tough judgements, financial statement users tend to attribute negative outcomes to auditors after the outcome has occurred. Ascribing of more or different responsibilities to auditors can also be witnessed in several statements by media and politicians alike (Sikka, 2009). However, Wright and Wright (2014) state that this attribution can be mitigated by disclosing the auditor judgment process.
Closing these gaps means an increase in audit reporting usefulness, which in turn has been indicated as a driver for audit quality (Knechel, et al., 2013). They cite prior literature which notes that the old auditor’s report is limited with regard to audit quality because of a lack of communicative value and an information gap between auditors and users (Church, et al., 2008; Mock, et al., 2013). This information gap comprises previously undisclosed audit information such as the materiality and the significant audit risks (Knechel, et al., 2013). By disclosing more information about the audit I expect the audit quality to go up. When auditors have indicated that a certain matter is within scope and within materiality as stated in their auditor’s report I expect them to be more critical in their auditor’s opinion. Or as stated in terms of DeAngelo (1981): I expect auditors to sooner report a discovered breach of account, when they have stated in their auditor’s report this account was within the scope- and within the materiality of the audit.
The Trust Doctrine, the New Auditor’s Report and Audit Quality
Closing these gaps not only increases audit quality through the informative and communicative value of the auditor’s report. It can also serve as an incentive for auditors to increase their efforts and thereby audit quality, through the expectation gap. By formally communicating to stakeholder what risks are identified and what materiality is used, auditors create expectations among financial statement users. Limperg mentions these expectations in his essays early in the previous century (1932). He states that the auditor is the confidant of society and is, therefore, bound by what he calls: “Trust Doctrine.” In this trust doctrine two important matters with regard to the public trust are noted. Firstly: the auditor should live up to the confidence given by the public. Secondly: the auditor should not invoke more trust in him than he can realize. Auditors who draft a more extensive auditor’s report raise expectations of the trust the public can put in them and, according to the trust doctrine, should therefore live up to this increased trust and perform more or better audit work delivering higher audit quality.
This is also pointed out in the results of prior research. Gray et al. (2011) find that auditors would likely use a lower internal materiality for their work, thus increasing audit quality through increased effort and stricter internal standards to which the audited financial statements must comply to be approved, just to be safe. This would mean audit quality goes up, because auditors will discover a breach more quickly when they increase their effort and extent, for instance, their sampling to be able to ensure these lower materiality levels (DeAngelo, 1981). Therefore, based on both communication theory and the trust doctrine, I expect there to be an association between the introduction of the new auditor’s report and the quality of audits, or stated formally:
The Trust Doctrine, the Quality of the New Auditor’s Report and Audit Quality
The newly implemented standard is principle- rather than rule based because the NBA wants auditors to be innovative in their implementation of the new auditor’s report (NBA, 2014). They want the new auditor’s report to be as organization-specific as possible and not bound by standard formats and boilerplate comments. This leads not only to the possibility of being innovative, it also results in sizeable differences in the application by various auditors (Deans & Fisher, 2014; Eumedion, 2014; Financial Reporting Counsil, 2015). According to these studies some auditor’s reports reflect the bare minimum of the newly required standard. These reports touch upon all the points, but with boilerplate explanations. Other reports in the tests, go beyond the requirements of the new standard and introduce more or improved parts in the new auditor’s report than the standard requires; notably the Rolls Royce report in the United Kingdom (Deans & Fisher, 2014).
These differences in new auditor’s reports are interesting in light of Limperg’s (1932) trust doctrine. By composing a more extensive auditor’s report than strictly necessary according to the new standard, auditor’s increase expectations of financial statement users beyond the expectations financial statements users had with the mere introduction of the new auditor’s report by itself. For instance, a more detailed auditor’s report can lead to more questions during the shareholders meeting in which auditors have to answer questions (NBA, 2012). Therefore, I expect that a higher quality and more extensive auditor’s report will be associated with even higher audit quality than the increase in audit quality following the general release of the newly required new auditor’s report. Consequently, my second hypothesis is:
Hypothesis 2. A higher quality auditor’s report is associated with higher audit quality, over and above the increase in audit quality following the general introduction of the new auditor’s report.
III. RESEARCH METHOD
Sample and Data
Table 1 contains the sample selection. The initial sample consists of 230 different companies that either deposited their annual reports at the Dutch AFM register, or companies registered in the Netherlands that can be found in the Compustat Global or Worldscope databases during the years 2010-2014. I excluded companies with missing fundamental data and special purpose vehicles with no actual activities other than providing capital for a parent company (Finance BV’s). Finally, I also excluded Financial Institutions (SIC codes 6000-6999), because for those companies many of the variables needed are fundamentally different which leads to a different procedure needed according to Frankel et al. (2002). For example cash flows, trade receivables, current-asset-ratio’s, quick-ratios and debt-ratios differ largely because of the business model of financial institutions which would lead to distortions in the data. All in all, these exclusions resulted in 112 firms in the final sample leading to a total of 503 (502) company years of data for the discretionary accruals method (unexpected audit fee method). To exclude impact of outliers, I winsorized all the variables at the 1st and the 99th percentiles.
Matching the firm’s annual report data from Worldscope with annual reports filed in conjunction with a new auditors report yields a sample of 100 observations of data with new auditor’s reports of which 26 observations from 2013 (pilot) and 74 observations from 2014 (first year legislated).
Model Specification
New Auditor’s report. To test the first hypothesis, I used data of whether new auditor’s reports
were issued in conjunction with issued annual reports of the years 2013 and 2014 by Dutch companies. Firstly, I utilized several sources mentioning the issuance of new auditor’s reports with annual reports over 2013 to get the number of reports issued during the pilot (Eumedion, 2014; PricewaterhouseCoopers, 2014). Secondly, I manually checked all the annual reports filed with the AFM registry to see if any report contained a new auditor’s report4.
For investigating the second hypothesis, I developed a specification of new auditor’s report quality. To compose a score for the new auditor’s reports on quality, I first of all made use of the
4 ‘Listed companies with the Netherlands as country of origin and with their securities allowed at a regulated European market are mandated to file their (semi) annual reports and interim statements at the ‘Loket AFM’ and this information is available in the AFM registry’ (Autoriteit Financiële Markten, 2015).
Table 1 Sample Description Panel A Total sample Full Companies (PADCA) Company years (PADCA) Full Companies (UnexpAF) Company years (UnexpAF) Total sample 230 1150 230 1150
Less: Special Purpose Vehicles (87) (445) (87) (445)
Less: Other Financial institutions (SIC codes 6000-6999)
(29) (138) (29) (138)
Less: Observations with missing data in Compustat or Worldscope
(2) (64) (3) (65)
Final Sample 112 503 111 502
Panel B
Total sample of company years with new auditor’s report
Company years (PADCA) Company years (UnexpAF) Total sample 159 159
Less: Special Purpose Vehicles (34) (34)
Less: Financial institutions (SIC codes 6000-6999)
(23) (23)
Less: Observations with missing data in Compustat or Worldscope
(2) (2)
Final Sample 100 100
New auditors report in 2013 26 26
New auditors report in 2014 74 74
In this table the sample selection is displayed. The columns give the sample selection per company and per company year of data for the two respective methods: discretionary accruals (PADCA) and unexpected audit fees (UnexpAF).
published statement of the NBA (2014) and several published statements from Eumedion5 and the Vereniging voor effectenbezitters6 (VEB) (Abma, 2014; Slagter, et al., 2014). I also used a publication by the FRC (2015) which comments on the experiences with the new auditor’s report after the first year in the United Kingdom. To capture the quality of the new auditor’s report I scored the reports on 17 items and ranked the resulting scores in percentile ranks (per decile) to alleviate measurement error, as suggested by Johnston (1984). The resulting score takes into account all of the new auditor’s report requirements according to the regulation, as well as the thoughts and requests that followed from the above stated documents. The complete development of the score and the final score itself are available in appendix A.
Audit quality estimation equations. Audit quality is a complex construct and looking at it through
the limited perspective of individual audit quality indicators is a potential problem according to Bedard et al. (2010). Similarly Balsam et al. (2003) state that no single auditor characteristic should be used by itself as a proxy, since auditor quality is multidimensional and inherently unobservable. To address this problem I applied two supplementary approaches to distinguish audit quality.
The first approach is based on the notion that audit quality can be measured through earnings management. Prior research argues that audit quality is inversely related to earnings management (e.g., Becker, et al., 1998; Lin & Hwang, 2010) which is in turn positively related to discretionary accruals (e.g., DeFond & Park, 2001; Jones, 1991; Levitt, 2010; Schipper & Williams, 1989). Asthana and Boone (2012) assert that discretionary accruals provide managers opportunities to manipulate earnings. They continue that, when auditors allow some of the manipulations to remain uncorrected, this adversely reflects on the audit quality.
Empirical evidence supporting the use of discretionary accruals as an indicator for audit quality comes for example from Menon and Williams (2004) who find that discretionary accruals are larger at companies that employ former audit partners as directors, which in turn has been found to lead to a higher chance of getting a favorable opinion (e.g., Firth, 1981; Lennox, 2005; Parlin & Bartlett, 1994). Furthermore, Balsam et al. (2003) find evidence that discretionary accruals are negatively related with their proxy for audit quality: auditor industry specialization.
To distinguish audit quality through earnings management I used performance adjusted discretionary accruals estimated with a cross-sectional modified Jones (1991) model as done in prior studies (Ashbaugh, et al., 2003; Lim, et al., 2008; Lim, et al., 2012). In this model performance adjusted discretionary accruals equal:
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡 = 𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡− 𝛼𝛼�1 − 𝛽𝛽̂1(𝛥𝛥𝛥𝛥𝛥𝛥 − 𝛥𝛥𝛥𝛥𝛥𝛥)𝑖𝑖,𝑡𝑡− 𝛽𝛽̂2𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖,𝑡𝑡−1− 𝛽𝛽̂3𝛥𝛥𝑅𝑅𝑃𝑃𝑖𝑖,𝑡𝑡−1 (1) where all variables, except PADCA and ROA, are deflated by lagged total assets to control for scale differences. CA is the current accruals per firm per year, defining current accruals as
5 ‘Eumedion represents the interests of the affiliated institutional investors in corporate governance and
sustainability. The aim of pension funds, insurance companies, investment companies and asset managers who are members of Eumedion is to improve listed companies’ performance in governance, environmental and social areas’ (Eumedion, 2015).
6 ‘Since 1924 the shareholders association VEB represents the interests of investors.The association makes its voice heard through shareholder meetings, in the media and campaigning with demonstrable abuses in listed companies’ (Vereniging voor effectenbezitters, 2015).
income before extraordinary items plus depreciation and amortization minus operating cash flows as in Lim et al. (2012); ΔRV is the measure of the change in revenues from the prior year to this year per firm; ΔTR is the change in trade receivables from the prior year to this year; IBE is the lagged income before extraordinary items per firm; and ROA is the lagged return on assets which, according to Kothari et al. (2005), eliminates the mechanical relation between performance and the current period’s discretionary accruals. 𝛼𝛼�, 𝛽𝛽̂1, 𝛽𝛽̂2, 𝛽𝛽̂3 in Eq. (1) are acquired from estimating 𝛼𝛼, 𝛽𝛽1, 𝛽𝛽2, 𝛽𝛽3 in Eq. (2) at the industry level. The variables in Eq. (2), with the exception of ROA, are also deflated by lagged total assets and εi,t is the residual term. To identify industry membership for Eq. (2) I used the two-digit SIC codes in Worldscope.
𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝛥𝛥𝛥𝛥𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽3𝛥𝛥𝑅𝑅𝑃𝑃𝑖𝑖,𝑡𝑡−1+ 𝜀𝜀𝑖𝑖,𝑡𝑡 (2) The second approach I applied to identify audit quality is through the use of audit fees. High auditor effort is assumed to result in higher audit quality, since more work done by the auditor (more effort) leads to the auditor being able to draw better conclusions. High audit fees are indicative of greater effort on the engagement and thus indicative of higher audit quality (Eshleman & Guo, 2014). According to Li et al. (2009) the audit effort can be measured using unexpected audit fees. Audit fees are normally dependent upon size, complexity and risk of the audit assignment; a larger more complex and riskier client requires more audit work which in turn results in higher audit fees. By capturing size, complexity and risk following the model of Ferguson et al. (2006), I can calculate the expected audit fee and the difference between actual and expected audit fees. The difference between expected audit fees and actual audit fees can then be used as a proxy of auditor effort not explained by the size, complexity and risk of the audit assignment. Unexpected audit fees are determined as a ratio of actual to expected audit fee, where the expected fee is the anti-log of the predicted fitted value from the OLS of the audit fee model Eq. (3) based on Ferguson et al. (2006): 𝐿𝐿𝑃𝑃𝐿𝐿𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐿𝐿𝑆𝑆𝐿𝐿𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝑃𝑃𝑃𝑃𝛥𝛥𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽4𝑄𝑄𝐿𝐿𝐼𝐼𝑃𝑃𝑄𝑄𝑖𝑖,𝑡𝑡+ 𝛽𝛽5𝑃𝑃𝐼𝐼𝑖𝑖,𝑡𝑡 + 𝛽𝛽6𝛥𝛥𝑅𝑅𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽7𝐿𝐿𝑅𝑅𝛥𝛥𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽8𝑅𝑅𝑃𝑃𝐼𝐼𝐸𝐸𝐼𝐼𝑅𝑅𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽9𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑖𝑖,𝑡𝑡 + 𝛽𝛽10𝑃𝑃𝐼𝐼𝑆𝑆𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽11𝑀𝑀/𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽12𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽13𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡 + 𝛽𝛽14𝑌𝑌𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝜀𝜀 (3)
where LAF is the natural log of audit fees; SIZE is the natural log of total assets; LSUB is the natural log of the investments in subsidiaries; CATA is the ratio of current assets to total assets; QUICK is the ratio of current assets (less inventories) to current liabilities; DE is the ratio of long-term debt to total assets; ROI is the ratio of earnings before interest and tax to total assets; FOREIGN is the proportion of foreign sales to total sales; ABSPADCA is the absolute value of the discretionary accruals model of Eq. (1); M/B is the market to book ratio, measuring growth; OPINION is an indicator variable, equaling to 1 for an other than unqualified audit report; YE is an indicator variable, corresponding to 1 for a non-December 31th year end; LOSS is an indicator variable, equaling to 1 if there is a loss in the past year; BIGN is an indicator variable indicating 1
for an audit done by any of the BIG4 firms and 0 for audits done by other audit firms; LITIG7 is an indicator variable equal to 1 indicating if a firm operates in a high litigation industry; finally I control for year and industry fixed effects and ε is the residual.
Size of the audit assignment is captured in SIZE while complexity is captured by looking at the investment in subsidiaries (LSUB), the proportion of foreign sales (FOREIGN) and growth (M/B), all expected to have a positive correlation with audit fees. Also included is the discretionary accruals variable (ABSPADCA) because I expect discretionary accruals to be coupled with ambiguity which leads to expect higher audit fees to cover this ambiguity. Furthermore the ratios CATA, QUICK, DE and ROI are used to indicate audit risk in which higher current-asset- debt- and return-on-investment ratios and lower quick ratios indicate higher audit risk according to Ferguson et al. (2003). The indicator variables OPINION, LOSS, LITIG, BIGN and YE are also included because a qualified audit opinion or a loss is expected to yield a higher audit fee. Furthermore, audits in high litigation industries and big 4 auditors are expected to charge higher fees. Finally audits outside busy season can be expected to have lower fees. The industry variables are also expected to cover differences in complexity while controlling for year effects is used to cover for potential differences in audit fees over time.
New auditor’s report and audit quality. To test my first hypothesis using the first approach, I
utilized the absolute value for measuring performance adjusted discretionary accruals to capture the combined effect of income-increasing and income-decreasing earnings management, following among others Frankel et al. (2002) and Reynolds and Francis (2000). This leads to the following model:
|𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃|𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝑃𝑃𝐿𝐿𝑅𝑅𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐿𝐿𝑃𝑃𝐸𝐸𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝐿𝐿𝐼𝐼𝛥𝛥𝑖𝑖,𝑡𝑡 + 𝛽𝛽4𝑀𝑀 𝐼𝐼⁄ 𝑖𝑖,𝑡𝑡 + 𝛽𝛽5𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼𝑖𝑖,𝑡𝑡 + 𝛽𝛽6𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑖𝑖,𝑡𝑡+ 𝛽𝛽7𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽8𝑃𝑃𝐿𝐿𝑀𝑀𝛥𝛥𝑖𝑖,𝑡𝑡 + 𝛽𝛽9𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝜀𝜀
(4)
The variable of interest is the coefficient of the dichotomous variable for the new auditor’s report (NEWAR), coded either 1 for a new auditor’s report and 0 for an old auditor’s report. A negative coefficient of NEWAR is consistent with the hypothesis that the new auditor’s report creates incentives for auditor’s to perform better audits thereby increasing audit quality, while an insignificant or positive coefficient suggests that the new auditor’s report has no effect on audit quality or a negative effect on audit quality. The first control variable I used is cash flow from operations scaled by lagged total assets (CFO), like Ashbaugh et al. (2003) and Lim et al. (2012). Following these studies I also controlled for lagged total current accruals (LAGAC). Furthermore I controlled for leverage, measured as the ratio of total liabilities to total assets (LEV), because it is found to be associated with discretionary accruals (DeFond & Jiambalvo, 1994; Becker, et al., 1998). I controlled for litigation risk (LITIG) and growth (M/B), because it is found that these factors can incentivize the use of earnings management (Matsumoto, 2002). I additionally controlled for firm size, measured as the natural log of total assets (SIZE), following prior researchers (Frankel, et al., 2002; Lim, et al., 2012). Likewise, I controlled for losses, measured as
7 LITIG denotes high litigation risk industries (SIC codes 2833-2836, 3570-3577, 3600-3674, 5200-5961, 7370-7374 and 8731-8734) (Francis, et al., 1994).
an indicator variable (LOSS) coded 1 for a loss and 0 otherwise, because it is found that firms reporting a loss are less likely to have applied earnings management (Brown, 2001). Finally I inserted the control variable AFMR to control for the critical report concerning auditors published by the Dutch financial market watchdog: AFM. This report stated that auditors have done a poor job improving their documentation over the last years and it came out the 25th of September 2014. This publication might have provoked a reaction of the audit firms to increase their audit efforts and improve their audit work. AFMR is coded 1 for annual reports over financial years ended on or after December 31th 20148. I furthermore control for year and industry fixed effects.
I’ll run the following model to test my first hypothesis using the second approach: 𝐿𝐿𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑃𝑃𝐿𝐿𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐿𝐿𝑆𝑆𝐿𝐿𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝑃𝑃𝑃𝑃𝛥𝛥𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽4𝑄𝑄𝐿𝐿𝐼𝐼𝑃𝑃𝑄𝑄𝑖𝑖,𝑡𝑡+ 𝛽𝛽5𝑃𝑃𝐼𝐼𝑖𝑖,𝑡𝑡
+ 𝛽𝛽6𝛥𝛥𝑅𝑅𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽7𝐿𝐿𝑅𝑅𝛥𝛥𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽8𝑅𝑅𝑃𝑃𝐼𝐼𝐸𝐸𝐼𝐼𝑅𝑅𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽9𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑖𝑖,𝑡𝑡 + 𝛽𝛽10𝑃𝑃𝐼𝐼𝑆𝑆𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽11𝑀𝑀/𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽12𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽13𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡 + 𝛽𝛽14𝛥𝛥𝑅𝑅𝛥𝛥𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽15𝑌𝑌𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽16𝑃𝑃𝐿𝐿𝑀𝑀𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝛽𝛽17𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝜀𝜀
(5)
where the variable of interest is again the coefficient of the variable for the new auditor’s report (NEWAR). This time a positive coefficient for NEWAR would be consistent with the hypothesis that the new auditor’s report is positively correlated with higher audit quality measured through auditor effort. A negative or insignificant coefficient would on the other hand suggests that the new auditor’s report is correlated with lower audit quality as measured through effort or there is no correlation at all. ROTA is an additional control variable to control for a change in auditor, coded 1 if switched from auditor after the previous audit. I inserted this variable because of new regulations requiring auditors to rotate after 10 years and it has been implicated that fees from the newly appointed auditor are likely to be substantially lower for the first year audit (Ghosh & Lustgarten, 2006; Simon & Francis, 1988). This might lead to rotation also capturing a large part of the difference between expected and actual audit fees.
New auditor’s report score and audit quality. To test hypothesis 2 I utilized the same approaches
as the tests for hypothesis 1. This time though, next to the single dichotomous variable NEWAR, I deployed an ordinal variable using the deciles of the acquired score on the new auditor’s report: NEWARSC. This lead to the slight adjustment of Eq. (4):
|𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃|𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝑃𝑃𝐿𝐿𝑅𝑅𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐿𝐿𝑃𝑃𝐸𝐸𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡 + 𝛽𝛽3𝐿𝐿𝐼𝐼𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝛽𝛽4𝑀𝑀 𝐼𝐼⁄ 𝑖𝑖,𝑡𝑡+ 𝛽𝛽5𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼𝑖𝑖,𝑡𝑡 + 𝛽𝛽6𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑖𝑖,𝑡𝑡+ 𝛽𝛽7𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽8𝑃𝑃𝐿𝐿𝑀𝑀𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝛽𝛽9𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑖𝑖,𝑡𝑡
+ 𝛽𝛽10𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑆𝑆𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝜀𝜀
(6)
Employing the second approach to test hypothesis 2 Eq. (5) is adjusted in:
8 I use yearend 2014 instead of the actual date of the report’s publication because changing procedures and audit work to respond to the critical report takes time to implement and the only company in my sample with a yearend between the report (September 25th) and December 31st had a yearend at September 30th.
𝐿𝐿𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑃𝑃𝐿𝐿𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐿𝐿𝑆𝑆𝐿𝐿𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝑃𝑃𝑃𝑃𝛥𝛥𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽4𝑄𝑄𝐿𝐿𝐼𝐼𝑃𝑃𝑄𝑄𝑖𝑖,𝑡𝑡+ 𝛽𝛽5𝑃𝑃𝐼𝐼𝑖𝑖,𝑡𝑡 + 𝛽𝛽6𝛥𝛥𝑅𝑅𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽7𝐿𝐿𝑅𝑅𝛥𝛥𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽8𝑅𝑅𝑃𝑃𝐼𝐼𝐸𝐸𝐼𝐼𝑅𝑅𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽9𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆𝑖𝑖,𝑡𝑡 + 𝛽𝛽10𝑃𝑃𝐼𝐼𝑆𝑆𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽11𝑀𝑀/𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽12𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸𝑖𝑖,𝑡𝑡+ 𝛽𝛽13𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡 + 𝛽𝛽14𝛥𝛥𝑅𝑅𝛥𝛥𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝛽𝛽15𝑌𝑌𝐼𝐼𝑖𝑖,𝑡𝑡+ 𝛽𝛽16𝑃𝑃𝐿𝐿𝑀𝑀𝛥𝛥𝑖𝑖,𝑡𝑡+ 𝛽𝛽17𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑖𝑖,𝑡𝑡 + 𝛽𝛽18𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑆𝑆𝑃𝑃𝑖𝑖,𝑡𝑡+ 𝜀𝜀 (7)
The variable of interest is now NEWARSC which measures the impact of the quality of the new auditor’s report above and beyond the impact of having a new auditor’s report on the discretionary accruals Eq. (6) and the unexpected audit fees Eq. (7) respectively. A negative (positive) coefficient for NEWARSC in Eq. (6) means that a higher (lower) quality new auditor’s report is consistent with a higher (lower) quality audit as measured by discretionary accruals. An insignificant coefficient for NEWARSC suggests that the quality of the new auditor’s report does not have a relation with the quality of the audit. For Eq. (7) a positive (negative) coefficient implies that a higher (lower) quality new auditor’s report is consistent with a higher (lower) quality audit as measured by auditor’s effort. An insignificant coefficient here would mean that the quality of the auditor’s report is not correlated with auditor’s effort as measured through audit fees.
Descriptive statistics. Table 2, Panel A reports descriptive statistics of the discretionary accruals
sample. The descriptive statistics for the unexpected audit fee sample are given in Panel B of table 2.
Table 2
Sample Descriptive Statistics Panel A Discretionary Accruals Tests
Continuous Variables
VARIABLES N Mean Std. D Min Max
CA 526 -0.0103 0.115 -0.592 0.331 RV 526 0.00147 0.0114 -0.0213 0.0907 TR 526 0.0328 0.296 -0.0946 2.778 IBE 526 -0.0393 0.855 -18.70 0.678 ROA 513 0.0810 31.16 -588.4 68.96 CFO 526 0.0742 0.131 -0.492 0.541 LAGAC 526 -0.0161 0.107 -0.564 0.331 LEV 526 0.546 0.228 0.00654 1.204 MB 526 1.876 2.814 -4.430 17.10 SIZE 526 6.529 2.745 -3.219 12.21 ABSPADCA 503 0.0691 0.0931 0.000697 0.516
Categorical Variables Mean or
VARIABLES N Percentage Std. D. Min Max
LOSS 526 0.260 0 1
AFMR 526 0.165 0 1
LITIG 526 0.245 0 1
NEWAR 526 0.190 0 1
Panel B Audit Fee Tests
VARIABLES N mean Std. D min max
SIZE 526 6.529 2.753 -3.324 12.38 LSUB 526 1.561 2.581 -2.996 8.934 CATA 526 0.471 0.207 0.0397 1 QUICK 526 1.162 0.987 0.118 6.808 DE 526 0.144 0.141 0 0.641 ROI 526 0.0392 0.182 -1.022 0.486 FOREIGN 526 0.0479 0.0379 0 0.100 ABSPADCA 503 0.0691 0.0931 0.000697 0.516 MB 526 1.932 3.186 -4.840 22.42 LnAF 502 6.690 1.836 0.490 10.66 LexpAF 526 6.608 1.633 1.584 10.42 UnexpAF 502 1.280 0.862 0 4.685
Categorical Variables Mean or
VARIABLES N Percentage Std. D. Min Max
YE 526 0.0399 0.196 0 1 BIGN 526 0.821 0.383 0 1 ROTA 526 0.279 0.449 0 1 LOSS 526 0.260 0.439 0 1 AFMR 526 0.165 0.372 0 1 LITIG 526 0.245 0.431 0 1 NEWAR 526 0.190 0.393 0 1 NEWARSC 526 0.0921 0.205 0 0.900
In this table summary statistics are given for both of the tests done in this research. Panel A depicts the statistics for the performance adjusted discretionary accruals method, while Panel B describes the summary statistics for the unexpected audit fee tests. The columns depict the number of observations (N), the mean, the standard deviation (Std. D), the minimum and the maximum scores per variable.
The variables are as follows: CA = the current accruals, defined as income before extraordinary items plus depreciation and amortization minus operating cash flows divided by lagged total assets; RV is the measure of the change in revenues from prior year to this year per firm divided by lagged total assets; TR is the change in trade receivables from the prior year to this year divided by lagged total assets; IBE is the lagged income before extraordinary items per firm divided by lagged total assets; ROA is the lagged return on assets; CFO = Cash flow from operations; LAGAC = lagged current accruals; LEV = leverage, equal to the ratio of total liabilities to total assets; M/B = growth, measured as the market-to-book-ratio; SIZE = company size, equal to the natural log of total assets; ABSPADCA = performance adjusted discretionary accruals estimated by using performance adjusted modified jones model; LOSS = indicating negative earnings; LITIG = indicating high litigation risk industry; AFMR = computed as an audit started after publication of the AFM report on audit quality
(December 31st, 2014); NEWAR = determined as an audit that resulted in a new auditor’s report; NEWARSC
= the results of the scoring mechanism measuring the quality of the new auditor’s report; LSUB = natural log of investment in subsidiary companies; CATA = ratio of current assets to total assets; QUICK = ratio of current assets (less inventories) to current liabilities; DE ratio of long-term debt to total assets; ROI = ratio of earnings before interest and tax to total assets; FOREIGN proportion of foreign sales to total sales; LnAF = natural log of audit fees; LnexpAF = natural log of expected audit fees where Expected Audit Fees are estimated by predicting the fitted value from the OLS of the audit fee model Eq. (3). UnexpAF = Audit Fees/Expected Audit Fees; OPINION = indicating a non-unqualified audit opinion; BIGN = indicating an audit done by a big 4
IV. RESULTS
New Auditor’s Report and Audit Quality
Discretionary accruals tests. My first test assesses the relation between audit quality, as measured
by discretionary accruals and the new auditor’s report. Table 3 Panels A, B and C reports statistics for the absolute value of performance adjusted discretionary accruals calculation: Eq. (4). The coefficient of interest NEWAR is the expected sign (negative) however the result is not significant (p-value = 0.176). Meaning that support for hypothesis 1 cannot be concluded.
Interesting is the sign for AFMR which is positive although also insignificant (p-value = 0.279). In model 2 the coefficient for AFMR is 0.00005 and highly insignificant (p-value = 0.996). The signs for the other year-dummies (not tabulated) are all negative and insignificant (p-value = 0.576 for 2011, p-value = 0.538 for 2012, p-value =0.503 for 2013).
Audit fee tests. My second test makes use of the unexpected audit fee model to evaluate the relation
between auditor effort, as a proxy for audit quality, measured through audit fees and the new auditor’s report. In Table 4 Panel A, B and C statistics are represented for the unexpected audit fee equation: Eq. (5). The relevant coefficient for this research is the coefficient for NEWAR. The sign of this variable is positive but marginally insignificant (p-value = 0.115). This coefficient is in line with my hypothesis but unfortunately not strong enough to support my hypothesis.
The other coefficients of interest are ROTA and AFMR. ROTA is, in line with the expectations, negative although insignificant (p-value = 0.156). AFMR is also negative, however this variable is significant (p-value = 0.014). The other year-dummy variables (not tabulated are all negative and respectively insignificant for 2011 (p-value = 0.589) and 2012 (p-value = 0.124) and significant for 2013 (p-value = 0.040).
New Auditor’s Report Quality and Audit Quality
Discretionary accruals tests. My third test examines the relation between the quality of the new
auditor’s report, as measured by NEWARSC, and audit quality, measured through discretionary accruals. Table 3 Panels D and E report statistics for the absolute value of performance adjusted discretionary accruals equation: Eq. (6). The coefficient of importance NEWARSC is the expected sign (negative) and marginally significant (p-value = 0.080) in model 4. When taken together with NEWAR in model 5 to measure the impact of the quality of the new auditor’s report above and beyond the impact of the new auditor’s report by itself both NEWAR and NEWARSC are insignificant again (p-value = 0.930 and p-value = 0.379 respectively). This is yet again consistent with my hypothesis but only very weak support can be concluded from model 4.
The sign for AFMR is again positive but still insignificant (p-value = 0.153 in model 4 and p-value = 0.155 in model 5). The other year-dummies are still all negative and insignificant (p-value = 0.576 for 2011, p-(p-value = 0.538 for 2012, p-(p-value =0.503 for 2013).
Audit fee tests. My fourth test measures the relation between the audit quality, proxied by audit
fees, and new auditor’s report quality, measured by NEWARSC. Table 4 Panel D and E describe statistics for the unexpected audit fee calculation: Eq. (7). I am interested in the coefficient for NEWARSC and these are the expected sign and both (marginally) significant (p-values = 0.053 in
model 4 and p-value = 0.094 in model 5). Outstanding is the coefficient for NEWAR in model 5, which is negative (p-value = 0.611), where it was positive in model 3. This negative effect is eliminated by the stronger positive effect for NEWARSC in model 5 leading to a total positive effect. These results are coherent with my hypothesis that the quality of the new auditors report is important in the relation between the introduction of the new auditor’s report and audit quality.
Table 3
Regressions Results Performance Adjusted Discretionary Accruals
|𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃| = 𝛼𝛼 + 𝛽𝛽1𝑃𝑃𝐿𝐿𝑅𝑅 + 𝛽𝛽2𝐿𝐿𝑃𝑃𝐸𝐸𝑃𝑃𝑃𝑃 + 𝛽𝛽3𝐿𝐿𝐼𝐼𝛥𝛥 + 𝛽𝛽4𝑀𝑀 𝐼𝐼⁄ + 𝛽𝛽5𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼 + 𝛽𝛽6𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆 + 𝛽𝛽7𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸 + 𝛽𝛽8𝑃𝑃𝐿𝐿𝑀𝑀𝛥𝛥 + 𝛽𝛽9𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥 + 𝛽𝛽10𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑆𝑆𝑃𝑃 + 𝜀𝜀
Panel A Panel B Panel C Panel D Panel E
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5
CFO -0.0189 -0.0189 -0.0207 -0.0195 -0.0196 (0.0635) (0.0635) (0.0636) (0.0641) (0.0641) LAGAC -0.0958 -0.0958 -0.0937 -0.0964 -0.0961 (0.0911) (0.0911) (0.0911) (0.0906) (0.0911) LEV 0.0611 0.0611 0.0614 0.0619 0.0618 (0.0397) (0.0397) (0.0388) (0.0386) (0.0388) M/B -0.00318 -0.00318 -0.00315 -0.00300 -0.00301 (0.00245) (0.00245) (0.00244) (0.00241) (0.00241) SIZE -0.0287** -0.0287** -0.0285** -0.0283** -0.0283** (0.0117) (0.0117) (0.0116) (0.0115) (0.0115) LOSS 0.0414** 0.0414** 0.0403** 0.0401** 0.0401** (0.0167) (0.0167) (0.0167) (0.0167) (0.0167) LITIG -0.0440 -0.0440 -0.0457 -0.0455 -0.0456 (0.0395) (0.0395) (0.0398) (0.0399) (0.0399) AFMR 5.25e-05 0.0219 0.0301 0.0303 (0.0112) (0.0201) (0.0210) (0.0212) NEWAR -0.0259 -0.00305 (0.0190) (0.0344) NEWARSC -0.0648* -0.0597 (0.0367) (0.0675) Constant 0.242** 0.242** 0.240** 0.238** 0.238** (0.0942) (0.0942) (0.0935) (0.0928) (0.0928) Observations 503 503 503 503 503 R-squared 0.380 0.380 0.382 0.383 0.383
Industry FE YES YES YES YES YES
Year FE YES YES YES YES YES
In this table the results from regression Eq. (4) as follows: Model 1, Model 2 (adding in AFMR), Model 3 (adding in NEWAR), Model 4 (adding in NEWARSC instead of NEWAR) and Model 5 (adding in both NEWAR and NEWARSC) are presented. The dependent variable in all models is ABSPADCA. The independent variables are CFO; LAGAC; LEV; M/B; SIZE; LOSS; LITIG;
AFMR; NEWAR; NEWARSC. Variables are defined in Table 2.
Coefficients estimates given in the columns. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 4
Regressions Results Unexpected Audit Fees
𝐿𝐿𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑃𝑃𝐿𝐿 = 𝛼𝛼 + 𝛽𝛽1𝑆𝑆𝐼𝐼𝑆𝑆𝐼𝐼 + 𝛽𝛽2𝐿𝐿𝑆𝑆𝐿𝐿𝐼𝐼 + 𝛽𝛽3𝑃𝑃𝑃𝑃𝛥𝛥𝑃𝑃 + 𝛽𝛽4𝑄𝑄𝐿𝐿𝐼𝐼𝑃𝑃𝑄𝑄 + 𝛽𝛽5𝑃𝑃𝐼𝐼 + 𝛽𝛽6𝛥𝛥𝑅𝑅𝐼𝐼 + 𝛽𝛽7𝐿𝐿𝑅𝑅𝛥𝛥𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸 + 𝛽𝛽8𝑅𝑅𝑃𝑃𝐼𝐼𝐸𝐸𝐼𝐼𝑅𝑅𝐸𝐸 + 𝛽𝛽9𝐿𝐿𝑅𝑅𝑆𝑆𝑆𝑆 + 𝛽𝛽10𝑃𝑃𝐼𝐼𝑆𝑆𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 + 𝛽𝛽11𝑀𝑀/𝐼𝐼 + 𝛽𝛽12𝐿𝐿𝐼𝐼𝛥𝛥𝐼𝐼𝐸𝐸 + 𝛽𝛽13𝐼𝐼𝐼𝐼𝐸𝐸𝐸𝐸 + 𝛽𝛽14𝛥𝛥𝑅𝑅𝛥𝛥𝑃𝑃 + 𝛽𝛽15𝑌𝑌𝐼𝐼 + 𝛽𝛽16𝑃𝑃𝐿𝐿𝑀𝑀𝛥𝛥 + 𝛽𝛽17𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥 + 𝛽𝛽18𝐸𝐸𝐼𝐼𝑁𝑁𝑃𝑃𝛥𝛥𝑆𝑆𝑃𝑃 + 𝜀𝜀
Panel A Panel B Panel C Panel D Panel E
VARIABLES Model 1 Model 2 Model 3 Model 4 Model 5
SIZE 0.0944* 0.0944* 0.0941* 0.0928* 0.0926* (0.0493) (0.0493) (0.0489) (0.0485) (0.0486) LSUB -0.0678* -0.0678* -0.0668* -0.0656* -0.0656* (0.0355) (0.0355) (0.0354) (0.0352) (0.0353) CATA -1.268*** -1.268*** -1.284*** -1.309*** -1.311*** (0.392) (0.392) (0.389) (0.389) (0.390) QUICK -0.0377 -0.0377 -0.0350 -0.0335 -0.0337 (0.0618) (0.0618) (0.0619) (0.0614) (0.0614) DE -1.357** -1.357** -1.306** -1.315** -1.327** (0.547) (0.547) (0.554) (0.545) (0.549) ROI -1.007*** -1.007*** -0.974*** -0.952*** -0.955*** (0.365) (0.365) (0.365) (0.362) (0.362) FOREIGN -0.348 -0.348 -0.333 -0.428 -0.451 (1.681) (1.681) (1.669) (1.668) (1.689) OPINION -0.00719 -0.00719 0.0122 0.0500 0.0539 (0.255) (0.255) (0.255) (0.258) (0.258) LOSS -0.0896 -0.0896 -0.0818 -0.0787 -0.0797 (0.110) (0.110) (0.109) (0.109) (0.109) ABSPADCA 0.313 0.313 0.346 0.362 0.359 (0.472) (0.472) (0.469) (0.464) (0.466) M/B 0.0359* 0.0359* 0.0362* 0.0355* 0.0353* (0.0190) (0.0190) (0.0189) (0.0189) (0.0189) LITIG -0.318 -0.318 -0.302 -0.301 -0.303 (0.242) (0.242) (0.240) (0.237) (0.238) BIGN 0.214 0.214 0.198 0.191 0.193 (0.173) (0.173) (0.173) (0.173) (0.174) ROTA -0.181 -0.181 -0.176 -0.169 -0.169 (0.123) (0.123) (0.123) (0.122) (0.122) YE -0.334 -0.334 -0.321 -0.306 -0.306 (0.241) (0.241) (0.247) (0.249) (0.249) AFMR -0.294** -0.459** -0.593*** -0.587*** (0.144) (0.183) (0.217) (0.216) NEWAR 0.187 -0.0794 (0.118) (0.156) NEWARSC 0.622* 0.755* (0.317) (0.446) Constant 1.771*** 1.771*** 1.774*** 1.790*** 1.793*** (0.478) (0.478) (0.474) (0.471) (0.473) Observations 502 502 502 502 502 R-squared 0.404 0.404 0.406 0.409 0.409
Industry FE YES YES YES YES YES
In this table the results from regression Eq. (5) as follows: Model 1, Model 2 (adding in AFMR), Model 3 (adding in NEWAR), Model 4 (adding in NEWARSC instead of NEWAR) and Model 5 (adding in both NEWAR and NEWARSC) are presented. The dependent variable in all models is UnexpAF. The independent variables are SIZE; LSUB; CATA; QUICK; DE; ROI; FOREIGN; OPINION; LOSS; ABSPADCA; M/B; LITIG; BIGN; ROTA; YE; AFMR; NEWAR; NEWARSC. Variables are defined in Table 2.
Coefficients estimates given in the columns. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Other interesting coefficients are the coefficient ROTA which has the expected sign (negative) although also insignificant (p-value = 0.169 in both model 4 and 5). The coefficient for AFMR is negative and significant (p-value = 0.007 in model 4 and p-value = 0.008 in model 5). This, in combination with the coefficients for the respective years 2011, 2012 and 2013 (not tabulated) which are also negative and are respectively insignificant for 2011 (p-value = 0.622) and 2012 (p-value = 0.139) and significant for 2013 (p-value = 0.025) shows a decline in unexpected audit fees over the years.
Additional Sensitivity Tests
Additional untabulated tests of Eq. (5) for total audit fees, as a proxy of audit quality according to (Yasin & Nelson, 2012), instead of unexpected audit fees yield fairly similar results as in Table 4. The most notable difference is the fact that ROTA is still negative but now significant (p-value = 0.032) when taken into relation with total audit fees instead of unexpected audit fees. In the meantime NEWAR, while still positive, loses significance and is now strongly insignificant (p-value = 0.284). AFMR is still negative and significant although less strong (p-value = 0.089) but the other year dummies also lose their significance.
The additional test tests of Eq. (7) for total audit fees instead of unexpected audit fees resulted in a loss of the significance for the variable NEWARSC, leaving it marginally insignificant in both model 4 (p-value = 0.130) and model 5 (p-value = 0.145). ROTA is in these models still significantly negative (p-value = 0.036). AFMR is also still negative and significant in this alternate specification (p-value = 0.028 for model 4 and p-value = 0.031 for model 5).
V. CONCLUSIONS AND DISCUSSION
This research provides evidence with regard to the association between audit quality, as proxied by performance adjusted discretionary accruals and unexpected audit fees, and the new auditor’s report. This relation is explored by looking at the new auditor’s report in two ways. Firstly as a dichotomous variable recognizing whether or not a new auditor’s report is provided with the audited financial statements, and secondly by taking into account the quality of these new auditor’s report. I find a positive and significant relation between the quality of the new auditor’s report and unexpected audit fees, while the relation between the new auditor’s report by itself and unexpected audit fees is also positive but marginally insignificant. This points at an audit quality increasing effect of implementing the new auditor’s report which is enhanced by higher quality audit reports in line with my hypotheses.
These results are supported by the discretionary accruals tests which present a negative association between both the new auditor’s report and the quality of the new auditor’s report and performance adjusted discretionary accruals. In these tests only the quality of the new auditor’s report is marginally significant in the model excluding the dichotomous variable new auditor’s report. The direction of the results thus supports the hypothesis that the new auditor’s report, and high quality versions of it more so, are accompanied by higher audit quality. The significance however, is weak to say at best. This also becomes evident when testing the robustness of the unexpected audit fee results by looking at total audit fees instead of unexpected audit fees. These results suggest the same directions as the main test, but the quality of the new auditor’s report is in these tests also insignificant.
My combined results support the postulation that the new auditor’s report closes, at least to some degree, the gaps identified by communication theory: “the expectation gap; the information gap; and the communication gap,” (Hronsky, 1998; Maijoor, et al., 2002; Mock, et al., 2013) thereby increasing audit quality. Since desired but previously undisclosed audit information as identified by research (e.g., Church, et al., 2008; Knechel, et al., 2013; Mock, et al., 2013) is now being disclosed it can be stated that audit reporting usefulness is increased. The results then further support Knechel et al. (2013) who state that an increase in audit reporting usefulness can be used as an indicated driver for audit quality, since I find a correlation between established indicators of audit quality and the new auditor’s report and especially new auditor’s report quality. The outcomes of this study also support the belief of Limperg (1932) that with higher invoked expectations, through more detailed description of previously unavailable audit information, higher quality is to be delivered. Since the results provide stronger and more significant results for the new-auditor’s-report-quality-variable than for the mere new-auditor’s-report-variable it can be stated that auditors who raise expectations by giving a higher quality new auditor’s report tend to live up to these expectations by delivering higher audit quality as well.
Interesting are the results from the ‘AFM report’ and ‘year’ dummies in the unexpected audit fee tests. The consecutive years 2011-2014 are all negatively correlated with unexpected audit fees, with the last two years being strongly significant. This means that unexpected audit fees, and with that auditor effort according to the theory (Eshleman & Guo, 2014), have decreased over the years. Herein lies also the most apparent reason why the AFM report is negatively related to unexpected audit fees. Because of limited data after the publication of the AFM report (2014) this variable equals the year dummy variable for the year 2014. It thus seems that the critical report did not inspire any change in auditor effort in the first year of audit assignments following it. The positive association between new auditor’s report and audit quality seems to cancel out the negative effect of the developments over time. This also becomes apparent when looking at model 2 and 3 in the discretionary accruals tests, where the tiny positive correlation of the AFM report on discretionary accruals is highly insignificant when taken without the new auditor’s report and rather more positive and significantly stronger when taken with the new auditor’s report. This lead to the conclusion that the AFM report has negative impact on audit quality in model 3 because the new auditor’s report better captures the positive effects in 2014 while the AFM report captures both positive and negative effects in 2014.
While my study contributes to the debate whether or not a more extensive auditor’s report needs to be mandatorily implemented and the results suggest that implementing it can lead to an
