! "! ! "#$!%$$&'()*($+!,!-&().&/0!1*.2*()!%$$&'()*($+!! 3*$'/)+!&4!5$&(&62$!*(7!8'92(:990!;(21:.92)+!&4!%69):.7*6! !
!
"#$%&'!%(&$)$*!!
!
!
!
-/2:()!<6=&.)*($:!*(7!5*.(2(>9!"*(*>:6:()?!@A:!
"&7:.*)2(>!B&/:!&4!-&.=&.*):!C&1:.(*($:!
! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !Student: Charlotte Bontje Student number: 10103783
Supervisor: Prof. dr. V.R. O’Connell Date: 22 June 2015
Education: MSC Accountancy and Control, Accountancy track Word count: 11,297
! #! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! Statement of Originality !
This document is written by student Charlotte Bontje 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.
! $!
Abstract: This study explores whether corporate governance moderates the association between
client importance and earnings management. The use of different fee specifications as proxy for client importance, and both audit committee governance and internal board governance as proxy for corporate governance. This study finds evidence that each measure of client importance is positively associated with the absolute value of discretionary accruals. However, only the client importance measure ln TOTALFEES has a positive relation with positive discretionary accruals. These results support the economic bonding theory of DeAngelo (1981). An economic bond between an auditor and a client can impair the auditors’ independence, and thus the audit quality. In addition, this study finds no evidence that audit committee governance or internal board governance moderates the relation between client importance and earnings management.
! %!
Table of Contents
1. Introduction 5
2. Literature review and hypotheses 7
2.1 Audit Quality 7
2.2 Earnings Management 7
2.3 Auditor Independence and Audit Fees 8
2.4 Corporate Governance 9
2.4.1 Audit committee 10
2.4.2 Internal board governance 11
3. Research design 12
3.1 Dependent variables 12
3.2 Independent variables 13
3.2.1 Audit fees 13
3.2.2 Audit committee governance 13
3.2.3 Internal board governance 14
3.3 Control variables 14
3.4 Empirical model 15
4. Evidence 17
4.1 Sample description 17
4.2 Descriptive statistics 18
4.3 Earnings management and client importance 21
4.4 Earnings management and the interaction between client importance and
audit committee governance 23
4.5 Earnings management and the interaction between client importance and
internal board governance 26
4.6 Earnings management and the interaction between client importance and
corporate governance 28
4.7 Comparison between the regression 30
5. Conclusion 30
Appendix 32
! &!
1. Introduction
The objective of this study is to provide empirical evidence regarding the influence of corporate governance on the association between client importance and the financial reporting quality. My motivation for this study is the recent accounting scandals, probably meaning that the
independence of the auditor is impaired. Future economic interests in a particular client (DeAngelo, 1981) and the provision of non-audit services (Simunic, 1984) may impair the independence of the auditor to a particular client. Prior studies have shown mixed results regarding the relation between (non-) audit fees and audit quality. Frankel et al. (2002) and Larcker and Richardson (2004) find evidence that the provision of non-audit services results in earnings management. In contrast, the studies of Ashbaugh et al. (2003), (DeFond et al., 2002) and Chung and Kallapur (2003) find no empirical evidence. In addition, the Sarbanes-Oxley Act (2002) emphasized the importance of effectiveness of the audit committee for the monitoring role of the financial reporting process. Prior studies have shown that independent directors and having at least one member with financial or corporate experience in the audit committee have a positive effect on the financial reporting quality (i.e. Klein, 2002; Xie et al., 2003; Bedard et al., 2004). However, there is little research done regarding the influence of corporate governance on the relation between client importance and audit quality.
The audit committee plays a central role in the monitoring function of the financial reporting process. The Blue Ribbon Committee (1999) has made several recommendations to improve the corporate audit committee effectiveness. These recommendations include strengthening the independence of the audit committee, making the audit committee more effective and addresses mechanisms for accountability among the audit committee, the outside auditors and management. Moreover, the internal board governance is primarily charged with monitoring of management to protect shareholders’ interest (Xie et al. 2003). Xie et al. (2003) argue that the board composition will influence whether or not an organization engages in earnings management. Prior studies found evidence that boards consistent of outside directors and directors with more financial expertise are more effective to monitor management, and thus are less likely to engage in earnings management (i.e Klein, 2002; Xie et al., 2003). In addition, evidence is found that more independent, expert and diligent boards are associated with higher audit fees (Carcello et al., 2002). In contrast, mixed empirical evidence is found regarding audit committee characteristics and audit fees. In particular, Abbott (2003b) argues that organization with an audit committee that meet at least four times a year and that consists of independent directors are likely to have a low ratio of non-audit fees to total audit fees.
! '! on the association between client importance and audit quality is scarce. The research of Larcker and Richardson (2004) found evidence that organizations with weak corporate governance have a more stringent negative relation between the level of fees and earnings management. Further, the research of Sharma et al. (2011) argues that the positive association between client importance and earnings management is less stringent when organizations have an effective audit committee. Based on prior researches, I predict that a more effect audit committee and internal board governance will ensure auditor independence. A more independent auditor can reduce the
economic bond between an auditor and client, which can lead to higher financial reporting quality. Therefore, my main research question is: ‘’Can corporate governance moderate the association between client importance and earnings management?’’
My sample consists of 6.834 firm observations for the years 2007 through 2013. Different audit fee variables are used as measure for client importance, and both absolute value of
abnormal accruals and positive discretionary accrual as proxy for audit quality. The results show a positive association between the absolute value of abnormal accruals and each of the client importance measures. However, only the client importance measure (ln TOTALFEES) has a positive relation with positive discretionary accruals. The results indicate that the auditor’s incentive is likely to be influenced by a client’s total economic importance to the auditor. This is in accordance with the economic bond theory of DeAngelo (1981). Furthermore, there is no significant effect on the interaction between client importance and corporate governance. This indicates that effective audit committee governance or internal board governance has no moderating effect on the relation between client importance and earnings management. This research makes several contributions to the literature. First, it contributes to the academic debate of auditor independence and the provision of non-audit services. In the past regulators have set limitations of the types of services auditors can provide to their client to ensure the independence of the auditor. My results find support for the concerns of the regulators. Second, this research provides empirical evidence of the influence of corporate governance on the relation between client importance and the financial reporting quality. And finally, for the measures of corporate governance I use a composite score of respectively audit committee characteristics and internal board characteristics. This composite score is an addition to the existing academic literature.
This research is constructed as follows. Section 2 provides the literature review and hypotheses development. Section 3 presents the research design. Section 4 presents the sample selection, descriptive statics and results. And finally, section 5 provides the conclusion, summary and limitations.
! (!
2. Literature review 2.1 Audit quality
The role of auditing is to reduce information asymmetries that exist between managers and firm stakeholder, which is done by allowing outsider to verify the validity of financial statements (Becker et al., 1998; Watts and Zimmerman, 1981). DeAngelo (1981a) defines the quality of audit services as ‘’the market-assessed joint probability that a given auditor will both 1) discover errors or breaches in the accounting system and (2) report the breach’’. Whether an auditor will discover and report a breach depends on the auditor’s technical skills. Furthermore, DeAngelo (1981b) argues that larger auditors, as measured by the number of clients, have less incentive to engage in earnings management, which results in a higher financial reporting quality. Larger auditors face higher reputation concerns in case of an audit failure (Reynold and Francis, 2002). A
consequence of an audit failure is that the financial statements give potentially misleading information to third parties (Francis, 2004).
2.2 Earnings management
Healy and Wahlen (1999) argues that because auditing is imperfect, it can create opportunities for earnings management. They define earnings management as the fact that ‘’managers choose reporting methods and estimates that do not accurately reflect their firms’ underlying
economics’’. However, the use of management discretion on financial reports can also provide more relevant and accurate information for third parties (Healy and Wahlen, 1999). In general, cash flow from operations and changes in working capital, two components of earnings, are used to achieve earnings management (Burgstahler and Dichev, 1997). A large body of academic literature has examined the various incentives for managers to engage in earnings management. For example, the research of Burgstahler and Dichev (1997) argue that managers engage in earnings management to lower the cost imposed on the firm in transaction with stakeholders. Moreover, they suggest that the prospect theory is the largest incentive to avoid reporting losses and earnings decreases. Furthermore, the research of DeGeorge et al. (1999) found evidence that behavioral thresholds drive earnings management. The identified behavioral thresholds were sustaining recent performance, meeting analyst forecast and the reporting of positive profits. The research of DeFond and Jiambalvo (1994) examines debt covenant restrictions and their relation to the choices of accounting. They found evidence of positive accounting manipulation in the year preceding and the year of violation. Furtermore, the research of Cahan (1992) found evidence that the incentive of managers to manage their discretionary accruals is in response to changes in potential political costs.
! )!
2.3 Auditor independence and audit fees
DeAngelo (1981a) define the level of auditor independence as the conditional probability that, given a breach has been discovered, the auditor will report the breach. Furthermore, he argues that the greater the observed economic interest in a particular client, the lower the level of auditor independence to that particular client. Firth (1997) argues that if an audit firm is economic dependent upon a client, the audit firm may be more willing to allow the client’s misrepresentation or misinterpretation of the financial statement. In contrast, the auditor’s
incentives to remain independence are loss of reputation and litigation risk (DeFond et al., 2002). Simunic (1984) argues that the provision of non-audit fees can strengthen the auditor’s economic bond with a particular client. If the non-audit service generates future economic rents, this may reduce the probability of truthful financial reporting. Regulators are concerned that for auditors the benefits of the provision of non-audit services exceed the costs of losing their independence to a particular client (DeFond et al., 2002). However, it may still be optimal for management to hire their auditor for non-audit service. The production of both services can provide knowledge externality or spillovers and cost savings (Simunic, 1984). Knowledge spillovers refer to the generation of information while performing non-audit service that can generate future economic rents, by reducing the fixed or setup costs of auditing services (DeFond et al., 2002).
A lot of researches have examined the association between auditor independence and financial reporting quality. The results from these researches are mixed. However, in the different researches different proxies for auditor independence and financial reporting quality are used. For example, the research of Frankel et al. (2002) find evidence of a positive association between the purchase of non-audit services and the magnitude of absolute discretionary accruals. This research uses three specifications for auditor fees: fee-ratio (non-audit fees to audit fees) and the percentile rank of total, non-audit or audit fees for each client for a given auditor. In contrast, the research of Ashbaugh et al. (2003) replicates this research of Frankel et al. (2002) to investigate whether their results are sensitive for the different fee specifications. This research takes the natural log of each fee variable as fee specification. Compared with the research of Frankel et al. (2002), they find no relation between the provision of non-audit fees and positive discretionary accruals. Furthermore, the researches of Chung and Kallapur (2003) and DeFond et al. (2002) provide no evidence of an association between non-audit fees and financial reporting quality. In addition, larger clients create greater economic fee dependence, which can impair the auditor independence. However, Reynold and Francis (2000) argue that economic dependence does not impair the independence of the auditor. They provided evidence that big audit firms do
! *! not treat larger clients differently compared with smaller clients. Specifically, large clients report lower discretionary accruals. Higher reputation concerns and litigation risks for larger clients may be an explanation of this reporting conservatism (Reynold and Francis, 2000).
To summarize, there is mixed empirical evidence on the provision of (non-) audit fees and the financial reporting quality. As discussed previously, auditor independence may be threatened when auditors provide non-audit service to their client. The provision of non-audit services can create an economic bond between the auditor and the client (Simunic, 1984). An auditor’s incentive to comprise independence to a client depends on the client importance. In line with the research of Frankel et al. (2002), I expect that the provision of non-audit services will impair auditor independence, and thus reduces the financial reporting quality. Therefore, I have tested the following hypothesis, stated in the null form:
H1: There is a positive relationship between earnings management and client importance
2.4 Corporate governance and earnings management
2.4.1 Audit committee
The audit committee have become an important mechanism for organizations to monitor the reliability of the financial reporting process (McMullen, 1996). The audit committee is responsible for the monitoring and selection of external auditors, and ensuring for effective and soundness internal control and accounting practices (Anderson et al. 2004). To improve the effectiveness of corporate audit committees the Blue Ribbon Committee (BRC, 1999) made several
recommendations. These recommendations include strengthening the independence of the audit committee, making the audit committee more effective and addresses mechanisms for
accountability among the audit committee, the outside auditors and management. The SEC acted on these recommendation and adopted new regulations in the Sarbanes Oxley Act of 2002 (2002) for audit committees.
Prior researches have examined the association between audit committee characteristics and the financial reporting quality. In line with the BRC’s recommendations, Carcello et al. (2002) argue that an independent audit committee led to a better monitor of the firm’s financial
accounting process. Benefits of this effective monitoring are unbiased accounting numbers, which lead to reliable and transparent financial statements and an active investment and trading markets (Klein, 2002). Furthermore, to make the audit more effective the Blue Ribbon
Committee recommends having an audit committee comprised of a minimum of three directors. In addition, at least one member of the audit committee needs accounting or related financial management expertise (BRC, 1999). The researches of Xie et al. (2002), Bedard et al (2004) and
! "+! Agrawal and Chadha (2005) suggest that an active audit committee with financial expertise
members is associated with a reduced level of earnings management. As a result, the audit committee may better serve as a financial monitor. With respect of the size of the audit committee, the research of Xie et al. (2002) finds no significant relation between size and discretionary accruals.
In addition, a few researches have examined the relation between audit committee characteristics and audit fees. For example, the results of the research of Abbott et al. (2003a) indicate that audit committees that consist of independent directors and at least one financial expert have higher audit fees. However, this result is in contrast with the research of Carcello et al. (2002). They argue that the audit committee characteristics provide no significant relation to audit fees when board characteristics are included. Furthermore, Abbott et al. (2003b) argue that organizations with an active and independent audit committee have incentives to lower the ratio of non-audit fees to total fees. They conclude that an audit committee is an important
mechanism to ensure auditor independence, which is in line with the Blue Ribbon Committee report.
To summarize, prior researches find a positive association between certain audit
committee characteristics and the financial reporting quality. A more effective audit committee is a better monitor of the financial reporting process. Thus, I expect that a more effective audit committee will ensure auditor independence, which could reduce the economic bond between the auditor and the client. I have tested the following hypothesis, stated in the null form:
H2: The association between client importance and earnings management is less pronounced when the audit committee meets best practice.
2.4.2 Internal board governance
Boards are primarily charged with monitoring management to protect shareholders’ interest (Xie et al., 2002). Fama and Jensen (1983) argue that the board is an effective device for decision control if it limits the decision discretion of individual top managers. Furthermore, they argue that outside directors need incentives to carry out their task.
An extensive body of academic researches has examined the association between board characteristics and financial reporting quality. Xie et al. (2002) argue that board composition will influence whether or not a company engages in earnings management. Their research indicates that independent outside directors monitor management more effectively than inside directors, and thus is less likely to engage in earnings management. However, inside directors are better
! ""! informed regarding firm projects than outside directors (Bushman et al., 2004). Therefore, the ability of outside directors to monitor effectively can be restricted by limited knowledge regarding firm projects or understanding of financial reporting details (Baber et al, 2012). Furthermore, Xie et al. (2002) found evidence that a director with corporate or financial background may be more familiar with the ways that earnings can be managed and therefore they can increase their monitoring effectiveness. Jensen (1993) argues that it is important to separate the chairman and the CEO positions for an effective board. Without this separation it is much more difficult for the board to perform its critical monitor function. In addition, he argues that small boards can help improve their performance. Smaller boards are more likely to function effectively and can control the CEO more easily. Finally, the research of Baber et al. (2012) examines corporate governance characteristics and their relation to restatements. For the corporate governance measure they included both internal and external characteristics. They found evidence, when interactions between the internal and external governance measures are considered as both main effects and interactions, that governance measures reduces the probability of restatements.
A few researches have examined the relation between audit fees and board governance. The research of Carcello et al. (2002) found that more expert, independent and diligent boards are associated with high audit fees. They suggest that more effective boards may demand differentially higher audit quality, primarily to protect the board’s own interest, to avoid legal liability and to protect its reputation capital. Furthermore, the research of Larcker and Richardson (2004) argue that that the association between audit fees and financial reporting quality is the most stringent when the client has weak corporate governance.
To summarize, certain board characteristics are positively related to financial reporting quality. A more effective board is less likely to engage in earnings management. Thus, I expect that a more effective board will ensure auditor independence, which could reduce the economic bond between the auditor and the client. Therefore, I have tested the following hypothesis, stated in the null form:
H3: The association between client importance and earnings management is less pronounced when internal board governance meets best practice.
Overall, I expect that a more effective corporate governance system within an organization will moderate the association between the economic bonding of an auditor to a client and the incentive of the auditor to allow earnings management. Therefore, I have tested the following hypothesis, stated in the null form:
! "#! H4: The association between client importance and earnings management is less pronounced when corporate governance meets best practice
3 Research design 3.1 Dependent variables.
In prior studies, discretionary accruals are used as a measure of earnings management, and thus audit quality. The research of Subramanyam (1996) suggests that discretionary accruals are priced by the stock market. Furthermore, he suggests that the pricing of discretionary accruals arises because managers use their discretion to manage earnings to better reflect fundamental value. According to Dechow et al. (1995), the usual starting point for the measurement of discretionary accrual is total accruals. Where, total accruals are the sum of non-discretionary and discretionary accruals. I have chosen to use the performance-adjusted Modified Jones model, which is used in the research of Cahan et al. (2011). In this research, the modified Jones model (1991) is estimated cross-sectionally and includes a correction for performance as suggested by Kothari et al. (2005). Consistent with the research of Cahan et al. (2011), the following model is estimated for all two-digit SIC industries with sufficient data and at least 10 observations in year t:
(1) TACC/A = b0 + b1 (1/A) + b2 (!SALES – !AR)/A + b3 (PPE)/A + b4 (LAGINC)/A + !
Where
TACC Total accruals calculated by income before extraordinary items less operating
cash flows
!SALES The change in sales revenue !AR The change in accounts receivables PPE Gross property, plant and equipment
LAGINC The lagged income before extra-ordinary items A Lagged total Assets
The residual from model (I) is an estimate of discretionary accruals. I have used in this research the absolute value of discretionary accruals (|DACC|) and the positive discretionary accruals (posDA) for the measure of earnings management. The absolute value of discretionary accruals is used to measure the combined effect of income-increasing and income-decreasing earnings management decisions (Becker et al., 1998; Chung and Kallapur, 2003). The positive
! "$! al., 2003) argue that managers have more incentives to overstate earnings and that auditors are more likely to be sued when earnings are overstated. This is also in line with the research of Cahan et al. (2011,) they argue that the asymmetric loss function leads auditors to scrutinize positive discretionary accrual more closely than negative discretionary accruals.
3.2 Independent variables.
3.2.1 Audit fees
As stated in the literature, DeAngelo (1981) argues that as the economic bond between the auditor and client increases, the independence of the auditor decreases. Different measures are used to measure the economic bond between the auditor and the client. I have used, in line with the studies of Frankel et al. (2002), Larcker and Richardson (2004) and Ashbaugh et al. (2002) the ratio of non-audit fees to total fees (FEERATIO), as proxy for auditor independence. However, Frankel et al. (2002) argue that this ratio is invariant to the scale of fees, and thus does not capture the financial importance of the client to the auditor. In addition, they suggest that the levels of both non-audit fees and audit fees drive cross-sectional variation in the ratio. In line with the research of Ashbaugh et al. (2003), to capture the financial importance of the client to the auditor, the natural log of audit fees (ln NONAUDIT) and the natural log of both non-audit and non-audit fees (ln TOTAL AUDIT) are used. The natural log of each fee variable is taken to normalize these variables’ distributions, allowing the cross-sectional aggregation of observations (Ashbaugh et al., 2003).
3.2.2 Audit Committee governance
The audit governance measure, designate A-index, focus primarily on the role of the audit committee. As stated in the literature, the audit committee primarily oversees the firm’s financial reporting process (Klein, 2002). Computation of A-index is in line with the recommendations of the Blue Ribbon Committee (BRC, 1999). These recommendations include strengthening the independence of the audit committee, making the audit committee more effective and addresses mechanisms for accountability among the audit committee, the outside auditors and
management. In particular, we use three audit committee attributes as audit governance measure to construct the audit governance measure. A-index increases by one if: (1) all audit committee members are independent auditors, (2) the audit committee size is equal to or greater than 3 directors, or (3) all audit committee members have financial expertise.
! "%! 3.2.3 Internal board governance
The internal governance measure, designated B-index, is based on the research of Baber et al. (2012). This measure focuses primarily on the role of director independence. Baber et al. (2012) used six board characteristics as internal governance measure to construct the internal
governance measure. B-index increases by one if: (1) greater than 2/3 of the board is composed of independent directors, (2) all audit committee members are independent directors, (3) all compensation committee members are independent directors, (4) the firm has a separate independent nominating committee, (5) the CEO is not the board chair, or (6) the board size is less than the median of the distribution for all firms (adjusted for firm size) (Baber et al., 2012).
3.3 Control variables
Because the quality of financial reporting, measured by discretionary accruals, can be affected by several factors I include a number of control variables in the empirical models. The dummy variable BIG 4 is included because, based on prior studies, Big 4 auditors are less likely to allow earnings management. Francis et al. (1999) argue that Big 6 auditors mitigate the involvement of firms’ earnings management, which is done by constraining aggressive or opportunistic reporting of accruals. The BIG 4 variable is a dummy variable, which is ‘one’ if the firm is audited by a big4 audit firm, otherwise ‘zero’.
Defond and Jaimbalvo (1994) conclude that debt covenant restrictions influence
accounting choices in the year preceding and the year of violation. To control for this effect, the variable LEVERAGE is included in the empirical model. The LEVERAGE variable is measured as total liabilities to total assets. In addition, to control for financially distressed firms the variable LOSS is included. This is done because financial distressed firms have incentives to use accrual to manage earnings (Brown, 2001). The LOSS variable is a dummy variable, which represents ‘one’ if the firm reports a negative net income, otherwise ‘zero’. The variable SIZE is included because large companies may have smaller accruals, which is caused by the correlation with operating activities (Reynolds and Francis, 2001). SIZE is measured as the natural logarithm of total assets.
To control for firm performance the MTB, ROA and CFO variables are included. The variable MTB, market to book ratio, is used to control for high growth firms. Skinner and Sloan (2002) argue that if managers of growth firms do not meet earnings forecast, they may have incentives to manage reported earnings to avoid negative earnings surprises. The variable CFO is included in the model to control for firms with high cash flow. Frankel et al. (2002) argues that firms with high cash flows may be more likely to beat earnings benchmarks and ca therefore more easily purchase outside consulting services. The variable CFO is defined as operating cash
! "&! flows scaled by total assets.
And finally, I control for firms that have a high litigation risk. Matsumoto (2002) reports that firms with a high litigation risk are more likely to be concerned about missing earnings benchmarks. The LITIGATION variable is a dummy variable, which is ‘one’ if the firm operates in a high-risk industry, otherwise ‘zero’. The high-risk industries are identified by the work of Francis et al. (1994).
3.4 Empirical model
To examine whether audit quality, measured by absolute value of discretionary accruals or positive discretionary accruals, is associated with client importance, I estimated the following empirical model:
(2) DACC = b0 + b1FEE + b2MTB +b3CFO + b4SIZE +b5ROA+b6LOSS + b7LITIGATION + b8LEVERAGE + b9 BIG 4 + !
To determine whether audit committee governance moderates the association between the absolute value of discretionary accruals and client importance, I estimated the following empirical model:
(3) |DACC| = b0 + b1FEE + b2A-index + b3 FEE * A-index +b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
To determine whether internal board governance moderates the association between the absolute value of discretionary accruals and client importance, I estimated the following empirical model: (4) |DACC| = b0 + b1FEE + b2B-index + b3 FEE * B-index +b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
To determine whether corporate governance moderates the association between the absolute value of discretionary accruals and client importance, I estimated the following empirical model: (5) |DACC|= b0 + b1FEE + b2AB-index + b3 FEE * AB-index +b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
! "'! |DACC| Absolute value of discretionary accruals measured by the performance
matched Jones model
as used in Kothari et al. (2005) and Cahan et al. (2008). DACC+ Positive discretionary accruals
FEE Different measures of client importance; (1) FEERATIO,
(2) ln NONAUDIT and (3) ln TOTALFEES
FEERATIO Non-Audit Fees to Total Fees
ln NONAUDIT The natural logarithm of Non-Audit Fees ln TOTAL FEES The natural logarithm of Total Fees
B-INDEX Internal corporate governance measure of Baber et al. (2012). Composite
score of 6 board
characteristics.
A-INDEX Audit committee governance measure. Composite score of 3 audit committee characteristics (independence, financial expertise, size).
AB-INDEX Sum of B-index and A-index
SIZE Natural logarithm of total assets at year t
BIG4 Dummy variable, if BIG 4 auditor dummy variable ‘one’, otherwise ‘zero’ CFO Net cash flows from operations scaled by total assets
MTB Market Value of Equity/ Book value of Equity
LOSS Dummy variable, ‘one’ if the firm reports an negative net income, otherwise ‘zero’
LEV The ratio of total liabilities to total assets
ROA Net income before extraordinary items scaled by total assets
LITIGATION Dummy variable, ‘one’ if the firm operates in a high litigation risk industry identified by Francis et al. (1994).
In model (2), positivecoefficients on FEERATIO or ln NONAUDIT are consistent with my first hypothesis; the provision of non-audit fees is positively related to earnings management. While a negative coefficient implies that the provision of non-audit services reduces the
magnitude of earnings management. In addition, a negative coefficient in model (3) on FEE * A-index is consistent with my second hypothesis; audit committee governance moderates the relation between client importance and the level of earnings management. A negative coefficient in model (4) on FEE * B-index is consistent with my third hypotheses; Internal board
governance moderates the relation between client importance and the level of earnings management. Finally, a similar interpretation applies for the coefficient FEE * AB-index in model (5). A negative sign is consistent with my fourth hypothesis; corporate governance moderates the relation between client importance and the level of earnings management.
! "(!
4. Evidence
4.1 Sample description.
The sample consists of firms that are included in the S&P 1500 for the years 2007 through 2013. This range is chosen due to the limited available data on the ISS database for the years prior to 2007. For the collection of data I have used three databases, Compustat, AuditAnalytics and ISS (formerly RiskMetrics). The database ’Compustat’ is used for the collection of financial data, which is used for the measure of discretionary accruals and control variables. Data regarding audit fees, non-audit fees and total fees are collected from the database ‘AuditAnalytics’. And finally, the corporate governance data is collected from the database ‘ISS’. The database
Compustat can be seen as central starting point for the collection of data. The Central Index Key (CIK) is used as unique identifier for the collection and merge of data from the database
AuditAnalytics. Furthermore, to collect and merge data from ISS, the unique identifier Cusip (CUSIP) is used. Finally, the software STATA is used to merge the different databases. The initial sample consists of 101,036 observations on the Compustat database with a filling date between January 1, 2006 and December 31, 2013. Consistent with previous researches (i.e. Frankel et al., 2002) I exclude, form the initial sample, firms filed by financial institutions (SIC codes 6000-6999). This is done because of the unique procedures required to estimate discretionary accruals for these firms (Frankel et al., 2002). I also exclude duplicates and
observations that do not disclose the required components of earnings management and control variables. The remaining sample consists of 45,369 observations. With the use of the Central Index Key (CIK) as identifier, I merge the Compustat data with data from the database
AuditAnalytics. After excluding duplicates and observations that are not matched with the unique identifier 34,482 observations are left. Another 213 observations were excluded because they do not disclose the required components of auditor fees. Finally, I merge the Compustat and ISS data with data from the ISS database. Matching the firms with the unique identifier Cusip results in a sample of 7,832 observations. Excluding unavailable data for the required components of the A-index and B-index and duplicates results in a final sample of 6,834 firm years. At the end, I winsorized the top and bottom 1% of the absolute abnormal accruals, fee ratio and the control variables; CFO, MTB, ROA, LEV and SIZE. This is done to decrease the effect of outliers in the sample regression (Chi et al., 2009).
! ")!
Table 1 Sample collection
4.2 Descriptive statistics
The industry distribution for the sample, based on the two-digit industry codes, is shown in table 2. The most represented industry is Manufacturing, which represents 42.68% of the total firm years sample. This is followed by the industry Services, which represents 16,22% of the total firm years sample and the Transportation, Communication, Electric, Gas and Sanitary industry, which represents 10.0% of the total firm years sample.
Table 2
Observations per industry
Industry SIC Number of
observations
Percentage
Agriculture, Forestry, & Fishing 0100-0999 19 0.26%
Mining 1000-1499 327 4.39%
Construction
Manufacturing 1500-1799 2000-3999 105 3304 1.4% 42.68%
Transportation, Communication, Electric, Gas and Sanitary 4000-4999 744 10.0% Wholesale Trade Retail Trade 5000-5199 5200-5999 295 645 3.96% 8.67% Services 7000-8999 1207 16.22% Total 6834 100
Information concerning the audit fees, non-audit fees, and total fees for the sample firms is summarized in table 3. In addition, table 4 summarizes the audit data for each of the BIG 4 audit firms, and for all other firms combined. Generally, there is an increase in the audit fees over the years. This can be the result of audit firms dealing with stricter regulations.
For example, after the SOX implementation in 2002, an increase of audit fees was expected because of increased audit effort and greater exposure to legal liability (Ghosh et al., 2009). However, the graphical representation in table 3 shows a drop in the audit fees, and thus total fees, in the year 2008-2009. This might be due to the financial crisis in 2008.
Based on previous researches (i.e. Francis et al., 1984), Big 4 audit firms billed on average
Selection criteria Observations
Compustat: definitive proxies filed between 1/1/06 and 31/12/13 101.036 Less: duplicates and the missing required financial data for the measure of
earnings management and control variables
61.455
Less: financial institutions 6000-6999 (1557)
45.369
Less: no match with CUMPUSTAT and AuditAnalytics (7.773)
Less: no match CUMPUSTAT and ISS (21.169)
Less: duplicates and unavailable audit fee and corporate governance date (998)
! "*! higher audit fees and non-audit fees compared with all other audit firms combined. In this research the means of audit fees and non-audit fees for Big 4 audit firms are respectively $3,54 billion and $ 835 million. Whereas the means of audit fees and non-audit fees for non-Big 4 audit firms are $ 1,09 billion and $177 million. Among the Big 4 audit firms, the means of total fees billed to the total sample firms range from $ 4,19 billion for Deloitte to just over $3,51 billion for Ernst & Young. The means of non-audit fees billed to the total sample firms range from $928 million for PricewaterhouseCoopers to just over $538 million for KPMG.
Table 3
Statistical and graphical representations of Audit fees, Non-Audit fees and Total fees
Table 4
Distribution Audit fees, Non-audit fees and Total fees for Big 4 auditors
Company N Percent Non-audit fees Audit fees Total fees
Deloitte 1,373 20,36 865.003 3.328.909 4.193.913
Ernst & Young 2.087 30,95 692.778 2.814.483 3.507.261
KPMG 1.172 17,38 538.438 3.041.909 3.580.348
PricewaterhouseCoopers 1.709 25,34 928.837 3.837.560 4.766.397
All other firms 402 5,97 154.473 990.406 1.144.880
Distributions of the internal board governance characteristics (B-index) and audit committee characteristics (A-index) between 2007-2013 are displayed in table 6. Generally, the internal board governance (B-index) increases substantially from 2007 to 2013. In particular, the independence of the board and of the sub-committees (audit, nominating and compensation) increase from 2007 to 2013. As example, means of percentages of board independence are 85,05 percent and 90,21 percent, and the percentages that the CEO is not the board chair are 46,56 percent versus 53,97 percent of the firms for 2007 and 2013 respectively (not tabled). The audit committee governance characteristics (A-index) are approximately stable over the years. For example, the means of percentages of audit committee independence are 97,35 percent versus 97,67 percent for the firms for 2007 and 2013 respectively. In addition, the financial expertise in the audit
Fiscal year
N Audit fees Non-audit fees Total fees 2007 756 3.155.621 686.542 3.843.267 2008 961 3.323.557 701.293 4.024.850 2009 988 2.948.829 638.580 3.587.409 2010 991 2.872.333 693.729 3.566.062 2011 992 3.011.375 718.396 3.729.771 2012 1017 3.142.163 799.252 3.941.415 2013 1032 3.345.501 848.299 4.182.800 +! "++++++! #++++++! $++++++! %++++++! &++++++! #++(! #++)! #++*! #+"+! #+""! #+"#! #+"$! ,-./0! 1223! 4567 ,-./0! 1223#!
! #+!
committees increases substantially from 2007 to 2008. In 2007 the audit committees does not consist of members that all have financial expertise. While, in 2008 20,26 percent of the firm years observations consist of audit committee members with financial expertise.
Table 5
Distribution of Corporate Governance measures
Panel A: Distribution of A-Index by Fiscal Year
A-index 2007 2008 2009 2010 2011 2012 2013 0 25 8 11 8 14 15 17 1 399 430 427 462 438 439 459 2 332 439 475 442 459 481 466 3 0 84 75 84 81 82 90 Means 1,41 1,62 1,62 1,60 1,61 1,62 1,61
Panel B: Distribution of B-Index by Fiscal Year
B-Index 2007 2008 2009 2010 2011 2012 2013 1 0 2 4 2 0 1 1 2 3 7 13 4 4 7 8 3 36 54 48 42 36 34 34 4 163 196 198 207 182 167 152 5 385 544 569 563 512 519 532 6 169 158 156 179 258 289 315 Means 4,9 4,82 4,80 4,87 4,99 5,03 5,08
Table 6 shows an overview of all variables used in this research. Furthermore, the correlation table is represented in the appendix. The main findings in the correlation table reveal that client importance measures (FEERATIO, ln NONAUDIT and ln TOTALAUDIT) and the BIG4 control variable are positively correlated. This will suggest that firms with high FEE are more likely to have a Big 4 auditor. Firms with a high FEE are larger than other firms, have a higher market to book ratio, have a higher composite score of internal governance board (B-index) and have a higher liability to asset ratio. In contrast, the client importance measures (FEE) and the audit committee governance (A-index) are negatively related. This will indicate that firms with a high composite score of audit committee governance, ‘strong’ audit committee mechanism, are more likely to have low audit and non-audit fees. Furthermore, the control variable SIZE and the internal board governance (B-index) are positively correlated. This suggests that larger firms are more likely to have ‘strong’ internal board governance.
The correlation table shows a high correlation between FEE and SIZE. To avoid a
mulitcollinearity problem, a variance inflation factor (VIF) test is used. Based on the results of this (VIF) test, I concluded that this high correlation would not lead to a mulitcollinearity
! #"!
problem. The variance inflation factor was below a level (<10) that would indicate a mulitcollinearity problem.
Table 6 Descriptive statistics
Variable Observations Mean Standard deviation Mix Max
|DACC| 6743 .00698075 .0876917 .0000122 1.554657 DACC+ FEERATIO 3213 6743 .00661109 .230884 .0889988 .2302814 .0000359 0.0004057 1.089155 1.51972 ln NONAUDIT 6743 12.48769 1.605648 6.620073 16.61358 ln TOTALAUDIT 6743 14.71366 .9236253 10.31361 17.8032 A-Index 6743 1.591576 .6464197 0 3 B-Index 6743 4.930743 .8102957 1 6 AB-index 6743 6.098472 .9337769 2 8 MTB 6743 2.7617 3.173392 -35.56398 41.51515 LOSS 6743 .138217 .3451536 0 1 LEV 6743 .5076808 .2109586 .034061 3.456335 BIG 6743 .9403826 .2367941 0 1 CFO 6743 .1096158 .0736648 -.9625774 .3693235 SIZE 6743 7.720891 1.404839 3.791165 10.8217
4.3 Earnings management and client importance
The first hypothesis examines the association between earnings management and audit fees. The findings are reported in Table 7. The first three columns are from empirical model (2a), where the absolute value of discretionary accruals is the dependent variable. I have performed three regressions; column (1) represents an OLS regression with FEERATIO, column (2) represents an OLS regression with ln NONAUDIT, and column (3) represents an OLS regression with ln TOTALFEES. The adjusted R square is respectively 7,12, 7,29, and 7,88 percent. This indicates that the OLS regression with ln TOTALFEES have a stronger relation with the magnitude of earnings management, compared with the other models. The difference in the models is the measure of client importance. The client importance measure FEERATIO is used as proxy for auditor independence and the client importance measures lnNONAUDIT and
lnTOTALAUDIT capture the financial importance of the client.
In table 7 we see that the coefficient on FEERATIO is 0.011 and is statistically significant. Further, we see that the coefficient on lnNONAUDIT is 0.003 and is statistically significant. The regression results for FEERATIO and ln NONAUDIT indicates that there is a positive association between absolute discretionary accruals and the provision of non-audit fees. These results are in line with the study of Frankel et al. (2002), regardless of the discretionary
! ##! accrual estimate used. Based on these results, I conclude that the provision of non-audit services can strengthen the economic bond between the auditor and the client. This conclusion is
consistent with the research of Simunic (1984). He argues that the provision of non-audit serves can increase the auditor’s incentive to acquiesce to client pressure, thereby including pressure to allow earnings management. Furthermore, we see in table 7 that the coefficient on
lnTOTALFEES is 0.014 and is statistically significant. This implies that high total fees are associated with higher earnings management. This results support the economic bond theory of DeAngelo (1981). Auditor’s incentive is likely to be influenced by a client’s total economic importance to the auditor relative to other clients.
The association between positive discretionary accruals and the measures of client
importance, FEERATIO and ln NONAUDIT, are insignificant. This indicates that the provision of non-audit fees has no effect on the magnitude of positive discretionary accruals. However, we see in table 7 that the coefficient on ln TOTALAUDIT is 0.101 and is statistically significant. With respect to the control variables, the market to book ratio has a positive and
significant effect on the magnitude of earnings management. This implies that firms with higher market to book value are more likely to engage in earnings management. The same result applies to the variable cash flow from operations. Firms with high cash flow from operations are more likely to involve in earnings management. In contrast, larger firms are less likely to engage in earnings management. The variable SIZE has a negative and significant effect. For firms with positive discretionary accruals, the LEV variable has a positive and significant effect. This indicates that firms with higher ratio of leverage increase the degree of earnings management. The same result applies to the LOSS variable. Firms that report a loss are more likely to engage in earnings management.
! #$! Table 7
OLS regression: The association between client importance and the quality of financial reporting
(2a) |DACC| = b0 + b1FEE + b2MTB +b3CFO + b4SIZE +b5ROA+b6LOSS + b7LITIGATION + b8LEVERAGE + b9 BIG 4 + !
(2b) DACC+ = b
0 + b1FEE + b2MTB +b3CFO + b4SIZE +b5ROA+b6LOSS + b7LITIGATION + b8LEVERAGE + b9 BIG 4 + !
4.4 Earnings management and the interaction between client importance and audit committee governance
The second hypothesis examines whether the audit committee moderates the association between earnings management and client importance. The findings are reported in table 8. The first three columns are from empirical model (3a), which includes the interaction between audit committee governance and client importance. The parameters on these interaction terms (FEERATIO * A-index, ln NONAUDIT * A-index and ln TOTALAUDIT * A-index) are not statistically significant. In particular, the parameter estimates (b2 and b3) are not statistically
Variable Absolute value of
abnormal accruals
Positive abnormal accruals
FEERATIO .0106 0.018 .0052 0.436 Ln NONAUDIT .0032 0.000 .0009 0.416 Ln TOTALAUDIT .0141 0.000 .01044 0.000 MTB .0018 0.000 .0017 0.000 .0016 0.000 .0019 0.000 .0019 0.000 .00186 0.000 LOSS .0022 0.565 .0021 0.594 -.0008 0.983 .0192 0.001 .0190 0.002 .01716 0.004 LEV -.0139 0.016 -.0139 0.015 -.0136 0.017 .0004 0.966 .0003 0.972 .00016 0.984 SIZE -.0058 0.000 -.0076 0.000 -.0128 0.000 -.0058 0.000 -.0063 0.000 -.0110 0.000 CFO .1681 0.000 .1714 0.000 .1809 0.000 -.2251 0.000 -.2239 0.000 -.2087 0.000 BIG 4 -.0069 0.129 -.0077 0.092 -.0096 0.035 -.0034 0.590 -.0036 0.576 -.0055 0.386 ROA -.2371 0.000 -.2383 0.000 -.2391 0.000 .2274 0.000 .2261 0.000 .2149 0.000 LITIGATION .0042 0.087 .0047 0.059 .0058 0.000 .0056 0.119 .0057 0.112 .0066 0.065 _Intercept .1127 0.000 .0894 0.000 -.0365 0.075 .1119 0.000 .1058 0.000 .0010 0.972 n 6743 6743 6743 3213 3213 3213 R square 7,24% 7,41% 8,0% 3,71% 3,71% 4,18%
* p-values of the estimated parameters are reported in italics. The variables are defined in section 3.4 Empirical model
! #%! significant, while the parameter estimate (b1)is statistically significant. A detailed description of the parameter estimates (b1, b2 and b3) is given below.
Empirical model (3a) is presented in the first three columns of table 8. This model includes the interaction between audit committee governance and each measure of client importance. We see in table 8 that the coefficient A-index has an insignificant negative association. This can be interpreted as firms with high composite score of audit committee characteristics lower the magnitude of earnings management. However, no conclusion can be drawn. In contrast, we see in table 8 that the coefficient ln NONAUDIT is 0.003 and that the coefficient ln TOTALAUDIT is 0.014, both coefficients are statistically significant. This suggests that firms with high non-audit fees and total fees are more likely to engage in earnings
management. This is in line with the economic bond theory of DeAngelo (1981). Finally, the coefficient FEE*A-index has an insignificant negative association on the interaction of audit committee governance and each measure of client importance. This can be interpreted as, an audit committee moderates the association between client importance and earnings management. However, no conclusion can be drawn.
To illustrate the relevance of the interaction between audit committee governance and client importance, the last three columns in table 8 are from empirical model (3b) where the main effects, but not the interaction term, is included. In this empirical model, the coefficient FEE is statistically significant. Compared with the first three columns, the client importance measure FEERATIO was not statistically significant. Furthermore, the coefficient A-index is not statistically significant. This result is similar for the first three columns.
To summarize, based on the results I cannot conclude that the audit committee governance index moderates the association between client importance and earnings
management. The interaction coefficient has the predicted negative sign, however this effect is not significant.
! #&! Table 8
OLS regression: The moderating effect of the audit committee
(3a) |DACC|= b0 + b1FEE + b2A-index + b3 FEE * A-index +b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
(3b) |DACC|= b0 + b1FEE + b2A-index + b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
Variable Estimate with A-index*Client Importance Interaction Estimate without A-index*Client Importance Interaction FEERATIO .0167 0.158 .0105 0.019 Ln NONAUDIT .0034 0.049 .0032 0.000 ln TOTAL .0139 0.000 .0141 0.000 A-index * Fee-ratio -.0039 0.573 A-index * ln NONAUDIT -.0002 0.875 A-index * ln TOTALAUDIT .0003 0.852 A-index -.0004 0.846 .0006 0.959 -.0062 0.801 -.0013 0.405 -.0013 0.430 -.0016 0.314 MTB .0018 0.000 .0017 0.000 .0016 0.000 .0017 0.000 .0017 0.000 .0018 0.000 LOSS .0021 0.578 .0019 0.605 -.0002 0.970 .0021 0.573 .0020 0.604 .0169 0.967 LEV -.0137 0.017 -.0138 0.016 -.0134 0.019 -.0137 0.017 -.0138 0.016 .0004 0.019 LITIGATION .0042 0.086 .0047 0.058 .0058 0.018 .0043 0.084 .0047 0.058 .0066 0.019 ROA -.2373 0.000 -.2384 0.000 -.2394 0.000 -.2372 0.000 -.2244 0.000 .2132 0.000 SIZE -.0058 0.000 -.0076 0.000 -.0128 0.000 -.0056 0.000 -.0076 0.000 -.0110 0.000 CFO .1684 0.000 .1716 0.000 .1814 0.000 .1683 0.000 -.2224 0.000 -.2073 0.000 BIG 4 -.0069 0.135 -.0077 0.091 -.0097 0.034 -.0069 0.127 -.0038 0.091 -.0057 0.034 _intercept .1132 0.000 .0883 0.000 -.0269 0.550 .1147 0.000 .0913 0.000 -.0067 0.095 R square 7,10% 7,42% 8,01% 7,25% 7,42% 8,01%
* p-values of the estimated parameters are reported in italics. The variables are defined in section 3.4 empirical model
! #'! 4.5 Earnings management and the interaction between client importance and internal board governance
The third hypothesis examines whether the internal board governance moderates the association between earnings management and client importance. The findings are displayed in table 9. The first three columns are from empirical model (4a), which includes the interaction between internal board governance and client importance. The parameters on these interaction terms are not statistically significant. Except for parameter estimate (b1) on client importance measure (ln TOTALAUDIT), parameter estimates (b1, b2 or b3) displayed in the first three columns are not statistically significant. A detailed description of parameter estimates (b1, b2 and b3) is given below. Empirical model (4a) is presented in the first three columns of table 9. This model
includes the interaction between internal board governance and each measure of client importance. We see in table 9 that the coefficient B-index has an insignificant positive association. This can be interpreted, as firms with high composite score of internal board governance characteristics are more likely to engage in earnings management. However, no conclusion can be drawn. The coefficients FEERATIO and ln NONAUDIT have an
insignificant positive association with the absolute value of discretionary accruals. Furthermore, we see in table 9 that the coefficient ln TOTALAUDIT is 0.0136, and is statistically significant. This indicates that auditor’s incentive is likely to be influenced by a client’s total economic importance to the auditor. Finally, the coefficient FEERATIO*B-index has an insignificant positive association with earnings management. This can be interpreted as firms with strong internal board governance increase the association between client importance and earnings management. However, no conclusion can be drawn.
To illustrate the relevance of the interaction between internal board governance and client importance, the last three columns in table 9 are from regression model (4b) that does not
include the interaction term. The coefficient FEE has a significant positive association with the magnitude of earnings management. Compared with the first three columns, only the client importance measure ln TOTALAUDIT has a significant positive effect. Furthermore, the coefficient B-index has a statistically positive significant association with the magnitude of earnings management, at a significance level of (p<0.10). However, in the first three columns there was no statistically significant effect. This indicates that firms with a high composite score of internal board governance are more likely to engage in earnings management.
To summarize, based on the results I cannot conclude that internal board governance moderates the association between client importance and earnings management. The parameters
! #(! on the interaction terms are not statistically significant, and in addition the parameter estimate (b3) has not the predicted negative sign.
Table 9
OLS regression: The moderating effect of the internal board governance
(4a) |DACC |= b0 + b1FEE + b2B-index + b3 FEE * B-index +b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
(4b) |DACC |= b0 + b1FEE + b2B-index + b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
Variable Estimate with B-index*Client Importance Interaction Estimate without B-index*Client Importance Interaction FEERATIO .0004 0.987 .0104 0.021 ln NONAUDIT .0023 0.572 .0032 0.000 ln TOTALAUDIT .0136 0.071 .0139 0.000 B-index * Fee-ratio .0020 0.715 B-index * ln NONAUDIT .0002 0.839 B-index * ln TOTALAUDIT .0003 0.963 B-index .0023 0.226 .0006 0.954 .0014 0.948 .0027 0.053 .0026 0.061 .0024 0.090 MTB .0018 0.000 .00172 0.000 .0016 0.000 .0019 0.000 .0019 0.000 .0018 0.000 LOSS .0023 0.588 .0020 0.605 -.0002 0.970 .0190 0.002 .0188 0.002 .0170 0.005 LEV -.0144 0.012 -.0138 0.016 -.0134 0.019 .0006 0.941 .0006 0.946 .0004 0.961 LITIGATION .0042 0.086 .0047 0.058 .0058 0.018 .0056 0.117 .0057 0.110 .0066 0.063 ROA -.2373 0.000 -.2384 0.000 -.2391 0.000 .2258 0.000 .2244 0.000 .2132 0.000 SIZE -.0058 0.000 -.0076 0.000 -.0128 0.000 -.0058 0.000 -.0063 0.000 -.0110 0.000 CFO .1684 0.000 .1716 0.000 .1810 0.000 -.2237 0.000 -.2224 0.000 -.2073 0.000 BIG 4 -.0069 0.135 -.0077 0.091 -.0097 0.034 -.0036 0.570 -.0038 0.555 -.0057 0.370 _intercept .1132 0.000 .0883 0.000 -.0269 0.550 .1152 0.000 .1005 0.001 -.0067 0.916 R square 7,29% 7,42% 8,01% 7.26% 7,36% 7,98%
* p-values of the estimated parameters are reported in italics. The variables are defined in section 3.4 Empirical model
! #)! 4.6 Earnings management and the interaction between client importance and corporate governance
The third hypothesis examines whether the corporate governance moderates the association between earnings management and client importance. The findings are displayed in table 10. The first three columns are from empirical model (5a), which includes the interaction between
corporate governance and client importance. The parameters on these interaction terms are not statistically significant. In particular, except for parameter estimate (b1) on client importance measure (ln TOTALAUDIT), parameter estimates (b1, b2 or b3) displayed in the first three columns are not statistically significant. A detailed description of the parameters (b1, b2 and b3) is given below.
Empirical model (5a) is presented in the first three columns of table 10. This empirical model includes the interaction between corporate governance and client importance. The
coefficient AB-index has an insignificant positive relation with earnings management. This can be interpreted, as firms with high composite score of corporate governance are more likely to engage in earnings management. However, no conclusion can be drawn. The coefficients FEERATIO and ln NONAUDIT have an insignificant positive relation with earnings
management. Furthermore, we see in table 10 that coefficient (ln TOTALAUDIT) is 0.02 and is statistically significant. This indicates that auditor’s incentive is likely to be influenced by a client’s total economic importance to the auditor. Finally, the coefficients FEERATIO*AB-index and ln TOTALAUDIT*AB-index have an insignificant positive relation with earnings management. This can be interpreted as firms with strong corporate governance increase the association between client importance and earnings management. However, no conclusion can be drawn. In contrast the coefficient ln NONAUDIT*AB-index has an insignificant negative relation with earnings management.
To illustrate the relevance of the interaction between corporate governance and client importance, the last three columns in table 10 are from regression model (5b) that does not include the interaction term. The coefficient FEE has a significant positive association with earnings management. Compared with the first three columns, only the client importance coefficient ln TOTALAUDIT has a significant positive effect. Furthermore, the coefficient AB-index has an insignificant positive relation with earnings management. This result is similar for the first three columns.
To summarize, based on the results I cannot conclude that corporate governance moderates the association between client importance and earnings management. The parameters
! #*! on the interaction terms are not statistically significant, and in addition the parameter estimate (b3) has not the predicted negative sign.
Table 10
OLS regression: The moderating effect of corporate governance
(5a) |DACC |= b0 + b1FEE + b2AB-index + b3 FEE * AB-index +b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
(5b) |DACC |= b0 + b1FEE + b2AB-index + b4 MTB +b5CFO + b6SIZE +b7ROA+b8LOSS + b9LITIGATION + b10LEVERAGE + b11 BIG 4 + !
Variable
Estimate with
AB-index*Client Importance Interaction
Estimate without AB-index*Client Importance Interaction FEERATIO .0102 0.734 .01056 0.019 ln NONAUDIT .0065 0.144 .0032 0.000 ln TOTALAUDIT .0195 0.015 .0140 0.000 AB-index * Fee-ratio .00005 0.991 AB-index * ln NONAUDIT -0.005 0.450 AB-index * ln TOTALAUDIT .0009 0.481 AB-index .00083 0.613 .0074 0.404 .0133 0.467 .0008 0.489 .0008 0.527 .0004 0.724 MTB .00175 0.000 .0017 0.000 .0016 0.000 .00175 0.000 .0017 0.000 .0016 0.000 LOSS .00225 0.559 .0021 0.590 -.0000 0.990 .0022 0.560 .0021 0.591 -.0001 0.986 LEV -.0141 0.014 -.0143 0.013 -.0139 0.015 -0.0141 0.014 -.0142 0.013 -.0137 0.016 LITIGATION .00415 0.092 .0046 0.062 .0057 0.021 .00415 0.092 .0046 0.063 .0057 0.020 ROA -.2371 0.000 -.2378 0.000 -.2387 0.000 -.2371 0.000 -.2383 0.000 -.2390 0.000 SIZE -.0061 0.000 -.0078 0.000 -.0129 0.000 -.0061 0.000 -.0078 0.000 -.0129 0.000 CFO .16838 0.000 .1715 0.000 .1809 0.000 .1684 0.000 .1716 0.000 .1809 0.000 BIG 4 -.0071 0.118 -.0081 0.077 -.0101 0.028 -.0071 0.118 -.0079 0.085 -.0097 0.033 _intercept .1096 0.000 .0462 0.395 -.1168 0.305 .1095 0.000 .0865 0.000 -.0378 0.069 R square 7,25% 7,42% 8,01% 7.25% 7,42% 8,00%
* p-values of the estimated parameters are reported in italics. The variables are defined in section 3.4 Empirical model
! $+!
4.7 Comparison between the regressions
My first hypothesis predicts that there is a positive association between earnings management and client importance. I found for each measure of client importance a significant positive association with the absolute value of discretionary accruals. However, only the client importance measure ln TOTALAUDIT has a significant positive association with positive discretionary accruals. These results support the economic bond theory of DeAngelo (1981). He argues that the provision of non-audit and audit services can create an economic bond between auditor and client, which gives incentives to the allowance of earnings management.
My second hypothesis predicts that effective audit committee governance moderates the association between client importance and earnings management. However, I found no empirical evidence that effective audit committee governance moderates the relation between client
importance and earnings management. The interaction coefficient has the predicted negative sign, however not significant. My third hypothesis predicts that internal board governance moderates the association between client importance and earnings management. However, I found no empirical evidence that internal board governance moderates the relation between client importance and earnings management. Similar results are provided for my fourth
hypothesis. I found no empirical evidence that corporate governance moderates the association between client importance and earnings management.
5. Conclusion
My objective of this research was to investigate the influence of corporate governance on the relation between client importance and earnings management. The provision of non-audit fees by auditors to their client can create an economic bond between the auditor and a particular client. This economic interest in a particular client can impair the auditor independence and thus a lower audit quality (Simunic, 1984; DeAngelo, 1981). Prior studies have provided mixed empirical evidence on the relation between the purchase of audit and non-audit service and audit quality. In addition, there has much less been written regarding the influence of corporate governance on the relation between client importance and earnings management. A more effective corporate governance mechanism will better serve as a financial monitor (Xie et al., 2002). Therefore, I expect that a more effective corporate governance mechanism will strengthen the auditor independence, and thus a higher financial reporting quality. The main research question of this study is: ‘’Can corporate governance moderate the association between client importance and earnings
management?’’
! $"! of earnings management. I found a significant positive relation for each of the client importance measure and the absolute value of abnormal discretionary accruals. However, only a significant positive association is found for the client importance measure (lnTOTALFEES) and positive discretionary accruals. The results support the economic bond theory of DeAngelo (1981). The auditor’s incentive is likely to be influenced by a client’s total economic importance to the auditor relative to other clients. Furthermore, I have examined the influence of audit committee
governance on the relation between client importance and earnings management. I found no evidence that audit committee governance has influence on the relation between client importance and earnings management. Similarly results are provided for the internal board governance and corporate governance.
This research has various limitations and room for improvement. First, the corporate governance indexes, A-index and B-index, may not capture the effectiveness of the corporate governance mechanism. This suggests that the composite score of the A-index or B-index may not accurate reflect the strength of respectively, the audit committee and internal board
governance. Second, the sample data is limited. This is due to the data limitations on the ISS database for the years prior to 2007. A data set with a wide range of years could provide more valuable information on changes in the market for audit and non-audit services, auditor incentives and quality of the audit. Third, the purchase of non-audit and audit services and the composition of the audit committee index and internal board governance index can create an endogeneity problem (Larker and Richardon, 2004; Xie et al., 2003). However, this problem is inherent in all previous studies that examine the relation between (non)-audit services, earnings management and corporate governance. Furthermore, for further research alternative measures of auditor independence and financial reporting quality could be used to test the hypotheses.
! ! ! ! ! ! ! ! ! ! ! ! ! ! !
