Audit quality: the effect of client importance on audit quality in Germany for second- and third-tier offices

(1)

Audit quality: the effect of client importance on audit quality in Germany

for second- and third-tier offices

Name: R. Boterman Student number: 1432788 University of Groningen

Faculty of Economics and Business Master Accountancy

Supervisor: dr. C.A. Huijgen Second assessor: dr. N. Hussain

(2)

2 Abstract

This study examines whether a high client importance influences audit quality in Germany.

Prior studies on the relation between client importance and audit quality show alternating

findings. Using discretional accruals and abnormal accruals as a proxy for audit quality, we

find evidence that for second-tier audit firms client importance at the audit office level has a

positive impact on audit quality. These results indicate that also second-tier auditors protect

their reputation and brand name at the audit office level.

(3)

3 1. Introduction

4 2. Literature review and hypotheses development

6 2.1 Audit quality

6 2.2 Client importance

8 2.3 Client importance and audit quality

9 2.4 Hypotheses development

10 3 Research Design

12 3.1 Sample and data

12 3.2 Audit quality

12 3.2.1 Modified Jones model

13 3.2.2 Defond and Park model

14 3.3 Client importance

15 3.4 Control variables

17 3.5 Regression model

18 4. Empirical Analysis and Results

19 4.1 Descriptive statistics

19 4.2 Regression results

21 4.2.1 Regressions hypotheses 1

21 4.2.2 Regressions hypotheses 2

23 4.2.3 Regressions hypotheses 3

25 4.3 Robustness checks

27 4.3.1. Defond and Park model

27 4.3.2. Adjusted classification first- and second tier firms

27 5. Conclusion

29 References

31

(4)

4 1. Introduction

Several noteworthy accounting scandals, including the WorldCom and Enron scandals, have

resulted in audit quality receiving a lot of attention from the media, politicians and

researchers. These well-known accounting scandals have partly arisen from aspects as

underreporting of costs, unreported loans, inflation of assets and misrepresentation of

revenues. As these scandals have resulted in huge losses for investors and society, new laws

were established to improve audit quality. To this aim, the Sarbanes-Oxley Act of 2002 was

established in the United States, while in the European Union the 8th EU Company Law

Directive serves this purpose (Braiotta & Zhou, 2008).

Auditors are entrusted with statutory audits by law. They fulfill an important role in giving an

assurance on whether the financial statements are without misstatements (Quick, 2012). The

opinion of the auditors that the financial statements are stated truly and fairly reduce the risk

of misstatement and subsequently reduce the costs of business failures. However, this can

only be achieved when a high-quality audit has been performed.

Therefore, according to DeAngelo’s (1981) definition, audit quality is a function of the

auditor’s ability to detect material misstatements (technical capabilities) and report the errors

(auditor independence). Unfortunately, these characteristics are largely unobservable.

Therefore, researchers have used different proxies to measure audit quality, such as

discretionary accruals, audit size, independence, audit fees, reputation, going concern

opinions, and litigation risks. According to Lennox (1999) researchers generally agree that

larger audit firms provide higher quality audits, resulting in greater credibility to clients'

financial statements. Therefore size of an audit firms is an adequate proxy of audit quality.

Francis (2004) identified a further wave of audit quality research which found that audit

quality is generally less homogenous within the Big Four firms. In his research, Francis

(2004) describe three sources of differences: institutional differences across countries,

differences across individual offices and differences due to industry specialization. This is in

contrast of the assumption of Clarkson and Simunic (1994), based on the first wave audit

quality research, that audit firms and auditors supply a single level of audit quality at a

moment in time.

According to DeAngelo (1981), auditors have an incentive to report in the interest of clients

because of their economic dependence. Big audit firms are less economically dependent

because they have multiple clients (DeAngelo, 1981). Reynolds and Francis (2001) define

(5)

5 economic dependence as the size of one client in relation to all audit clients of the audit firm.

Shafer, Morris, and Ketchland (1999) find evidence that the economic dependence endangers

the objectivity of the auditors.

In this thesis, we will explore audit quality for second- and third-tier offices in Germany and

whether client importance influences the audit quality in the German context. The German

audit market is more differentiated than markets in other countries and therefore very suitable

for investigating the quality of smaller audit firms. This finding is especially important as Big

Four companies are able to charge higher fees than non-Big Four auditors (Choi et al., 2008).

In this research we address the following main question:

What is the effect of client importance on audit quality in Germany for second- and third-tier

audit firms?

Prior research on client importance and the effect on audit quality mainly focused on Big Four

versus Non Big Four firms. In this study we will examine the national and regional audit

firms, mainly classified as Francis et al. (1999), in Germany. Gaver and Paterson (2007) find

that the important clients of Big Five firms in the U.S. apply less earnings management. In

this study we examine the negative impact of client importance on audit quality for the

second- and third tier auditors, and try to determine if they exhibit the same behavior as the

international audit firms regarding their important clients.

In Chapter 2, I will discuss the literature on audit quality and client importance. Based on the

literature discussed, we will also develop the hypotheses. Chapter 3 contains the methodology

that I use in my research. The results of my study are discussed in Chapter 4. In Chapter 5, I

will draw conclusion based on the findings.

(6)

6 2 Literature review and hypotheses development

2.1 Audit quality

The agency theory assumes that there is a separation between the owners (principal) and the

board (agent) of a firm leading to the principal-agent problem, as it is assumed that the owners

and the directors may have conflicting interests. Furthermore, according to the agency theory,

it is difficult for the owners (principal) to monitor the board (agent). In the agency theory,

corporate governance is the way to better control the company executives.

The agency theory supposes that the principals’ interests and the agent’s interest may not be

the same and thus monitoring of the management will decrease the agency costs (Jensen &

Meckling, 1976). Watts and Zimmerman (1983) conclude that an independent, external

auditor will decrease these costs further. The annual report of a company is an important

source of information for the shareholders. Becker, DeFond, Jiambalvo, and Subramanyam

(1998) indicate that auditing reduces the information asymmetry between the managers and

the stakeholders of the company, especially through an independent audit of the annual report

of the manager.

Knechel et al. (2013) suggest that stakeholders have different views regarding what

constitutes audit quality. Users of financial reports can view high audit quality as the absence

of material misstatements in the financial statements. For an auditor, high audit quality can

mean complying with the tasks assigned by the audited company. Society, however, may

associate high audit quality with the absence of economic problems for a company. Since

stakeholders have different views on audit quality, it is difficult to state what the one defining

criterion of audit quality is (Knechel et al., 2013). DeAngelo (1981) recognizes in her

definition of audit quality two key components, namely, the competence and independence of

the auditor.

According to DeAngelo (1981), “the quality of audit services is defined to be the

market-assessed joint probability that a given auditor will both discover a breach in the client's

accounting system; and report the breach”. The probability of detecting misstatements

depends on the competence of the auditor. Whether these misstatements are reported,

however, depends on the independence of the auditor.

Previous research suggests that when an auditor performs both audit and advisory services for

a client, there can be a knowledge transfer which leads to cost savings for the auditor.

(7)

7 According to Simunic (1984), this implies that the auditor is economically less objective

regarding the accounting choices of the management. Furthermore, performing audit as well

as advisory services can create a conflict of interest as there is an incentive to not report

consulting deficiencies observed during the audit to avoid the erosion of its consulting brand

name (Simunic, 1984). DeAngelo (1981) also addresses the audit firm’s economic

dependence on the client and claims that the auditor counts on the quasi-rents it earns when it

retains the relationship with the client.

Audit quality can be related to the type of audit firm performing the audit. According to

DeAngelo (1981), audit quality is higher when a big audit firm performs the audit because a

big audit firm is not economically dependent on the fees of a single client (Reynolds &

Francis, 2001). Furthermore, the auditor risks to loose more if a misstatement occurs in the

audit, because this can lead to the loss of other audit clients.

Narayanan (1995) and Francis, Reichelt, and Wang (2005) also indicate that audit quality

should be assessed at the office level instead of at the firm level. Other studies show that there

are significant differences between audit firms’ local offices (Gaver & Paterson, 2007), which

supports Francis (2004) conclusion. Additionally, Francis and Yu (2009) conclude that the

larger Big Four offices provide higher quality audits, in the way of the likelihood of issuing

going concern audit reports and less aggressive earnings management in the audited reports.

In most research on audit quality, audit firms are divided into two groups: “Big Five” firms

(PWC, KPMG, Deloitte, Arthur Andersen and EY) and non-Big Five firms. After 2002, the

Big Five became the Big Four because Arthur Andersen disappeared due to its involvement in

the Enron scandal.

To protect their reputation, big audit firms perform audits with higher quality (Becker et al.,

1998). Moreover, big audit firms are able to charge higher audit fees compared to the non-Big

Five (or Four) firms (Choi et al., 2008). Due to higher audit fees, big audit firms have the

ability to perform higher quality audits. Gul et al. (2009) conclude that industry specialization

of auditors positively affects the earnings quality, a proxy for audit quality.

Another conclusion, according to Francis (2004), could be that clients with high-quality

financial statements would like to be audited by one of the big audit firms because of the

reputation of these firms. This implies that it is not only the auditor that ensures the high

quality of financial statements.

(8)

8 2.2 Client importance

Audit firms are economically dependent on their clients and therefore auditors have an

incentive to report more favorably in the interest of the clients, in order to retain them

(DeAngelo, 1981). This can be described as economic dependence. When an audit firm has

multiple clients, there is less economic dependence on each client. DeAngelo (1981)

concluded that the size of the audit firm is a proxy for the firm’s independence and audit

quality because larger audit firms have more clients than smaller firms have.

Large audit clients ensure economic dependence that could endanger the auditor’s

independence, causing the auditor to report in accordance with its clients’ interests. Economic

dependence can be defined as the size of one client in relation to all audit clients of the audit

firm (Reynolds & Francis, 2001). Shafer, Morris, and Ketchland (1999) find indications that

economic dependence endangers the objectivity of the auditors, as they tend to report more

favorably for large clients. Large clients also increase the auditor’s risk, since auditors may be

held personally liable for uncorrected misstatements leading to reputational damage for the

firm (Shafer et al., 1999). Auditors become more economically dependent on their clients, and

do not want to dissatisfy and thereby lose their clients. Furthermore, auditors are reluctant to

be critical about the reporting of the client, despite the risks of lawsuits.

According to other researchers, the risks of reputational damage and litigation to remain

independent are influenced by other factors. Carey and Simnett (2006) find evidence that long

audit partner tenure is negatively correlated with audit quality. Jackson, Moldrich, and

Roebuck (2008) have studied audit firm tenure and concluded that there are benefits of

mandatory audit firm rotation.

The independence of auditors is also threatened by client importance (DeAngelo, 1981), due

to the economic dependence of the audit firm on the earned fees, for both audit and non-audit

services. According to Simunic (1980), the level of audit fees depends on the size of the client

measured by year-end total assets, but also on the complexity of the client measured by the

number of consolidated subsidiaries, the number of Standard Industrial Classification System

and the ratio of foreign assets and total assets. Furthermore, Simunic (1980) concludes that

the ratios of receivables to assets and inventories to assets are determinants for the size of

audit fees.

Taylor and Simon (1999) find that, besides clients’ characteristics, litigation and regulation

are also determinants of audit fees. Their research within twenty countries shows that

(9)

9 increased risk of litigation, increased client disclosure, and increased regulation results in

higher audit fees.

According to Li (2009), client importance is normally measured as the ratio of the audit fee

for a client to the total audit fees earned by an audit firm or office. When data for audit fees

are not available, other variables are used such as the ratio of client’s sales to total sales

audited by the audit firm or office (Reynolds & Francis, 2001), or the ratio of client’s assets to

the total assets audited by the audit firm or office (Chen et al., 2010).

2.3 Client importance and audit quality

There are different methods used to measure audit quality. Lennox (1999) focuses on the

relationship between the incorporation of a going concern opinion in the audit report and audit

quality. Lennox (1999) concludes that the auditor often wrongly incorporates the going

concern opinion. In addition, his findings show that when a company has gone bankrupt, the

auditor often has failed to include a going concern opinion in the audit report. Francis (2011)

studied the number of audit failures mentioned in court cases against the auditor and the

findings of the Securities and Exchange Commission.

Most researchers use earnings management as a proxy to measure the quality of the audit.

Previous research has concluded that high audit quality mitigates the risk of abusing the

flexibility in reporting rules (Barth et al., 2008). Lower discretionary accruals indicate higher

quality audits. Gunny and Zhang (2013) show that the inspected auditors engagements that

receive a deficient or seriously deficient report by the Public Company Accounting Oversight

Board are associated with significantly higher discretionary accruals.

Previous research has shown that audit fees are correlated with client importance. Frankel et

al. (2002) find that the size of audit fees is associated with earnings management. They

provide evidence that non-audit fees are associated with earnings surprises and discretionary

accruals. Therefore, they claim that due to additional fees auditors have become less

independent from their clients. Other researchers, including Ashbaugh, LaFond, and Mayhew

(2003), do not find evidence which supports the association found by Frankel et al. (2002)

between non-audit fees and earning management that would suggest auditors compromise

their independence by offering additional services, such as consultancy or tax advice.

Like Frankel et al. (2002), Larcker and Richardson (2004) also examine the relation between

audit and non-audit fees on the one hand and accruals on the other hand. However, they find a

(10)

10 negative relation between the level of fees and accruals. Higher fees were associated with

smaller accruals, especially for clients with a weak corporate governance. The result is

consistent with the auditors’ concern for reputation, when clients have the ability to report

unusual accruals.

According to Francis (2004), analyzing specific offices of large accounting firms provides

more insight than studying the whole firm as one entity. The rationale for this approach is that

individual audit clients are managed by a local office clients partner. Earlier research by

Reynolds and Francis (2001), however, did not show evidence that offices of big audit firms

report more favorably for larger clients in their office portfolio. Auditors of big audit firms

report more conservatively for larger clients, which indicates that they protect their brand

name.

Narayanan (1995) and Francis, Reichelt, and Wang (2005) also indicate that audit quality

should be assessed at an office level instead of a firm level. Other studies show that there are

significant quality differences between audit firms local offices (Gaver & Paterson, 2007),

supporting Francis’ (2004) conclusion. Gaver and Paterson (2007) find evidence that

financially struggling insurance clients normally tend to under-reserve; however, when this

client is important for an audit office, there is less of a tendency to under-reserve.

2.4 Hypotheses development

Most of the previous studies on audit quality examine the relation between discretionary

accruals and client importance, explaining their findings by referring to reputation protection

or economic dependence (Reynolds & Francis, 2001). The discussed literature draws

conclusions about big audit firms. Based on these conclusions, we developed hypothesis 1,

which assumes that client importance is not associated with audit quality for large audit firms

and offices.

Hypothesis 1:

Client importance is not associated with audit quality for first-tier audit firms and offices in

Germany.

Francis et al. (1999) recognize that the commonly used separation between big audit firms and

non-big audit firms will not fully cover the audit landscape. Francis et al. (1999) divide the

population into three groups. The first tier consists of large international firms. National audit

firms make up the second tier and the third tier comprises local and regional audit firms.

(11)

11 Partly based on a further breakdown of these firms, I have compiled the following hypotheses,

in which we assume that the second- and third-tier audit firms provide lower audit quality

(Francis et al., 1999) because the client importance is higher.

Hypothesis 2:

Client importance is negatively associated with audit quality for second- and third-tier audit

firms in Germany.

Francis, Reichelt, and Wang (2005) and Narayanan (1995) also declare that audit quality

should be assessed at the audit office level instead of at the audit firm level. Choi et al. (2010)

also find a clear relation between audit quality measured by discretionary accruals and local

office size. Therefore, we developed the following hypothesis.

Hypothesis 3:

Client importance on the local office level is negatively associated with audit quality for

second- and third-tier audit offices in Germany.

(12)

12 3 Research Design

3.1 Sample and data

For this research, we analyze data from the years 1999 to 2009 for German public companies.

The total data set includes 5013 firm years for which the data are complete with respect to

accounting variables, audit firms and audit offices. The data do not contain companies with an

SIC code between 6000 and 6999. These companies are financial institutions. They are

removed from the data due to their nature and accounting policies, which makes it difficult to

measure discretionary accruals (Choi et al., 2010).

3.2 Audit quality

The quality of the audit can be measured by the degree of earnings management. The absolute

value of discretionary accruals is used as a proxy for earnings management. The extent to

which the company has the ability to influence earnings can be measured by using different

models (Jones et al., 2008). In the earnings management literature, the accrual component of

earnings is commonly calculated from balance sheet and income statement items (Dechow et

al., 1995). Furthermore, Dechow et al. (1995) conclude that the Modified Jones model

provides the most powerful test of earnings management. A later study of Jones, Krishnan,

and Melendrez (2008) identifies nine competing models that are commonly used to capture

earnings management. They also find that the cross sectional Modified Jones model provides

the best results. According to them, the second-best model is the Industry model developed by

Dechow and Sloan (1991). However, to increase the robustness of our outcomes, we will use

the discretionary working capital accruals model used by DeFond and Park (2001), because

this model performs well with non-US data (Peek et al., 2013).

For the models used in this research, we identify the amount of the discretionary accrual

component. Discretionary accruals provide management the opportunity to apply earnings

management. Managers have to make forecasts, estimates and judgments (Dechow &

Schrand, 2004), and therefore they influence earnings. Dechow and Schrand (2004) give

several examples of accruals with a high discretion:

• Accounts receivable

Managers forecast expected product returns and the proportion

of customers who will not pay.

• Inventory

Managers capitalize some costs in inventory and expense other

costs as periodic expenses. They forecast expected demand in

order to determine future sales prices and whether a write-down

is necessary.

(13)

13 • Other current assets

This account is typically a catchall category for capitalized

costs.

• Property, plant, and

Managers capitalize a multitude of costs and depreciate them in

equipment (PP&E)

arbitrary ways. Managers must also forecast future demand to

determine whether an impairment has occurred.

• Pension liabilities and

Managers must forecast the expected return on pension plan

post-retirement benefits

assets and obtain actuarial assumptions on life expectancies.

3.2.1 Modified Jones model

The Modified Jones model is an adjustment to the original Jones model made by Dechow,

Sloan, and Sweeney (1995). According to Dechow et al. (1995), the original Jones model

does not account for the possibility of earnings management regarding revenues. In the

Modified Jones model, revenues are corrected for credit sales in determining the

non-discretionary accruals.

The discretionary accruals are calculated by measuring the non-discretionary accruals as a

portion of the total accruals in the Modified Jones model.

𝐷𝐴

_𝑡

= 𝑇𝐴

_𝑡

− 𝑁𝐷𝐴

_𝑡

𝐷𝐴

_𝑡

= Discretionary accruals in year 𝑡,

𝑇𝐴

𝑡

= Total accruals in year 𝑡,

𝑁𝐷𝐴

𝑡

= Non-discretionary accruals in year 𝑡,

The following equation is used to measure total accruals:

𝑇𝐴

_𝑡

= ∆𝐶𝐴

_𝑡

− ∆𝐶𝑎𝑠ℎ − ∆𝐶𝐿

_𝑡

+ ∆𝐷𝐶𝐿

_𝑡

− 𝐷𝐸𝑃

_𝑡

𝑇𝐴

𝑡

= Total accruals in year 𝑡,

∆𝐶𝐴

𝑡

= Change in current assets in year 𝑡,

∆𝐶𝑎𝑠ℎ

= Change in cash and cash equivalents in year 𝑡,

∆𝐶𝐿

𝑡

= Change in current liabilities in year 𝑡,

∆𝐷𝐶𝐿

𝑡

= Change in short-term debt included in current liabilities in year 𝑡,

𝐷𝐸𝑃

𝑡

= Depreciation and amortization expense in year 𝑡.

The non-discretionary accruals can be estimated as follows:

𝑁𝐷𝐴

_𝑡

𝑇𝐴

𝑡−1

= 𝛽

1

1 𝑇𝐴

𝑡−1

+ 𝛽

2

(∆𝑅𝐸𝑉

_𝑡

− ∆𝑅𝐸𝐶

_𝑡

)

𝑇𝐴

𝑡−1

+ 𝛽

3

𝑃𝑃𝐸

_𝑡

𝑇𝐴

𝑡−1

(14)

14 ∆𝑅𝐸𝑉

𝑡

= Change in revenues in year 𝑡,

∆𝑅𝐸𝐶

𝑡

= Change in net receivables in year 𝑡,

𝑃𝑃𝐸

𝑡

= Gross property, plant and equipment in year 𝑡,

𝑇𝐴

𝑡−1

= Total assets in year 𝑡 − 1,

β

1

, β

2

, and β

3

= Estimated parameters, obtained from the original Jones model

For the estimation of the industry-specific coefficients, we use the industry classification used

by Barth et al. (1998) instead of the two-digit SIC code classification. Furthermore, we need

at least 10 observations per industry classification per year. Table 1 provides an overview of

the classification of industry groups and the number of observations classified per year. This

shows that the durable manufacturers represented the largest part of the total number of

observations for measuring discretionary accruals and there is a fairly even distribution for the

different periods.

Table 1 Data Sample

Industry description according

Barth et al., (1998) SIC codes N Year N

1. Mining and construction 1000–1999, except 1300–1399 191 1999 399

2. Food 2000–2111 243 2000 355

3. Textiles, printing and publishing 2200–2799 364 2001 431

4. Chemicals 2800–2824, and 2840–2899 166 2002 478

5. Pharmaceuticals 2830–2836 135 2003 508

7. Durable manufacturers 3000–3999, except 3570–3579, and 3670–3679 1515 2004 495

8. Computers 7370–7379, 3570–3579, and 3670–3679 845 2005 481 9. Transportation 4000–4899 240 2006 441 10. Utilities 4900–4999 170 2007 452 11. Retail 5000–5999 531 2008 475 14. Services 7000–8999, except 7370–7379 613 2009 498 Total 5013 Total 5013

3.2.2 Defond and Park model

For robustness of this study we use the Defond and Park model (2001), which measures the

discretionary working capital accrual scaled by lagged assets and can be used as a proxy for

earnings management. In this model the discretionary working capital accrual is calculated as

the difference between expected accruals and working capital accruals.

The discretionary working capital accrual can be calculated as follows:

𝐷𝑊𝐶𝐴

𝑡

𝑇𝐴

_𝑡−1

= W𝐶

t

− (

(𝑊𝐶

𝑡−1

)

𝑅𝐸𝑉

_𝑡−1

∗ 𝑅𝐸𝑉

t

)

𝐷𝑊𝐶𝐴

𝑡

= discretionary working capital accrual in year t

W𝐶

t

= (current assets – cash and cash equivalents) - (current liabilities – the current

(15)

15 𝑅𝐸𝑉

t

= Change in revenues in year 𝑡

𝑇𝐴

𝑡−1

= Total assets in year 𝑡 − 1,

3.3 Client importance

Client importance is normally measured based on audit fees and non-audit fees. When these

data are not available, other variables have to be used. Reynolds and Francis (2001) use the

ratio of the client’s sales to the total sales audited by the audit firm or office as a proxy for

audit fees. Chen et al. (2010) use the ratio of the client’s assets to the total assets audited by

the firm or office. Fleischer and Goettsche (2012) find in their research that, for all clients,

size based on total assets is a predictor of audit fees.

For the measurement of client importance on the audit firm level, we use the following

equations:

CI

−

FI_TA

t

=

Ln (ASSET𝑆

_𝑡

)

∑

(Ln (ASSET𝑆

𝑡

)

𝑘 𝑛=1

or CI

−

FI_REV

t

=

Ln (REV

_𝑡

)

∑

(Ln (REV

𝑡

)

𝑘 𝑛=1

CI

₋

FI_TA

_t

= Client importance based on the natural logarithm of client’s assets as a fraction of

the sum of the natural logarithm of all client’s assets on firm level in year t

CI

₋

FI_REV

_t

= Client importance based on the natural logarithm of client’s revenues as a fraction

of the sum of the natural logarithm of all client’s revenues on firm level in year t

ASSET𝑆

𝑡

= Total assets in year t

RE𝑉

_𝑡

= Revenues in year t

For the measurement of client importance on the audit office level, we use the following

equations:

CI

−

OFF_TA

t

=

Ln (ASSET𝑆

_𝑡

)

∑

(Ln (ASSET𝑆

𝑡

)

𝑘 𝑛=1

or CI

−

OFF_SAL

t

=

Ln (REV

_𝑡

)

∑

(Ln (REV

𝑡

)

𝑘 𝑛=1

CI

₋

OFF_TA

_t

= Client importance based on the natural logarithm of client’s assets as a fraction of

the sum of the natural logarithm of all client’s assets on office level in year t

CI

₋

OFF_REV

_t

= Client importance based on the natural logarithm of client’s revenues as a fraction

of the sum of the natural logarithm of all client’s revenues on office level in year t

ASSET𝑆

𝑡

= Total assets in year t

RE𝑉

_𝑡

= Revenues in year t

(16)

16 findings of Fleischer and Goettsche (2010) on audit pricing in Germany. They find that audit

fees, used as a proxy for client importance, are related to size measured by total assets or

revenues. For smaller clients, the leverage ratio and past losses are also determinants of audit

fees. Clients with the lowest ranking based on total assets, total sales, leverage and previous

year losses receive the dummy variable 1.

For the measurement of client importance on the audit firm level, we use the following

equation:

CI

₋

FI_RANK

_t

= RANK(CI

₋

FI_TA

_t

) + RANK(CI

₋

FI_REV

_t

) + RANK(

(𝑇𝐿𝐷

𝑡

+ 𝐿𝐷𝐼𝑆𝐿

𝑡

)

𝑇𝐴

_𝑡

)

+ LOSS

t−1

CI

₋

FI_TA

_t

= Client importance based on the natural logarithm of client’s assets on firm level in

year t ranked from high to low.

CI

₋

FI_REV

_t

= Client importance based on the natural logarithm of client’s revenues on firm level

in year t ranked from high to low

𝑇𝐿𝐷

_𝑡

+ 𝐿𝐷𝐼𝑆𝐿

_𝑡

= Total long-term debt + current portion of the long term-debt in the short-term

liabilities ranked from high to low based on firm level

𝑇𝐴

𝑡

= Total assets in year t

LOSS

_t−1

=Loss in year t-1 Is a dummy variable with a value of 0 for losses and 1 for profits.

For the measurement of client importance on the audit office level, we use the following

equation:

CI

₋

OF_RANK

_t

= RANK(CI

₋

OFF_TA

_t

) + RANK(CI

₋

OFF_REV

_t

)

+ RANK(

(𝑇𝐿𝐷

𝑡

+ 𝐿𝐷𝐼𝑆𝐿

𝑡

)

𝑇𝐴

_𝑡

) + LOSS

t−1

CI

₋

FI_TA

_t

= Client importance based on the natural logarithm of client’s assets on office level in

year t ranked from high to low.

CI

₋

FI_REV

_t

= Client importance based on the natural logarithm of client’s revenues on office

level in year t ranked from high to low

𝑇𝐿𝐷

𝑡

+ 𝐿𝐷𝐼𝑆𝐿

𝑡

= Total long-term debt + current portion of the long term-debt in the short-term

liabilities ranked from high to low based on office level

𝑇𝐴

_𝑡

= Total assets in year t

(17)

17 3.4 Control variables

Audit quality is measured by discretional accruals, which are calculated based on the

Modified Jones model or as discretionary working capital accruals which are calculated based

on the Defond and Park model. However, there are other variables that may affect the level of

audit quality. As previous studies have done, we also add commonly used control variables to

our regression models. The control variables we use in our models are total assets (TA),

revenues (REV), leverage (LEV), return on assets (ROA), book–to-market ratio (BM) growth

(GROW), loss (LOSS) and cash flow from operations (CFO).

Variable total assets is measured as the natural logarithm of total assets (Simunic, 1980) and

variable revenues is measured as the natural logarithm of total revenues (Reynolds & Francis,

2001). Variable total assets and variable revenues have a negative influence on the

discretional accruals due to better financial reporting systems (Reynolds & Francis, 2001).

Leverage is measured by dividing total long-term debt plus debt in current liabilities by total

assets (Johnson et al., 2002). Leverage has a positive influence on the discretional accruals

because companies will manage their earnings to comply with the terms of creditors (Johnson

et al., 2002).

Return on assets is measured by the absolute value of income before extraordinary items

divided by average total assets in the previous year (Lawrence et al., 2011). Companies with

extreme performance, either positive or negative, are likely to engage in earnings

management (Kothari et al., 2005). Therefore, return on assets has a positive relationship with

discretionary accruals.

Book-to-market ratio is measured as book value of equity divided by the market value of the

equity. Companies with a low book-to-market ratio have high growth opportunities and are

more likely to apply earnings management (Skinner & Sloan, 2002). Book-to-market value

thus has a negative relationship with discretionary accruals.

Growth is measured as the percentage of growth based on revenues with respect to the

revenues in the previous year. Matsumoto (2002) shows that companies manage their earnings

upwards to fulfill the expectation of the market. Growth has a positive relationship with

discretionary accruals.

Loss is measured as dummy variable 1 if the company reports a net loss in the current year

and as 0 for a current year profit. Variable loss has both a negative as well a positive

(18)

18 relationship with discretionary accruals. Companies which reported losses are more likely to

overstate their costs in favor of next year profits or report a small earnings surprise to prevent

reporting earnings that miss analyst estimates (Brown 2001).

Cash flow from operations is measured as cash flow from operations divided by the lagged

total assets. According to Kothari, Leone and Wasley (2005), accruals are positively related

with companies’ performance. Cash flow from operations has thus a positive correlation with

discretionary accruals.

3.5 Regression model

Based on previous research, we use the absolute value of discretional accruals as a proxy for

audit quality. To test the hypotheses, in which the relationship between client importance and

audit quality is examined, we use the following regression models:

DA

_ABS

= β

₀

+ β

₁

CI

₋

FI_TA

_t

+ β

₂

TA

_t

+ β

₃

LEV

_t

+ β

₄

ROA

_t

+ β

₅

BM

_t

+ β

₆

GROW

_t

+ β

₇

LOSS

_t

+ β

₈

CFO

_t

DA

_ABS

= β

₀

+ β

₁

CI

₋

FI_REV

_𝑡

+ β

₂

REV

_t

+ β

₃

LEV

_t

+ β

₄

ROA

_t

+ β

₅

BM

_t

+ β

₆

GROW

_t

+ β

₇

LOSS

_t

+ β

₈

CFO

_t

DA

_ABS

= β

₀

+ β

₁

CI

₋

OFF_TA

_t

+ β

₂

TA

_t

+ β

₃

LEV

_t

+ β

₄

ROA

_t

+ β

₅

BM

_t

+ β

₆

GROW

_t

+ β

₇

LOSS

_t

+ β

₈

CFO

_t

DA

_ABS

= β

₀

+ β

₁

CI

₋

OFF_REV

_t

+ β

₂

REV

_t

+ β

₃

LEV

_t

+ β

₄

ROA

_t

+ β

₅

BM

_t

+ β

₆

GROW

_t

+ β

₇

LOSS

_t

+ β

₈

CFO

_t

DA

_ABS

= β

₀

+ β

₁

CI

₋

FI_RANK

_t

+ β

₂

TA

_t

+ β

₂

REV

_t

+ β

₃

LEV

_t

+ β

₄

ROA

_t

+ β

₅

BM

_t

+ β

₆

GROW

_t

+ β

₇

LOSS

_t

+ β

₈

CFO

_t

DA

_ABS

= β

₀

+ β

₁

CI

₋

OF_RANK

_t

+ β

₂

TA

_t

+ β

₂

REV

_t

+ β

₃

LEV

_t

+ β

₄

ROA

_t

+ β

₅

BM

_t

+ β

₆

GROW

_t

+ β

₇

LOSS

_t

+ β

₈

CFO

_t

We will run the regression separately for the first-tier, second-tier and third-tier audit firms.

Furthermore, we run this regression with the DWCA_ABS instead of DA_ABS to verify the

robustness of our outcomes.

(19)

19 4. Empirical Analysis and Results

4.1 Descriptive statistics

Table 2 provides an overview of the variables used in the regression model from 1999 to

2009. The average of the absolute value of discretionary accruals is 0.0849. In the research

conducted by Chen et al. (2010), this value is quite similar (0.10). The difference between

these two values may be due to the difference in the number of years and observations in the

sample. The average of the client importance at firm level (CI_FI_TA and CI_FI_SAL ) is

lower than at office level (CI_OFF_TA and CI_OFF_SAL), which means that the client

importance at office level is generally higher than at firm level, due to fewer clients at office

level. Additionally, for the dichotomous client importance (CI_FI_RANK and

CI_OFF_RANK), there is the same difference between firm level and office level. All

continuous variables are winsorized at the 1

st

_{and 99}

th

_{percentiles of their distribution,}

reducing the influence of outliers. The actual values of the client importance variables are

lower because 21% of the total sample consists of firms with fewer than four clients per year.

Table 2 Descriptive Statistics

N

Mean Median

Std.

Deviation Minimum Maximum

Valid Missing DA_ABS 4990 23 0,085 0,056 0,091 0,000 0,474 DWCA_ABS 5003 10 0,081 0,046 0,100 0,000 0,571 TA 5013 0 5,043 4,829 2,132 -0,740 12,477 REV 4996 17 5,105 4,932 2,166 -0,043 10,968 CFO 5013 0 0,063 0,075 0,171 -0,560 0,634 LOSS 5013 0 0,313 0,000 0,464 0,000 1,000 LEV 5001 12 0,582 0,602 0,227 0,045 1,147 BM 5013 0 0,759 0,609 0,684 -0,881 3,394 ROA 5013 0 0,089 0,051 0,114 0,000 0,654 GROW 4978 35 0,051 0,030 0,288 -0,709 1,262 CI_FI_TA 5013 0 0,590 0,535 0,260 -0,089 1,000 CI_FI_SAL 4996 17 0,593 0,549 0,267 -2,805 1,000 CI_FI_RANK 5011 2 0,216 0,000 0,412 0,000 1,000 CI_OFF_TA 5013 0 0,739 0,769 0,244 -0,098 1,000 CI_OFF_SAL 4996 17 0,737 0,776 0,256 -2,805 1,000 CI_OFF_RANK 5011 2 0,387 0,000 0,487 0,000 1,000

ABS_DA: absolute value discretionary accruals: DWCA_ABS: absolute value discretionary working capital accruals; TA: natural logarithm of total assets; REV: natural logarithm of total revenues; CFO: cash flow from operations divided by total assets; LOSS: dummy variable 1 if the company reports a net loss in the current year and as 0 for profits; LEV: leverage ratio, measured as total liabilities divided by total assets; BM: book-to-market ratio, measured as book value divided by the market value of equity; ROA: net income before extraordinary items divided by net assets; GROW: Percentage change of net sales; CI_FI_TA: client importance at firm level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by a firm in a given year; CI_FI_SAL: client importance at firm level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by a firm in a given year; CI_FI_RANK; dummy variable for client importance at firm level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by a firm in a given year; CI_OFF_TA: client importance at office level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by an office in a given year; CI_OFF_SAL: client importance at office level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by an office in a given year; CI_OFF_RANK: dummy variable for client importance at office level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by an office in a given year.

(20)

20 The mean values of the control variables (CFO, LEV, BM, ROA, GROW) indicate that the

companies are, in general, financially healthy in the period 1999 to 2009. Furthermore, of all

5,013 companies’ years, 69% of the years were profitable.

Table 3 shows the allocation of the audit firms to the first-, second- and third- tier group. It is

noteworthy that the current Big Four firms provide higher audit quality, as shown in

Appendix 1 by the lower absolute discretional accruals measured by the Modified Jones

model (DA_ABS) and the Defond and Park model (DWCA_ABS).

Table 3 Classification of audit firms

Tier Audit Firm Name N DA_ABS DWCA_ABS

First tier Arthur Andersen 112 0,1001 0,0993

EY 851 0,0830 0,0802

KPMG 864 0,0732 0,0686

PWC 772 0,0826 0,0801

Total First tier 2599 0,0804 0,0771

Second tier BDO 339 0,0902 0,0808

Deloitte 239 0,0697 0,0647

Ebner Stolz Mönning Bachem GmbH & Co. KG 130 0,1004 0,0750

MAZARS GmbH 14 0,1091 0,1363

RSM Hemmelrath GmbH 40 0,0939 0,0932

Susat & Partner 90 0,0866 0,0797

Total Second tier 852 0,0861 0,0768

Third tier AWT Horwath 58 0,0822 0,0732

Dr. Kleeberg & Partner GmbH 32 0,0879 0,0763

Grant Thornton GmbH 24 0,0741 0,0621

Pannell Kerr Forster GmbH 66 0,1174 0,0987

Rödl & Partner 66 0,0686 0,0821

Rölfs WP Partner 48 0,0686 0,0751

Röverbrönner GmbH & Co 25 0,0797 0,0811

RP Richter GmbH 12 0,0726 0,0769

SüdTreu Süddeutsche Treuhand AG 34 0,0801 0,0542

WAPAG 16 0,0867 0,0491

Warth & Klein 57 0,1108 0,0985

Wirtschaftstreuhand GmbH 25 0,1059 0,1431

Wollert-Elmendorff Deutsche Industrie Treuhand

GmbH 40 0,0778 0,0720

Total Third tier 503 0,0874 0,0823

Third tier with less than 4 clients 1059 0,0937 0,0916

Total 5013 0,0849 0,0806

ABS_DA: absolute value discretionary accruals: DWCA_ABS: absolute value discretionary working capital accruals.

Table 4 presents the Pearson pairwise correlations among variables included in the regression

equations. The Pearson correlations are provided for two different significant levels at 5% and

1%, respectively. Based on auditors’ tendency to protect their brand name and reputation, we

expect a negative correlation between the discretionary accruals and client importance. These

expected correlations are significant for the continuous variables of client importance, but not

for the dichotomous variable for client importance.

(21)

21

Table 4 Pearson pairwise correlations

₍₁₎ ₍₂₎ ₍₃₎ ₍₄₎ ₍₅₎ ₍₆₎ ₍₇₎ ₍₈₎ ₍₉₎ ₍₁₀₎ ₍₁₁₎ ₍₁₂₎ ₍₁₃₎ ₍₁₄₎ ₍₁₅₎ ₍₁₆₎ ABS_DA (1) 1,00 DWCA_ABS (2) ,691** 1,00 TA (3) -,202** -,161** 1,00 REV (4) -,211** -,164** ,953** 1,00 CFO (5) -,034* ,032* ,146** ,180** 1,00 LOSS (6) ,096** ,056** -,272** -,299** -,386** 1,00 LEV (7) ,035* ,042** ,232** ,300** -0,01 ,134** 1,00 BM (8) -,072** -,092** -,01 -0,02 -,045** ,132** -,210** 1,00 ROA (9) ,181** ,103** -,304** -,315** -,350** ,371** -,108** -,091** 1,00 GROW (10) ,092** ,228** ,059** ,084** ,074** -,184** -,076** -,085** -,094** 1,00 CI_OFF_TA (11) -,122** -,104** ,556** ,524** ,078** -,215** ,082** ,035* -,199** ,084** 1,00 CI_OFF_SAL (12) -,131** -,105** ,537** ,594** ,108** -,238** ,157** 0,011 -,207** ,104** ,937** 1,00 CI_OFF_RANK (13) -0,02 -0,02 ,191** ,203** ,01 -,084** ,121** -0,01 -,069** ,035* ,671** ,664** 1,00 CI_FI_TA (14) -,162** -,126** ,734** ,701** ,098** -,225** ,179** 0,025 -,251** ,076** ,672** ,632** ,352** 1,00 CI_FI_SAL (15) -,178** -,129** ,718** ,784** ,143** -,261** ,257** 0,006 -,278** ,103** ,625** ,690** ,343** ,936** 1,00 CI_FI_RANK (16) -,036* -0,01 ,146** ,154** -,02 -0,03 ,144** 0,002 -,059** 0,003 ,257** ,252** ,372** ,462** ,448** 1,00

Notes: *, ** indicate two-tailed statistical significance at the 5%, and 1% level, respectively.

ABS_DA: absolute value discretionary accruals: DWCA_ABS: absolute value discretionary working capital accruals; TA: natural logarithm of total assets; REV: natural logarithm of total revenues; CFO: cash flow from operations divided by total assets; LOSS: dummy variable 1 if the company reports a net loss in the current year and as 0 for profits; LEV: leverage ratio, measured as total liabilities divided by total assets; BM: book-to-market ratio, measured as book value divided by the market value of equity; ROA: net income before extraordinary items divided by net assets; GROW: Percentage change of net sales; CI_OFF_TA: client importance at office level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by an office in a given year; CI_OFF_SAL: client importance at office level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by an office in a given year; CI_OFF_RANK: dummy variable for client importance at office level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by an office in a given year; CI_FI_TA: client importance at firm level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by a firm in a given year; CI_FI_SAL: client importance at firm level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by a firm in a given year; CI_FI_RANK; dummy variable for client importance at firm level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by a firm in a given year.

The correlations shown in Table 3 between the control variables which we will use in a single

regression are low. The variables of CFO, ROA and LOSS have above average correlations of

around 0.35 due to the same assumptions for measuring these variables. In general, however,

Table 3 shows no multicollinearity problems for our regressions.

4.2 Regression results

4.2.1 Regressions hypothesis 1

Table 5 and Table 6 present the multiple regressions resulting from testing hypothesis 1,

which was set out in Chapter 2. Table 5 shows the outcomes on firm level, while Table 6

shows the outcomes on office level. Table 5 and Table 6 show that there is an adjusted

R-square for these models of 9% and 11%, respectively, which means that 9% and 11%,

respectively, of the variance in discretionary accruals is explained by the independent

variables. Based on Table 5 and Table 6, we can conclude that most control variables affect

the value of the discretionary accruals and therefore affect audit quality. Hypothesis 1

assumes that client importance is not associated with audit quality for the first-tier audit firms

and offices. Tables 5 and 6 also show the relationship between the discretionary accruals as

proxy for audit quality in relation to the client portfolio of the firm and office. Reynolds and

Francis (2001) do not find a relation between large clients who provide economic

(22)

22 dependence to the auditors cause auditor’s dependency to be compromised. Furthermore, they

find negative coefficients, which can be expected on the basis of the previous study by Becker

et al. (1998) on auditors’ behavior to protect their brand name. Based on Table 4, we can

conclude that the client’s importance based on assets, sales, and ranking does not affect the

audit quality for first-tier firms.

Table 5 Regression client importance for first tier firms and discretionary accruals

Coef-ficients t Sig. Coef-ficients t Sig. Coef-ficients t Sig. Intercept 0,101 10,732 0,000 0,104 11,383 0,000 0,097 10,652 0,000 LEV (+) 0,036 4,290 0,000 0,048 5,625 0,000 0,054 6,077 0,000 BM (-) -0,005 -1,933 0,053 -0,004 -1,572 0,116 -0,004 -1,462 0,144 ROA (+) 0,089 4,706 0,000 0,088 4,677 0,000 0,091 4,821 0,000 TA (-) -0,006 -1,769 0,077 0,006 2,456 0,014 REV (-) -0,005 -1,504 0,133 -0,015 -5,746 0,000 GROW (+) 0,032 5,128 0,000 0,035 5,625 0,000 0,036 5,780 0,000 LOSS (?) 0,010 2,319 0,020 0,007 1,596 0,111 0,006 1,367 0,172 CFO (+) 0,049 4,363 0,000 0,053 4,756 0,000 0,056 4,955 0,000

YEAR (controlled) (controlled) (controlled)

CI_FI_TA (-) -0,021 -0,459 0,647 CI_FI_SAL (-) -0,053 -1,280 0,201 CI_FI_RANK (-) -0,011 -0,795 0,427 F 14,332 16,062 15,477 Adjusted R2 0,086 0,096 0,097 Sig. 0,000 0,000 0,000

LEV: leverage ratio, measured as total liabilities divided by total assets; BM: book-to-market ratio, measured as book value divided by the market value of equity; ROA: net income before extraordinary items divided by net assets; TA: natural logarithm of total assets; REV: natural logarithm of total revenues; GROW: Percentage change of net sales; LOSS: dummy variable 1 if the company reports a net loss in the current year and as 0 for profits; CFO: cash flow from operations divided by total assets; CI_FI_TA: client importance at firm level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by a firm in a given year; CI_FI_SAL: client importance at firm level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by a firm in a given year; CI_FI_RANK; dummy variable for client importance at firm level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by a firm in a given year.

At the office level, however, we find a positive relationship between the alternative

measurement of client importance based on ranking of office level and audit quality. For

client importance based on assets and sales, we find positive coefficients instead of negative

coefficients, although these are not statistically significant. As only the alternative

(23)

23

Table 6 Regression client importance for first tier offices and discretionary accruals

Coef-ficients t Sig. Coef-ficients t Sig. Coef-ficients t Sig. Intercept 0,079 7,681 0,000 0,083 8,287 0,000 0,084 8,364 0,000 LEV (+) 0,046 4,909 0,000 0,057 6,065 0,000 0,064 6,562 0,000 BM (-) -0,003 -1,109 0,268 -0,003 -0,871 0,384 -0,002 -0,652 0,514 ROA (+) 0,119 5,665 0,000 0,115 5,550 0,000 0,121 5,858 0,000 TA (-) -0,008 -4,998 0,000 0,011 4,014 0,000 REV (-) -0,010 -5,680 0,000 -0,020 -6,979 0,000 GROW (+) 0,034 4,936 0,000 0,036 5,238 0,000 0,039 5,623 0,000 LOSS (?) 0,010 2,150 0,032 0,007 1,520 0,129 0,006 1,341 0,180 CFO (+) 0,071 5,653 0,000 0,075 5,973 0,000 0,083 6,568 0,000

CI_OFF_TA (-) 0,015 0,886 0,376 CI_OFF_SAL (-) 0,018 0,961 0,337 CI_OFF_RANK (-) 0,013 2,079 0,038 F 12,567 14,442 14,932 Adjusted R2 0,096 0,110 0,119 Sig. 0,000 0,000 0,000

LEV: leverage ratio, measured as total liabilities divided by total assets; BM: book-to-market ratio, measured as book value divided by the market value of equity; ROA: net income before extraordinary items divided by net assets; TA: natural logarithm of total assets; REV: natural logarithm of total revenues; GROW: Percentage change of net sales; LOSS: dummy variable 1 if the company reports a net loss in the current year and as 0 for profits; CFO: cash flow from operations divided by total assets; CI_OFF_TA: client importance at office level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by an office in a given year; CI_OFF_SAL: client importance at office level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by an office in a given year; CI_OFF_RANK: dummy variable for client importance at office level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by an office in a given year.

4.2.2 Regressions hypothesis 2

Table 7 and Table 8 present the multiple regression results of testing hypothesis 2, which test

the relationship between client importance for second- and third-tier audit firms and audit

quality. Here, we see an adjusted R-square of 8% and 16%, respectively, for these models,

which means that about of 8% and 16%, respectively, of the variance in discretionary accruals

is explained by the differences in variables. Based on the findings of Francis et al. (1999), we

predicted more earnings management, measured by discretionary accruals, and consequently

less audit quality, due to a higher client importance. For second-tier audit firms (Table 7), we

see positive, but not significant, coefficients which indicate the predicted relationship between

client importance and audit quality. We predicted that auditor independence would be

impaired due to fewer clients for second- and third-tier firms and therefore higher economic

dependence.

(24)

24

Table 7 Regression client importance for second tier firms and discretionary accruals

Coef-ficients t Sig. Coef-ficients t Sig. Coef-ficients t Sig. Intercept 0,082 4,259 0,000 0,088 4,698 0,000 0,089 4,626 0,000 LEV (+) 0,041 2,658 0,008 0,048 3,088 0,002 0,049 3,141 0,002 BM (-) 0,002 0,308 0,758 0,001 0,235 0,814 0,001 0,239 0,811 ROA (+) 0,117 3,706 0,000 0,116 3,703 0,000 0,120 3,814 0,000 TA (-) -0,013 -3,719 0,000 0,006 1,295 0,196 REV (-) -0,014 -3,823 0,000 -0,017 -3,408 0,001 GROW (+) 0,045 4,068 0,000 0,046 4,190 0,000 0,047 4,271 0,000 LOSS (?) 0,009 1,166 0,244 0,006 0,807 0,420 0,006 0,741 0,459 CFO (+) 0,031 1,543 0,123 0,034 1,704 0,089 0,037 1,832 0,067

CI_FI_TA (+) 0,043 1,321 0,187 CI_FI_SAL (+) 0,039 1,124 0,261 CI_FI_RANK (+) 0,014 1,043 0,297 F 4,827 5,371 5,169 Adjusted R2 0,077 0,087 0,087 Sig. 0,000 0,000 0,000

LEV: leverage ratio, measured as total liabilities divided by total assets; BM: book-to-market ratio, measured as book value divided by the market value of equity; ROA: net income before extraordinary items divided by net assets; TA: natural logarithm of total assets; REV: natural logarithm of total revenues; GROW: Percentage change of net sales; LOSS: dummy variable 1 if the company reports a net loss in the current year and as 0 for profits; CFO: cash flow from operations divided by total assets; CI_FI_TA: client importance at firm level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by a firm in a given year; CI_FI_SAL: client importance at firm level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by a firm in a given year;

CI_FI_RANK; dummy variable for client importance at firm level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by a firm in a given year.

The coefficients for third-tier audit firms, however, are the opposite of second-tier firms as

shown in Table 8. We expected a positive relationship between client importance and audit

quality (Francis et al., 1999). However, we do not find statistically significant relationships at

a 95% confidence level between the dependent and independent variables and therefore

hypothesis 2 should be rejected.

(25)

25

Table 8 Regression client importance for third tier firms and discretionary accruals

Coef-ficients t Sig. Coef-ficients t Sig. Coef-ficients t Sig. Intercept 0,121 5,233 0,000 0,114 5,170 0,000 0,106 4,275 0,000 LEV (+) 0,021 1,093 0,275 0,025 1,182 0,238 0,023 1,119 0,264 BM (-) 0,003 0,450 0,653 0,003 0,435 0,664 0,003 0,452 0,651 ROA (+) 0,126 3,189 0,002 0,127 3,209 0,001 0,132 3,326 0,001 TA (-) 0,003 0,619 0,536 -0,003 -0,461 0,645 REV (-) 0,001 0,138 0,890 -0,017 -3,408 0,001 GROW (+) 0,071 5,077 0,000 0,072 5,113 0,000 0,070 4,956 0,000 LOSS (?) -0,018 -1,673 0,095 -0,019 -1,674 0,095 -0,018 -1,631 0,104 CFO (+) -0,128 -4,895 0,000 -0,128 -4,884 0,000 -0,131 -4,973 0,000

CI_FI_TA (+) -0,076 -1,795 0,073 CI_FI_SAL (+) -0,049 -1,207 0,228 CI_FI_RANK (+) -0,018 -1,331 0,184 F 5,191 5,097 4,836 Adjusted R2 0,164 0,161 0,159 Sig. 0,000 0,000 0,000

LEV: leverage ratio, measured as total liabilities divided by total assets; BM: book-to-market ratio, measured as book value divided by the market value of equity; ROA: net income before extraordinary items divided by net assets; TA: natural logarithm of total assets; REV: natural logarithm of total revenues; GROW: Percentage change of net sales; LOSS: dummy variable 1 if the company reports a net loss in the current year and as 0 for profits; CFO: cash flow from operations divided by total assets; CI_FI_TA: client importance at firm level, measured as: the natural logarithm of the total assets of a client to the sum of the natural logarithm of total assets of all clients audited by a firm in a given year; CI_FI_SAL: client importance at firm level, measured as: the natural logarithm of the total revenues of a client to the sum of the natural logarithm of total revenues of all clients audited by a firm in a given year;

CI_FI_RANK; dummy variable for client importance at firm level, coded as 1 if a client belongs to the top ranking based on assets, sales, loss and leverage otherwise 0 by a firm in a given year.

4.2.3 Regressions hypothesis 3

Narayanan (1995) and Francis et al. (2005) propose that audit quality should be assessed at

the audit office level. Table 9 and Table 10 present the multiple regressions results of testing

hypothesis 3, in which we test the relationship between client importance for second- and

third-tier audit offices and audit quality. In Table 9 and Table 10, we present the findings of

our regression. Based on the findings of Choi et al. (2010), we expect a negative relationship

between client importance on audit office level for second-tier audit firms and audit quality.

For the commonly used client importance proxies based on assets and sales, there is a

negative, but insignificant relationship for client importance and discretionary working capital

accruals. This indicates that higher client importance leads to higher audit quality. For the

alternative ranking method based on assets, sales loss and leverage, there is a positive

relationship for second tier offices; however, it is not significant.

(26)

26

Table 9 Regression client importance for second tier offices and discretionary accruals

Coef-ficients t Sig. Coef-ficients t Sig. Coef-ficients t Sig. Intercept 0,070 2,750 0,006 0,085 3,431 0,001 0,084 3,304 0,001 LEV (+) 0,041 2,113 0,035 0,054 2,785 0,006 0,059 2,983 0,003 BM (-) 0,002 0,269 0,788 0,000 0,003 0,998 -0,001 -0,180 0,857 ROA (+) 0,166 3,962 0,000 0,158 3,830 0,000 0,163 3,972 0,000 TA (-) -0,004 -0,969 0,333 0,024 3,419 0,001 REV (-) -0,008 -1,822 0,069 -0,036 -5,185 0,000 GROW (+) 0,053 3,247 0,001 0,054 3,344 0,001 0,055 3,408 0,001 LOSS (?) 0,021 2,129 0,034 0,015 1,539 0,124 0,013 1,359 0,175 CFO (+) 0,077 2,865 0,004 0,082 3,072 0,002 0,085 3,198 0,001

CI_OFF_TA (+) -0,047 -1,191 0,234 CI_OFF_SAL (+) -0,045 -1,148 0,251 CI_OFF_RANK (+) 0,016 1,387 0,166 F 4,552 5,393 5,894 Adjusted R2 0,110 0,133 0,153 Sig. 0,000 0,000 0,000

S

i

milarly, for the third-tier offices we found negative coefficients between client importance

and the discretionary accruals; however, they were not statically significant. Based on the

regression presented in Table 9 and Table 10, we reject hypothesis 3.

Table 10 Regression client importance for third tier offices and discretionary accruals

Coef-ficients t Sig. Coef-ficients t Sig. Coef-ficients t Sig. Intercept 0,135 4,805 0,000 0,133 5,013 0,000 0,121 4,006 0,000 LEV (+) -0,006 -0,248 0,804 0,000 0,006 0,995 0,005 0,203 0,839 BM (-) 0,002 0,173 0,863 0,002 0,242 0,809 0,002 0,174 0,862 ROA (+) 0,092 1,889 0,060 0,091 1,879 0,062 0,098 1,984 0,048 TA (-) -0,002 -0,244 0,808 0,007 0,732 0,465 REV (-) -0,004 -0,657 0,512 -0,010 -1,110 0,268 GROW (+) 0,072 3,897 0,000 0,072 3,899 0,000 0,072 3,931 0,000 LOSS (?) 0,000 0,025 0,980 -0,001 -0,082 0,935 -0,002 -0,179 0,858 CFO (+) -0,075 -2,186 0,030 -0,076 -2,223 0,027 -0,078 -2,250 0,025

CI_OFF_TA (+) -0,020 -0,373 0,710 CI_OFF_SAL (+) -0,005 -0,112 0,911 CI_OFF_RANK (+) -0,007 -0,437 0,662 F 3,058 3,112 2,980 Adjusted R2 0,132 0,135 0,134 Sig. 0,000 0,000 0,000

Audit quality: the effect of client importance on audit quality in Germany for second- and third-tier offices