The effect of the new auditor’s report in the UK on the audit quality

(1)

Amsterdam Business School

MSc thesis

Name: Ivana Einmahl Student number: 10868216 Thesis supervisor: dr. A. Sikalidis Date: 19-06-2016

Word count: 14025

MSc Accountancy & Control, Accountancy

(2)

2

Statement of Originality

This document is written by student Ivana Einmahl who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3 Abstract

The Financial Reporting Council (FRC) introduced new requirements for auditor’s reports for companies subject to the UK Corporate Governance Code, International Standard on Auditing (ISA) 700, which are effective for the audits of financial statements from October 1, 2012. The FRC states that the new requirements will increase the audit quality, but does not have evidence for this. Therefore this is the first study explicitly examining the impact of the new auditor’s report in the UK and how this affects audit quality. To measure audit quality the proxies accrual-based earnings management and real activities manipulation are used, measured by the absolute value of discretionary accruals and by the level of real activities manipulation. I found, however, that there is not enough evidence to prove my research hypothesis, based on two research methods that compare the audit quality for companies subject to the UK Corporate Governance Code for the period 2010-2011 (before the implementation of the new auditor’s report) and for the period 2013-2014 (after the implementation of the new auditor’s report). That means that the audit quality does not significantly increase through the introduction of the new auditor’s report in the UK. Next to this, I performed two additional analyses. Interestingly, the first additional analysis revealed that the firms engage less in income-increasing earnings management (positive discretionary accruals) after the introduction of the new auditor’s report in the UK, when measuring audit quality with this proxy I found a significantly increased audit quality. On the other hand, the introduction of the new auditor’s report in the UK has no significant effect on negative discretionary accruals. Further, the second additional analysis revealed the robustness of the results of the main analysis when taking the absolute value of abnormal working capital accruals (AWCA) as an alternative proxy for audit quality.

Keywords: ISA 700, audit quality, discretionary accruals, real activities manipulation, regulatory changes

(4)

4

Acknowledgment

I would like to sincerely thank the following persons and organization who have contributed to this MSc thesis. First, I would like to express my deepest gratitude to my thesis supervisor dr. A. Sikalidis for supporting me. He assisted me during this process, shared his insights, and referred to interesting literature. Our discussions gave me more clarity on how to take the next steps in the various stages of my research. Second, I am highly thankful to PwC who gave me the golden opportunity to write my MSc thesis at the Amsterdam office in the department Audit & Assurance. Next to ameliorating my theoretical skills, I acquired more practical experience and I developed myself during this internship, wherefore my gratitude goes out to my company supervisor mr. M. Kip, MSc, from PwC. Besides that, I want to thank him for his valuable guidance.

(5)

5

“I am pleased that auditors as well as investors have given their enthusiastic support for the proposal in the Consultation Paper to supplement the binary pass/fail model of audit report. The provision of a fuller description of the work the auditor has undertaken will give far more insight to investors than the binary pass/fail model of the current audit report. The improved report will be a better basis for engagement by investors with companies, and we encourage auditors and companies to work together to develop succinct communication to do so”, according to Nick Land, Chairman of the FRC’s Audit and Assurance Council (Financial Reporting Council, 2013, p. 1).

The audit expectation gap exists for too long, the differences between what users of the financial statement expect an audit is delivering and what the auditor believes an audit is providing (Mock et al., 2013). During the financial crisis and the related attention to the role of the auditor, the gap had become bigger. For example, the European Commission discussed the role of the auditor during the financial crisis in its Green Paper, ‘Audit policy: Lessons from the crisis’ (European Commission, 2010). The European Commission made suggestions for improvements of the role and position of the audit, but also for adaptation of the European market for audit services. Also in the aftermath of the 2008 financial crisis the auditor’s report had been criticized. One of the points of criticism is that the auditor’s report is not very useful for the investors, because it is a binary pass/fail model that gives hardly to no information on the work that the auditors have performed (Financial Reporting Council, 2013). Because of this criticism the FRC suggested revisions for the auditor’s report and conducted an extensive consultation on these revisions, including a public meeting and an outreach program with UK and US investors, the accounting Institutes and audit firms. The revisions received strong support during the consultation process. Especially from investors, because they supported the provision of information on the audit scope. Because of this strong support the FRC has adopted the revisions for the auditor’s report for companies subject to the UK Corporate Governance Code with only minor amendments, which are effective for the audits of financial statements from October 1, 2012 (International Standard on Auditing (ISA) 700). The auditor’s report has to meet the audit standard ISA 700, making the report more useful for the investors (close any audit expectation gap).

The results of Gold, Gronewold, and Pott (2012) indicated a persistent expectation gap related to auditor’s responsibilities. Hence, the explanations of the ISA 700 auditor’s report do not close any audit expectation gap. Hossain (2013) tested the effect of CLERP 9 on the audit quality. In this research the proxy for audit quality is performance-adjusted discretionary

(7)

7

accruals and the auditor’s propensity to issue a going-concern opinion for a financially distressed company. The results showed an improvement of audit quality after the implementation of CLERP 9. Given the societal concern for the effectiveness of regulatory changes, this study will investigate the impact of the new auditor’s report in the UK on the audit quality. Therefore, the research question of this thesis is: to what extent was the audit quality improved after the introduction of the new auditor’s report in the UK?

Following Kim, Park, and Wier (2012), I computed the absolute value of discretionary accruals and the level of real activities manipulation. Discretionary accruals are calculated through the cross-sectional Modified Jones Model adjusted for performance. Next to this, in my main analysis I took the absolute value of the discretionary accruals, because earnings can be managed either upwards or downwards (Dechow & Dichev, 2002). However, in the additional analysis I performed regressions with signed discretionary accruals to gain more insight. Discretionary accruals and real activities manipulation are used as proxies for audit quality. The sample consists of companies of the London Stock Exchange with a premium listing. Next to this, the sample period consists of the years 2010 through 2014 and is divided into two periods: the period before the implementation of the new auditor’s report (2010-2011) and the period after the implementation of the new auditor’s report (2013-2014). The results of the two research methods indicate that the audit quality does not significantly increase through the introduction of the new auditor’s report in the UK. Explanations for these findings could be: the already high level of audit quality, a persistent expectation gap after the introduction of the new auditor’s report, limited data available (only two years after the introduction of the new auditor’s report), and existing literature suggests that new rules do not always indicate more audit quality (Smith, 2012).

Next to this, I performed two additional analyses. In contrast to the main analysis, the first additional analysis revealed that the firms engage less in income-increasing earnings management (positive discretionary accruals) after the introduction of the new auditor’s report in the UK. Thus measuring audit quality with this proxy I found a significantly increased audit quality. On the other hand, the introduction of the new auditor’s report in the UK has no significant effect on negative discretionary accruals. Further, the second additional analysis revealed that the results of the main analysis are robust to the alternative proxy for audit quality: the absolute value of abnormal working capital accruals (AWCA).

The study of this paper will contribute to the existing literature in two ways. First, the research will contribute to the existing literature from a scientific point of view. This research question will especially expand existing literature about regulatory changes and audit quality.

(8)

8

My study can possibly give new scientific viewpoints on the topic, which maybe give the existing literature more power or just decrease the credibility of the existing literature or give new findings. Next to this, this study will contribute to the existing literature that examines a proxy for audit quality, because I use two different proxies for audit quality: (1) discretionary accruals and (2) real activities manipulation. Hence, I will examine a broad set of financial reporting characteristics that represent opportunistic financial reporting. To my best knowledge, this is the first study explicitly examining the impact of the new auditor’s report in the UK on the audit quality. There are only studies about ISA 700 in other countries than the UK. In this way, this research can possibly be a starting point for future research about ISA 700 in the UK.

Second, the research will contribute to the existing literature from a societal point of view. The societal contribution is the usefulness of this study for the FRC and the investors in the UK. The FRC could use the findings of this study to review the implementation of the changes to UK Auditing Standards. Furthermore, the findings may indicate that the additional requirements increase the informativeness of the reports for the investors.

The structure of the paper is as follows. Section 2 describes the new requirements for auditor’s reports in the UK and Ireland (ISA 700), audit quality, the most relevant literature on the topic ISA 700 (in other countries), and relevant papers on regulatory changes and their effect on audit quality. This section also develops the hypothesis. Section 3 describes the methodology, which includes the sample selection, the two proxies for audit quality (discretionary accruals and real activities manipulation), and the two research methods (regression model for relating audit quality to control variables in combination with a Paired Samples test / regression model with a dummy for the introduction of the new auditor’s report in the UK). Section 4 first outlines the descriptive statistics and the correlations, second the results of the two research methods, third the additional analyses that were performed, and finally a summary of the analyses. In Section 5 the conclusion is presented and recommendations for future research are provided.

(9)

9 2. Literature review and hypothesis development

This section will provide the existing literature and the hypothesis. First, the new requirements for auditor’s reports in the UK and Ireland (ISA 700) and audit quality will be described. Second, the most relevant literature on the topic ISA 700 (in other countries) and relevant papers on regulatory changes and their effect on audit quality will be discussed. Finally, the hypothesis for this study will be developed.

2.1 New requirements for auditor’s reports

The FRC responded to calls from investors to make audits more transparent (Financial Reporting Council, 2013). Namely, the FRC introduced new requirements for auditor’s reports on companies subject to the UK Corporate Governance Code (ISA 700 [UK and Ireland] “The independent auditor’s report on financial statements”), which are effective for the audits of financial statements from October 1, 2012. The revisions require auditors to provide greater transparency about the work the auditors have performed on the audit. Greater transparency was addressed by requiring auditors to disclose on the following three areas:

a) A description of those assessed risks of material misstatement that were identified by the auditor and which had the greatest effect on the overall strategy; the allocation of resources in the audit; and directing the efforts of the engagement team;

b) An explanation of how the auditor applied the concept of materiality; and

c) A summary of the audit scope, including an explanation of how the scope was responsive to the assessed risks of material misstatement described in (a) and the concept of materiality as described in (b).

These revisions focus more on the ‘inputs’ to the scope of the audit, rather than the ‘outputs’. An example of a new auditor’s report with the disclosed three areas can be found in Appendix 1.

2.2 Audit quality

There is no clear definition for audit quality, but the most common definition is: the market-assessed joint probability that an auditor will detect a breach in the financial statements and will

(10)

10

report this breach (DeAngelo, 1981, p. 186). This depends on two components. The first component is the auditor’s technological capabilities, the used audit procedures, and the extent of sampling, etc. (detect the breach). The second component is the independence of an auditor (report the breach). Independence of an auditor according to the Securities and Exchange Commission is a mental state of objectivity and a lack of bias (Securities and Exchange Commission, 2001).

The General Accounting Office defined audit quality as providing the audited financial statements and related disclosures in accordance with laws and regulations and free from material errors (General Accounting Office, 2003). This definition assumes that through laws and regulations the financial statements and related disclosures provide reasonable assurance. According to Francis (2004) audit quality is seen as a theoretical continuity which varies from low audit quality to high audit quality. An error in the audit can be explained by two situations (Francis, 2004). The first situation is when an auditor cannot ensure that the client acts in accordance with auditing standards. In the second situation an error occurs when an auditor does not succeed in forming a judgment in circumstances that call for such judgments. Francis (2004) calls this an audit report failure. Other studies provide a similar definition for audit quality: auditors perform high quality audits when auditors detect errors in financial statements (Behn, Choi, & Kang, 2008; Chan & Wong, 2002; Chang, Dasgupta, & Hilary, 2009; Gul, Lynn, & Tsui, 2002).

Knechel, Krishnan, Pevzner, Shefchik, and Velury (2013) argued that the two components of DeAngelo (1981) of audit quality are not observable. Given this fact, this study will use accrual-based earnings management and real activities manipulation as proxies for audit quality. The absolute value of discretionary accruals and the level of real activities manipulation provide a measure to indirectly capture earnings management, because they indicate the active involvement of the management in reporting of earnings.

2.3 Regulatory changes

This section will discuss the key papers on the topic ISA 700 (in other countries) and relevant papers of regulatory changes and their effect on audit quality. This can help to better understand the impact of the new requirements for auditor’s reports in the UK on the audit quality (ISA 700).

(11)

11

2.3.1 ISA 700

Gold et al. (2012) checked the effectiveness of explanations of the ISA 700 auditor’s report in making the audit expectation gap smaller. This was tested by an experiment with German auditors and users of the financial statements. These two groups read a summary of a firm’s financial statements and an auditor’s report manipulated to include the explanations of the ISA 700 requirements versus an audit opinion-only version. The findings showed a persistent expectation gap pertaining to the auditor’s responsibilities. So, the explanations of the ISA 700 auditor’s report do not reduce the audit expectation gap. The findings also showed that the auditors and the financial statement users reached a reasonable belief consensus about management’s responsibilities and the reliability of the financial statements and that the audit opinion alone may indicate enough useful information to financial statement users.

Auditors periodically publish audit reports that reinforce the financial communication and reliability of accounting information; the content of the audit reports and the quality of the audit results is defined by the ISA (Fakhfakh, 2015). The audit standardization must shape understandable and clear audit reports for their users. When auditors comply with linguistic principles the structure of the audit reports will improve (Fakhfakh, 2015). But these audit reports are not effective, because the reports could be unreadable for users. Because of this criticism the International Federation of Accountants had revised ISA 700: “Forming an opinion and reporting on financial statements” to improve the content of the audit reports. Fakhfakh (2015) discussed the linguistic problems regarding to the preparation of audit reports according to the International Standards on Auditing. The findings of this study indicated that when there are conflicting requirements (auditing standards and linguistic principles), auditors have to optimize the presentation of the audit reports.

Al-Matarneh (2011) examined the extent to which auditors include the elements and content of auditor’s reports correspondingly with the auditor and the Audit Office's requirements. The results showed that the auditors comply with the preparation standards for the auditor’s report at the rate of 97% and that the auditors comply with formal elements of these reports at the rate of 89%. The results also indicated that there is a compliance with the content of the auditor’s report, as in the standard at the rate of 98.3%. Al-Matarneh (2011) recommended auditors to take into account the auditor's requirements when preparing the reports, including all elements and noting that the report's opinion was in the light of the information and explanations provided. Besides that, ISA could reconsider laws and regulations relating to audit standards and could require auditors to fulfill the audit standard ISA 700 at a rate of 100%.

(12)

12

2.3.2 Effect on audit quality

Hossain (2013) tested the impact of CLERP 9 on audit quality. CLERP 9 was introduced by the Australian Securities and Investments Commission to promote the confidence of the investors in the reports of listed companies in Australia (Lee & Shailer, 2008, p. 123). In the study of Hossain (2013) audit quality is measured by performance-adjusted discretionary accruals and the auditor’s propensity to issue a going-concern opinion for a financially distressed company. The findings of this study indicated an increase in audit quality after the implementation of CLERP 9. So, this study provides evidence for the effectiveness of regulatory changes (CLERP 9) on the improvement of audit quality.

Smith (2012) investigated the impact of two important regulatory changes on the perception of investors of audit quality. The first regulatory change examined is from a bottom-up coverage-based approach for auditing internal controls to a more top-down risk-based approach. The second regulatory change is a litigation reform further limiting auditor liability exposure following an alleged audit failure. The evidence of this study indicated that investors perceived a decrease in audit quality regarding to the two regulatory changes. The findings also suggested that the investor’s perceived decrease in audit quality will lead to less investments of resources in internal controls by management. The observed impact is mediated by considering efficiency for the new auditing standard and by a change in the auditor’s economic incentives following the proposed litigation reform.

Through the global consequences of the Enron scandal and the introduction of the Sarbanes–Oxley Act (SOX) in the USA, the UK regulatory regime for audit and corporate governance has changed (Beattie, Fearnley, & Hines, 2013). Features of the new regime are independent investigations of audit firms and an enhanced role for audit committees. Beattie et al. (2013) surveyed in 2007 three groups: chief financial officers of UK-listed companies, audit committee chairs, and audit partners. The purpose of this survey was to acquire views of the effect of 36 economic and regulatory factors on the audit quality for Post-SOX. Different audit committee interactions with auditors regarding to the other aspects are rated as the biggest improvement for audit quality. Though, ISAs and the audit inspection regime regarding to aspects of the ‘standards-surveillance-compliance’ regulatory system are rated as less effective. Next to this, respondents are concerned about many aspects of the new regime to influence audit quality introduced by Post-SOX regulations, because they think it is largely process and compliance driven, with more costs than benefits.

Ernst & Young released a report, the Sarbanes-Oxley Act at 10, about the regulation and the impact (Review in The Corporate Board, 2012). "At this anniversary, it is important to

(13)

13

acknowledge one of the greatest successes of SOX: the alignment of the interests of shareholders, with independent audit committees, audit oversight authorities, and auditors. Increased transparency is critical to improving audit quality, maintaining investor confidence, and ensuring the strength and competitiveness of U.S. capital markets." The Ernst & Young report also suggested that investors and U.S. capital markets have benefited from SOX. Through stronger alignment of independent auditors, audit committees, audit oversight authorities, and public company shareholders audit quality has increased (Review in The Corporate Board, 2012). Especially audit quality has increased through Public Company Accounting Oversight Board inspections and standard setting. So, the Ernst & Young report indicated governance and audit quality have increased ten years after SOX.

The global financial crisis and corporate scandals let people doubt about the audit quality (Holm & Zaman, 2012). In response to this, the FRC in the UK codified the audit quality in its ‘Audit Quality Framework’. Holm and Zaman (2012) analyzed if audit firms, professional bodies, and investors think that the FRC proposals are sufficient for increasing the audit quality. To analyze stakeholder responses they used impression management and legitimacy as a framework. They focused on different drivers than identified by the FRC, namely on transparency, expertise, professionalism, and commercialization of the audit. In contrast to the FRC, audit firms and professional bodies have mostly focused on problems which might not bring the commercial interest of audit firms in jeopardy. The findings of this study suggested that regulatory and professional bodies involved in image management and the improvement of audit quality try to restore tarnished image and to increase their legitimacy and standing. If regulatory bodies want to restore trust and legitimacy, they have to reconcile complicated often contradictory stakeholder demands (Holm & Zaman, 2012).

2.4 Hypothesis

Although the literature review in the previous section gives mixed views of the effect of regulation changes on audit quality, I believe that the new requirements for auditor’s reports in the UK have a positive effect on the audit quality. First, the auditor’s report is now more extended for investors (increased disclosure on risks, materiality, and scope of the audit). Hence, the auditors provide significantly more disclosure around their audit work and are moving away from the traditional binary pass/fail model of the past. The results of Review in The Corporate Board (2012) also indicated that an increase of transparency (more disclosure) and investor confidence is critical to improving audit quality. Second, Hossain (2013) showed evidence of an increase in audit quality after the implementation of CLERP 9. CLERP 9 is

(14)

14

rather similar to ISA 700 in the UK, because these regulatory changes both focused on improved reports for investors. Third, the Ernst & Young report suggested that ten years after SOX, governance and audit quality have improved (Review in The Corporate Board, 2012). They found that standard setting and a stronger alignment of independent auditors, audit committees, audit oversight authorities, and public company shareholders are two of the causes of increased audit quality. These two causes are also valid for the new auditor’s report in the UK, because the FRC has introduced ISA 700 in response to calls from investors to make the audits more transparent and auditors provide these new requirements. So, there is a strong alignment between independent auditors, the FRC, and investors. This leads to the following hypothesis:

Hypothesis 1: The audit quality increases after the introduction of the new auditor’s report in the UK.

(15)

15 3. Data and Method

This section will provide the research method: the sample selection will be described and the research design will be discussed. To examine the effect of the new auditor’s report in the UK on the audit quality, discretionary accruals and real activities manipulation will be used as proxies for audit quality. Next to this, I will use two research methods. First, a regression model for relating audit quality to control variables in combination with a Paired Samples test. Second, I will not exploit the paired structure and will only use a regression model with a dummy for the introduction of the new auditor’s report in the UK.

3.1 Sample selection

The FRC introduced new requirements for the auditor’s report for companies subject to the UK Corporate Governance Code (ISA 700), which are effective for the audits of financial statements from October 1, 2012. Given this fact, I have chosen the companies of the London Stock Exchange with a premium listing as initial sample (see Table 1). I excluded 2 financial institutions (SIC codes 6000-6999), because characteristics of accruals differ in these firms. The period that I want to investigate is the period before and after the changes for the auditor’s report in the UK are implemented.

Table 1: Number of companies 2010-2014

London Stock Exchange (premium listing) in Compustat Global – Fundamentals Annual (2010-2014)

Excluded (missing information for accrual-based earnings management/ real activities manipulation and control variables)

Financial institutions

661

351

2

Total number of companies retained 308

For the above 308 companies with a premium listing on the London Stock Exchange I will compare the audit quality before (2010-2011) and after (2013-2014) the implementation of the new auditor’s report. I will use the audit quality of these companies, to perform a Paired Samples test (after the audit quality is adjusted for control variables) and regression analyses.

(16)

16

3.2 Research design

Researchers use various proxies to measure audit quality and to determine an increase or decrease in audit quality. Most of the audit differentiation research focuses on the quality of the client’s financial statements (Lawrence, Minutti-Meza, & Zhang, 2011, p. 261). Hence, I will investigate financial reporting characteristics that represent opportunistic financial reporting. Specifically, I will use two different proxies for audit quality: (1) discretionary accruals and (2) real activities manipulation. Following Kim et al. (2012), I will compute the absolute value of discretionary accruals and the level of real activities manipulation. If firms want to manage their reported earnings they often use discretionary accruals as well as real activities manipulation. Another possibility is that firms choose the earnings management method that is less costly (Cohen, Dey, & Lys, 2008; Zang, 2012).

3.2.1 Proxy 1 for audit quality: discretionary accruals

Accruals are often used as a proxy for audit quality, because they reflect the auditor’s constraint over management’s reporting decisions. The benefit of this proxy is that it reflects the auditor’s enforcement of accounting standards (Lawrence et al., 2011, p. 261). So, the proxy accrual-based earnings management is appropriate for audit quality and I will use this as proxy 1 for audit quality. The most popular five models to measure accruals are the DeAngelo Model (DeAngelo, 1986), the Healy Model (Healy, 1985), the Jones Model (Jones, 1991), the Modified Jones Model (Dechow, Sloan, & Sweeney, 1995), and the Industry Model (Dechow & Sloan, 1991). For the proxy accrual-based earnings management I will use the Modified Jones Model.

Modified Jones Model

Dechow et al. (1995) showed that the Modified Jones Model has the highest statistical power in detecting earnings management. Comparing to the original Jones Model, the change in revenues is adjusted for the change in receivables in the event year. I will utilize a cross-sectional version of the modified Jones model, because of its superior specification and less restrictive data requirements (DeFond & Subramanyam, 1998). The variables I will use for this model will be scaled by lagged total assets to reduce the presence of heteroscedasticity in the regression errors (Kothari, Leone, & Wasley, 2005). I will take the absolute value of the discretionary accruals, because earnings can be managed either upwards or downwards (Dechow & Dichev, 2002). The higher the absolute value of the discretionary accruals, the

(17)

17

lower the audit quality. To extend the modified Jones Model, I will include ROAt−1 to control

for the impact on measured discretionary accruals of performance (Kothari et al., 2005). Next to this, it will increase the reliability of inferences from discretionary accrual estimates. The model to determine discretionary accruals is as follows:

TAit / Ait−1 = α0 + α1(1 / Ait−1) + α2ΔREVit / Ait−1 + α3PPEit / Ait−1 + α4IBXIit−1 / Ait−1 + εit, (1)

where:

TAit =total accruals for firm i in year t;

ΔREVit =change in net revenues for firm i in year t from year t−1;

PPEit = gross property, plant and equipment for firm i in year t;

IBXIit−1 = income before extraordinary items for firm i in year t−1;

Ait−1 = total assets for firm i in year t−1;

εit = residual, which indicates the raw discretionary accruals (RDA) for every firm-year.

I will use the cash flows statement approach, superior in certain circumstances to the balance sheet approach, to compute the total accruals (Hribar & Collins, 2002). They argued that using a balance sheet approach to detect earnings management is associated with measurement errors in accruals estimates.

TAit = IBXIit − CFOit, (2)

where:

IBXIit = income before extraordinary items for firm i in year t;

CFOit = cash flow from operations for firm i in year t.

I can calculate the discretionary accruals by:

DAit = RDAit + a1ΔRECit / Ait−1 (3)

where:

ΔRECit =change in net receivables for firm i in year t;

RDAit is as in equation (1);

(18)

18

3.2.2 Proxy 2 for audit quality: real activities manipulation

Real activities manipulation means that management actions deviate from normal business practices undertaken for purposes of meeting or beating certain earnings thresholds (Roychowdhury, 2006). Real activities manipulation is also important to use as a proxy for audit quality, because recent studies suggest that firms use real activities manipulation as an alternative tool for earnings management and trade-off real activities manipulation and accrual-based earnings management as substitutes (Badertscher, 2011; Cohen et al., 2008; Cohen & Zarowin, 2010; Roychowdhury, 2006; Zang, 2012). So, the proxy real activities manipulation is appropriate for audit quality and I will use this as proxy 2 for audit quality. Note that the variables in the equations (4), (6), and (7) will be scaled by lagged total assets to reduce the presence of heteroscedasticity in the regression errors (Kothari et al., 2005).

According to prior studies (Badertscher, 2011; Cohen et al., 2008; Cohen & Zarowin, 2010; Roychowdhury, 2006; Zang, 2012), I will use the following four measures to detect real activities manipulation: abnormal levels of operating cash flows (AB_CFO), abnormal production costs (AB_PROD), abnormal discretionary expenses (AB_EXP), and a combined measure of real activities manipulation (COMBINED_RAM). A positive value of AB_CFO indicates less real earnings management. A positive value of AB_PROD indicates more real earnings management. A negative value of AB_EXP indicates more real earnings management. Finally, if the proxy COMBINED_RAM decreases, a firms engage in more aggressive earnings management through real activities. Hence, when firms engage in real activities manipulation, the audit quality is lower.

Sales manipulation causes lower current-period operating cash flows (Cohen et al., 2008; Roychowdhury, 2006).The model to determine the normal level of operating cash flows is as follows (Roychowdhury, 2006):

CFOit / Ait−1 = α0 + α1(1 / Ait−1) + β1(Sit / Ait−1) + β2(∆Sit / Ait−1) + εit, (4)

where:

CFOit = cash flow from operations for firm i in year t;

Ait−1 = total assets for firm i in year t−1;

Sit = net sales for firm i in year t;

∆Sit= Sit − Sit−1;

εit = residual, which indicates abnormal cash flow from operations (AB_CFO) for every

(19)

19

Hence, the first measure of real activities manipulation is abnormal levels of operating cash flows (AB_CFO). The second measure of real activities manipulation is abnormal production costs (AB_PROD). Production costs are the cost of goods sold plus the change in inventory during the year and they express expenses as a linear function of contemporaneous sales (Badertscher, 2011; Cohen et al., 2008; Roychowdhury, 2006; Zang, 2012). Hence, I could determine production costs as follows:

PRODit = COGSit + ∆INVit, (5)

where:

COGSit= cost of goods sold for firm i in year t;

∆INVit = change in inventory for firm i in year t.

Finally, I can define normal production costs as follows:

PRODit / Ait−1 = α0 + α1(1 / Ait−1) + β1(Sit / Ait−1) + β2(∆Sit / Ait−1) + β3(∆Sit−1 / Ait−1) + εit, (6)

where:

εit = residual, which indicates abnormal production cost (AB_PROD) for every firm-year.

Abnormal discretionary expenses (AB_EXP) is the third measure of real activities manipulation. The model to determine the normal level of discretionary expenses is as follows (Badertscher, 2011; Cohen et al., 2008; Roychowdhury, 2006; Zang, 2012):

DISEXPit / Ait−1 = α0 + α1(1 / Ait−1) + β(Sit−1 / Ait−1) + εit, (7)

where:

DISEXPit = discretionary expenses for firm i in year t consist of SG&A expenses;

εit = residual, which indicates abnormal discretionary expenditure (AB_EXP) for every

firm-year.

Taking into account the direction of the three real activities manipulation elements, the combined measure of real activities manipulation (COMBINED_RAM) is included.

(20)

20

COMBINED_RAM is computed by AB_CFO minus AB_PROD plus AB_EXP. This is the fourth measure of real activities manipulation.

3.2.3 Research method 1: regression model for relating audit quality to control variables in combination with a Paired Samples test

This research method exists of two parts. First, I will perform the regression model for relating audit quality to control variables. Second, I will perform a Paired Samples test with the residuals obtained from the regression model for relating audit quality to control variables.

Regression model for relating audit quality to control variables

I consider two time periods: 2010-2011 (before) and 2013-2014 (after). Such a period is denoted with the index p. For equations (8) and (9), I will calculate for each company and each two-year time period the average value for each proxy for audit quality and the average value of each of the seven control variables. This yields 308×2=616 data points as input for the regression analysis for equations (8) and (9). Then I perform the regression analyses and calculate for every point the residual.

Firm characteristics could have an impact on the level of discretionary accruals and on the level of real activities manipulation, hence control variables will be used to control these factors. The choice of control variables is based on prior studies that examine my proxies for audit quality. The following models will be used:

ABS_DAip =  + 1COMBINED_RAMip + 2LNTAip + 3GROWTHip + 4LEVip + 5ROAip

+ 6LOSSip + 7SCFOip + ip, (8)

RAM_PROXYip =  + 1ABS_DAip + 2LNTAip + 3GROWTHip + 4LEVip + 5ROAip +

(21)

21

where:

ABS_DAip = absolute value of discretionary accruals for firm i in period p, where discretionary

accruals are calculated through the cross-sectional Modified Jones Model adjusted for performance;

RAM_PROXYip = AB_CFOip, AB_PRODip, AB_EXPip or COMBINED_RAMip;

AB_CFOip = the level of abnormal cash flows from operations for firm i in period p;

AB_PRODip= the level of abnormal production costs for firm i in period p, where

production costs are the COGS plus inventory change;

AB_EXPip= the level of abnormal discretionary expenses for firm i in period p, where

discretionary expenses are the SG&A expenses;

COMBINED_RAMip = AB_CFOip – AB_PRODip + AB_EXPip;

LNTAip = natural log of total assets for firm i in period p;

GROWTHip = annual growth rate of total assets for firm i in period p;

LEVip = total liabilities divided by total assets for firm i in period p;

ROAip = earnings before interest and taxes divided by total assets for firm i in period p;

LOSSip = set to 2 if firm i for both years has a negative income, set to 1 if firm i is has for one

of the two year a negative income, and 0 otherwise;

SCFOip = cash flows from operations divided by total assets for firm i in period p.

The two earnings management methods have a substitutive nature (Cohen et al., 2008), to control for this a proxy for real activities manipulation is included (COMB_RAM) in the accrual-based earnings management regression and a proxy of accrual-based earnings management (ABS_DA) is included in the real activities manipulation regression. LNTA is included to control for the client size effect on earnings management (Roychowdhury, 2006). GROWTHis used, because high growth opportunities of firms are related to earnings management (Carey & Simnett, 2006). LEVcould influence the level of discretionary accruals and the level of real activities manipulation, because when firms operate with a high leverage this could be an indication of a bad financial condition which may increase management’s incentives to manage earnings (Watts & Zimmerman, 1990). ROAis used, to

(22)

22

control for the effects of firm performance on earnings management (Dechow et al., 1995). LOSSis included because firms with bad financial conditions have more incentives to engage in earnings management (Cahan, Emanuel, Hay, & Wong, 2008; Frankel, Johnson, & Nelson, 2002). Finally, SCFO is also included to control for the effects of firm performance on earnings management. Firms with high cash flow from operations are associated with less earnings management (Frankel et al., 2002).

Paired Samples test

To examine the effect of the new auditor’s report in the UK on the audit quality (measured by the absolute value of discretionary accruals and the level of real activities manipulation), I want to compare the 308 matched pairs of residuals obtained above. These will be the inputs for my Paired Samples test. Then will I compute for each of the 308 companies the difference of the residuals “before” and “after” (for both proxies of audit quality) and denote them with Di, i=1,

…, 308. I will use the following test statistic:

D

̅

S

_D

_{/ √𝑛}

_D

, (10)

where:

nD = number of pairs;

D̅= sample mean of the Di;

SD = sample standard deviation of the Di.

The number of pairs is 308 (nD>30) and hence it can be assumed that the test statistic has a

standard normal distribution under the null hypothesis. Recall that the larger the absolute value of the discretionary accruals, the lower the audit quality. The audit quality is also lower when firms engage in real activities manipulation. But the change in level (before minus after) for the four measures of real activities manipulation have a different meaning in terms of real earnings management. For example, a high positive value of AB_CFO indicates less real earnings

(23)

23

management and a high positive value of AB_PROD indicates more real earnings management. Further, a high negative value of AB_EXP indicates more real earnings management. Finally, if the proxy COMBINED_RAM decreases, a firms engage in more aggressive earnings management through real activities. Hence, D̅ should be significantly larger than 0 for ABS_DA and AB_PROD; similarly D̅ should be significantly lower than 0 for AB_CFO, AB_EXP, and COMBINED_RAM. More precisely, taking a significance level of 5%, I will check if the value of the test statistic is larger than 1.645 or lower than −1.645 respectively.

3.2.4 Research method 2: regression model with a dummy for the introduction of the new auditor’s report in the UK

Next to the regression model for relating audit quality to control variables in combination with a Paired Samples test, I will use the following empirical models to examine the effect of the new auditor’s report in the UK on the audit quality (measured by the absolute value of discretionary accruals and the level of real activities manipulation):

ABS_DAip =  + 1IAR_UKp + 2COMBINED_RAMip + 3LNTAip + 4GROWTHip +

5LEVip + 6ROAip + 7LOSSip + 8SCFOip + ip, (11)

RAM_PROXYip =  + 1IAR_UKp + 2ABS_DAip + 3LNTAip + 4GROWTHip + 5LEVip

+ 6ROAip + 7LOSSip + 8SCFOip + ip, (12)

where:

IAR_UKp = set to 1 if the period is after the introduction of the new auditor’s report in the UK

(2013-2014) and set to 0 if the period is before the introduction of the new auditor’s report in the UK (2010-2011);

other variables are as in equations (8) and (9).

My prediction is that Hypothesis 1 will be supported by the data. That means that audit quality significantly increases through the introduction of the new auditor’s report in the UK: there is

(24)

24

a positive relationship between audit quality and the introduction of the new auditor’s report in the UK. More precisely, I expect a negative coefficient 1 for IAR_UK if the dependent variable

is ABS_DA or AB_PROD and a positive coefficient 1 for IAR_UK if the dependent variable

(25)

25 4. Results

This section will represent the empirical results. First, the descriptive statistics and correlations will be presented. Second, the results of the two research methods will be analyzed. Finally, two additional analyses will be described.

4.1 Descriptive statistics

Table 2 presents the descriptive statistics of the selected variables for both research methods. To be certain that the impact of the extreme observations is reduced, 98% Winsorization is used. This means that all data below the 1th percentile are set equal to the 1th percentile and data above the 99percentile are set equal to the 99 percentile. First, Table 2 shows a mean value of 0.0454 for the absolute discretionary accruals before the introduction of the new auditor’s report and a mean value of 0.0438 for the absolute discretionary accruals after the introduction of the new auditor’s report. Hence, the mean value of the absolute discretionary accruals has slightly decreased. The mean values of ABS_DA for both periods are similar to those of Barua, Davidson, Rama, and Thiruvadi (2010), who found a mean of 0.0530. The mean value of AB_CFO is lower after (−0.0026) the introduction of the new auditor’s report than before (0.0026) the introduction of the new auditor’s report, hence this indicates more real earnings management. AB_PROD also indicates more real earnings management, because the mean value of AB_PROD is higher after (0.0029) the introduction of the new auditor’s report than before (−0.0029) the introduction of the new auditor’s report. The mean value of AB_EXP decreases after the introduction of the new auditor’s report (0.0047 and −0.0047), hence this indicates more real earnings management. This is also the case for COMBINED_RAM (0.0102 and −0.0102), which indicates that the sample firms engage in more aggressive earnings management through real activities after the introduction of the new auditor’s report.

Concerning the control variable LOSS (values 0, 1, and 2), I found a mean value of 0.2484 and this means that most firms report a positive net income. Table 2 also shows that the mean size of the firms, measured by the natural log of total assets (LNTA), is 6.6637. Next to this, the mean value of GROWTH is 0.0699. Furthermore, I found a mean value of 0.5662 for LEV (a value of 0.500 or less is ideal), this indicates that approximately half of the company’s assets are financed by debt. The mean value of ROA is 0.0949, which is usually considered good. Finally, the mean value of SCOF is 0.0965, this indicates that firms are on average not that efficient in their business.

(26)

26

Table 2: Descriptive statistics of selected variables

N Mean Median Std. Deviation Min. Max. 25% 75%

Dependent variables ABS_DA before 308 .0454 .0364 .0351 .0018 .1926 .0195 .0610 ABS_DA after 308 .0438 .0333 .0347 .0016 .2177 .0183 .0596 AB_CFO before 308 .0026 −.0069 .0741 −.2070 .2221 −.0439 .0413 AB_CFO after 308 −.0026 −.0124 .0747 −.2168 .2249 −.0516 .0395 AB_PROD before 308 −.0029 .0431 .2708 −1.5378 .7387 −.1427 .1659 AB_PROD after 308 .0029 .0428 .2758 −1.6713 .7412 −.1284 .1652 AB_EXP before 308 .0047 −.0406 .2192 −.6703 .8175 −.1188 .0985 AB_EXP after 308 −.0047 −.0448 .2167 −.5882 .8322 −.1243 .0710 COMBINED_RAM before 308 .0102 −.0897 .5031 −1.4563 2.2515 −.2973 .2645 COMBINED_RAM after 308 −.0102 −.0915 .5040 −1.2895 2.3193 −.2968 .2187 Control variables LOSS 616 .2484 .0000 .5573 .0000 2.0000 .0000 .0000 LNTA 616 6.6637 6.6080 1.8770 1.9553 11.1489 5.4458 7.9628 GROWTH 616 .0699 .0428 .1501 −.2027 .7272 −.0168 .1184 LEV 616 .5662 .5625 .2195 .0966 1.2713 .4210 .7046 ROA 616 .0949 .0843 .0706 −.1569 .3024 .0580 .1304 SCFO 616 .0965 .0890 .0697 −.0934 .3062 .0542 .1353 Variable definitions:

ABS_DAip = absolute value of discretionary accruals for firm i in period p,

where discretionary accruals are calculated though the cross-sectional Modified Jones Model adjusted for performance; AB_CFOip = the level of abnormal cash flows from operations for firm i in period p;

AB_PRODip= the level of abnormal production costs for firm i in period p,

where production costs are the COGS plus inventory change;

AB_EXPip= the level of abnormal discretionary expenses for firm i in period p, where discretionary expenses are the SG&A expenses;

COMBINED_RAMip = AB_CFOip – AB_PRODip+ AB_EXPip;

LOSSip = set to 2 if firm i for both years has a negative income, set to 1 if firm i is has for one of the two year a negative income, and 0 otherwise;

LNTAip = natural log of total assets for firm i in period p;

GROWTHip = annual growth rate of total assets for firm i in period p;

LEVip = total liabilities divided by total assets for firm i in period p;

ROAip = earnings before interest and taxes divided by total assets for firm i in period p;

SCFOip = cash flows from operations divided by total assets for firm i in period p.

4.2 Correlations

Table 3 shows the Pearson correlations between the various variables of the two research methods; I use a significance level of 5%. The variable ABS_DA is significantly negatively correlated with LNTA, which indicates that the use of discretionary accruals will decrease with the size of the firm. Furthermore, it can be observed in Table 3, that ABS_DA is also significantly negatively correlated with ROA, which indicates a decrease of the discretionary accruals and hence an increase in audit quality when ROA increases. Furthermore, the variable ABS_DA is significantly positively correlated with LOSS, which indicates that firms with bad financial conditions use more discretionary accruals.

(27)

27

which means a decrease of aggressive earnings management through real activities (the audit quality increases), when the size of the firm also decreases. Furthermore, it can be observed in Table 3, that the variable COMBINED_RAM is significantly negatively correlated with LOSS, which indicates that firms with bad financial conditions engage in more aggressive earnings management through real activities. Next to this, the variable COMBINED_RAM is significantly positively correlated with ROA, which indicates a decrease of aggressive earnings management through real activities when ROA increases. Finally, it can be observed in Table 3, that there is a significantly positive correlation between COMBINED_RAM and SCFO, which means that a decrease of aggressive earnings management through real activities means an increase in SCFO.

The variable of interest, IAR_UK (research method 2), is not correlated with proxy 1 or proxy 2 for audit quality. This suggests that the introduction of the new auditor’s report is not related to the level of audit quality, which is not in line with my hypothesis.

To check for multicollinearity, I computed the Variance Inflation Factor (VIF). The VIF scores for all the variables of the two research methods are below 3 and hence below 10, meaning that I can safely assume that there is no multicollinearity of the control variables.

Table 3: Correlations among audit quality proxies, period of the introduction of the new auditor’s report, and other selected variables

1 2 3 4 5 6 7 8 9 10 11 12 1. ABS_DA 1 2. AB_CFO .059 1 3. AB_PROD −.020 −.397**₁ 4. AB_EXP .016 .201** _−.866**₁ 5. COMBINED_RAM .027 .450** _−.976**_.933** ₁ 6. LOSS .221** _−.300**_.062 _{−.038 −.094}*₁ 7. LNTA −.225**_.031 _.147** _−.164**_−.146**_−.096*₁ 8. GROWTH .078 .120** _−.043 _.087* _.078 _−.217**_{−.036 1} 9. LEV −.043 −.019 .053 −.104**_−.077 _.021 _.224**_−.171**₁ 10. ROA −.087*_.708** _−.343**_.185** _.371** _−.485**_.125**_.083* _.095*₁ 11. SCFO .018 .965** _−.381**_.202** _.437** _−.339**_.093*_.057 _{.024 .771}**₁ 12. IAR_UK −.023 −.035 .011 −.022 −.020 .114** _.033 _−.076 _{.003 −.061 −.033 1}

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

(28)

28

4.3 Results for research method 1: regression model for relating audit quality to control variables in combination with a Paired Samples test

Recall, that I compared the audit quality before (2010-2011) and after (2013-2014) the implementation of the new auditor’s report. First, I calculated for each company (the same 308 companies before and after) and each two-year time period the average value for each proxy for audit quality (absolute value of discretionary accruals and the level of real activities manipulation) and the average value of each of the seven control variables. Hence, I have 308×2=616 data points as input for the regression analysis for the equations (8) and (9). Second, I performed these regression analyses and computed for every point the residual. Third, I calculated for each of the 308 companies the difference of the residuals “before” and “after” (for both proxies of audit quality) and denote them with Di, i=1, …, 308. Fourth, I computed

D̅ (sample mean of the Di) and SD (sample standard deviation of the Di). Fifth, I calculated

subsequently the value of the test statistic for every proxy of audit quality. In Table 4 the statistics required for calculating the value of the test statistic and the calculated value of the test statistic itself can be observed.

Table 4: Paired Samples test for discretionary accruals and real activities manipulation

nD D̅ SD Test statistic ABS_DA 308 .0024 .0376 1.1202 AB_CFO 308 .0001 .0218 .0805 AB_PROD 308 .0021 .1060 .3477 AB_EXP 308 .0034 .0917 .6507 COMBINED_RAM 308 .0013 .1773 .1287

*. The value of the test statistic is significant at the 0.05 level (1-tailed).

The calculation for the values of the test statistics for the Paired Samples test for proxies 1 and 2 for audit quality are as follows:

ABS_DA: 0.0024/(0.0376/√308)=1.1202 AB_CFO: 0.0001/(0.0218/√308)=0.0805 AB_PROD: 0.0021/(0.1060/√308)=0.3477 AB_EXP: 0.0034/(0.0917/√308)=0.6507

(29)

29

Recall that D̅ should be significantly larger than 0 for ABS_DA and AB_PROD if the audit quality significantly increases through the introduction of the new auditor’s report in the UK and D̅ should be significantly lower than 0 for AB_CFO, AB_EXP, and COMBINED_RAM if the audit quality significantly increases through the introduction of the new auditor’s report in the UK. So more exactly, I finally checked if the value of the test statistic is larger than 1.645 or lower than −1.645, respectively. Since the critical values are 1.645 or −1.645, respectively, none of the values of the test statistics lead to rejection of the null hypothesis, that is, for all proxies of audit quality there is not enough evidence to prove my research hypothesis. Hence, the audit quality does not significantly increase through the introduction of the new auditor’s report in the UK.

4.4 Results for research method 2: regression model with a dummy for the introduction of the new auditor’s report in the UK

In this section I will describe the results of the regression analyses with a dummy variable. First I will look at the regression results for the absolute value of the discretionary accruals and second at the regression results for the four measures of real activities manipulation. All the statistical tests are performed at a significance level of 5%.

Proxy 1 for audit quality: discretionary accruals

Table 5 summarizes the results of the regression analysis for proxy 1: discretionary accruals. The R-Squared is 0.124, which indicates that 12.4% of the total variation of the absolute value of discretionary accruals (ABS_DA) is explained by the regression model. The p-value is 0.000, which shows that the model is significant.

It can also be observed in Table 5, that four of the control variables of the regression model are significant. The variables LNTA, GROWTH, LOSS, and SCFO have significant impact on the absolute value of discretionary accruals. However, I am interested in the variable IAR_UK. The coefficient of this variable is negative (−0.002) and not significant. The p-value

is namely 0.359/2=0.180 (p-value>0.05). This means that the introduction of the new auditor’s report in the UK has no significant effect on the absolute value of discretionary accruals (proxy 1 for audit quality).

This is not in line with what I expected: I cannot prove my research hypothesis. Hence, the audit quality does not significantly increase through the introduction of the new auditor’s report in the UK.

(30)

30

Table 5: Regression results of ABS_DA for the periods before and after the introduction of the new auditor’s report

ABS_DAip =  + 1IAR_UKp + 2COMBINED_RAMip + 3LNTAip + 4GROWTHip +

5LEVip + 6ROAip + 7LOSSip + 8SCFOip + ip

SS df MS

R Square Adjusted R Square F Sig. F Model .093 8 .012

.124 .113 10.770 .000 Residual .656 607 .001 Total .749 615 Unstandardized Coefficients N=616 B t Sig. (Constant) .059 9.597 .000 IAR_UK −.002 −.917 .359 COMBINED_RAM −.003 −1.025 .306 LNTA −.004 −5.347 .000 GROWTH .029 3.164 .002 LEV .003 .543 .587 ROA −.048 −1.484 .138 LOSS .016 5.576 .000 SCFO .104 3.365 .001

Proxy 2 for audit quality: real activities manipulation

Table 6 summarizes the results of the regression analyses for proxy 2: real activities manipulation. The R-Squared for the measures of real activities manipulation are 0.943, 0.201, 0.093, and 0.244 respectively. For the first measure of real activities manipulation, this indicates that 94.3% of the total variation of the level of abnormal cash flows from operations (AB_CFO) is explained by the regression model. On the other hand, 20.1% of the total variation of the level of abnormal production costs (AB_PROD) and 9.3% of the total variation of the level of abnormal discretionary expenses (AB_EXP) are explained by the regression model. Finally, 24.4% of the total variation of the combined measure of real activities manipulation (COMBINED_RAM) is explained by the regression model. The p-value is 0.000 for all the four regressions, which indicates that the model is significant for all four measures of real activities manipulation.

It can also be observed in Table 6, that four of the control variables of the regression model of the dependent variable AB_CFO are significant. The variables LNTA, GROWTH, ROA, and SCFO have significant impact on the level of abnormal cash flows from operations, but also in this case, the variable IAR_UK is the variable of interest. The coefficient of IAR_UK is negative (−6.191E−005) and not significant. Namely 1−0.966/2=0.517 (p-value>0.05). This

(31)

31

means that the introduction of the new auditor’s report in the UK has no significant effect on the level of abnormal cash flows from operations (measure 1 for proxy 2 for audit quality).

For the regression model of the dependent variable AB_PROD four of the control variables are significant. The variables LNTA, ROA, LOSS, and SCFO have significant impact on the level of abnormal production costs. However, the coefficient of IAR_UK is negative (−0.002) and not significant. The p-value is namely 0.914/2=0.457, which indicates that the introduction of the new auditor’s report in the UK has no significant effect on the level of the abnormal production costs (measure 2 for proxy 2 for audit quality).

Only three control variables are significant in the regression model of the dependent variable AB_EXP. The variables LNTA, ROA, and SCFO have significant impact on the level of abnormal discretionary expenses. Though, the coefficient for the variable of interest IAR_UK is negative (−0.003) and not significant, namely 1−0.839/2=0.581. This suggests that the introduction of the new auditor’s report in the UK has no significant effect on abnormal discretionary expenses (measure 3 for proxy 2 for audit quality).

Finally, for the regression model of the dependent variable COMBINED_RAM four of the control variables are significant. The variables LNTA, ROA, LOSS, and SCFO have significant impact on the combined measure of real activities manipulation. However, IAR_UK has a negative coefficient (−0.001) and is not significant, namely 1−0.970/2=0.515. This means that the introduction of the new auditor’s report in the UK has no significant effect on the combined measure of real activities manipulation (measure 4 for proxy 2 for audit quality).

In summary, these results are not in line with what I expected; I cannot prove my research hypothesis. Hence, the audit quality does not significantly increase through the introduction of the new auditor’s report in the UK.

(32)

32

Table 6: Regression results of RAM_PROXY on IAR_UK

RAM_PROXYip =  + 1IAR_UKp + 2ABS_DAip + 3LNTAip + 4GROWTHip +

5LEVip + 6ROAip + 7LOSSip + 8SCFOip + ip,

Panel A: Regression results of AB_CFO for the periods before and after the introduction of the

new auditor’s report

SS df MS

R Square Adjusted R Square F Sig. F Model 3.207 8 .401

.943 .942 1258.217 .000 Residual .193 607 .000 Total 3.401 615 Unstandardized Coefficients N=616 B t Sig. (Constant) −.088 −24.645 .000 IAR_UK −6.191E−005 −.043 .966 COMBINED_RAM .030 1.368 .172 LNTA −.002 −4.331 .000 GROWTH .033 6.569 .000 LEV −.005 −1.557 .120 ROA −.074 −4.230 .000 LOSS .003 1.860 .063 SCFO 1.095 66.690 .000

Panel B: Regression results of AB_PROD for the periods before and after the introduction of

the new auditor’s report

SS df MS

R Square Adjusted R Square F Sig. F Model 9.220 8 1.153

.201 .191 19.093 .000 Residual 36.640 607 .060 Total 45.861 615 Unstandardized Coefficients N=616 B t Sig. (Constant) −.017 −.347 .729 IAR_UK −.002 −.108 .914 COMBINED_RAM .328 1.081 .280 LNTA .027 4.855 .000 GROWTH −.052 −.750 .454 LEV .047 .999 .318 ROA −.833 −3.467 .001 LOSS −.066 −3.074 .002 SCFO −1.088 −4.813 .000

(33)

33

Panel C: Regression results of AB_EXP for the periods before and after the introduction of the

new auditor’s report

SS df MS

R Square Adjusted R Square F Sig. F Model 2.706 8 .338

.093 .081 7.758 .000 Residual 26.464 607 .044 Total 29.170 615 Unstandardized Coefficients N=616 B t Sig. (Constant) .092 2.208 .028 IAR_UK −.003 −.203 .839 COMBINED_RAM −.260 −1.008 .314 LNTA −.020 −4.312 .000 GROWTH .102 1.729 .084 LEV −.072 −1.796 .073 ROA .434 2.124 .034 LOSS .034 1.871 .062 SCFO .430 2.239 .026

Panel D: Regression results of COMBINED_RAM for the periods before and after the

introduction of the new auditor’s report

SS df MS

R Square Adjusted R Square F Sig. F Model 37.990 8 4.749

.244 .234 24.479 .000 Residual 117.753 607 .194 Total 155.743 615 Unstandardized Coefficients N=616 B t Sig. (Constant) .021 .242 .809 IAR_UK −.001 −.038 .970 COMBINED_RAM −.557 −1.025 .306 LNTA −.049 −4.928 .000 GROWTH .187 1.504 .133 LEV −.125 −1.472 .141 ROA 1.193 2.769 .006 LOSS .103 2.677 .008 SCFO 2.613 6.449 .000 4.5 Additional analyses

In this section I will perform two additional analyses. First, I will test the impact of the introduction of the new auditor’s report in the UK on positive discretionary accruals and negative discretionary accruals, separately. Second, I will test the impact of the introduction of

(34)

34

the new auditor’s report in the UK on the absolute value of abnormal working capital accruals (AWCA). Again, all the statistical tests are performed at a significance level of 5%.

4.5.1 Positive/negative discretionary accruals

In the main analysis I took the absolute value of the discretionary accruals. However, earnings can be managed either upwards or downwards (Dechow & Dichev, 2002), therefore in this additional analysis I am going to perform regressions with signed discretionary accruals to gain more insight. I will divide my sample into two groups: positive discretionary accruals (group one) and negative discretionary accruals (group two). The following empirical models will be used to test whether the introduction of the new auditor’s report in the UK has an impact on one of those groups:

POSITIVE_DAit =  + 1IAR_UKt + 2COMBINED_RAMit + 3LNTAit + 4GROWTHit +

5LEVit + 6ROAit + 7LOSSit + 8SCFOit + it, (13)

NEGATIVE_DAit =  + 1IAR_UKt + 2COMBINED_RAMit + 3LNTAit + 4GROWTHit +

5LEVit + 6ROAit + 7LOSSit + 8SCFOit + it, (14)

where:

POSITIVE_DAit = positive discretionary accruals for firm i in year t, where discretionary

NEGATIVE_DAit = negative discretionary accruals for firm i in year t, where discretionary

IAR_UKt = set to 1 if the period is after the introduction of the new auditor’s report in the UK

(2013 or 2014) and set to 0 if the period is before the introduction of the new auditor’s report in the UK (2010 or 2011);

COMBINED_RAMit = AB_CFOit – AB_PRODit + AB_EXPit;

The effect of the new auditor’s report in the UK on the audit quality

Contents

D

̅

S

D

/ √𝑛

D

_D

_{/ √𝑛}

_D