Audit quality, auditor size and legal environments

A

B

F

M

A

TRACK ACCOUNTANCY

AUDIT QUALITY, AUDITOR SIZE AND LEGAL ENVIRONMENTS

Student Jimé Princen

Student number 10679847

Course Master Thesis Accountancy & Control Course code 6314M0044Y

A

I first would like to express my gratitude to my supervisor Vincent O’Connell for the useful comments, remarks and nice chats during the learning process of this master thesis. With your support I’ve managed to find an interesting subject after a stressing period of topic exploration. Furthermore, I would like to thank KPMG for offering me a motivating and helpful environment for writing this thesis. Finally, I would like to thank my girlfriend and of course my parents for their financial support and their inexhaustible patience and trust during my student years.

June, 2014 Jimé Princen

A

This paper examines the audit quality of first, second and third tier auditors and additionally investigates the influence of different legal environments on audit quality during the period 2008-2012. Audit quality is measured by abnormal accruals, which is considered an adequate measure for determining the auditors’ ability to restrain the client from managing earnings. As large auditors have a greater litigation risk and reputation to lose it is expected they provide the highest audit quality. In addition, it is expected that auditors in countries with strong investor protection provide the highest audit quality. The findings seem to be consistent with the first conjecture, yet no significant difference is found for the latter.

Keywords: Audit quality, big 4 accounting firms, second tier accounting firms, accruals, investor protection, legal environments, common law, civil law.

T

1. INTRODUCTION ... 5

2. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT ... 7

2.1 DEFINING AUDIT QUALITY ... 7

2.1.1 THE DEMAND FOR HIGH QUALITY AUDITS ... 8

2.2 AUDITOR SIZE AND AUDIT QUALITY ... 9

2.2.1 ABNORMAL ACCRUALS AS A MEASURE FOR AUDIT QUALITY ... 11

3.3 LEGAL ENVIRONMENT ... 12

3. METHODOLOGY ... 14

3.1 SAMPLE SELECTION ... 14

3.2 AUDIT QUALITY:ACCRUALS-BASED EARNINGS MANAGEMENT ... 15

3.3 AUDIT FIRM SIZE ... 17

3.4 COMMON AND CIVIL LAW LEGISLATION ... 18

3.5 CONTROL VARIABLES ... 19 4. EMPIRICAL FINDINGS ... 21 4.1 DESCRIPTIVE STATISTICS ... 21 4.2 CORRELATION ANALYSIS ... 23 4.3 REGRESSION RESULTS ... 25 4.4 ADDITIONAL ANALYSIS ... 29 5. CONCLUSION ... 30 5.1 LIMITATIONS ... 31 6. REFERENCES ... 32

1.

I

Right after the turn of the century the world was shocked by several substantial cases of accounting fraud. The Enron and Arthur Anderson debacle proved the necessity and requirement for high quality audits once again. DeAngelo (1981) formulated a two-dimensional definition of audit quality as: detecting misstatements and errors in financial statements, and reporting these material misstatements and errors. Due to the lack of direct measurement possibilities for these characteristics, prior literature came up with several surrogates like: audit fees and hours, reputation, litigation risk, auditor size and abnormal accruals. This research will focus on the latter three, formulated in two hypotheses. The first part examines the relation between the audit quality an auditor provides to its clients and the size of the auditor, where audit quality is measured by the client’s abnormal accruals and auditor size by a three-tier classification. Prior literature has found evidence suggesting that the main drivers of audit quality are reputational loss and litigation cost. In particular, DeAngelo (1981) states that audit firm size is an important determinant of audit quality. Consistent with her work, Palmrose (1988) and Simunic and Stein (1987) argue that due to ‘deeper pockets’ and extensive investing in brand names, large audit firms have higher incentives to minimize litigation risk and protect their reputation by providing high quality audits. DeAngelo (1981) further claims that large auditors posses more financial resources for training and technology and are less dependent on an individual client.

The second part focuses on the association between audit quality and the legal regime a firm is governed by. According to La Porta et al. (1998), Europe accommodates three fundamentally different legal environments; English common law, German civil law and French civil law. Based on their investor protection, and with that the litigation risk for auditors, the audit quality is assumed to differ between these countries (Francis and Wang, 2008). Earlier research found evidence of greater financial transparency in countries with high investor protection (Bhattacharya et al. 2003 and Bushman et al, 2004). Ball et al. (2000) documented that in these highly protective countries earnings are less managed and more value relevant.

Consistent with prior literature the results of this research on auditor size suggest that larger audit firms show greater ability in restraining the clients’ abnormal accruals. According to Jones (1991), abnormal accruals are a valid measure for earnings quality and should therefore give a fair representation of the quality an auditor provides. Next to this, an additional analysis is

done focusing merely on positive abnormal accruals. Evidence for this suggests that no significant difference exist between the audit quality provided by first, second and third tier auditors. Further, the analysis on audit quality in different legal environments showed no significant difference between the assessed common and civil law countries. This contradicts the argument of La Porta et al. (1998) on diverging investor protectionism in European legal regimes. The additional test for positive abnormal accruals presented similar results. Based on conflicting results in prior literature on the influence of legal regimes on diverging audit quality between Big Four/non-Big Four auditors, an additional analysis was undertaken to clarify this issue. The results however don’t seem to concur, as for all legal regimes no significant difference is found in audit quality provided by first, second or third tier.

This research contributes to present academic literature in several ways. First, audit firm size has proved itself a valid measure for audit quality. However, nearly all-previous studies on this topic focus on the Big N/non-Big N dichotomy. This study uses a three-tier classification, which gives better insight in the audit quality distribution over different sized auditors. Where Francis et al. (1999) chose for a three-tier classification based on level of operations (Big 6/national/local), this research will base the distinction on net income from audit practices. In accordance with DeAngelo’s (1981) argument that the value of an audit firm is determined by the present value of its future quasi rents, I believe this classification is more consistent with the established theory on audit quality and should gives a better representation of the auditors’ incentives. Put differently, the higher the earnings of the auditor, the more is lost when a breach is not detected and reported. Second, Gore et al (2001) identified a growing interest in the impact of different legal regimes and economic environments on accounting attributes. Further, the outcome of this research contributes to the current discussion on the comparability of financial statement and earnings in the European Union. Third, this study will extend the existing academic literature by further examining the relation between both audit quality and auditor size, and audit quality and legal environments as earlier studies often found conflicting results.

The paper proceeds as follows, the next section will further elaborate on established literature and the development of the hypotheses. Section 3 clarifies the methodology used for data collection and analysis. The results will be presented in Section 4, followed by the conclusion and limitations in Section 5.

2.

L

In the following section I will elaborate on the different concepts underlying this research and provide insights with respect to the relation between the size of the auditor, the legislation under which the auditor operates and the audit quality it provides to its clients. Starting off with section 2.1, in which the term audit quality will be described and an explanation will be given for the need for high quality audits. Section 2.2 focuses on the relation between auditor size and audit quality. Section 2.3 will discuss the possible influence a country’s legal environment has on audit quality. The hypotheses will be build from the literature in these three sections.

2.1 D

DeAngelo (1981) defines the quality of audit services as the market-assessed joint probability that a given auditor will both discover a material misstatement in the client’s accounting system and subsequently report this misstatement. DeAngelo (1981) further states that the probability that a given auditor will find a misstatement depends on a combination of the firm’s technological capabilities, procedures employed on an audit, the sampling method and so on. DeAngelo (1981) and Watts and Zimmerman (1990) describe an auditors’ independence as the conditional probability that a discovered breach will be reported. They further argue that the ex ante value of an audit depends on the auditor’s incentive to disclose selectively ex post.

Palmrose (1988) focuses more on ‘assurance’ in her definition of audit quality. She argues that the intention of an audit is to provide assurance on the financial statements. This leads to audit quality being the likelihood that a firm’s financial statements are free of material misstatements. Note that this definition emphasized the quality of the end result of an audit. In a framework of audit quality developed by the IFAC (2007), quality is the most fundamental characteristic of international auditing standards, and should be capable of consistent interpretation and unambiguous translation, enforceable and designed to achieve a high quality audit. Francis (2004) seems to have more of a pragmatic perspective by defining audit quality as the auditors’ ability to meet legal and professional requirements.

2.1.1 T

In contemporary business it’s almost impossible for a company to not find itself caught in a web of interconnected business relationships, with a wide range of stakeholders and contractual parties pursuing their own interests. Because of these diverging interests it’s not realistic to assume unified objectives and actions by stakeholders. To align the interests of all parties and moderate possible conflicts, a company can voluntarily opt to hire an external auditor. As the role of the external auditor has proven its usefulness over the years, many countries made the hiring of an auditor mandatory for firms meeting certain criteria. So, the need for an auditor can be found in either the firms’ signaling needs (Titman and Trueman, 1986) or its need of restraining agency costs (Jensen and Meckling, 1976).

Further, Wallace (1980) explained the demand for an external auditor by three hypotheses. The first is ‘The Stewardship Hypothesis’ that connects the agency theory to the external audit by arguing that managers are keen on showing the stakeholders their willingness to enhance the transparency of their actions (Jensen and Meckling, 1976). Jensen and Meckling (1976) also suggested that the shareholder protecting his interest could explain the hiring of an external auditor. As a reaction to the moral hazard problem they might discount the value of their initial investment and reduce management compensation. This is also in accordance with one of the agency theory features that DeFond (1992) defined as the deviation in preferences of manager and owner concerning the managers’ actions. The second feature concerns the owners’ lack of observability of the manager. Another alternative to the misalignment of interest is the implementation of performance compensation contracts, however this also doesn’t go without additional monitoring expenses. The second hypothesis defined by Wallace (1980) is ‘The Information Hypothesis’. This hypothesis emphasizes the importance of high quality financial information disclosed by the companies. Investors request financial information about the company’s performance, as they don’t have the ability to evaluate it themselves. As most investment decisions are based on these disclosed financial statements, investors request the hiring of an auditor with the goal of improving the quality of the information. The third and so-called ‘Insurance Hypothesis’ argues that the auditor provides a certain amount of insurance to the company’s stakeholders, and more specifically the ones that base their investment decisions on the audited financial statements. When a decision turns out bad and a misstatement in the financial statement is to blame, investors will turn to the auditor for retribution.

The complexity of the environment in which a company operates has significantly increased over the years. Both the diversification of the relationships with stakeholders as the increasing globalization has played a great role. One of the consequences of an expanding firm is a decrease in the managers’ ability to hold control over the firm’s processes. Knechnel et al. (2008) identify this problem and further argue the need for monitoring and how the auditor can improve the internal control by increasing efficiency and improving the interest alignment of shareholders, managers and employees. Watts and Zimmerman (1983) concur with this and stated that audit services act as a monitoring device and reduce agency costs with debt- and stockholders.

2.2 A

In his research Francis (2004) refers to two waves of auditor differentiation research. The first focusing on the Big N/non-Big N dichotomy based on the audit quality they provide. The second wave disregards the Big Four1 as a homogenous group and concentrates on the potential sources of differentiation in audit quality. This research investigates the audit quality of different size auditors and hence builds on the first principle.

Over time many studies have used the size of an accounting firm as a proxy for audit quality. Consumers of financial statement information incur cost when evaluating audit quality. As these costs often outweigh the benefits, a potential response is the development of less costly surrogate measures for audit quality (DeAngelo, 1981). Francis and Wilson (1988) and Simunic and Stein (1987) argued that large accounting firms have a greater reputation to uphold and are therefore more incentivized to provide high quality audits. DeAngelo (1981) states that large audit firms have a greater reputation to lose and will therefore have lower incentives to engage in misreporting to preserve a specific client relation. Ultimately, by giving an unqualified opinion on a misstated financial report the firms puts its whole clientele base at stake. However, this doesn’t insinuate that every Big Four firm provides better audit quality for every audit. Unfortunately large audit failures still occur too often with all the consequences that entail for both the individual firms and the audit profession as a whole. DeAngelo (1981) explains this incentive to ‘cheat’ by the potential loss of future quasi-rents, which are lost when the auditor is terminated.

Palmrose (1986) and later Moizer (1997) found evidence of a positive relation between audit fee and audit quality, which is explained by both audit effort (more hours billed) and the use of more expertise (higher hourly billing rates). Further, Simunic (1980) stated that, relative to other audit firms, Big Four auditor charge their clients a premium. Francis (2004) then correctly brings up the issue of a differential demand for audit quality and raises the question; why would firms voluntarily pay more for a high quality audit when lower-priced and legal alternatives exist? Prior literature found several explanations for this phenomenon. Beatty (1989) and Willenborg (1999) argue that firm with a high information gap between management and investors are more incentivized to hire a high quality auditor to reduce the information asymmetry and communicate on the firm’s financial condition. Additionally, Francis and Wilson (1988), DeFond (1992) and Francis et al. (1999) elaborated on this by arguing that firms are more likely to hire a Big Four auditor when agency cost rise and monitoring needs are high. When moving the scope to the organization of the auditor, DeAngelo (1981) highlights some points of interest. First, Watts and Zimmerman (1983) argued that large audit firms are more capable of monitoring the individual auditor; DeAngelo (1981) notes that if this is the case, the benefits to auditor size are greater than specified here. When looking at the partner level of a large firm, the partners share proportionately in the firms’ profits. The greater the amount of clients, the less a partner will lose when an individual client is let go. Therefore, the greater the likelihood a breach will be reported when discovered. However, a greater number of partners might give individuals incentive to shirk (Alchain and Demsetz, 1972).

Over the years researchers have identified numerous diverging proxies for the measurement of audit quality. As mentioned earlier a significant amount of researchers have based their study on the Big Four dichotomy for the measurement of audit quality, either individually or combined with other proxies. Palmrose (1988) and Feroz et al. (1991) found results indicating Big Four firms endure fewer legal proceedings by private or institutional parties and are sanctioned less by the Securities and Exchange Commission. Other research focuses on the output of an audit in the form of an audit report or audited financial statement, where the results consistently show higher quality audits for Big Four auditors. When looking at pre-IPO audit reports, Weber and Willenborg (2003) argued that large audit firms provide greater accuracy for predicting future earnings. Lennox (1999) finds consistent results for the post-IPO

period. Additionally, Big Four auditors show greater conservatism in giving an unqualified audit opinion due to higher auditing thresholds (Francis and Krishan, 1999).

2.2.1 A

In the previous section I argued that researchers seem to have found consensus on the high audit quality providence of Big Four auditors. However, the proxy that will be used in this research and which hasn’t been discussed yet distinguishes through its ex ante measurement possibilities. The model developed by Jones (1991) and its later modification by Dechow et al. (1995) are widely adopted measures for audit quality, and measure the discretionary or abnormal part of the total accruals. This part suggests discretion by management for personal gains and clouds the fair representation of the financial statement. Research on this showed that firms audited by Big Four auditors have a lower amount of abnormal accruals, meaning less aggressive earnings management and with that high earnings quality (Becker et al., 1998; Francis et al., 1999). The literature in this section led to the formulation of the following hypothesis:

H1: Audit quality, measured by abnormal accruals, is the highest for first tier auditors,

followed by second and third tier auditors.

Prior literature has shown various ways of classifying auditors for the measurement of audit quality. The earlier mentioned second wave aimed at the sources of differentiation. Examples for this are industry expertise (Solomon et al., 1999; Francis et al., 2005 and DeFond et al. (2000), cross-office differences in audit quality (Ferguson et al, 2003; Francis et al., 2005 and Reynolds and Francis), auditor tenure (Johnson et a., 2002 and Meyers et al., 2003), non-audit fees (Ashbaugh et al., 2003; Chung and Kallapur, 2003 and Frankel et al., 2002) and board independence (Carcello and Neal, 2000; Dechow et al., 1996 and Klein, 2002). The first wave determined audit quality based on the dichotomy of small and large audit firms. Although most researchers adopt the Big N1/non-Big N dichotomy, I will partially follow Francis et al. (1999) and focus on auditor size in three levels. Francis et al. (1999) however made a separation between Big Six, national and local offices. As the income level of the Big Four and the subsequent two/three is substantial I chose to categorize those as second tier and all others as third.

1_{Big N signifies the top four (Francis and Yu, 2009), six (Francis et al. 1999) or eight (Teoh and Wong, 1993)} audit firms used for research.

3.3

L

As can be seen from the previous sections, an auditor’s behavior with respect to pricing and quality is the result of various factors. In addition to the earlier discussed reputational issue, the second hypothesis will focus on the legal environment an auditor is governed by. Over the last decades market-wide convergence has a high priority on the agenda of the European Union. Next to the labor market, international trade and social issues, there have been several attempts to harmonize the audit profession and its services. The Eighth Directive initiated by the European Union covers the issue of auditor independence only to a limited extent and researchers have criticized the national culture and accounting traditions for interfering with this harmonization process. La Porta et al. (1997) argues that besides the cultural differences a primary distinction can be made. Ball et al. (2000) further elaborate on this as the systematic difference of the demand for accounting earnings for common-law countries and civil-law countries. The first is described by Jaggi and Low (2000) as a country of which their legislation is formed by judge’s decisions on specific cases. The second, are countries that are part of the scholar and legislator-made civil law tradition (La Porta et al., 1997). Both La Porta et al. (1997) and Giner and Rees (2001) even further diversify this into three legal traditions; English common law, German civil law and French civil law. This research will put its focus on these legal regimes, as they seem to differ distinctively with regard to their protectionism over investors.

Leuz et al. (2003) investigate earnings management in different legal environments and find evidence that dispersed ownership structures, developed equity markets, strong legal enforcement and investor rights negatively influence earnings management in companies. Jaggi et al. (2000) found evidence that difference in cultural values of the common- and civil-law countries influence the willingness of multinationals to disclosure financial statement information. Consistent with Leuz et al. (2003), Jaggi et al. (2000) also found higher levels of debt and more ownership dispersion with multinationals located in common law countries. As mentioned above, La Porta et al. (1997) investigated investor protection in different legal environments and conclude that investors in civil law countries have weaker legal rights than investors in common law countries. Further, they found that shareholders and creditors get the strongest protection in common law countries followed by German-civil-law countries and French-civil-law countries. The quality of law enforcement is highest in German-civil-law countries and again the lowest in French-civil-law countries. Although common law countries

have strong law enforcement they fall between the other two legal traditions. The prior literature described above leads me to assume that audit firms operating in a legal environment of high investor protection have a high motivation to provide high quality audit in order to lower litigation risk. Hence, the second hypothesis is formulated as;

H2: Audit quality, measured by abnormal accruals, is highest for companies operating under

English common law, followed by companies under German-civil-law and French-civil-law.

In their research Francis and Wang (2008) discuss three alternative scenarios regarding the international behavior of Big Four firms. The first possibility only concerns the US market and is therefore left out of consideration. The second viewpoint argues that Big Four audit firms are international organizations with global operations and therefore have incentives to develop and uphold a global standard with regard to their reputation. The third, which is considered the most plausible by the authors, states that Big Four behavior is not internationally uniform and varies systematically with incentives in different institutional environments. They conjecture that Big Four auditors are more conservative towards a client’s financial reports as a response to high litigation risk caused by stricter investor protection.

3.

M

In the following section I will elaborate on the research method used for testing the hypotheses. Starting off with section 3.1, where the sample selection procedures are explained. Section 3.2 elaborates on the dependent variable, audit quality. Section 3.3 will then explain the independent variables included for both the auditor size and legal rules hypothesis. Section 3.4 will discuss the control variables and finally in section 3.5 the empirical model will be presented in order to test the hypotheses.

3.1 S

For this research a quantitative archival-based research approach is chosen. The data is retrieved from the DataStream database and afterwards tidied in Microsoft Excel. The sample consist of listed companies from the United Kingdom, Germany and France, as these companies should give a fair representation of the differences in common and civil law legislation (La Porta et al. 1997). The sample has an observation period of five years, from 2008 till 2012. Prior literature showed very dispersed periods of observations from three years (Boone et al., 2010) until nineteen years (Francis et al., 1999). As mentioned, this research will cover five years, mainly due to a substantial amount of missing values before 2007.

Sample selection and composition

Table 1a.

Table 1b.

Data cleansing steps United Kingdom Germany France Total

Raw data 11.346 10.290 10.968 32.604

Missing values (7.338) (8.634) (9.324) (25.296)

Lagged year (2007) (668) (276) (274) (1.218)

<10 observations per industry (435) (130) (135) (700)

Utilities and financial institutions (180) (55) (170) (405)

Total firm year obs 2725 1195 1065 4985

Auditor Firm year observations

First tier/ Big Four 3.462

Second tier 914

Third tier 609

After removing all firms with missing values and outliers at the 1st and 99th percentile, the sample contained 4985 firm year observations.

In addition, utilities, financial and governmental institutions are excluded from the sample because of their specific regulatory and operating characteristics, which makes it hard to estimate their discretionary accruals (Fields et al., 2004; Frankel et al., 2002). To calculation the industry specific modified Jones (1991) model, firms with fewer than 10 client-year observations available per industry (based 2-digit SIC code) were also excluded. The subsequent variable calculations were done in Microsoft Excel and exported to STATA for regressions. Next to normal linear regressions, I follow Boone et al. (2010) and conduct a truncated regression for all observations with abnormal accruals higher than zero. The reason for this is the assumption that it’s more common for firms to manage their earnings upwards causing greater concern for the auditor (Heninger, 2001).

3.2 A

:

A

-

Earlier studies on audit quality have found results suggesting a relationship between audit quality and earnings management (Becker et al., 1998; Francis et al., 1999). As discussed earlier and building on DeAngelo (1981) and Palmrose (1988) it is argued that large audit firms have shown a greater capability of constraining a firm in manipulating earnings. One of the most frequently used methods to measure earnings management, and with that the audit quality, are the discretionary total accruals methods developed by Jones (1991) and its’ later modification by Dechow et al. (1995). To estimate the discretionary accruals as a proxy for earnings management this model divides the total accruals into a discretionary and nondiscretionary part. The first allows managers to use discretion in their accounting estimates and decisions. In accordance, Healy (1985) and Dechow et al. (1995) found evidence that firms use these discretionary accruals in managing their earnings. Due to accrual differences caused by the nature and industry of a firm the Jones (1991) model requires the sample to be subdivided into industry groups with a minimum of ten firms in every group. As suggested by Kothari et al. (2005) industry groups with less than ten firms are deleted from the sample to enhance the reliability of the estimates. As mentioned above, the subdivision is based on a 2-digit SIC code.

The calculation of the total accruals is done by either deducting the cash flow from operations from the net income; or measuring the difference between working capital accruals and depreciation. The total accruals function as a dependent variable in a regression, with the independent variables acting as proxies for normal accruals (nondiscretionary accruals) and gross fixed assets. These normal accruals allow for working capital needs of a firm and the latter represents the normal depreciation.

The model:

TACCit = 𝛽0 + 𝛽1 (1/Ait-1) + 𝛽2 (ΔREVit – ΔRECit) / At-1 + 𝛽3 (PPEit + At-1) +

ε

(1) Where

TACCit Total accruals deflated by total asset at t-1 and calculated by income before

extraordinary items minus cash flow from operations for year t and firm i.

Ait-1 Total assets of year t-1 for year t and firm i.

ΔREVit Change in revenue from prior year to year t for year t and firm i.

ΔRECit Change in accounts receivable from prior year to year t for year t and firm i.

PPEit Gross property, plant and equipment for year t and firm i.

ε

Model (1) shows the regression model for the total accruals. The discretionary accruals can then be estimated by comparing the total accruals with the non-discretionary accruals. The residual or error term is considered as the discretionary part (2). Becker et al. (1998) and Chung and Kallaput (2003) explain the absolute nature of the residual by the possibility for managers to manipulate the earnings both upwards and downwards, therefore the combination of negative and positive accruals should be brought along in the estimation.

DACCit = TACCit - 𝛽1 (1/Ait-1) + 𝛽2 (ΔREVit – ΔRECit) / At-1 + 𝛽3 (PPEit + At-1)

(2) DACCit Discretionary accruals for year t and firm i.

3.3 A

As concluded in prior literature and discussed earlier there is a significant relationship between the size of an audit firm and its ability to restrain accrual-based earnings management. The more effective the auditor is in controlling managerial opportunism, the greater the credibility of the reported earnings. The first hypothesis tests whether the audit quality provided by different tier audit firms differs, where first tier or Big Four firms provide the highest quality, than second tier and finally third tier audit firms. As there is no established method on how to classify the different tier audit firms I chose to distinguish them based on their income from audit practices. The first tier contains the Big Four audit firms, the second tier the three to four following firms, and the remaining firms are classified as third tier.

Audit tier categorization

Table 2

Audit tiers United Kingdom1 _Germany2 _France2

First tier PwC PwC Deloitte

Deloitte KPMG KPMG

KPMG Ernst & Young Ernst & Young

Ernst & Yong Deloitte PwC

Second tier Grant Thornton UK BDO

BDO

Nexia International

Fiducial Praxity Baker Tilly Grant Thornton Mazars

- Moore Stephens -

Third tier Other Other Other 1 _{Ranking retrieved from ICAEW website (2009)}

2 _{Ranking retrieved from ESCP Europe’s Study on the effects of the implementation of the acquis on statutory audits}

of annual and consolidated accounts including the consequences on the audit market by J. Le Vourc’h and P.

Morand published in 2011. The ranking concerns 2009.

The model is composed of three dummy variables representing the different tiers and six control variables that will be discussed in the next section. Please note, that firms can only be classified as either first, second or third tier and multiple classifications are not possible. In a regression the dummy variables gives an indication on the presence or absence of a categorical effect that may be expected to influence the outcome (Draper and Smith, 1998). It is expected that the first tier audit firms have the lowest corresponding discretionary accruals and the third tier audit firms the highest. The model is formulated as:

DACCit = 𝛽0 + 𝛽1 D1Tit + 𝛽2 D2Tit + 𝛽3 D3Tit + 𝛽4 SIZEit + 𝛽5 MTBit +

𝛽6 DISTRESSit + 𝛽7 CFFOit + 𝛽8 GROWTHit + 𝛽9 LEVit + εit

Where

D1Tit = 1 if auditor is a first tier auditor; 0 if auditor is other tier for year t and firm i.

D2Tit = 1 if auditor is a second tier auditor; 0 if auditor is other tier for year t and firm i.

D3Tit = 1 if auditor is a third tier auditor; 0 if auditor is another tier for year t and firm i.

SIZEit Natural log of total assets for year t and firm i.

MTBit Market-to-book ratio for year t and firm i.

DISTRESSit Financial distress measure based on Altman’s (2000) z-score for year t and firm i.

CFFOit Cash flow from operations deflated by total assets for year t and firm i.

GROWTHit Change in sales from prior year to year t for year t and firm i.

LEVit The ratio of debt to total assets year t and firm i.

ε

3.4

The second hypothesis investigates the relationship between the audit quality an audit firm provides and the country in which it is located. As mentioned in the literature section La Porta et al. (1997) investigated investor protection in common and civil law countries. Their results indicated that shareholders and creditors have the best legal protection in English common law, followed by German and French civil law. In addition, German civil law showed the strongest legal enforcement just followed by English common law. French civil law showed the poorest legal enforcement. From this it can be assumed that audit firms operating in a legal environment of high investor protection have a high motivation to provide high quality audit to lower litigation risk. Hence, I expect firms under English common law to provide the highest audit quality, firms under German common law to take the middle position and firm under French common law to have the poorest audit quality. Similar to model (3) this model will contain three dummy variables all representing a different legal environment and six control variables.

The model is formulated as:

DACCit = 𝛽0 + 𝛽1 DUKit + 𝛽2 DBDit + 𝛽3 DFRit + 𝛽4 SIZEit + 𝛽5 MTBit +

𝛽6 DISTRESSit + 𝛽7 CFFOit + 𝛽8 GROWTHit + 𝛽9 LEVit +

ε

Where

DUKit = 1 if auditor is subject to English common law; 0 if auditor subject to other

legislation for year t and firm i.

DBDit = 1 if auditor is subject to German civil law; 0 if auditor subject to other

legislation for year t and firm i.

DFRit = 1 if auditor is subject to French civil law; 0 if auditor subject to other

legislation for year t and firm i.

SIZEit Natural log of total assets for year t and firm i.

MTBit Market-to-book ratio for year t and firm i.

DISTRESSit Financial distress measure based on Altman’s (2000) z-score for year t and firm i.

CFFOit Cash flow from operations deflated by total assets for year t and firm i.

GROWTHit Change in sales from prior year to year t for year t and firm i.

LEVit The ratio of debt to total assets year t and firm i.

ε

3.5 C

Earlier research has shown that including control variables in a statistical regression enhances the internal validity of the primary variables of interest. By adding these control variables to the regression one can control for their individual or aggregated effect on the dependent variable. All continuous variables are deflated by total assets to control for firm size. Ashbaugh et al. (2003) and Frankel et al. (2002) have found evidence that a relation exist between SIZE and earnings management. Lang and Lundholm (1993) argue that larger firms have a higher incentive to report high quality earning due to the presence of litigation risk. In addition, Myers et al. (2003) state that larger firms have greater accrual stability. However, Watts and Zimmerman (1990) observe the opposite relation based on the political cost hypothesis. This hypothesis predicts that large firms rather than small firms are more likely to engage in earnings manipulation to reduce reported profits. Hence, no prediction is offered on the nature of this relationship.

According to Ashbaugh et al. (2003) the market-to-book ratio (MTB) represents the inverse of the firm’s growth opportunities; therefor a negative relationship is expected. The third control variable is DISTRESS, which, according to DeFond and Jiambalvo (1994), Reynolds and Francis (2000) and Sweeney (1994), has a positively correlated with abnormal accruals.

However, consensus hasn’t been found yet, as DeAngelo et al. (1994) found evidence that firms in financial distress manipulate their earnings downwards to facilitate debt renegotiations. Financial distress is measured by Altman’s (2000) Z-Score.

Z = 0,012WCit/TAit + 0.014REit/TAit + 0.033EBITit/TAit + 0.006MVit/TLit +

0.999SALESit/TAit

(5) Where

Z Overall index that indicates the likelihood of a firm going to bankruptcy. The lower (higher) the score, the more (less) likely a firm is heading to bankruptcy, where 1.3 (bankruptcy) and 3.0 (safe zone) are threshold values.

WCit Working capital for year t and firm i.

TAit Total assets for year t and firm i.

REit Retained earnings for year t and firm i.

EBITit Earnings before interest and taxes for year t and firm i.

MVit Market value of equity for year t and firm i.

TLit Book value of total liabilities for year t and firm i.

SALESit Total sales for year t and firm i.

CFFO, meaning the cash flow from operation, is expected to be negatively associated with a firm’s abnormal accruals (Ashbaugh et al. 2003; Chung and Kallapur, 2003). Barth et al. (1999) and Dechow and Skinner (2000) argue the rapidly growing (GROWTH) firms are more sensitive to managing earnings as they continuously try to meet or beat expectations to sustain the stock price. Hence, a positive relation is expected. Finally, leverage (LEV) is included as a control variable as prior studies by Ashbaugh et al. (2003), Frankel et al. (2002) and Kim et al (2012) found a relation between the extent of leverage and the use of abnormal accruals. As the nature of the relation is disputable, no assumption is made.

4.

E

4.1 D

The descriptive statistics displayed in table 3 give an overview of the distribution of firm-year observations over the different tiers and legal environments. As can be seen from the table, the Big Four audit firms hold a dominant position in the market of audits. In the United Kingdom, 69,42% of the firms in the sample are audited by a Big Four auditor. In Germany the proportion is almost the same with a percentage of 69,46 and in France 69,48% of the firms hire a Big Four auditor. Due to greater investor protection one would expect the highest rate of Big Four audits in the United Kingdom, however the findings seem to disagree

Descriptive statistics

Table 3

Table 4a and 4b provide descriptive statistics on the dependent and independent variable for respectively the legal environments and audit size. The explanatory variable AAC, or abnormal accruals, is scaled on total assets and functions as a measure for reported earnings. The higher this number is, the higher the reported earnings, which leaves more room for earnings manipulation. When looking at the mean AAC of table 4a, the ranking of the mean AAC is consistent with the expectations, where first tier auditors have the lowest abnormal accruals, followed by second tier and third tier auditors. Table 4b shows that firms under English civil law show the lowest abnormal accruals, followed by firms under German common law and firms under French common law. These results are aligned with the findings of La Porta et al. (1997, 1998), who argued that the investor protection is the highest in the United Kingdom and lowest in France.

Firm year distribution United Kingdom Germany France Total

First tier 1892 37,96% 830 16,65% 740 14,84% 3462 69,45% Second tier 500 10,02% 219 4,39% 195 3,92% 914 18,33% Third tier 333 6,68% 146 2,93% 130 2,61% 609 12,22 Total firm 2725 54,66% 1195 23,97% 1065 21,37% 4985 100%

Descriptive statistics (auditor size) Table 4a.

1 _{Abnormal accruals deflated by total assets.}

D1T, D2T, D3T: 1 if auditor is respectively a first, second or third tier auditor. SIZE: natural log of total assets. MTB:

market-to-book ratio DISTRESS: financial distress measure based on Altman’s (2000) z-score. CFFO: cash flow

from operations deflated by total assets. GROWTH: change in sales from prior year to year t. LEV: the ratio of debt to

total assets.

Descriptive statistics (legal environments) Table 4b

DUK, DBD, DFR: 1 if auditor is subject to respectively English, German or French law. SIZE: natural log of total

assets. MTB: market-to-book ratio DISTRESS: financial distress measure based on Altman’s (2000) z-score. CFFO:

cash flow from operations deflated by total assets. GROWTH: change in sales from prior year to year t. LEV: the ratio

of debt to total assets.

Mean Median Std. Dev. Mean Median Std. Dev. Mean Median Std. Dev.

United Kingdom Germany France

AAC1 _{-0,002 0,004 0,114 -0,001 -0,001 0,099 0,003 0,006 0,068} D1T 0,721 1 0,449 0,700 1 0,458 0,620 1 0,486 D2T 0,152 0 0,359 0,160 0 0,367 0,290 0 0,454 D3T 0,127 0 0,333 0,140 0 0,347 0,090 0 0,287 SIZE 5,024 4,923 1,041 5,528 5,340 0,970 5,698 5,548 0,911 MTB 2,543 1,540 21,529 1,9790 1,500 3,424 1,693 1,310 1,644 ZSCORE 3,919 2,677 12,701 3,185 2,570 3,246 2,502 2,046 3,221 CFFO 0,040 0,072 0,224 0,066 0,0740 0,129 0,726 0,074 0,094 GROWTH 1,554 0,072 58,247 0,062 0,050 0,289 0,064 0,047 0,240 LEV 20,770 8,730 42,850 23,070 17,165 56,419 23,789 21,500 18,914

Mean Median Std. Dev. Mean Median Std. Dev. Mean Median Std. Dev.

First tier Second tier Third tier

AAC1 _{-0,002 0,002 0,095 -0,001 0,007 0,106 0,011 0,009 0,130} DUK 0,568 1 0,496 0,453 0 0,498 0,568 1 0,496 DBD 0,242 0 0,428 0,209 0 0,407 0,274 0 0,447 DFR 0,191 0 0,393 0,338 0 0,473 0,158 0 0,3647 SIZE 5,355 5,283 1,022 5,175 5,097 1,021 5,081 4,881 1,132 MTB 2,194 1,500 18,175 2,281 1,410 6,341 2,321 1,450 12,836 ZSCORE 3,505 2,478 10,618 3,311 2,307 7,469 3,269 2,602 6,231 CFFO 0,057 0,075 0,179 0,049 0,069 0,167 0,035 0,070 0,229 GROWTH 1,113 0,055 51,599 0,294 0,058 3,676 0,422 0,081 5,637 LEV 22,918 14,800 45,582 18,968 12,780 22,489 21,057 11,720 50,096

4.2 C

Table 5a and b show the result of the Pearson correlations of respectively the auditor size and legal environment model. For the first table there are some important observations. First is the significant correlation between SIZE on the three different tiers, which can be interpreted as large firms being more likely to hire a first tier auditor. The control variables in both tables further show an appreciable relation between SIZE and CFFO (27,6%), LEV and SIZE (17,2%) and LEV en ZSCORE (14,1%). All don’t appear to be much of a surprise.

A large firm, relative to a small one, possesses more economies of scale, which leads to a higher cash flow from operations. Further, due to more extensive resources it’s relatively easier for a large firm to attract capital, which explains the positive correlation between the two variables. Last is the negative association between LEV and ZSCORE. The latter describes the probability a firm will finds itself in financial distress in the following firm year. Previous literature has shown that highly leveraged firms are more likely to make default in their financial obligations, leading them in financial distress.

The collinearity diagnostics are displayed in table 6a and b. According to Manfields and Helms (1982) the main concerns of multicollinearity are the large variances of the least squares estimators of coefficients of variables. The VIF value indicates how many times larger the variance of multicollinear data is compared to orthogonal data (VIF value of 1.0). If the VIF is close to 1.0 the effect of multicollinearity is inappreciable. Calculation the VIF gives a better indication of how estimates of variables are individually influenced by multicollinearity. Although table 5a and 5b show some correlating variables, the VIF values (all < 2,6) suggest no significant threats of collinearity biases.

Pearson correlations (auditor size) Table 5a.

* correlation is significant at the 0.05 level (2-tailed) **correlation is significant at the 0.01 level (2-tailed)

D1T, D2T, D3T: 1 if auditor is respectively a first, second or third tier auditor. See table 4a for other variable

definitions.

Pearson correlations (legal environment) Table 5b.

* correlation is significant at the 0.05 level (2-tailed) **correlation is significant at the 0.01 level (2-tailed)

DUK, DBD, DFR: 1 if auditor is subject to respectively English, German or French law. See table 4b for other

variable definitions. Collinearity diagnostics

Table 6a Table 6b

Auditor size Legal environment Tolerance VIF Tolerance VIF

D1T 0,486 2,06 DUK 0,578 1,73 D2T 0,489 2,05 DBD 0,618 1,62 SIZE 0,874 1,14 SIZE 0,817 1,22 CFFO 0,895 1,12 CFFO 0,895 1,12 LEV 0,941 1,06 LEV 0,941 1,06 ZSCORE 0,961 1,04 ZSCORE 0,960 1,04 MTB 0,996 1,00 MTB 0,995 1,00 GROWTH 0,999 1,00 GROWTH 0,999 1,00

Notes: VIF denotes the variance inflation factor , the tolerance is the reciprocal of the VIF.

D1T, D2T, D3T: 1 if auditor is respectively a first, second or third tier auditor. DUK, DBD, DFR: 1 if auditor is

subject to respectively English, German or French law. See table 4a and 4b for other variable definitions.

D1T D2T D3T SIZE MTB ZSCORE CFFO GROW LEV

D1T 1,000 D2T -0,714** 1,000 D3T -0,563** -0,177** 1,000 SIZE 0,097** -0,052** -0,074** 1,000 MTB -0,003 0,002 0,002 0,004 1,000 ZSCORE 0,010 -0,006 -0,007 -0,069** 0,045** 1,000 CFFO 0,034* -0,009 -0,037** 0,276** -0,026 -0,112** 1,000 GROWTH 0,008 -0,006 -0,004 -0,014 -0,000 0,005 0,011 1,000 LEV 0,033* -0,033* -0,008 0,172** -0,033* -0,141** 0,070** -0,005 1,000

DUK DBD DFR SIZE MTB ZSCORE CFFO GROW LEV

DUK 1,000 DBD -0,617** 1,000 DFR -0,572** -0,293** 1,000 SIZE -0,280** -0,1292** -0,205** 1,000 MTB -0,022 0,009 0,017 0,004 1,000 ZSCORE 0,054** -0,015 -0,051** -0,069** 0,045** 1,000 CFFO -0,080** -0,041** -0,056** 0,276** -0,026 -0,112** 1,000 GROWTH 0,017 -0,011 -0,010 -0,014 -0,000 0,005 0,011 1,000 LEV 0,031* 0,014 0,118 0,172** -0,033* -0,141** 0,070** -0,005 1,000

4.3 R

In the following section the regression results will be discussed. Table 7a Panel A show the results of the linear regression of the auditor size for the full sample. When looking at the variables for the first, second and third tier auditors, significant differences in the coefficients (with a confidence level of 99% for D1T and confidence level of 95% for D2T) suggest a difference in audit quality they provide. With the third tier operating as a base level, one can interpret that firms audited by second tier auditors have lower abnormal accruals (-,0121) than firms audited by third tier auditors. In addition, the table shows that firms audited by first tier or Big Four auditors even have lower abnormal accruals (-,0130). These results are consistent with the first hypothesis and indicate that large auditors are more successful in lowering a client’s discretionary accruals and with that providing a higher audit quality. The most generally accepted explanation for this is given by DeAngelo (1981) and Simunic and Stein (1987) who argues that large audit firms have a greater reputation to uphold and are therefore more incentivized to provide high quality audit.

The significant associations of the control variables CFFO and MTB are consistent with the expectations. However, note that for the latter the coefficient is extremely small, which is consistent with the findings of Francis and Yu (2009). Analogous to the results of Boone et al. (2010) no significant relation is be found for the control variable GROWTH. Because of conflicting results from prior literature, no assumption was made for the three remaining control variables. The variable SIZE is positively associated, which is consistent with the findings of Watts and Zimmerman (1990). Finally, both the variables ZSCORE and LEV show a weak but positive relationship, consistent with prior literature (Boone et al. 2010 and DeAngelo et al., 1994).

For the second hypothesis on legal environments the null hypothesis cannot be rejected due to insignificant results. This implies that there’s no significant difference in abnormal accruals of firms under the three assessed legislations. Although the outcome is not aligned with the expectations, an explanation can be found in the emerging international convergence of accounting practices caused by the general implementation of IFRS. As mentioned earlier, Francis and Wang (2008) refer to three scenarios regarding international behavior of Big Four auditors. One of these viewpoints discusses the international character and global operations of

Panel A: Full sample

Linear regression (auditor size) Table 7a

Predicted sign

Coefficient Std. Error t-statistic p-Value 95% Conf. Interval Intercept ? -0,0300 0,0083 -3,59 0,000 -0,0463 -0,0136 D1T - - -0,0130 0,0044 -2,94 0,003 -0,2162 -0,0043 D2T - -0,0121 0,0052 -2,32 0,020 -0,0224 -0,0019 D3T1 _{0 0 (omitted)} SIZE ? 0,0091 0,0015 6,27 0,000 0,0063 0,0120 MTB - -0,0003 0,0001 -3,65 0,000 -0,0005 -0,0001 ZSCORE ? 0,0008 0,0001 5,72 0,000 0.0006 0,0011 CFFO - -0,1046 0,0082 -12,82 0,000 -0,1206 -0,0886 GROWTH + 0,0000 0,0000 0,01 0,993 -0,0000 0,0006 LEV ? 0,0001 0,0000 -4,99 0,000 -0,0002 -0,0001 Nobs Adj. R2 4981 4,06%

D1T, D2T, D3T: 1 if auditor is respectively a first, second or third tier auditor. SIZE: natural log of total assets. MTB:

market-to-book ratio DISTRESS: financial distress measure based on Altman’s (2000) z-score. CFFO: cash flow

from operations deflated by total assets. GROWTH: change in sales from prior year to year t. LEV: the ratio of debt

to total assets.

1_{STATA used this group is as base level.}

Linear regression (legal environment) Table 7b

Predicted sign

Coefficient Std. Error t-statistic p-Value 95% Conf. Interval Intercept ? -0,0359 0,0089 -4,04 0,000 -0,0534 -0,0185 DUK - - -0,0030 0,0037 -0,81 0,421 -0,0103 0,0043 DBD - -0,0035 0,0042 -0,84 0,402 -0,0118 0,0047 DFR1 _{0 0 (omitted)} SIZE ? 0,0086 0,0015 5,71 0,000 0,0056 0,0116 MTB - -0,0003 0,0001 -3,62 0,000 -0,0005 -0,0001 ZSCORE ? 0,0008 0,0001 5,71 0,000 0,0006 0,0011 CFFO - -0,1049 0,0082 -12,85 0,000 -0,1210 -0,0889 GROWTH + -0,0000 0,0000 -0,00 0,998 -0,0001 -0,0001 LEV ? -0,0002 0,0000 -4,98 0,000 -0,0002 -0,0001 Nobs Adj. R2 4981 3,91%

DUK, DBD, DFR: 1 if auditor is subject to respectively English, German or French law. SIZE: natural log of total

assets. MTB: market-to-book ratio DISTRESS: financial distress measure based on Altman’s (2000) z-score. CFFO:

cash flow from operations deflated by total assets. GROWTH: change in sales from prior year to year t. LEV: the

ratio of debt to total assets.

Panel B: Only observations with AAC>0 Truncated regression (auditor size) Table 7c

Predicted sign

Coefficient Std. Error t-statistic p-Value 95% Conf. Interval

Intercept ? 2,2764 1,7735 1,28 0,199 -1,1996 5,7524 D1T - - -0,8995 0,6819 -1,32 0,187 -2,2361 0,4372 D2T - -0,8539 -0,7036 -1,21 0,225 -2,2331 0,5252 D3T 0 (omitted) SIZE ? -1,8302 1,3804 -1,33 0,185 -4,5357 0,8754 MTB - 0,0053 0,0090 0,60 0,551 -0,0123 0,0231 ZSCORE ? 0,0190 0,0142 1,33 0,183 -0,0090 0,0469 CFFO - -2,4688 1,4813 -1,67 0,096 -5,3721 0,4346 GROWTH + 0,0106 0,0294 0,36 0,720 -0,0471 0,0682 LEV ? -0,0063 0,0078 -0,81 0,418 -0,0216 0,0090 Nobs Adj. R21 _0,007% 2616

1 _{A rough estimate of the adjusted R}2 _{is calculated by correlating the abnormal accruals with the predicted value and} squaring the result.

D1T, D2T, D3T: 1 if auditor is respectively a first, second or third tier auditor. SIZE: natural log of total assets. MTB:

market-to-book ratio DISTRESS: financial distress measure based on Altman’s (2000) z-score. CFFO: cash flow

from operations deflated by total assets. GROWTH: change in sales from prior year to year t. LEV: the ratio of debt to

total assets.

Truncated regression (legal environment) Table 7d

Predicted sign

Coefficient Std. Error t-statistic p-Value 95% Conf. Interval

Intercept ? 0,8055 1,5387 0,52 0,601 -2,2104 3,8214 DUK - - 1,2371 1,3761 0,90 0,369 -1,4601 3,9342 DBD - 2,7582 2,7988 0,99 0,324 -2,7274 8,2437 DFR 0 0 (omitted) SIZE ? -2,5800 2,5669 -1,01 0,315 -7,6111 2,4511 MTB - 0,0103 0,0150 0,69 0,491 -0,0191 0,0398 ZSCORE ? 0,0252 0,0251 1,01 0,315 -0,0239 0,0744 CFFO - -3,3702 2,8634 -1,18 0,239 -8,9824 2,2420 GROWTH + 0,0241 0,0439 0,55 0,584 -0,0619 0,1100 LEV ? -0,0066 0,0107 -0,62 0,537 -0,0276 0,0144 Nobs Adj. R21 _0,002% 2616

1 _{A rough estimate of the adjusted R}2 _{is calculated by correlating the abnormal accruals with the predicted value and} squaring the result.

DUK, DBD, DFR: 1 if auditor is subject to respectively English, German or French law. SIZE: natural log of total

assets. MTB: market-to-book ratio DISTRESS: financial distress measure based on Altman’s (2000) z-score. CFFO:

cash flow from operations deflated by total assets. GROWTH: change in sales from prior year to year t. LEV: the

Big Four firms, and how this incentivizes them to maintain a uniform reputation worldwide (Simunic and Stein, 1987). They further argue that this can be achieved by standardized training of personnel, knowledge sharing practices, and the use of uniform audit methodologies and technology.

For the coefficient of the control variable LEV the nature of the relation changed. As mentioned, before prior literature showed conflicting results. For example, by Ashbaugh et al. (2003) who found a negative association between LEV and abnormal accruals when accruals are positive and a positive association when accruals are negative.

Please note that for both regressions of Panel A the explanatory power of 4,06% and 3,91% is very limited. However, similar studies also didn’t manage to find significantly higher adjusted R2-statistic. Ashbaugh et al.’s (2003) results showed an adjusted R2-statistic of 18% with approximately 400 observations more and seven additional control variables. Boone et al. (2010) had similar results with an adjusted R2_{-statistic of 14,59%, three additional control}

variables and more than twice as many observations.

Furthermore, Panel B of table 7 describes the analyses of income-increasing abnormal accruals. This is done by excluding all observations with AAC < 0. As discussed earlier and also stated by Heninger (2001), it is more common for firm to managers their earning upwards. Also the concerns these upward manipulations bring for the auditor is a reason for a separate analysis. For this additional analysis only a proportion of the sample’s observation is used (AAC>0). If this is the case, Greene (1997) prescribes a truncated regression approach for estimating the model. As seen in table 7 Panel B the independent variables on national legislation and auditor size are not significant. This indicates that, after controlling for several firm characteristics that could influence a client’s abnormal accruals, no difference is found in the ability of different sized audit firms to constrain the client from managing earnings upwards. Although the Big N/non-Big N method used by Francis and Yu (2009) and Boone et al. (2010) slightly differs from the three-stage method used in this research, their results lead to a similar conclusion. That is, for firms with positive abnormal accruals,. no significant difference is found between Big N/non-big N auditors, suggesting small audit firms show similar capability of restraining the client in manipulating reported earnings upwards.

Finally, table 7d Panel B contain the results of the truncated regressions on legal environments for all firms with positive abnormal accruals. As the table shows insignificant results for the independent dummy variables on national legislation the same conclusion can be drawn as for the full sample. Specifically, there is no difference in the audit quality provided by auditors governed under different legal environments.

4.4 A

Both Francis and Wang (2008) and Choi et al. (2008) investigate audit quality of Big Four/non-Big Four auditors in difference legal regimes but found conflicting results. Choi et al. (2008) found evidence that Big Four audits in ‘weak’ legal regimes are of higher quality than non-Big Four audits; however no quality differentiation was found in ‘stronger’ regimes. In contrast, Francis and Wang (2008) found no differences in the earnings quality of Big Four and non-Big Four clients in weak legal regimes, whereas the earnings quality of Big Four clients seemed to increase, relative to non-Big Four clients, as the legal regimes became stricter. It is worth mentioning that although both studies reach contradictory conclusions, for the measurement of earnings quality Francis and Wang (2008) examine client earnings characteristics, where Choi et al. (2008) chose audit fee as proxy. In response to these conflicting results an additional analysis is done examining audit quality of the three tiers in three legal regimes using separate regressions. This should tell us if the effect is apparent across all three countries or specific to one country. Even though this research uses a similar proxy for earnings quality as Francis and Wang (2008), the results suggest no difference in audit quality. That is, in all three legal regimes no significant difference is found in the audit quality provided by first, second and third tier auditors.

5.

C

The purpose of this research was to examine a possible difference in audit quality between first, second and third tier auditors. As first tier or Big Four auditors have a greater reputation to lose it is expected they provided the highest audit quality. In addition, I examined if audit quality differs in countries with diverging degrees of investor protection. According to the ‘Deep Pockets theory’ first tier audit firms have more financial resources, which results in a higher litigation risk and with that a greater incentive to provide high quality audits. To test these conjectures, data is collected for a sample of 997 firms from one common law country and two civil law countries over a period of five years (2008 – 2012). More specifically, the sample consists of firms governed by English common law, German civil law and French civil law.

The first hypothesis stated that first tier auditors provide the highest quality audits, followed by second tier auditors and third tier auditors. Audit quality is measured by discretionary accruals, which represent the discretion by management for personal gains. Consistent with the first hypothesis and established prior literature, evidence is found that large audit firms provide higher quality audits by showing a lower amounts of abnormal accruals with their clients, meaning less aggressive earnings management and high earnings quality (Becker et al., 1998; Francis et al., 1999). Also an additional analysis is done on all observations with positive abnormal accruals (Francis and Yu, 2009). Insignificant results suggest that for firms with positive abnormal accruals there is no difference in the audit quality provided by the three tiers, which is consistent with the findings of Francis and Yu (2009) and Boone et al. (2010).

The second hypothesis argues that due to litigation risk, audit firms located in countries with high investor protection provide higher audit quality than audit firms located in countries with weak investor protection. La Porta et al. (1998) found evidence that investor protection is the highest in English common law and the lowest in French civil law, with German civil in the middle. Although for audit quality such an order of merit was expected, results show no significant differences. A possible explanation could be found in the emerging international convergence of accounting practices caused by the general implementation of IFRS. The additional analysis on positive abnormal accruals showed similar results.

Based on conflicting results of Francis and Wang (2008) and Choi et al. (2008) on the influence of legal regimes on diverging audit quality between Big Four/non-Big Four auditors, an additional analysis was undertaken to further investigate this issue. Choi et al. (2008) found

evidence that Big Four audits in ‘weak’ legal regimes are of higher quality than non-Big Four audits; however no quality differentiation was found in ‘stronger’ regimes. Where Francis and Wang’s (2008) results show the opposite, this research failed to find significant differences in audit quality provided by first, second or third tier, suggesting an apparent effect across the three countries.

5.1 L

The results of this study are subject to the following limitations. First, is a difficulty arising in many cross-national samples. That is, uncertainties exist about to what extent the differences in audit quality are really influenced by differences in national legislations. For example, research has shown the influence of economic and institutional factors on abnormal accruals. Although I included several control variables, Leuz et al. (2003) acknowledges the difficulty to fully control for possible institutional factor. This is caused by their complementary nature, which makes in hard to disentangle them. In addition, I acknowledge that next to the five controls I’ve adopted, others potential incentives for audit quality may exist.

An issue regarding the tier classification is the personal judgment on the determination of the dividing line between the second and third tier. Francis and Wang (2008) make a division based on Big Six, national and local offices, but due to both insufficient data on local offices and the earlier discussed motivation (section 1) I chose to base the classification on net income from audit practices. As no clear benchmark on this is present, I used personal judgment on where a dividing line between second and third tier should be drawn (Big Four is classified as first tier). Please note, that for this research a specific dividing line is not considered relevant. Hence, I wanted to test whether large auditors provide a higher audit quality than middle-sized auditors, and, in turn, middle-sized auditors provide higher audit quality than small auditors. Finally, I am aware of the limit sample size of this research and how this might influence the results.