Master thesis Final draft
Are the Big Four audit firms homogeneous? Evidence from Europe
Name: Yvonne Meijer Student number: 0605158 Date: July 2009 First supervisor: Dr. I. Goncharov Second supervisor: Dr. G. Georgakopoulos Universiteit van Amsterdam Amsterdam Business School Accountancy Section MSc in Accountancy & Control, Accountancy track
Abstract This thesis examines whether the assumption is valid that the Big Four audit firms are homogeneous in audit quality, both at a firm-level, as well as at a country-level. More specifically, the Big Four audit firms are compared on the extent to which their auditees exercise discretion in reporting earnings. In addition, the influence of the investor protection environment on audit quality is examined. The results provide evidence that the Big Four auditors are heterogeneous in audit quality, as E&Y is the only auditor performing better than the non-Big Four auditors. The industry analysis has not proved to be a uniform factor providing incremental audit quality. Finally, Big Four auditees have a lower magnitude of accruals in a strong investor protection regime, compared to a weak protection regime, indicating that the legal environment is dominant over Big Four audit quality.
Table of contents 1. Introduction
1
2. Audit quality
2
2.1. Big Four versus non-Big Four
5
2.2. Moving beyond the Big Four/ non-Big Four distinction
6
3. Link between audit quality and earnings quality
10
3.1. Conservatism and timeliness
11
3.2. Value relevance
14
3.3. Earnings management
16
4. Legal and institutional environment shaping audit quality
20
4.1. Audit environment
20
4.2. Investor protection regime
22
5. Research design
24
5.1. Hypotheses
24
5.2. Research method
27
5.3. Data
31
6. Empirical results
34
7. Conclusions, implications and limitations
39
References
41
Appendix A
45
1. Introduction In the academic literature, much research in audit quality has focused on the comparison between Big Four auditors and non-Big Four auditors. Thereby adding the observations from all Big Four firms together (Becker et al., 1998; Dechow, Sloan, and Sweeney, 1996; Maijoor and Vanstraelen, 2006). In this thesis, I argue that this dichotomy might be invalid if there are audit quality differences among the Big Four. Therefore, in this thesis I examine whether there are audit quality differences among the Big Four, responding to a call from Fuerman (2004). Next, I will also examine the influence of the investor protection regime on audit quality. A metric of earnings management is used, specifically the ratio of the absolute value of firms’ accruals scaled by the absolute value of firms’ cash from operations, due to the requirement that it allows for a firm-level examination, and its direct relation to audit quality. This thesis documents significant audit quality differences between the Big Four and non-Big Four audit firms, in the benefit of the Big Four auditors. However, this result is driven by E&Y, which is associated with a lower level of accruals than Deloitte, KPMG, and PwC. The coefficients at Deloitte, KPMG, and PwC are not significant enough to suggest that they are of higher audit quality than the non-Big Four auditors, which is an interesting finding. The industry analysis indicates that industry specialization does not provide Deloitte, E&Y, and PwC with incremental audit quality. In contrast, the audit quality of KPMG is enhanced by its specialist’s knowledge in construction and wholesale. However, the possibility arises that an entity chooses to engage a particular auditor for its specialist’s knowledge, which would mean that the auditor’s industry specialization is already captured in the auditor dummy variable, rather than appear in the interaction term. Consistent with prior research, this thesis documents that earnings are of higher quality in countries with stronger investor protection regimes; however, this holds only for Big Four auditees. This finding suggests that Big Four audit firms are concerned with the risk of litigation in a strong investor protection environment, and perform their audit more thorough than in a weak investor protection regime. The effect of the strong investor protection regime is approximately the same across all Big Four auditors, which indicates that the limits placed on the magnitude of accruals by Big Four audit firms are not uniform across countries.
1
This thesis contributes to existing literature through the following ways. First, this study is the first research to date that has examined the audit quality differences among the Big Four. Previous literature had examined audit quality differences among the Big Six (Fuerman, 2004) or only compared the individual Big Four audit quality to the average Big Four audit quality (Tilis, 2005), while lacking to compare Big Four audit firms to their other Big Four competitors. Second, this research has made a methodological contribution by showing that pooling Big Four audit firms might not be a valid approach in studies using an indicator variable for auditor firm size, since some Big Four audit firms are not found to perform a significantly better audit than non-Big Four auditors, which makes them heterogeneous. The remainder of this thesis is organized into six sections. In the next section, I will take an in-depth review of the audit quality literature. In Section three, the earnings quality literature is linked to audit quality. The effect of the legal and institutional environment on audit quality is examined in Section four. Specifically, the audit environment and investor protection regime are examined. Section five outlines the research design to be used in the empirical analyses. Empirical tests and results are presented in Section six. Section seven concludes.
2. Audit quality Auditors are a valuable source of mitigating information asymmetries between managers and firm stakeholders by verifying the validity and reliability of financial statements (Becker et al., 1998). The perceived quality of the audit affects users of financial statements, because it influences the credibility that external parties place on the financial information flowing from the reporting-on entity, which, in turn, influences the decisions of external parties with regard to investing, lending, and underwriting. Audit quality also influences the management of the reporting entity, board members, and audit committee members, because their beliefs concerning the audit quality of a given CPA firm influences decisions with regard to a potential engagement of that CPA firm to perform audits, and how much fee to grand them (Fuerman, 2004). The effectiveness of auditing varies with the quality of the auditor, which depends on auditors’ propensity to detect and report material omissions or misstatements in the clients’ financial statements (DeAngelo, 1981). High-quality auditors are more likely 2
to detect material misstatements, and to constrain the use of questionable accounting practices by auditees than low-quality auditors. However, there are difficulties to determine whether an auditor can be classified as a low-quality auditor or a highquality auditor, because the only observable outcomes of the audit process are the audit report and the audited financial statements. In some cases, it is evident to determine audit quality, e.g. with audit failures, because audit failures are inversely related to audit quality; the higher the failure rate, the lower the quality of the auditor. However, audit failures are relatively infrequent occurrences, so there are other factors determining the conceived audit quality (Francis, 2004; Becker et al., 1998). An important determinant of audit quality is auditor independence. In the academic literature, audit quality and auditor independence are often used interchangeably, due to their positive association; the higher the independence state of the auditor, the higher the quality of the audit. The value of the audit depends heavily on the public’s perception of the independence of auditors, because the public’s reliance on the audit report will be damaged when they perceive that the auditor does not have an unbiased viewpoint. Hence, auditors need to be independent in fact (unbiased attitude towards the audit), but must also be independent in appearance (others’ interpretations of the auditors’ independence), because the latter determines for a large part the value of the audit function. Auditor independence can be undermined by the audit firm’s economic dependence on the client, which reduces the auditor’s willingness to mitigate client-induced biases in financial statements. However, auditor independence is not directly observable, as this is a mental state. Therefore, other proxies of audit quality have been optioned, which are alleged to suggest that audit independence or audit quality is affected (Arens, Elder and Beasley, 2008; Khurana and Raman, 2006). First, audit quality is partly determined by the knowledge and practical experience of the auditor. All auditors must meet minimum legal and professional requirements to receive the right to practice as an auditor. In addition, they must continue professional education during their career as an auditor to maintain this right. However, beyond these minimum requirements, high-quality auditors are more likely to follow extra courses to perform a higher quality audit. In addition, high-quality audits are more likely to be performed by auditors who are experienced in their audit clients’ industries, because it enables them to make more accurate judgments (Arens, Elder and Beasley, 2008; Francis, 2004).
3
Second, a research stream in accounting examines the claim that audit tenure, which is the length of the relationship between a company and an audit firm, affects audit quality. Proponents of mandatory auditor rotation argue that a longer audit tenure is detrimental to audit independence. The underlying reason suggests that as the audit tenure lengthens, the economic bond between the auditor and the client becomes stronger which will impair independence (Frankel, Johnson and Nelson, 2002; Myers, Myers and Omer, 2003). The counter-argument is that the auditor has strong economic incentives to remain independent and internal procedures are sufficient to maintain the independent mental attitude of auditors, e.g. rotation of engagement personnel. In addition, opponents of mandatory auditor rotation even suggest that audit quality is enhanced with extended audit tenure, because auditors will gain more experience and a deeper knowledge of the client than a new auditor possesses (Francis, 2004). Third, it is argued that the relative size of non-audit fees to total fees is damaging to audit quality. It is alleged that the increasing provision of non-audit services to audit clients may strengthen the economic bond with the client, thereby impairing independence in fact and in appearance as the auditor becomes more financially dependent on their clients. In addition, it is suggested that the consulting nature of many non-audit services puts the auditor in a managerial role, which can be threatening the audit as the risk increases that the auditor will have to perform selfcontrolling activities (Dee, Lulseged and Nowlin, 2002; Frankel, Johnson and Nelson, 2002; DeFond, Raghunandan and Subramanyam, 2002; Ruddock, Taylor and Taylor, 2006; Francis, 2004; 2006). This has motivated the Securities and Exchange Commission (SEC) to issue revised auditor independence rules consistent with the Sarbanes-Oxley Act of 2002. The revised rules require entities to disclose in their financial statements the fees billed by the auditor in the last two fiscal years, with the fees broken down into four categories; audit fees, tax fees, audit-related fees, and all other fees. The SEC has also restricted the number of non-audit services that auditors are allowed to provide for audit clients (SEC, 2004). Thus, the practical response suggests that the provision of non-audit services to audit clients significantly impairs audit quality. In the following subsection, literature is examined on the alleged differences between the often used Big Four versus non-Big Four dichotomy in the literature. Empirical differences will be discussed in Chapter 3 on earnings quality. 4
2.1. Big Four versus non-Big Four The Big Four consists currently of Deloitte, Ernst & Young (E&Y), KPMG, and PricewaterhouseCoopers (PwC). The number of Big auditors had been eight prior to 1989; however, due to mergers and the collapse of Arthur Andersen, the former Big Eight has shrunk to the Big Four (Tilis, 2005)1. As mentioned in the previous paragraph, accounting literature distinguishes between Big Four audit firms and non-Big Four audit firms. This is due to the documented association between auditor size and audit quality. DeAngelo (1981) has shown that larger audit firms have less incentive to behave opportunistically, and thus they are perceived as being of higher quality. More specifically, it is argued that Big Four auditors face a higher risk of damaging their brand name and reputation when they fail to discover and report breaches in the client’s accounting system or when they issue a wrong audit report (DeAngelo, 1981). The potential loss from lawsuits against auditors can be significant, as suggested in the insurance hypothesis or ‘deep pocket theory’2, which incents Big Four auditors to perform a more thorough audit and be more independent than their non-Big Four counterparts (Arens, Elder and Beasley, 2008). In addition, it is argued that bigger audit firms are more likely to be independent of their clients, because no single client is driving their earnings. This can act as an effective catalyst to allow less discretion to managers, and the ability to detect questionable accounting practices, and when detected, object to their use, in comparison to non-Big Four auditors. In contrast, smaller audit firms with only a small number of large clients have more to lose due to the strong economic bond created between them, e.g. if the client terminates the audit engagement, smaller audit firms lose a significant part of their earnings, which creates incentives to go along with client reporting. This will be reinforced when audit tenure lengthens; the economic bond between the client and the audit firm becomes stronger, which makes it more likely that smaller audit firms are less independent in order to retain their client, compared to Big auditors (Becker et al., 1998; Francis, 2004; DeAngelo, 1981). Next, Big-Four auditors are paid a higher audit fee than their non-Big Four 1
The literature studies mentioned in this review are performed in auditor eras when the dominant
auditor group was the Big Eight, Big Six, Big Five or Big Four. In this thesis, I use the term Big Four to refer to all these groupings for convenience purposes. 2
The deep pocket theory implies that due to auditors’ “deep pockets” relative to an (almost) bankrupt
corporation that cannot pay, investors turn to auditors in the case of a financial loss or business failure.
5
competitors, which is alleged to be a proxy for audit quality (more audit effort); however, it can also be explained by the insurance hypothesis (Fuerman, 2004). Finally, Big Four auditors have more resources and tools available to train their auditors and to perform a qualitative better audit, which is expected to enhance audit quality. Thus, due to the mentioned reasons, auditor firm size is often used as an indicator variable to distinguish between Big Four and non-Big Four in studies on audit quality. The obvious exception to this association is Arthur Andersen, which collapsed after the Enron, Sunbeam and Waste Management controversies (Fuerman, 2004). However, using the Big Four versus non-Big Four as a proxy for audit quality implicitly assumes that the Big Four audit firms are homogeneous, and that they have a uniform level of independence with respect to each client (Dee, Lulseged and Nowlin, 2002). In Chapter 3, the empirical evidence is presented about the audit quality differences between the Big Four and non-Big Four. Although most studies point into one direction, namely documenting that the Big Four is of higher quality than non-Big Four auditors, there are also a number of studies presenting mixed evidence. The mixed evidence suggests that the Big Four auditors provide not a uniform level of audit quality and that it is possible that there are audit quality differences among the Big Four, which is offset in research between the Big Four and non-Big Four. In the following paragraph, the theoretical differences among Big Four auditors are examined in-depth.
2.2. Moving beyond the Big Four/ non-Big Four distinction As mentioned in the preceding section, accounting literature implicitly assumes that the Big Four audit firms are homogeneous in audit quality. In this thesis, it is examined whether this dichotomy between Big Four and non-Big Four is justified, or whether there are differences with regard to audit quality, which makes pooling Big Four audit firms together invalid. This section reviews the literature on the differences among the Big Four auditors. Francis (2004) provides a useful start for examining differences among the Big Four auditors. Three potential sources of differentiation have been optioned: industry specialization, differences across individual practice offices (cross-city differences), and cross-country differences. The first two sources are dealt with in this chapter; the 6
third will be discussed in Chapter 4. In addition, the audit quality factors discussed earlier in this chapter will be applied in order to explain differences among auditors. First, differences in audit quality among Big Four auditors are alleged to be due to industry specialization. In most countries, the industry market shares are unevenly distributed among the Big Four audit firms, e.g. in the Netherlands, most financials are audited by KPMG and E&Y. The linkage between industry specialization and audit quality is as follows: industry specialists have more expertise and experience in their audit clients’ industries than nonindustry specialists which enables them to offer a qualitative better audit due to the more accurate judgments that can be made (Francis, 2004; Carcello and Nagy, 2004). This might be the reason why most audit firms have organized their audit practices along industry lines and why the Big Four auditors promote themselves by referring to their alleged industry expertise. This is apparent in the following extracts from the Big Four websites:
“Our professionals focus on targeted industries and develop in-depth understanding of specific business sectors and the technologies and disciplines that drive them. Multifunctional teams representing each of our service areas create client specific solutions that add value across an organization”. (Deloitte, 2009).
(…). “That’s why we’ve invested in dedicated Global Industry Centers around the world — centers that serve as virtual hubs for sharing industry-focused knowledge and experience. Our commitment of time and resources means that we can anticipate market trends, identify implications and develop clear points of view on relevant industry issues”. (Ernst & Young, 2009).
“An intense emphasis on discrete industry groups provides KPMG firms with important business insights that are reflected in our work with clients. It equips us with an understanding of the challenges facing clients and the competitive and regulatory environments in which they operate. It enhances the relevance and quality of client communication”. (KPMG, 2009).
“We organise around industries to share the latest research and points of view on emerging industry trends, develop industry-specific performance benchmarks based upon global best practices, and share methodologies and approaches in complex areas such as financial instruments and tax provisioning. (…). We invest significant resources in acquiring, refining and sharing these capabilities to further benefit our clients”. (PWC, 2009).
7
Hence, industry specialization seems to be the strategy to achieve competitive advantage over the other Big Four offices. The empirical literature generally confirms the notion that industry specialization enhances audit quality. Ohwoso, Messier and Lynch (2002) find that industry specialists outperform non-industry specialists in detecting errors. Carcello and Nagy (2004) build upon this research by documenting a significant negative relation between industry specialization and financial fraud. This result is even stronger for clients of Big Four auditors. Consistent with these findings, Balsam, Krishnan and Yang (2003) find evidence that clients of industry specialists have lower levels of discretionary accruals and higher earnings response coefficients (ERCs) than clients of non-industry specialized auditors. The ERC measures the change in market return given the change in accounting earnings. Hence, this result indicates that industry specialized auditors provide higher earnings quality and likely a higher quality audit. Taken together, these results suggest that the empirical results from the current research can be influenced by the industry specialization of the Big Four auditors, which must be taken into account when drawing inferences from the results. Second, accounting literature usually examines the Big Four firms as a whole; however Francis (2004) has argued that it might be more insightful to examine specific offices of an audit firm. This is explained by the argument that large clients create for a Big Four as a whole less financial independence than for individual practice offices of this audit firm, because relatively larger clients are no significant source of firm-wide revenue; however, for a specific office a larger client can represent a significant part of the office’s revenue. The empirical results do not suggest that Big Four auditors treat larger clients in their offices more favourably than smaller clients (Reynolds and Francis, 2001). In addition to industry specialization and cross-city differences, other sources which possibly cause differences among Big Four auditors are size of the audit firm and the relative size of non-audit fees. Appendix A provides descriptive statistics about the Big Four auditors about the year 2007, in order to compare the Big Four firms in terms of size and the relative size of non-audit fees to total fees. The data presented in panel A suggest that there are differences in size among the Big Four audit firms, in terms of global revenues. However, the size differences between PwC and KPMG, the audit firms with the largest and smallest revenues respectively, are
8
not that large to indicate any significant weakened audit quality on the side of KPMG; however, this minor size effect can have an incremental effect on the results. Next, the relative size of non-audit fees to total fees can cause audit quality differences among the Big Four audit firms, as this might not be the same across the Big Four auditors. In panel B, it is documented that the proportion of audit fees to total fees vary within a range of 47 to 57 percent, with E&Y and PwC being the only Big Four auditors in 2007 gaining more than 50% of its global fees from its audit services. Deloitte and KPMG gain both far less relative global revenues from audit services than E&Y. This suggests that the provided audit quality of both Deloitte and KPMG can be affected by the large proportion of non-audit fees to total fees, which, in turn, suggests no homogeneity among the Big Four auditors (Deloitte, 2009; E&Y, 2009; KPMG, 2009; PwC, 2009). From the above discussion, it appears that the Big Four audit firms are not homogeneous. However, there have not been many studies to date that actually examine the differentiation among Big Four auditors. Fuerman (2004) provides a first examination of tests of audit quality between each of the Big Four auditors and nonBig Four auditors, using outcomes of auditors in civil and criminal litigation and administrative proceedings. By looking at these private securities class actions, Fuerman deals only with the most extreme cases, which could potentially bias the results. The results show that Deloitte, E&Y and KPMG are all of higher quality than non-Big Six auditors. However, Fuerman’s research is performed in the Big Six era, which means that the audit quality of PwC is split up into their predecessors Coopers & Lybrand and Price Waterhouse. Although both auditors show a higher quality than non-Big Six auditors, nothing can be said about the audit quality of PricewaterhouseCoopers after the merger. Therefore, Fuerman calls for research after the audit quality of the Big Four audit firms. The current research responds to this call by providing a comparison between the Big Four auditors. In this chapter, the determinants of audit quality were listed, together with the alleged differences among Big Four auditors and the reasons from prior literature to contrast Big Four to non-Big Four auditors. In the next chapter, the empirical literature is examined on the association between audit quality, measured by auditor size, and earnings quality, the proxy to be used to make inferences about audit quality.
9
3. Link between audit quality and earnings quality As argued in the previous chapter, in order to assure users of the veracity of financial statements, managers hire auditors to opine on the reliability of the financial statements. The effectiveness of the audit is highly dependent of the quality of the auditor. However, because audit quality is unobservable, a direct examination is impossible. Therefore, the audit process to come to high audit quality is treated as a black box, and only the output of that black box is investigated, see Figure 1. As can be seen, the outcomes of the audit process are the audit report and the audited financial statements. An audit report provides only marginally an indication about the audit quality due to the limited content of the rather general template. On contrary, the audited financial statements do provide an indication about the quality of the audit process. Therefore, the current research asserts that auditees’ earnings quality can be used to draw inferences about audit quality. This is motivated by the notion that when audit quality is high, auditors constrain client’s management in their extreme reporting decisions to present the financial position of the firm, constrain management in applying aggressive assumptions, and encourage management to be consistent across periods, leading to higher earnings quality (Myers, Myers and Omer, 2003; Becker et al., 1998; Ruddock, Taylor and Taylor, 2006). In most studies on audit quality, proxies of earnings quality are used as the dependent variable, and often an indicator variable is used to distinguish between Big Four auditors and non-Big Four auditors to find whether the accounting numbers under a Big Four auditor are of significant higher quality. This section will follow these studies and discusses the concept of earnings quality in relation to audit quality. More specific, this section focuses on audit quality differences between auditors at a firm-level, rather than at a country-level, which will be discussed in Chapter 4. Earnings quality can be decomposed into several components which are alleged to increase the quality of earnings; conservatism, timeliness and value relevance. In addition, earnings quality is inversely related to earnings management, which includes earnings smoothing and the level of discretionary accruals. In the accounting literature, the terms earnings quality and earnings management are used interchangeably. This thesis follows prior research that interprets earnings as being of higher quality when they exhibit less earnings management (Barth, Landsman and Lang, 2007; Barth et al., 2006).
10
Timeliness
Conservatism
Value relevance
Earnings management
Earnings quality
Audited financial statements
Audit report
Legal and institutional environment
Relation non-audit fees/total fees
Audit quality
Audit tenure
Knowledge and experience
Auditor independence
Auditor independence
Figure 1: examination of audit quality.
3.1. Conservatism and timeliness This section deals with the concepts conservatism and timeliness, as these concepts are closely related. Timeliness is defined as the extent to which earnings reflect the variance in returns contemporaneously (Basu, 1997). Conservatism, also called asymmetric timeliness, is defined in the framework by Basu (1997, p. 4) as follows:
“Capturing accountants' tendency to require a higher degree of verification for recognizing good news than bad news in financial statements”. Thus, earnings reflect bad news more quickly than good news, which causes differences in the timeliness of earnings in bad news periods and good news periods (Basu, 1997). Both concepts are viewed as properties of high quality earnings, because they improve the transparency of financial statements. Further, conservative
11
earnings have more useful characteristics. First, it improves contracting efficiency. Conservative accounting arises in debt covenants, compensation and other contracts to efficiently address moral hazard problems caused by parties having asymmetric information to the entity. These contracts often require more verification for recognizing gains than losses in order to constrain managerial opportunistic payments to themselves and shareholders. Secondly, it is argued that litigation produces asymmetric payoffs with regard to the overstatement or understatement of assets. Overstating the firm’s assets is more likely to generate litigation costs for the firm than understating assets. Hence, managers have more incentives to report conservative earnings to reduce litigation costs. Third, conservative accounting enables managers of profitable firms to defer tax payments. Managers can recognize expenses immediately, while delaying recognition of revenues, which reduces the present value of taxes and increases the current value of the firm. Finally, standard setters and regulators favor the use of conservative accounting, because they face more criticism if firms overstate net assets than if they understate net assets, which cause asymmetric definitions in the accounting standards (Ball, Kothari and Robin, 2000; Jenkins and Velury, 2008; Watts, 2003). For example, IFRS standard IAS 37 reports that “an entity should not recognise a contingent liability. An entity should disclose a contingent liability, unless the possibility of an outflow of resources embodying economic benefits is remote”. While the same standard also reports that “an entity shall not recognise a contingent asset. However, when the realisation of income is virtually certain, then the related asset is not a contingent asset and its recognition is appropriate” (emphasis added; IASB, 2009). From these definitions, it is shown that the IASB applies a stricter test for the recognition of contingent assets than for the recognition of contingent liabilities, which can be due to concerns regarding overstatements of assets or political pressure from lobby groups.
Conservatism and timeliness are directly linked to audit quality. As argued in Section 4.2, auditors will face more litigation risk if outside parties to the firm have incurred losses due to overstatement of assets rather than forgone gains as a result of understated assets. Hence, high-quality auditors are more likely to apply a lower threshold for overstatements than for understatements (Watts, 2003). This implies that if Big Four auditors are of higher quality than non-Big Four auditors, then Big Four auditees should report more conservative earnings than clients of non-Big Four
12
auditors, which is generally confirmed in the academic literature (Becker et al., 1998; Francis and Wang, 2008). Francis and Krishnan (1999) examined whether high-accrual firms are more likely to receive modified audit reports compared to low-accruals firms by auditors. Highaccrual firms are classified as such on the basis of the magnitude of accumulated accruals. A high level of accruals, either positive or negative, leads to an increased divergence between cash flows and reported earnings, and is therefore not conservative. However, a consistent predominance of negative accruals over time, when net income falls below cash flow from operations, is an indication of accounting conservatism (Givoly and Hayn, 2000). The results show that auditors are more likely to issue modified audit reports to high-accrual firms than to low-accrual firms, which suggests that auditors report more conservative with respect to high accrual firms. Additional tests show that Big Four auditors are consistently conservative with respect to financial reporting of high-accruals firms, while non-Big Four auditors show no such effect. Given that 85% of the observations were audited by Big Four auditors, this article documents that Big Four auditors apply a lower threshold for issuing a modified report than non-Big Four auditors to compensate for the uncertainty in high accrual situations, which adds to the belief that Big Four auditors are of higher audit quality than non-Big Four auditors (Francis and Krishnan, 1999). Hence, the direct evidence is Big Four auditors report more conservative than non-Big Four auditors with respect to high-accrual firms, which suggests higher audit quality on the side of Big Four auditors. In addition, the audit quality factors mentioned in Chapter 2 can be applied to indirectly examine the relation between Big Four auditors and reporting conservatism. Jenkins and Velury (2008) documented the association between audit tenure and the reporting of conservative earnings. The results indicate that conservatism increases between short and medium audit tenure. There is no significant difference found between medium and long audit tenure. Hence, the results imply that as conservatism does not deteriorate over long tenure, mandatory audit rotation adversely affects conservative earnings. Next, Ruddock, Taylor and Taylor (2006) examined whether the provision of nonaudit services is associated with conservatism. As argued in the previous chapter, the provision of non-audit services might compromise independence, which in turn might impair earnings conservatism. This is due to the auditor’s acceptance of more 13
aggressive accounting practices in order to continue the sale of non-audit services. The results show that the provision of non-audit services is not reducing earnings conservatism. If anything, the opposite is documented. Similar results are found when the tests are performed for the sample divided into Big Four auditors and non-Big Four auditors, which indicates that the results of the full sample are not affected by a Big Four effect. The authors also estimated the regression for each of the Big Four audit firms to investigate possible audit firm effects. The results show that the offices of the Big Four are generally conservative. Although the authors did not calculate the statistical significance of the differences in coefficients between the offices, it is shown that overall KPMG and PwC auditees show a higher coefficient (significant at 1 percent) than Deloitte and E&Y (insignificant and significant at 10 percent respectively), which implies that the former are more conservative (Ruddock, Taylor and Taylor, 2006). In this section, it was shown that both the direct and indirect evidence point into one direction, namely that Big Four auditees are more conservative than non-Big Four auditees. In addition, it was documented that audit quality differences exist among the Big Four. In the next section, the association between value relevance and audit quality is examined.
3.2. Value relevance The usefulness of accounting information can also be captured by the concept of value relevance. Value relevance is measured by the ability of financial statement information (e.g. book value and net income) to reflect the underlying economic value of the firm, measured through contemporaneous stock prices. Value relevance is also a proxy of earnings quality that can be linked to audit quality. The expected quality of reported earnings depends on the perceived quality of the auditor. If investors belief that the auditor is of high quality, then they will consider the reported earnings of the reporting entity more useful for decision making, which will increase the value relevance (Cho, Han and Brown, 2006). Following the theory of DeAngelo, if auditor size is positively related to audit quality, then it is expected that investor will value the reported earnings higher when the earnings are audited by a Big Four audit firm, than audited by a non-Big Four audit firm. Consistent with these expectations, Teoh and Wong (1993) document a relation between auditor size and the credibility that investors place on the earnings 14
reports. Specifically, Big Four auditors are associated with a higher ERC than nonBig Four auditors. From the above discussion, it appears that the accounting literature confirms the expectation that value relevance is enhanced when the earnings are audited by a Big Four auditor. In addition to the direct evidence presented about the effect of the Big Four auditors on the value relevance, the audit quality factors can be applied to find whether these affect the effect of the Big Four auditors on the usefulness of accounting information. Gul, Tsui and Dhaliwal (2006) test the effect of non-audit services on the stock return/ earnings relation. As argued in Chapter 2, the provision of non-audit services to an audit client by an auditor can increase the auditor’s economic dependence on the client firm. This implies that auditor’s independence is impaired, which could lower the perceived quality of the firm’s reported earnings by investors. The perceived bias and noise in reported earnings increase the information risk associated with them, which, in turn, reduces the value relevance of reported earnings. The results show that firms with a higher level of non-audit services fees are associated with lower ERCs. Thus, it can be concluded that the provision of nonaudit fees has a negative effect on earnings quality, and thus automatically on audit quality. Evidence is also found that the decline in the ERC for firms with increasing levels of non-audit services fees is weaker for clients of Big Four auditors than for clients of non-Big Four auditors (Gul, Tsui and Dhaliwal, 2006). These results suggest that investors belief that Big Four auditors are more likely to maintain their audit independence, which leads to better reported earnings of their auditees. They are not compromised by increasing non-audit service fees. However, Cho, Han and Brown (2006) find opposite results to the ones found by Gul, Tsui and Dhaliwal (2006). They find that the non-audit fees ratio has a positive impact on the association between stock prices and accounting numbers. Cho, Han and Brown (2006) argue that their results can be explained by the “value enhancement hypothesis”, which predicts that non-audit services improve earnings quality and thus investors will be more inclined to rely on the accounting information. Although the above results show that Big Four auditors can influence the value relevance of accounting information, the results with respect to the effect of non-audit fees on the value relevance of the stock return/ earnings relation are mixed. This provides an opportunity for future research.
15
3.3. Earnings management This section will discuss earnings management. Earnings management is a comprehensive concept. Therefore, only some key components are discussed, specifically discretionary accruals and earnings smoothing. In addition, the literature is examined on linkages between earnings management and audit quality. In general, earnings can be decomposed into cash flow from operations, nondiscretionary accruals, and discretionary accruals. Non-discretionary accruals are used for accounting adjustments to the firm’s cash flows required by accounting setting bodies, such as the IASB, FASB or SEC. In contrast, discretionary accruals are used for accounting adjustments to the firm’s cash flows selected by the manager (Healy, 1985). Because discretionary accruals are subject to judgment by managers, the earnings management literature often examines discretionary accruals rather than nondiscretionary accruals. Due to information asymmetry between managers and investors, managers have a deeper understanding of the business of the company than outsiders. Hence, they are in a better position to make judgments about how to reflect the transactions made by the company to make the financial statements more relevant to outside stakeholders, thereby using accruals (Palepu et al., 2007). This judgment can be very valuable if it is used in the interest of the entity. For example, accruals can be used to reduce noise in cash flows, which makes the earnings variable more relevant to measure firm performance on a periodic basis. It is also used to enhance timeliness and conservatism of earnings. Timely gain and loss recognition must occur around the time of a revision in expectations of future cash flows, which are likely to occur before the actual realization of the cash flows. Hence, accounting accruals are needed (Ball and Shivakumar, 2006). However, the discretion given to managers can also lead to misstated or omitted accounting data or material facts; earnings management. Generally used earnings management activities are ‘big bath accounting’, where managers understate current year’s earnings to create a turnaround in later periods, small loss avoidance, when insiders use their accounting discretion to avoid reporting small losses, or ‘income smoothing’, which occurs when managers decide to incorporate economic gains and losses gradually in accounting income over periods to reduce the variability of reported earnings. Earnings smoothing is made through the creation of reserves, by building up provisions, or by recognizing excessive depreciation of assets. This way, an entity builds up a reserve against current profit, 16
which is released in future years to income (Ball, Kothari and Robin, 2000; Palepu et al., 2007; Leuz, Nanda and Wyosocki, 2003; Van Tendeloo and Vanstraelen, 2005). Managers have incentives to engage in earnings management for contracting purposes, especially in situations where contracts between the entity and their stakeholders are based on accounting income numbers, e.g. bonus schemes and debt covenants. Bonus schemes create incentives for managers to select accounting procedures and accruals to maximize their bonus awards. Debt covenants often contain contractual obligations which require the firm to meet certain financial ratios or minimum net worth. If entities are close to violating these obligations, then managers have incentives to make accounting adjustments to reduce the probability of debt covenant violation (Healy, 1985; Palepu et al, 2007). Although these explanations are widely examined in the academic literature as being most likely to induce managers to manipulate earnings, the most common motivation for earnings manipulation for firms subject to accounting enforcement actions by the SEC is to lower the cost of external financing, while the earnings-based bonus motivation is ranked at a third position and the debt covenant motivation is not mentioned at all (Dechow, Sloan and Sweeney, 1996). The link between audit quality and earnings management is quite straightforward. The effectiveness of auditing to detect earnings management is dependent on the quality of the auditor. In other words, earnings management has an inverse relation with audit quality. High-quality auditors are more likely to detect and report on errors in the client’s financial statements, which, in turn, will mitigate earnings management as management faces reputation damages and reduced firm value when misreporting is revealed by the auditor (Becker et al., 1998). Hence, auditees of high-quality auditors will display a lower level of discretionary accruals and less earnings smoothing compared to clients of low-quality auditors. The academic literature in this area is extensive. As argued before, the empirical literature often uses the Big Four as a control variable to find whether they mitigate earnings management, measured by earnings smoothing, magnitude of discretionary accruals or accrual quality. Becker et al. (1998) examine the relation between audit quality and earnings management. They use the level of income-increasing discretionary accruals as the proxy for earnings management. The results show that non-Big Four auditors allow more income-increasing discretionary accruals than Big Four auditors. This result is 17
consistent with Big Four auditors being of higher audit quality, as they apparently allow less accounting flexibility to auditees. Tilis (2005) builds upon this research by examining differences in audit quality between Big auditors and Small auditors across separate ‘auditor eras’. Each auditor era reflects the number of Big auditors in the industry. From the results, it appears that in all auditor eras (Big Eight, Mergers, Big Six, Big Five and Big Four) Big auditors show a higher audit quality, measured by accrual quality. Accrual quality and audit quality have an inverse relation; when accrual quality is high, audit quality is likely to be low. Further, Tilis (2005) examined differences among Big Four auditors. The findings suggest that during the Big Four era, there are significant differences among the Big Four auditors in audit quality. Specifically, Deloitte & Touche and PwC show better than average results in the Big Four era. In contrast, E&Y and KPMG show worse than average overall quality (Tilis, 2005). Taken together, these findings by Becker et al. (1998) and Tilis (2005) suggest that the Big Four auditors are of higher quality than non-Big Four auditors; however, by pooling Big Four auditors together, the results are biased, because the Big Four auditors provide no homogeneous audit quality. In the following paragraphs, the earnings literature is examined on the effect of the (Big Four) audit firms through the audit quality factors mentioned in Chapter 2. Dee, Lulseged, and Nowlin (2002) investigated the influence of non-audit fees on earnings quality. Due to an extreme growth in non-audit services by audit firms to audit clients, their prediction is that auditors’ independence might be impaired. The authors find that client firms paying higher proportions of non-audit fees to their external auditors have significantly higher (income-increasing) total and discretionary accruals. Consistent with Dee, Lulseged and Nowlin (2002), Frankel, Johnson and Nelson (2002) report evidence that entities that purchase more audit services have a higher level of absolute discretionary accruals and a small earnings surprise. These results suggest that the more fees the auditor derives from clients, the more likely an inherent economic dependency between the auditor and the client will be created. This may reduce auditors’ incentives to mitigate earnings management, which suggests less audit quality. In contrast, DeFond, Raghunandan and Subramanyam (2002) find no evidence that auditor independence is affected by the level of nonaudit fees as measured by the auditor’s propensity to issue a going concern opinion. This would mean that the economic bond theory is not valid, as the auditor seems to 18
be able to remain objective and withstand any pressures to issue a clean audit report. In addition, DeFond, Raghunandan and Subramanyam (2002) find that Big Four auditors are more likely to issue a going-concern opinion to a financially distressed client, than non-Big Four auditors. This suggests that the Big Four auditors are more independent than their non-Big Four counterparts. In combination with the results found in the previous sections, the association between non-audit fees and earnings quality is confusing. The findings suggest that the provision of non-audit services creates more conservatism; however, mixed evidence is found on the association with value relevance and earnings management. Hence, this association provides a opportunity for future research. Myers, Myers, and Omer (2003) examine the relation between audit tenure and earnings quality. Abnormal and absolute current accruals are used as proxies for earnings quality, with a control variable for auditor type (Big Four or non-Big Four). Myers, Myers and Omer (2003) argue that as the audit-client relationship lengthens, auditors place greater constraints on extreme management decisions (incomeincreasing and income-decreasing accruals) in the financial statements. However, there is no significant Big Four effect, placing more constraints on discretionary and current accruals than non-Big Four auditors (Myers, Myers and Omer, 2003). Johnson, Khurana and Reynolds (2002) add to this research by examining the earnings quality in three different audit tenures, short (2-3 years), medium (4-8 years) and long (9 or more years). Their results show that a short audit tenure is associated with a lower reporting quality, as measured by higher absolute levels of unexpected accruals and accruals that are less persistent in subsequent earnings, compared to medium audit tenure. Both the medium and long tenures show no significant differences in the properties of the reported accruals by clients. These empirical research findings suggest that longer audit tenure is not automatically associated with impaired auditor independence and audit quality; in fact it strengthens audit quality. These results are consistent with the results found in Section 3.1, where it was found that longer audit tenure did not deteriorate conservatism, it actually enhanced conservatism. From the empirical review, it can be seen that Big Four auditors are in most studies of higher quality than non-Big Four auditors, because they show less signals of earnings management. Hence, it is expected that non-Big Four auditees are more subject to accounting enforcement actions by the SEC for alleged violation of GAAP. 19
However, Dechow, Sloan and Sweeney (1996) show no significant difference between the SEC and control firms’ use of a Big Four auditor. This finding, in combination with the studies providing evidence that the Big Four and non-Big Four are of equal audit quality, weakens the evidence on an overall Big Four audit quality effect. Further, the results by Tilis suggest that there are indeed differences between the Big Four auditors, which imply that pooling observations across auditors together can bias findings, and shows the relevance of the current study. A limitation of Tilis’ research is that it only provides a comparison of the several Big Four auditors against the pooled Big Four auditors, while lacking to make a relative comparison of the Big Four auditors against each other, which makes the statistical differences unknown. In addition, Tilis lacks to examine what influence the audit environment has on the results. This thesis will overcome these limitations and extend Tilis’ research by examining the relative audit quality differences between the Big Four auditors both cross-sectional and cross-country.
4. Legal and institutional environment shaping audit quality This chapter deals with the effect of the legal and institutional environment on audit quality. This is important, because as argued in Section 2.2, the audit quality among the Big Four auditors can vary due to cross-country differences, which is examined in the research of Hypothesis 2. Hence, this chapter takes a more in-depth examination of the country-level variation in Europe causing audit quality differences within each Big Four audit firm, rather than looking at a firm-level examination. The most commonly used starting point for dividing countries according to their legal system is the distinction in common-law and code-law countries. However, this distinction is very broad and includes more perspectives to examine reporting incentives. In this chapter, reporting incentives will be examined via the following perspectives; audit environment and investor protection regime.
4.1. Audit environment Since in this thesis only countries in Europe are examined, one would expect that audit quality is equal across countries, due to similar accounting standards (IFRS) and auditor firms. The Big Four audit firms each use their own audit approach, audit manuals and audit programs globally, to ensure that auditors can work within any 20
team around the world. In addition, the Eight Directive, issued by the European Parliament and of the Council, aims to increase the harmonization of statutory auditing in the European Community. However, this harmonization has not been completed yet. This causes differences in the national audit environment like auditor independence regulation, which, in turn, makes that audit quality can vary across countries (EU, 2006). Hence, in this section, the influence of the audit environment on audit quality is examined. In the accounting literature, the audit environment is viewed in terms of auditor independence rules. Auditor independence rules are requirements imposed on auditors in order to ensure audit quality to the users of financial statements, and for which there are cross-country differences. Auditor independence rules includes, but is not restricted to: length of audit tenure, audit partner rotation requirement, joint audit requirement, disclosure of (non)audit fees, advertising allowed, review by regulators, and peer review (Maijoor and Vanstraelen, 2006). In a study of the effect of the member state audit environment on earnings management for three countries in Europe, Maijoor and Vanstraelen (2006) find that the magnitude of earnings management, measured by abnormal working capital accruals, is the highest in Germany, followed by the UK and France. France has a stricter audit environment in terms of independence rules and auditor liability than both Germany and the UK, which causes the increased earnings quality. In general terms, Maijoor and Vanstraelen (2006) find that a flexible audit regime is associated with a higher level of abnormal working capital accruals compared to a strict audit regime. These differences across EU member state audit environments remain significant in the presence of a Big Four auditor. Therefore, it can be concluded that the limits placed by Big Four audit firms on earnings management are not uniform across countries (Maijoor and Vanstraelen, 2006). In addition to the aforementioned auditor independence rules, cross-country variation can also be caused by size differences across countries. It can be argued that if market shares are not evenly distributed between the Big Four auditors in every country, then the probability increases that the audit quality factor size affects the results of the current thesis. Panel C of Appendix A provides the relative revenues of each Big Four auditor per region. Panel D of Appendix A makes a comparison of the market shares of the Big Four auditors per region. Taken together, it is shown that the relative size of the Big Four auditors differ significantly in the Americas. Deloitte 21
earns over 31% of the total Big Four fees in that region, while KPMG earns only 18%. In Europe, Middle East and Africa (EMA), the range between the auditors taking the largest and smallest market shares, PwC and Deloitte respectively, is not that large (29 versus 22 percent) to indicate any weakened audit quality on the side of Deloitte. This suggests that the relative size of the Big Four auditors is not significantly different in Europe; however, since not all information about the revenues per European country is available, a calculation of the market shares of the Big Four auditors per country in Europe can not be made. In the Asian region, all Big Four auditors earn about the same amount of revenues, with a small size advantage for PwC (Deloitte, 2009; E&Y, 2009; KPMG, 2009; PwC, 2009). In the next section, the effect of the investor’s protection regime on audit quality is examined.
4.2. Investor protection regime The investor protection regime is alleged to be an important institutional factor affecting reporting incentives. This is motivated by the argument that strong investor protection regimes provide outsiders with the right to discipline insiders, e.g. replacing insiders or file a claim against managers for their losses. Hence, these rights are a tool to limit insiders to manage earnings (Leuz, Nanda and Wyosocki, 2003). In addition, the investor protection regime also affects the reporting incentives of auditors. The information asymmetry between corporate insiders and outsiders makes that external auditors are relied upon to provide assurance on the reliability of the financial statements. When auditors fail to detect or report on errors in the client’s financial statements, they face litigation risk by the investors, because they want to be compensated for the loss they have incurred by relying on the audited financial statements. Litigation can be very costly, which provides an incentive for auditors to deliver high quality audits. This is especially true for Big Four auditors since they have more to lose, in terms of reputation and their perceived deep pockets (Palmrose, 1988). Hence, there is an inverse relation between audit quality and litigation rate; auditors with relatively low (high) litigation activity represent higher (lower) audit quality suppliers (Palmrose, 1988). In general, auditors are more likely to be sued in a lawsuit if financial statements are overstated rather than understated. The underlying reason suggests that if the auditor fails to attenuate income-increasing manipulations, then the financial statements may not provide an adequate warning for potential 22
financial problems and portrays an overly favorable impression. This will result in a higher litigation risk than if auditors fail to detect income-decreasing accrual manipulation (Heninger, 2001). Therefore, a strong investor protection system makes that auditors will be more inclined to assure themselves that the clients’ financial statements contain no material omissions or misstatements, leading to higher audit quality (Leuz, Nanda and Wyosocki, 2003). In a study on the litigation activity among the largest independent audit firms, Palmrose (1988) finds that Big Four auditors experience less litigation occurrences than non-Big Four auditors. Khurana and Raman (2004) document in their study on the audit quality of Big Four auditors that litigation concerns rather than reputation protection drives the perceived higher audit quality and financial reporting quality. Heninger (2001) builds upon this research by studying the relation between a firm’s income-increasing abnormal accruals and the risk of auditor litigation. He documents a significant relation between a firm’s positive abnormal accruals and auditor litigation. However, the probability of litigation risk is lower for Big Four auditors (Heninger, 2001). Consistent with their expectations, Leuz, Nanda and Wyosocki (2003) conclude that a strong investor protection regime is more effective in mitigating earnings management than a weak investor protection regime. Specifically, they find a negative relation between earnings management (measured by earnings smoothing, correlation between changes in accounting accruals and operating cash flows, magnitude of accruals, and small loss avoidance) and outsider rights and legal enforcement. This is supported by the argument that a strong investor protection limits insiders’ ability to enjoy private control benefits and hence lowers the incentives to mask firm performance. Ball, Kothari and Robin (2000) reinforce these findings by documenting that countries with stronger investor protection environments experience greater earnings conservatism. Also Francis and Wang (2006) find that earnings are of higher quality in countries with strong investor protection, as measured by abnormal accruals and the likelihood of reporting a loss. However, their results show that this only holds for Big Four auditees, which is due to differences in enforcement by Big Four and non-Big Four auditors. Firms prefer discretion to manage earnings and require auditors to allow the reporting of low quality earnings at the risk of auditor dismissal. However, in a stronger investor protection and enforcement environment, Big Four auditors are less willing to allow 23
such discretion, because the consequences of client misreporting are larger for Big Four auditors than for non-Big Four auditors, due to their greater reputation capital at risk. In contrast, countries with a weak investor protection show no differences in earnings quality between Big Four clients and non-Big Four auditees. (Francis and Wang, 2006). Maijoor and Vanstraelen (2006) extend these findings by documenting that Big Four auditors are associated with a lower absolute value of abnormal working capital accruals compared with non-Big Four auditors. However, this result is driven by the UK, which has a much stronger investor protection environment compared to France and Germany, which leads to the growth in Big Four auditor conservatism. In the weaker investor protection environments France and Germany, there are less differences between the Big Four audit firms and non-Big Four audit firms. Taken together, these results suggest that the Big Four auditors are of higher quality than their non-Big Four counterparts, because they appear less in litigation occurrences, which, in turn, is caused by their concerns about potential costly litigation. Furthermore, the investor protection regime seems to be dominant over the presence of a Big Four auditor in affecting audit quality.
5. Research design This section provides the groundwork for the empirical analysis in Chapter 6. More specifically, the theoretical basis and development of the hypotheses is explained in-depth, together with the research method to be used and the data selection.
5.1. Hypotheses As argued in the previous chapters, this thesis tests the relative audit quality of the Big Four offices compared to their Big Four competitors. To date, literature has mainly examined audit quality differences between the Big Four and non-Big Four auditors, because of size differences. However, by pooling the Big Four auditors together, these studies implicitly assume that the audit quality of the Big Four is homogeneous. The research of Tilis (2005) has documented differences among the Big Four auditors, which implies that pooling Big Four auditors together is invalid. This research will extend Tilis’ (2005) research by examining the audit quality among
24
the Big Four auditors and trying to explain the expected differences. Therefore, the following hypothesis is proposed: H1a: The relative audit quality of Big Four auditors differs among Big Four firms. In Section 2.2, some evidence was provided about the causes of audit quality variation among the Big Four auditors. More specifically, it was shown that differences among Big Four auditors could be caused by industry specialization. Industry expertise is expected to enhance the earnings quality due to the more accurate judgments that can be made. Therefore, auditors’ industry expertise should be positively related to earnings quality and its ability to mitigate reporting aggressiveness. Further, the audit quality factors can cause differences among the Big Four audit firms. It was shown in Appendix A that the size of the Big Four auditors differs in terms of revenues. However, the difference between the largest and smallest audit firm is not large enough to indicate a significant weakened audit quality on the side of the smallest Big Four auditor, KPMG. Hence, size variation among the Big Four auditors is expected only to have an incremental effect on the results, rather than the dominant factor causing audit quality differences among the Big Four audit firms. In addition, the relative size of the non-audit fees to total fees was suggested in Section 2.2 as being a cause of heterogeneity among the Big Four audit firms. It appeared that E&Y and PwC are the only Big Four auditors obtaining more than 50 percent of their global revenues from audit services in 2007. This raises the possibility that the reporting aggressiveness of KPMG and Deloitte can be affected by their possibly impaired independence. However, because the differences were not that large, it is expected to have only a minor effect on the results. Based on the above discussion, the current paper aims to provide an explanation for any observed variation among the Big Four audit firms. The current thesis expects industry specialization to be the most dominant audit quality factor causing audit quality differences among the Big Four audit firms. This leads to the following hypothesis: H1b: The observed relative audit quality differences are caused by industry specialization.
25
The main issue in this paper is the examination of audit quality among the Big Four audit firms; however, this thesis will also examine whether audit quality within each Big Four audit firm is consistent across countries, or whether audit quality within each audit firm is affected by the local environment. In Section 4.1 it was shown that the audit environment differs across countries, which affect audit quality, even in the presence of a Big Four auditor. This suggests that cross-country variation in audit environments can also influence audit quality differences of a Big Four firm across countries. However, testing the effect of the audit environment on audit quality is a very complicated task. The audit environment consists of many audit independence rules, which makes the separation between a strong audit environment and a weak environment not easily drawn. Therefore, it won’t be used in this thesis. The proxy that will be used is investor protection, uncorrelated, but rather complementary to the audit environment as showed by Maijoor and Vanstraelen (2006). Maijoor and Vanstraelen (2006) examined the influence of both the audit environment and the investor protection regime on earnings management in France, UK, and Germany. Maijoor and Vanstraelen (2006) documented that the magnitude of earnings management decreased in the presence of more audit independence rules. However, the UK, ranked second on the strictness of the audit environment, showed significant differences in audit quality between the Big Four and non-Big audit firms. This was driven by the fact that the UK has a high risk of auditor litigation. In contrast, the differences between the Big Four and non-Big Four audit firms were smaller in both France (strict audit environment, but weak audit liability) and Germany (weak audit environment and auditor liability). In Section 4.2, it was argued that the investor protection environment will influence audit quality. In a strong investor protection regime, the risk of litigation for auditors is higher when the audited financial statements contain material omissions or misstatements, compared to a weak investor protection regime. Thus, a strong investor protection regime will enhance audit quality, as auditors, especially Big Four auditors, will be more inclined to allow less reporting discretion to their clients due to a greater reputation and capital at risk. Therefore, it is important to document whether the audit quality of the Big Four auditors will be affected similarly in the transition of a strong investor protection regime to a weak regime. If this is the case, then is can be concluded that the investor protection regime is dominant over the quality of the 26
auditor. However, if this is not the case, then other audit quality factors might be causing the differences, e.g. the market shares of the Big Four audit firms can be distributed unequally across European countries. If size of the audit firm is correlated with the investor protection regime, then the results will be affected. Appendix A showed the market shares of the Big Four audit firms in three world regions. In Europe, the focus of this thesis, the market shares were distributed quite evenly among the Big Four auditors, which imply an equal audit quality. However, due to unavailable information of the revenues per European country, it is hard to make inferences about the possible audit quality differences of a Big Four audit firm across European countries, caused by unevenly distributed market shares. Since there are no reasons to expect that auditor size and investor protection regime are correlated, the above reasoning leads to the following hypothesis: H2: The level of investor protection will be dominant over the Big Four audit quality. Specifically, a strong investor protection environment will strengthen audit quality, whereas a weak investor protection regime will weaken the audit quality, relative to the pooled sample.
5.2. Research method This section describes the research model to test the hypotheses formulated in the previous section. In Chapter 3, it was shown that findings with respect to audit quality are quite similar across the metrics conservatism, value relevance, and earnings management; therefore, the choice of the model used is not likely to affect the results. The current paper will use a measure of earnings management, because it is believed that this is the most appropriate metric to use due to its strong and direct relation with audit quality. Auditors have a strong objective to mitigate earnings management both to enhance their audit quality, and to avoid costly litigation occurrences, whereas the link between audit quality and conservatism is weakened by other sources, e.g. accounting standards prescribing conservative behavior to auditees. With respect to the measure value relevance, it is expected that the usefulness of accounting information increases in the presence of Big Four auditors compared to non-Big Four auditors due to the perceived higher quality of the Big Four auditors. However, I would not expect there to be significant differences in value relevance among the Big Four auditors. Hence, using the proxy value relevance for earnings quality to make 27
inferences about audit quality among the Big Four auditors will not lead to a valid approach; therefore it won’t be used in this thesis. Another reason for using the earnings management metric is the requirement that it allows for a firm-level examination. With the metrics conservatism and value relevance this would be more difficult. Drawing on Leuz, Nanda and Wyosocki (2003), the earnings management measure used in the current paper is magnitude of accruals. This measure is appropriate, because it captures the degree to which insiders use their discretion to engage in earnings management activities. Hence, the magnitude of accruals indicates the reporting aggressiveness of entities. The measure is computed as an entity’s absolute value of accruals scaled by the absolute value of firms’ cash flow from operations. A higher score on this measure indicates more earnings management (Leuz, Nanda and Wyosocki, 2003). Consistent with previous research (Dechow, Sloan and Sweeney, 1996; Leuz, Nanda and Wyosocki, 2003), accruals are calculated as follows: Accrualsit = (∆CAit - ∆Cashit) - (∆CLit - ∆STDit - ∆TPit) - Depit
(1)
Where ∆CAit is the change in total current assets, ∆Cashit is the change in cash and cash equivalents, ∆CLit is the change in current liabilities, ∆STDit is the change in short-term debt, ∆TPit is the change in income taxes payable, and Depit is the depreciation, depletion and amortization expense for entity i in year t. As mentioned earlier, the computed accruals are scaled by cash flow from operations, obtained directly, in order to control for differences in firm size and performance (Leuz, Nanda and Wyosocki, 2003; Burgstahler, Hail, Leuz, 2006). Hence, the scaled measure forms the dependent variable in all regressions. Prior research uses auditor firm size as an indicator variable to distinguish between Big Four and non-Big Four. The evidence mostly points into one direction, namely Big Four audit firms are of higher quality than non-Big Four auditors. However, this thesis also mentioned some studies providing mixed evidence. Therefore, the following regression tests whether the Big Four audit firms are of higher quality than their non-Big Four counterparts:
28
|Acc| / |CFO| = α0 + α1BIG4it + α2YEARit + α3LEVit +
(2)
α4GROWTHit + α5SIZE + εit. BIG4 is a dummy variable that equals 1 if the entity is audited by a Big Four auditor and 0 otherwise. A significant negative coefficient on BIG4 is expected, and provides evidence that the Big Four auditors are of higher quality than non-Big Four auditors, because they allow less reporting aggressiveness. YEAR represents a vector of dummy variables to control for year effects. LEV is measured as the ratio of total liabilities to total assets. GROWTH is the revenue growth rate, defined as the revenues in year t minus revenues in year t-1 and scaled by revenues in year t-1. SIZE is the natural logarithm of total assets in year t. The regression in (2) pools the Big Four audit firms. However, Hypothesis 1a states that pooling Big Four audit firms may be misleading, due to audit quality differences among the Big Four auditors. Therefore, the Big Four dummy variable from the previous regression is split into four dummy variables for each of the Big Four audit firms. It is believed that Big Four auditors are of higher audit quality than non-Big Four auditors, which means that α1, α2, α3, and α4 are all significantly negative, and that there exists at least some significant variation among the Big Four auditors, which would mean that α1 ≠ α2 ≠ α3 ≠ α4. |Acc| / |CFO| = α0 + α1Deloitteit + α2E&Yit + α3KPMGit +
(3)
α4PwCit + α5YEARit + α6LEVit + α7GROWTHit + α8SIZE + εit. Deloitteit, E&Yit, KPMGit, and PwCit are all dummy variables, which turn 1 if the firm observation is audited by that particular auditor and 0 otherwise. As argued in the preceding section, industry specialization is expected to be the most dominant factor causing the audit quality differences as expected to be observed in (2); therefore, the industry specialization of each of the Big Four auditors is calculated based on Standard Industrial Classification (SIC)-codes. For each auditor, the two industries with the highest market shares are determined, and taken into account as being the industry specialization. If a specialist’s knowledge of the industry is better able to detect earnings management, then I expect the coefficient on the interaction terms α3, α5, α8, and α11 in regression (4) to be significantly negative, which would indicate that industry specialization provides incremental audit quality 29
for the Big Four auditor. The following regression includes the industry specialization: |Acc| / |CFO| = α0 + α1Deloitteit + α2INDDeloitte + α3Deloitteit*IND +
(4)
α4E&Yit + α5INDE&Y + α6E&Yit*IND+ α7KPMGit + α8INDKPMG + α9KPMGit*IND + α10PwCit + α11INDPwC + α12PwCit*IND + α13YEARit + α14LEVit + α15GROWTHit + α16SIZE + εit. INDDeloitte is a dummy variable that turns 1 if a firm has its main activities in the industries mining and construction, and zero otherwise. INDE&Y is the dummy variable that turn 1 if the auditee has its main activities in mining and services, and zero otherwise. INDKPMG is a dummy variable turning 1 if the observation performs its activities in construction and wholesale, and zero otherwise. INDPwC is a dummy variable that turns 1 if an entity performs its main activities in the industries retail and wholesale. The interaction terms Deloitteit*IND, E&Yit*IND, KPMGit*IND, and PwCit*IND capture then the industry specialization of Deloitte, E&Y, KPMG, and PwC respectively. In order to capture the influence of the investor protection regime on audit quality, the examination of the second hypothesis will use the anti-self-dealing index of Djankov et al. (2008). The anti-self-dealing index represents a measure of legal protection of minority shareholders against expropriation by corporate insiders. The index is composed of factors that determine the structure of the regulation of selfdealing in different countries covering both private and public enforcement, and stock market development (Djankov et al., 2008). The index’ median is determined, after which the countries above (under) the median are classified as being a high (low) protection regime. Thus, if the legal protection of shareholders is high, the auditors’ risk of litigation is also likely to be high, which is expected to lead to incremental audit quality on the interaction terms α3, α5, α7, and α9. This leads to the following regression: |Acc| / |CFO| = α0 + α1Deloitteit + α2IPit + α3Deloitteit*IP + α4E&Yit +
(5)
α5E&Yit*IP + α6KPMGit + α7KPMGit*IP + α8PwCit + α9PwCit*IP + + α10YEARit + α11LEVit + α12GROWTHit + α13SIZE + α14INDit + εit.
30
The dummy variable IPit turns 1 if the investor protection regime can be classified as being high, and 0 otherwise. The interaction terms Deloitteit*IP, E&Yit*IP, KPMGit*IP, and PwCit*IP capture the incremental audit quality of Deloitte, E&Y, KPMG, and PwC in a strong investor protection regime. It is believed that these interaction terms are all significantly negative, because in a strong investor protection regime, the litigation risk is higher for auditors, which makes it more likely that they will be more inclined to perform a thorough audit.
5.3. Data The sample data is obtained from the Datastream files for the period 2005-2007. Due to the transition of the European Union from local GAAP to IFRS in 2005, all observations apply the same accounting standards, which make a combination and comparison of numbers reliable. The Datastream files contain among other indices the DJSTOXX 600, which is used in this thesis. The DJ STOXX 600 is derived from the Dow Jones STOXX Total Market Index (TMI) and a subset of the Dow Jones STOXX Global 1800 Index. It consists of 600 components, representing large, mid, and small capitalization firms across 18 European countries. This is especially useful in this thesis, since Hypothesis 2 examines the influence of cross-country differences across the European Union on audit quality (Stoxx, 2009). Fiscal year 2008 was not included in this sample, because due to the global financial crisis, these numbers may not be representative and can therefore influence the results. Initially, the sample comprised of 1800 firm-year observations; however, financial institutions (SIC-codes between 6000-6999) are excluded from the sample, because their reporting requirements and accrual process differ significantly from the other industries in the sample, which prevents them from having a comparable dependent variable with the other observations (Goncharov and Zimmermann, 2006; Van Tendeloo and Vanstraelen, 2005). Next, I have excluded the observations that had missing data for either the dependent variable or the independent variables. Finally, I have excluded outliers in the top and bottom of the dependent variable. Table 2, panel A summarizes the sample selection process. Details of the samples and variables used in each of the tests are summarized in panel B, C, and D.
31
Table 2 Descriptive statistics of the sample Panel A: Sample selection Number of observations obtained from DATASTREAM for the years 2005-2007 Less observations that were financial institutions (SIC-code) Less unavailable data for either the dependent or the independent variables Less observations that were on the basis of data analysis not representative Final number of observations used in the tests
Panel B: Pooled Sample Statistics (2005-2007) Variables |Acc| / |CFO| Big4 IP YEAR 2006 YEAR 2007 LEV GROWTH SIZE
Panel C: Auditor statistics Variables Deloitte Ernst & Young KPMG PwC Non-Big Four auditors
Mean
N
Median 0,54800 0,45503 0,95308 1 0,48094 0 0,34213 0 0,36852 0 2,46844 1,70267 0,13460 0,09236 15,52103 15,37984
Mean 207 192 260 316 48
1800 408 187 182 1023
Median 0,20235 0,18768 0,25415 0,30890 0,04692
0 0 0 0 0
(Table 2 is continued on the next page)
32
Table 2 (continued) Panel D: Industry statistics
Industry (SIC-code) Agriculture (0000-0999) Mining (1000-1499) Construction (1500-1999) Manufacturing (2000-3999) Transportation (4000-4999) Wholesale (5000-5199) Retail (5200-5999) Financial (6000-6999) Services (7000-9999)
Deloitte N
Total
0 24 12 89 46 0 11 0 25
% 35,82% 31,58% 19,02% 23,35% 0,00% 19,30% 21,74%
207
Panel E: Investor protection statistics N Mean Country Austria 16 0,41614 Belgium 25 0,64573 Denmark 28 0,49246 Finland 33 0,57949 France 137 0,62494 Germany 129 0,58298 Greece 13 0,47167 Ireland 12 0,38471 Italy 56 0,51890 Netherlands 45 0,56720 Norway 26 0,53055 Portugal 16 0,60605 Spain 51 0,61209 Sweden 65 0,51021 Switzerland 75 0,48427 United Kingdom 296 0,52210 Total 1023
Big Four auditors Ernst & Young KPMG N % N 0 0 16 23,88% 10 0 0,00% 13 87 18,59% 136 44 22,34% 48 6 18,18% 14 7 12,28% 11 0 0 32 27,83% 28 192
Median 0,35914 0,42579 0,47305 0,51712 0,57303 0,53027 0,29551 0,32783 0,43784 0,46219 0,52073 0,55170 0,56027 0,40183 0,37900 0,41472
%
PwC N
14,93% 34,21% 29,06% 24,37% 42,42% 19,30% 24,35%
260
Anti-self-dealing index 0,21 0,54 0,46 0,46 0,38 0,28 0,22 0,79 0,42 0,20 0,42 0,44 0,37 0,33 0,27 0,95
Total % 0 17 13 156 59 13 28 0 30 316
25,37% 34,21% 33,33% 29,95% 39,39% 49,12% 26,09%
0 67 38 468 197 33 57 115
100,00% 100,00% 100,00% 100,00% 100,00% 100,00% 100,00%
975
Investor protection Low High High High Low Low Low High High Low High High Low Low Low High
(Table 2 is continued on the next page)
33
Table 2 (continued) Variable definitions |Acc| / |CFO| is a ratio of the absolute value of accruals and the absolute value of cash flow from operations. Big4 equals 1 if the firm is being audited by a Big Four auditor, 0 else. Deloitte, Ernst & Young, KPMG, and PwC are all dummy variables, which turn 1 if the firm observation is audited by that particular auditor and 0 otherwise. IP is a dummy variable that equals 1 if the investor protection regime of the observation can be classified as high, and 0 otherwise. YEAR 2006 and 2007 are both dummy variables that turn 1 if the observation is obtained from that particular year, and 0 else. LEV is the ratio of total liabilities to total assets. GROWTH is the revenue growth rate, calculated as the revenues in the current year minus the revenues in the prior year and divided by revenues in the prior year. SIZE is the natural logarithm of total assets.
6. Empirical results In order to provide evidence on potential heterogeneity among the Big Four audit firms, I start with a general Big Four versus non-Big Four earnings management analysis. This approach will provide information about which Big Four firms drive the expected audit quality differences between the Big Four and non-Big Four, and how the audit quality ranking of the Big Four will be. The auditor analyses, using the absolute value of auditees’ accruals scaled by the absolute value of auditees’ cash flow from operations as the earnings management proxy, are reported in Table 3. Panel A of Table 3 present the results of the regressions that examine the extent to which insiders exercise discretion in reporting earnings between the Big Four and non-Big Four auditees. The findings show that the coefficient on the Big Four variable is significantly negative (p