Faculty of Economics and Business
Master thesis:
The relationship between firms earnings management
choice, auditor-quality and litigation risk.
Student Gert-Jan Rotteveel
Studentnummer 10259392
Inleverdatum 17 juni 2015 Final version
Word count: 11659
MSc Accountancy and Control, variant Accountancy 2014-2015
Faculty of Economics and Business, University of Amsterdam
2
Statement of Originality
This document is written by student Gert-Jan Rotteveel who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Abstract
I predict that firms that have an incentive to manage earnings and who employ a high-quality auditor make less use of accrual based earnings management. Additionally I expect that litigation risk increases this negative relationship. Furthermore I predict that that managers of firms with an incentive to manage earnings make more use of real earnings management. Additionally I expect that litigation risk increases this positive relationship. Using
discretionary accruals as a proxy for accruals based management, and abnormal cash flow from operations and abnormal discretionary expenses as a proxy for real earnings
management I find that (1) the presence of a high-quality auditor constrains the use of accrual based earnings management, (2) litigation risk increases this negative relationship, and (3) there is a weak positive relationship between the use of real earnings management and the presence of a high-quality auditor. I find mixed evidence regarding the effect of litigation risk with regard to the use of real earnings management and the presence of a high-quality auditor.
This study contributes to the existing literature on accrual based earnings management by confirming results in prior literature. Secondly this study contributes to the literature on real earnings management by providing (weak) empirical evidence on the negative
relationship between real earnings management and the presence of a high-quality auditor. Thirdly this study contributes to the literature on earnings management tradeoff-decisions, because I provide empirical evidence on the decisions managers of firms make with regard to the use of accrual based earnings management or real earnings management.
4
Content
Statement of originality...2
Abstract………...3
1 Introduction……….5
2 Literature review and hyphotheses... 9
2.1 Earnings management...9
2.2 High-quality auditor characteristics... 10
2.3 High-litigation risk characteristics………...11
2.4 Tradeoff decision between using ABEM and REM... 12
2.5 Hypotheses development... 13 3 Research design...14 4 Evidence………... 20 4.1 Sample………...20 4.2 Descriptive statistics………... 20 4.3 Results………...………...24 5 Conclusion...32 References…………...34
5
1 Introduction
The objective of this study is to gain more knowledge on firms earnings management choice and whether this choice is influenced by the presence of a high quality auditor and whether there is a difference between firms operating in a high- or low-litigation risk industry. I provide empirical evidence regarding the relationship between accrual based earnings
management (real earnings management) and the presence (absence) of a high-quality auditor for firms operating in a high- (low-) litigation risk industry. My motivation for this study is the study of Cohen & Zarowin (2010). They empirically show that managers of firms use more real earnings management around seasoned equity offerings when their firm is audited by a Big-8 auditor. Additionally the use of real earnings management (instead of accrual based earnings management) is potentially more harmful in the long run for the firm and its shareholders (Roychowdhury, 2006). Cohen & Zarowin (2010) emphasize that studying the tradeoff decision between accrual based earnings management and real earnings management in a more general setting (so not around SEO's or IPO's) is an important topic for future research.
Managers of some firms have an incentive to manage earnings. For example as described above, managers of firms that are about to issue new equity have an incentive to report higher earnings to boost share price. In a more general setting managers can have an incentive to manage earnings for example to avoid reporting a loss, or reach a certain earnings benchmark. Managers of firms who have an incentive to manage earnings have two methods to achieve reaching an earning benchmark. They can thus engage in accrual based earnings management and/or in real earnings management. Prior literature has shown that these two earnings management methods function as substitutes (Zang, 2012). So when management of firms who have an incentive to manage earnings are not able to use one of the methods, they are more likely to engage in the other earnings management method. Prior literature shows that firms who have an earnings management incentive resort to real earnings management when their ability to use accrual based earnings management is constrained (Cohen & Zarowin, 2010).
There has been a large stream of research about earnings management and factors that constrain earnings management. Prior research shows that Big 8 audit firms constrain
earnings management through discretionary accruals (e.g Francis et al., 1999). Earnings management through discretionary accruals is known as accrual based earnings management. Kanagaretnam et al. (2010) investigate the relation between auditor reputation and earnings management. In their research auditor reputation consists of type auditor and auditor industry
6
specialization. They want to examine if auditor reputation constrains accrual based earnings management in an industry with strict regulatory oversight. The setting of this research is therefore the banking industry. They find that both variables (type auditor and auditor industry specialization) have a moderating effect on benchmark beating. Kwon et al. (2007) also investigate the effect of auditor industry specialization on audit quality. They examine if this effect differs between countries with either a weak or a strong legal environment. They find that auditor industry specialization improves audit quality in both weak and strong legal environments but auditor industry specialization is more effective in improving audit quality in countries with a weak legal environment compared to countries with a strong legal
environment. Audit quality is measured by discretionary accruals used by the managers of the firms. So they find that auditor industry specialization constrains accrual based earnings management.
Recent research is getting more involved with the use of real earnings management by managers of firms who have an incentive to manage earnings. Next to accrual based earnings management, managers can also use real earnings management which is potentially more harmful to the firm and its shareholders in the long run. Vladu & Cuzdriorean (2014) argue, in their overview paper about earnings management research, that there is much research about accrual based earnings management but that only a few studies asses real earnings
management. When managers of a firm are constrained in using accrual based earnings management they have thus an incentive to use real earnings management. As described above Cohen & Zarowin (2010) investigate the tradeoff decision managers of firms who have an earnings management incentive make between using accrual based earnings management and real earnings management around seasoned equity offerings. They find that managers use more real earnings management around seasoned equity offerings when their firm is audited by a Big-8 auditor. Furthermore they find that managers also use more real earnings
management when their firm is in a high-litigation industry. In addition Chi et al. (2011) investigate whether managers of firms use more real earnings management when their ability of using accrual based earnings management is constrained by a high-quality auditor. Their sample consist of a mix of firms that meet or just beat earnings benchmarks in a more general setting and firms around seasoned equity offerings. They find that firms audited by a high-quality auditor have higher levels of real earnings management, however the results are weak. Contrary to Chi (2011) and Cohen & Zarowin (2010), Zang (2012) finds that the presence of a Big-8 auditor constrains management of firms who have an incentive to manage earnings in the use of accrual based earnings management, but she doesn't find that the use of real
7
earnings management is increasing with the presence of a Big-8 auditor. Zang studies the tradeoff decision between accrual based earnings management and real earnings management in a more general setting (so not around equity offerings).
The research question I try to answer with my research is the following: Does the presence of a high-quality auditor in a high/low litigation risk industry have influence on firms earnings management choice? Based upon prior literature I predict a negative
relationship between the use of accrual based earnings management by management of firms who have an incentive to manage earnings and the presence of a high-quality auditor.
Additionally I expect that the use of accrual based earnings management is more constrained when a firm operates in a high-litigation risk industry and is audited by a high-quality auditor. Furthermore I expect a (weak) positive relationship between the use of real earnings
management by managers of firms who have an incentive to manage earnings and the
presence of a high-quality auditor. In addition I predict this positive relationship to be stronger for earnings management incentive firms that operate in a high-litigation risk industry and which are audited by a high-quality auditor.
To provide empirical evidence to test my predictions I performed regressions and analysis on the variables high-quality auditor, if a firm is operating in a high- or low-litigation risk industry, and if a firm is operating in a high-litigation risk industry and is audited by a high-quality auditor. I find an association between these variables and the earnings
management methods. quality auditor is determined by being a Big-4 auditor. High-litigation risk industry is determined by SIC-codes of the firms. The use of accrual based earnings management is measured by discretionary accruals (Jones, 1991; Zang 2012). The use of real earnings management is measured by two real earnings management methods. The first method is determined by abnormal cash flow from operations and the second method is determined by abnormal discretionary expenses (Roychowdhury, 2006). Furthermore I included 5 control variables: firm-size, return on assets in year t-1, market-to-book ratio in year t-1, leverage in year t-1 and auditor tenure.
My sample consists of firms with an incentive to manage earnings from the S&P 500 for the years 2004 through 2013. Total sample size includes 617 firm-year observations. This small number of firm-year observations is due to the way I try to construct a sample of larger and more stable firms, and also because I eliminate firms without a clear incentive to manage earnings. In my research I find support for the negative relationship between the use of accrual based earnings management by management of firms who have an incentive to manage earnings and the presence of a high-quality auditor. Additionally I find support that
8
the use of accrual based earnings management is more constrained when a firm operates in a high-litigation risk industry and is audited by a high-quality auditor. Furthermore I find weak support for the positive relationship between real earnings management and the presence of a high-quality auditor. Finally I find mixed evidence for the relationship between the use of real earnings management by firms that operate in a high-litigation risk industry and which are audited by a high-quality auditor.
With this study I contribute to the literature in 3 ways. Firstly I contribute to the existing literature on accrual based earnings management. I show empirically that managers of firms that employ a high-quality auditor are constrained in the use of this earnings
management method, thereby confirming prior research. I also show that litigation risk increases this negative relationship between accrual based earnings management and the presence of a high-quality auditor. Secondly I contribute to the literature on real earnings management by providing (weak) empirical evidence on the negative relationship between real earnings management and the presence of a high-quality auditor. Thirdly I contribute to the literature on earnings management tradeoff-decisions, because I provide empirical
evidence on the decisions managers of firms that have an incentive to manage earnings make with regard to the use of accrual based earnings management or real earnings management. I explicitly contribute to the research of Cohen & Zarowin (2010) because they emphasize that studying the tradeoff decision between accrual based earnings management and real earnings management in a more general setting is an important topic for future research.
From a societal point of view the research on the tradeoff decision between accrual based and real earnings management is important for shareholders and stakeholders of firms. This is because the use of real earnings management as a substitute of accrual based earnings management is potentially more harmful to the firm and its shareholders in the long run. Cohen & Zarowin (2010) find evidence that the negative effects of using real earnings management on subsequent operating performance are greater than the effects of using accrual based earnings management, at least in the seasoned equity offerings context. As this effect on operating performance may be the same when managers use real earnings
management in a more general setting (so not around seasoned equity offerings) it is
interesting to investigate the tradeoff decision between the earnings management methods in a more general setting and find what factors constrain or, on the opposite, increase the use of real earnings management.
The rest of the paper is organized as follows: in section 2 relevant literature is
9
hypothesis. In section 4 the sample selection process is explained, provide the descriptive statistics and discuss the results. Section 5 provides the conclusion.
2 Literature review and hypotheses 2.1 Earnings management
The motivation for managers to engage in earnings management can come from different sources (Chen & Tsai, 2010). Motivation can come from attitudes and beliefs. According to Chen & Tsai this refers to the personal perceptions and behavioral tendencies to earnings management. Beliefs and attitudes includes altruistic, selfish and behavior convictions. Altruistic means that managers engage in earnings management for the benefit of the firm, so for example they want to prevent a drop in share prices. Selfish refers to managers engaging in earnings management for their own personal gain, for example to obtain a bonus. Behavior convictions refers to the assessment of earnings management behavior and expectations of the potential pros and cons(Chen & Tsai, 2010). According to Chen & Tsai motivation to engage in earnings management can also come from pressure from affiliated parties. This parties are for example supervisors, shareholders, creditors or financial analysts.
Another theory which can explain earnings management behavior of managers stems from agency theory. There are incentives for managers to present earnings that will maximize the wealth of the firm and/or maximize the wealth of the manager (Becker et al., 1998). This is because managers are mostly at least partially compensated with compensation plans that are based on reported earnings. After the manager is appointed and the compensation plan is determined, a moral hazard problem arises. According to agency theory this moral hazard problem arises because the agent (manager) will behave in his own best interest and not in the best interest of the principal (owners of the firm). When a manager is compensated with a bonus when he reaches a certain earnings benchmark, he will try to reach this earnings benchmark. If he is not able to reach the benchmark in the normal course of the business he can engage in earnings management to reach the earnings benchmark.
No matter if the choice of managers to engage in earnings management stems from selfish behavior, is in the best interest of the firm, or comes from pressure from affiliated parties there are two earnings management methods he can use to meet his goal. The following definition of earnings management comes from Healy & Wahlen (1999):
Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some
10
stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depend on reported accounting numbers.
From this we can see that the first earnings management method a manager can use has to do with using judgment in financial reporting, ABEM. When a manager makes the financial report he has to use his judgment to estimate future economic events, for example the expected lives of long term assets. In this example with different expected lives the
depreciation expenses will be different and so the manager can use this to manage earnings. From the definition of Healy & Wahlen we can see that the second earnings
management method a manager can use has to do with structuring transactions, REM. A manager has to use judgment in working capital management, for example inventory levels. A manager can influence the cost of goods sold for a year through producing more at the end of the year. Also a manager has to make decisions in discretionary expenses, such as R&D. When he defers R&D expenses at year-end to the next year this will have a positive effect on earnings for this year.
2.2 High-quality auditor characteristics
According to DeAngelo (1981) audit quality is the joint probability of detecting and reporting material financial statement errors. She argues that auditors can opportunistically lower audit quality to retain a particular client. This is because some aspects of audit
technology gives incentives to do so. Audit technology is characterized by significant client-specific startupcosts (DeAngelo, 1981). The present auditor of a firm possess cost advantages in future audits of this firm over potential competitors, and according to DeAngelo they can earn client-specific quasi-rents because they can set future audit fees above the avoidable costs of producing audits. This gives an incentive to lower audit quality, so not report any material misstatement errors for a particular client, in order to retain this client in future periods.
DeAngelo (1981) also argues that larger audit firms supply audits of higher quality than small audit firms. This is because larger audit firms have incentives to supply high-quality audits. Because larger audit firms have more clients and with this more aggregate client-specific quasi-rents they can be more independent. If they report a material
misstatement and therefore loses the client, they only lose a small portion of client-specific quasi-rents. Also it is assumed by DeAngelo that an audit firm will lose clients if a low-quality audit becomes known. For larger audit firms this gives an incentive to produce
high-11
quality audits because they have more clients to lose. So according to DeAngelo large audit firms provide higher quality audits than small audit firms.
Another theory why larger audit firms provide higher quality audits comes from the brand name investment model of Klein & Leffler (1981). They argue that audit-firms are motivated to develop and maintain a strong brand name for high-quality audits so that they can obtain and protect quasi-rents from clients. According to Klein & Leffler firms first develop a strong brand name for high-quality audits and from that they can earn a price that is higher than the price for a minimum-quality audit. Firms with a strong brand name have an incentive to keep on delivering high-quality audits because of the potential loss of brand name. According to Francis & Wilson (1988) the big-8 audit firms are brand name producers in the U.S. audit market. So according to the brand name investment model large audit firms provide higher quality audits than small audit firms.
Next to the size of the audit-firm as an indicator of a high-quality auditor, prior literature suggests that auditor industry specialization is also an indicator of a high-quality auditor. Simunic and Stein (1987; in: Kwon et al., 2007) argue that audit firms that are industry-focused are more likely to invest in for example technologies, that improve the quality of the audits in the focal industries of the firm. In addition Reichelt & Wang (2010) argue that auditor industry specialists have greater knowledge of the accounting practices used in the industry. As a result they can better detect earnings management techniques used by managers in this industry. Also in line with the above mentioned brand name investment model of Klein & Leffler, Reichelt & Wang (2010) argue that auditor industry specialists have developed a reputation for industry expertise, so they have an incentive to keep on delivering high-quality audits because of the potential loss of industry expert reputation. So according to Reichelt & Wang auditor industry specialist provide higher quality audits than non-industry specialists.
2.3 High-litigation risk characteristics
According to Kim & Skinner (2012) almost all research in accounting literature that investigates the effect of litigation risk uses an industry-based proxy to measure litigation risk. Litigation risk is the risk of securities class action lawsuits. A common proxy is based on membership in the biotechnology, computers, electronics, and retail industries (Kim & Skinner, 2012). According to prior literature litigation risk can have an effect on corporate decisions. Kim & Skinner argue that when the industry based proxy is supplemented with firm characteristics such as size, growth and growth opportunities the predictive ability of the
12
litigation risk measure improves considerably. Also larger potential amounts to be paid when a firm loses a lawsuit is more attractive to plaintiffs' attorneys. According to Kim & Skinner this is strongly associated with litigation risk. So this explains that litigation risk is probably higher for larger firms and firms with more stable growth because these long-lasting firms have more money that can be paid out when plaintiffs win a lawsuit. Also from deep pockets theory we know that Big-4 firms have higher litigation risk because they have 'deeper pockets' (Lee & Mande, 2003). This means that in case of a lawsuit larger accounting firms will have more resources to pay the damages to plaintiffs.
2.4 Tradeoff decision between using ABEM and REM
Recent research on earnings management (Cohen & Zarowin, 2010; Zang, 2012) shows that managers use ABEM and REM as substitutes to manage earnings to meet certain earnings benchmarks. The survey of Graham et al. (2005; in: Cohen & Zarowin, 2010) finds that managers prefer to use REM instead of ABEM to manage earnings. One reason for this is that the use of ABEM to meet or beat an earnings benchmark is likely to draw more attention from the auditor than the use of REM. Empirical evidence of managers using REM to meet or beat earnings benchmarks comes for example from Bushee (1998). He finds in his research on institutional investors, when it is important for a manager to meet or beat an earnings
benchmark, that managers reduce R&D expenditure.
According to prior research the cost of using ABEM and REM influences managers tradeoff decision between using ABEM or REM (Cohen & Zarowin, 2010; Zang, 2012). This is empirically shown in Cohen et al. (2008), where they find that because of the heightened scrutiny of accounting practice after the passage of SOX, managers of firms use less ABEM but more REM. Cohen & Zarowin (2010) argue that the cost of using ABEM includes the scrutiny provided by auditors. Zang (2012) argues that managers of firms face different constraints for using ABEM and REM. She argues that when discretion is more constrained for one of the earnings management methods, the manager of the firm will make more use of the other method. Zang argues that both earnings management methods are costly activities and that the tradeoff decision between using ABEM and REM depends on the relative
costliness of the two methods, which is influenced by the firms accounting environment. The influence of the accounting environment is seen in another cost of using ABEM. According to Cohen & Zarowin (2010), in addition to scrutiny provided by auditors, the cost of ABEM includes also potential litigation penalties.
13
2.5 Hypotheses development
According to DeAngelo (1981) present audit-firms at a client can earn client-specific quasi-rents. Because it is assumed by DeAngelo that when a low-quality audit becomes known the audit firm loses clients, she argues that larger audit firms have incentives to supply high-quality audits, because they have more clients and therefore have more clients to lose. Klein & Leffler (1981) argue that audit-firms are motivated to develop and maintain a strong brand name for high-quality audits so that they can obtain and protect quasi-rents from clients. According to Francis & Wilson (1988) the big-8 audit firms are brand name producers in the U.S. audit market, so they are incentivized to keep on delivering high-quality audits because of the potential loss of brand name.
As audit firms that are industry-focused are more likely to invest in for example technologies, it is argued by Simunic and Stein (1987; in: Kwon et al., 2007) that the quality of the audits in the focal industries of the firm will be of higher quality. Also because auditors that are an expert in the industry have greater knowledge of the accounting practices used in the industry, they can better detect earnings management techniques used by managers in this industry (Reichelt & Wang, 2010). According to Reichelt & Wang (2010) auditor industry specialists also have an incentive to keep on delivering high-quality audits because their reputation of industry expert is at stake.
There is a lot of empirical evidence that large audit firms constrain ABEM. For
example Becker et al.(1998) find that firms that are audited by a non-big-6 audit firm have, on average, significant higher levels of discretionary accruals than clients audited by a big-6 audit firm. Also for auditor industry specialists there is empirical evidence that these industry specialists constrain ABEM. For example Kwon et al.(2007) finds evidence that firms that are audited by an industry specialist have lower discretionary accruals than firms that are audited by a non-industry specialist.
According to Burnett et al. (2012) high-quality auditors are associated with
constrained ABEM. They constrain opportunistic accrual choices because of the higher risk that ABEM will be discovered. Burnett et al. argue that auditors are not responsible for evaluating the motivations of managers related to real activities, so therefore they do not scrutinize REM. Based on the aforementioned the first hypotheses is:
14
When audited by a high-quality auditor managers are constrained in their ability to use
ABEM. As was argued by Burnett et al. (2012) that auditors are not responsible for evaluating the motivations of managers related to real activities, there is an opportunity for managers to engage in REM to meet or beat earnings benchmarks. As recent research on earnings
management (Cohen & Zarowin, 2010; Zang, 2012) shows that managers use ABEM and REM as substitutes to manage earnings, I hypothesize that:
H1b: The presence of a high-quality auditor is positively associated with REM.
In prior literature it is argued that managers of firms face different constraints for using ABEM and REM (Zang, 2012). She argues that when discretion is more constrained for one of the earnings management methods, the manager of the firm will make more use of the other method. The tradeoff decision between is influenced by the cost of using ABEM and REM. Cohen & Zarowin (2010) find for example that managers make less use of ABEM after the passage of SOX. Because of the heightened scrutiny of auditors after SOX, using ABEM becomes more costly. Another cost that is included in the use of ABEM are potential
litigation penalties. From deep pockets theory we also know that Big-4 firms have higher risk of having to pay damages to plaintiffs in a lawsuit. Therefore I hypothesize that:
H2a: Litigation risk increases the negative association between the presence of a high-quality auditor and ABEM.
Because of the aforementioned that mangers use ABEM and REM as substitutes in managing earnings, I predict that:
H2b: Litigation risk increases the positive association between the presence of a high- quality auditor and REM.
3. Research Design
Measure of accrual based earnings management
Following prior literature (e.g. Jones, 1991; Zang 2012), I use the modified Jones model to estimate discretionary accruals as a proxy for ABEM. Managers can use their discretion in choosing the level of discretionary accruals and Jones (1991) suggests that a high
15
discretionary component of total accruals is an indicator of accrual based earnings management (ABEM).
I use the following model to estimate coefficients α1, α2 andα3:
Accrualst/At-1 = α0 + α1(1/At-1) + α2(ΔSt/At-1) + α3(PPEt/At-1) + εt (a)
Where:
Accrualst = Total accruals in year t (earnings before extraordinary items and
discontinued operations minus the operating cash flows reported in the statement of cash flows in year t (Zang, 2012)
At-1 = Total assets in year t-1
ΔSt = Change in net sales from year t-1 to year t
PPEt = Gross property, plant, and equipment
Coefficients estimates of model (a) are used to estimate the firm specific 'normal' level of accruals for my sample. Measure of discretionary accruals (DiscrAccruals) is the difference between actual accruals and the firm specific 'normal' level of accruals. DiscrAccruals are then a proxy for ABEM. Because I focus on upward-earnings management suspect firms higher positive values of discretionary accruals are an indication of earnings management. I expect the sign to be negative form firms that employ a high-quality auditor.
Measures of real earnings management
In prior literature (f.e. Roychowdhury, 2006) three real activities manipulation methods, which are a measure of REM, are investigated. Managers of firms can use these methods to meet or beat earnings benchmarks. Roychowdhury investigates cash flow from operations (CFO) because the first method to manipulate real activities according to Roychowdhury is the manipulation of sales, for example through unsustainable price discounts at year-end. This will lead to more sales being made at year-end which will increase earnings in the current year. The second method is overproduction which reduces costs of goods sold in the current period. This will lead to lower costs of the products that are sold this year and thus also increase earnings in the current year. And the third method is to reduce discretionary expenses, for example the reduction of R&D expenses. These expenses are generally expensed in the period when they are incurred, so decreasing discretionary expenses at year-end will also lead to higher earnings in the current year.
I have chosen to include two of the three real activities manipulation methods in my research. I investigate CFO and the reduction of discretionary expenses, because prior
16
research indicates that managers predominantly use this real earnings management methods (f.e. Rowchowdury, 2006; Chi, Lisic & Pezner, 2011). To find out if managers use these manipulation strategies I follow Roychowdhury and use the model in Dechow et al. (1998) to derive normal levels of CFO and normal levels of discretionary expenses. Deviations from the normal levels are termed abnormal CFO and abnormal discretionary expenses and these are an indicator for REM.
The model for CFO to estimate coefficients α1, α2 andα3:
CFOt/At-1= α0 + α1(1/At-1) + α2(St/At-1) + α3(ΔSt/At-1) + εt (b)
Where:
CFOt = cash flow from operations in year t (represents cash flow from operations as
reported in the statement of cash flows) At-1 = total assets at the end year t-1
St = sales during period t
ΔSt = St-St-1
For every firm-year the 'normal' level of CFO is then calculated using estimated coefficients form model (b). Measure of abnormal CFO (ABN_CFO) is then the actual CFO minus the 'normal' CFO. ABN_CFO are then a proxy for real earnings manipulation. Because I focus on upward-earnings management suspect firms I expect the sign on ABN_CFO to be negative for firms with an high-quality auditor (Chi et al., 2011).
The model for discretionary expenses to estimate coefficients α1 andα2:
DISEXPt/At-1= = α0 + α1(1/At-1) + α2(St-1/At-1) + εt (c)
Where:
DISEXPt = discretionary expenses in year t (sum of advertising expenses, R&D
expenses, and selling, general and administrative (SG&A) expenses) At-1 = total assets at the end year t-1
St = sales during period t
For every firm-year the 'normal' level of discretionary expenses is then calculated using estimated coefficients form model (c). Measure of abnormal discretionary expenses
17
expenses. ABN_DISEXP are then a proxy for real earnings manipulation. . Because I focus on upward-earnings management suspect firms I expect the sign on ABN_DISEXP to be negative for firms with an high-quality auditor (Chi et al., 2011).
Measure of High Quality Auditor (HQA)
According to prior literature large audit-firms supply higher quality audits than small audit firms (f.e. DeAngelo, 1981, Becker et al., 1998). Also firms who are industry experts supply higher quality audits than firms who are not experts in the industry (f.e. Reichelt & Wang, 2010). Because industry experts are almost always Big-4 firms (f.e. 96,8% in the sample of Chi et al (2011) in their investigation of the effect of industry expertise on audit quality consists of Big-4 firms) I choose to only incorporate a Big-4 firm variable as a proxy of high quality auditor.
Measure of High Litigation Risk Industry (HLIT)
A common proxy used in literature to investigate the effect of litigation risk is based on an industry-based proxy. According to Kim & Skinner (2012) being in the biotechnology, computers, electronics, and retail industries will increase litigation risk. These industries are SIC codes 2833–2836, 8731–8734, 7371–7379, 3570–3577, and 3600–3674. But Kim & Skinner argue there are more factors explaining litigation risk. Larger firms and more stable firms will probably have a higher change of being involved in lawsuits because plaintiffs think there is a better chance of getting money from firms which are larger and/or firms that will be in business for a longer time. I tried to incorporate these kind of firms in my sample. This is further explained in the sample selection. Also from deep pockets theory we know that Big-4 firms have higher litigation risk because they have 'deeper pockets' (Lee & Mande, 2003). This means that in case of a lawsuit larger accounting firms will have more resources to pay the damages to plaintiffs. Therefore also firms that operate in a high litigation risk industry that employ a Big-4 auditor (HQAHLIT) are investigated.
Control variables
I included five control variables in my study. The first control variable is SIZE (logarithm of market value of equity). Firm size is included as large firms have high political costs and thus are more likely to manage earnings (Watts and Zimmerman, 1986 in: Sun & Liu, 2012).
18
beginning period total assets). This variable is included to control for the effect of firm performance on the earnings management measures (Chi et al,2011).
Thirdly, MTB (market-to-book ratio, measured by market value of equity divided by book value of equity) is included as a control variable. This variable is included because firms with high MTB ratios are suggested to have high growth opportunities, and are more likely to engage in earnings management according to Skinner and Sloan (2002; in: Sun & Liu, 2012).
The fourth control variable included in the analysis is LEVERAGE (short-term debt plus long-term debt divided by average total assets). This control variable is included because prior literature (f.e. Dechow et al, 1996) find that there is a positive association between leverage and earnings management (Sun & Liu, 2012).
The last control variable included is TENURE (logarithm of number of years the auditor has audited the company’s financial statements). This control variable is included because in literature there is a lot of debate about the consequences of long auditor tenure with a client. Longer auditor tenure could lead to higher audit quality because the auditor becomes more familiar with the auditing environment of the client, compared to short audit tenure. On the other hand, longer auditor tenure could lead to lower audit quality because there is the risk of the auditor becoming too friendly with management of the company. Empirical evidence is mixed. For example Johnson et al. (2002) find that longer tenure leads to higher audit quality, where Davis et al. (2009) find that longer tenure leads to lower audit quality.
Empirical models
To test my hypotheses to following empirical models are tested:
To test H1a the following regression is run:
DiscrAccruals= β0 + β1HQA + β2SIZE + β3ROAt-1 + β4 MTBt-1 + β5LEVERAGE +
β6TENURE (1)
To test H2a the following regression is run:
DiscrAccruals= β0 + β1HQA + β2HLIT + β3HQAHLIT + β4SIZE + β5ROAt-1 +
β6 MTBt-1 + β7LEVERAGE+ β8TENURE (2)
To test H1b the following regressions are run:
19
To test H2b the following regression is run:
REM= β0 + β1HQA + β2HLIT + β3HQAHLIT + β4SIZE + β5ROAt-1 + β6 MTBt-1 +
β7LEVERAGE+ β8TENURE (4)
Where:
DiscrAccruals = measure of accrual based earnings management
REM = one of the measures of real earnings management
HQA = firm audited by high-quality auditor (dummy variable, 1=yes, 0=no)
HLIT = firm operating in high-litigation industry
HQAHLIT = firm audited by high-quality auditor and operating in a high litigation risk industry (dummy variable, 1=yes, 0=no
SIZE = size of the firm measured by logarithm of market value of equity
ROA = return on assets measured by income before extraordinary items divided by beginning period total assets
MTB = market-to-book ratio measured by market value of equity divided by book value of equity
LEVERAGE = measured by short-term debt plus long-term debt divided by average total assets
TENURE = length of auditor with the firm measured by logarithm of number of years the auditor has audited the company’s financial statements
20
4 Evidence 4.1 Sample
I start with a sample of firms from the S&P 500 in the years from 2004-2013. Because I argued in section 3 that larger and more stable firms will probably have a higher change of being involved in lawsuits because plaintiffs think there is a better chance of getting money from firms which are larger and/or firms that will be in business for a longer time I require that the firms in the sample have data for all 10 years (2004-2013). Requiring that firms have data for all ten years leads to a sample of firms that are in business in the whole sample period and thus this probably will be relatively stable and larger firms. Starting with this requirement leads to a sample of 14.959 firm-year observations. Following prior literature
(Roychowdhury, 2006; Zang, 2010) I eliminate firms in regulated industries (SIC codes between 4400 and 5000) and banks and financial institutions (SIC codes between 6000 and 6500). This results in the loss of 2.036 year observations. Requiring that for the firm-year observations all data is available to calculate the ABEM and the REM measures (discretionary accruals and discretionary expenses) reduces the sample by 3.987 firm-year observations. Requiring that for all observations data is available to calculate the control variables (SIZE, ROA, MTB, LEVERAGE, TENURE) reduces the sample by another 103 observations. This results in a final sample of 8.833 observations. This sample is used to run the regression models (a), (b) and (c)toestimate the coefficients for calculating 'normal' levels of accruals, 'normal' CFO and 'normal' discretionary expenses. Prior literature suggests that firm-years with earnings right at or just above important benchmarks are likely to manage earnings to meet these important benchmarks (Zang, 2012). Following Roychowdhury (2006) I concentrate on earnings management suspect firm-years to increase the power of my tests. An earnings management suspect firm-year is a year where net income scaled by total assets is greater than or equal to zero but less than 0.005 (Roychowdhury, 2006). Because I focus on these firms that have incentives to manage earnings, I eliminate another 8.224 observations. This results in the final sample of 617 firm-year observations, which is used to run the
regression models (1), (2), (3) and (4). The sample selection procedure is presented in table 1.
4.2 descriptive statistics
In table 2 are the number of observations per industry for the final sample of 617 firm-years. Manufacturing years represent almost half of the sample with almost 47% of total firm-years. Wholesale and retail estate represents 24% and services 25%.
21
TABLE 1 Sample selection
Firm-year observations S&P 500 2004-2013 firms with data for all 10 years 14.959
Regulated industries, banks/financial institutions (2.036) Missing data to calculate ABEM & REM measures (3.987) Missing data to calculate control variables (103) Observations with data available for regression
models
8833
Firms without earnings management incentives (8.216) Final sample of earnings management suspect
firm-year observations
617
TABLE 2
Observations per industry
Industry SIC # Observations
Agriculture, Forestry and Fishing 0100 – 0999 11
Agricultural, Production-Livestock and Animal Specialties 1000 - 1999 13
Manufacturing 2000 - 3999 289
Transportation, Communication, Electric, Gas and Sanitary 4000 – 4999 -
Wholesale and retail estate 5000 – 5999 149
Services 7000 – 8999 155
22
In table 3 information is summarized concerning the distribution of firm-years with respect to type auditor and type litigation industry. In the final sample 481 firm-years are audited by a Big 4 firm which is 78% of total firm-years. In 136 firm-years the financial statements are audited by other than Big 4 auditors. This thus means that 78% of the firm-years are audited by a high-quality auditor and 22% of the firm-years are audited by an auditor of lower quality. Furthermore 137 firm-year observations are of firms which operate in a high-litigation risk industry (SIC codes 2833–2836, 8731–8734, 7371–7379, 3570–3577, and 3600–3674), which is 22,2%. The other 77,8% of observations are of firms operating in a non-high-litigation risk industry. Furthermore 111 firm-year observations are of firms which operate in a
high-litigation risk industry and are audited by a high-quality auditor. This is 18% of the total sample.
To summarize the largest part of firms are audited by a high-quality auditor (78%). Most firms don't operate in a high litigation risk industry (77,8%). And for the firms that operate in a high-litigation risk industry most of them employ a high-quality auditor. From the total of 137 firm-year observations in a high-litigation risk industry, 111 are audited by a high-quality auditor, which is 81%. So firms that operate in a high-litigation risk industry employ a slightly larger percentage of high quality auditors compared to the total sample (81% versus 78%).
TABLE 3
Distribution of firms audited by High Quality Auditor (HQA), operating in a High Litigation Risk Industry (HLIT), and audited by HQA and operating in HLIT
(HQAHLIT)
Firm audited by/ operating in # Observations Percent
HQA 481 78% Not HQA 136 22% Total 617 HLIT 137 22,2% Not HLIT 480 77,8% Total 617 HQAHLIT 111 18% Not HQAHLIT 506 82% Total 617
23
In table 4 are the descriptive statistics of the sample of 617 firm-years. The mean values of DiscrAccruals and ABN_DISEXP are negative. The mean value of ABN_CFO is positive. The negative value of DiscrAccruals and the positive value of ABN_CFO are in line with prior literature (Becker et al., 1998; Chi et al., 2011). The negative value on ABN_DISEXP is not consistent with prior literature. The signs and values of the mean values of the control variables are in line with prior literature (Chi et al., 2011). Mean logarithm of size is 6,87. But the dispersion is pretty wide. The sample selection did not only filter out large firms.
Although there is also reasonable dispersion in return on assets, mean return on assets is around 4%, which is a steady growth rate. Mean market-to book-ratio is around 3, which according to Kim & Skinner indicaties firms with high growth oppurtunities. Kim & Skinner (2012) argue that firms with high growth oppurtunities have higher litigation risk because they will probably engage more in earnings management. Mean tenure is 8,62 years, which in most literature is medium-term tenure (f.e. Johnson, Khurana & Reynolds, 2002).
TABLE 4 Descriptive statistics
Variable Minimum Maximum Mean Standard deviation
DiscrAccruals -,3251 3,809 -,0168 ,1719 ABN_CFO -,7414 2,470 ,0209 ,2823 ABN_DISEXP -1,971 ,9974 -,2051 ,2887 SIZE ,4466 12,36 6,8693 2,264 ROAt-1 -,4005 ,3137 ,0425 ,0594 MTBt-1 -17,27 158,0 2,957 7,953 LEVERAGEt-1 ,0000 1,489 ,2260 ,2077 TENURE (log) 0,000 2,944 1,891 ,7345 TENURE 1 19 8,62 4,777
24
In table 5 correlations between the dependent and independent variables are presented. DiscrAccruals is positively and significantly associated with ABN_CFO and also
DiscrAccruals is positively associated with ABN_DISEXP, but not at the conventional 95% level. Also ABN_CFO and ABN_DISEXP are positively and significantly correlated. These correlations are in line with my hypotheses, as all dependent variables are expected to move in the same direction. Turning to the variables of interest to test my hypotheses,
DiscrAccruals is negatively and significantly correlated with HQA and HQAHLIT which is in line with H1a and H2a. Furthermore DiscrAccruals is slightly positively associated with HLIT, but this is not significant. The measures for REM (ABN_CFO and ABN_DISEXP) are negatively associated with HQA which is in line with H1b and H2b, but both are not
statistically significant. Furthermore both measures are positively and significantly associated with HQAHLIT, for which my expectation was that the sign would be negative. This can maybe be explained that the correlation between DiscrAccruals and HQA has a heavier weight in measuring correlations between the variables. Also there is a positive and
significant correlation between the two REM measures (ABN_CFO and ABN_DISEXP) and HLIT. These correlations suggest that firms that are audited by an high quality auditor are associated with less accrual based earnings management and more real earnings management. Furthermore the correlations suggest that when firms operate in a high litigation risk industry and are audited by a high-quality auditor firms engage less in accrual based earnings
management. These suggestions are in line with H1a, H1b and H2a. The correlations suggest that firms operating in a high-litigation risk industry and are audited by a high-quality auditor also engage less in real earnings management. This is conflicting with H2b. However, these are merely univariate correlations. I rely on the multivariate regressions analysis to test my hypotheses.
4.3 results
In table 6 the results of the OLS regression on accrual based earnings management are
presented. In column 1 the results are presented with respect to the regression including HQA as variable of interest. In column 2 next to HQA also HLIT and HQAHLIT are included. The R-squared of model 2 is 0,108 which means almost 11 percent of the variation in
discretionary accruals (the measure for accrual based earnings management) is explained by the variables of interest (HQA, HLIT, HQAHLIT) and the 5 control variables. The Anova F-value is 9,211. This means that the model is significant and fits the data.
25
TABLE 5 Correlations
DiscrAccruals ABN_CFO ABN_DISEXP HQA HLIT HQAHLIT
DiscrAccruals 1,000 ABN_CFO ,773*** (,000) 1,000 ABN_DISEXP ,075* ,278*** (,062) (,000) 1,000 HQA -,201*** -,018 -,031 (,000) (,650) (,446) 1,000 HLIT ,010 ,156*** ,211*** ,039 (,805) (,000) (,000) (,327) 1,000 HQAHLIT -,111*** ,084** ,172*** ,249*** ,877*** (,006) (,036) (,000) (,000) (,000) 1,000 * significant at 90% level ** significant at 95% level *** significant at 99% level
The coefficient on HQA in model 1 is negative and significant on the 95% confidence
interval. This implies that when the financial statements of a firm are audited by an auditor of high quality, management of these firms can make less use of discretionary accruals to manage earnings. Auditors of high quality constrain management in the use of discretionary accruals. This observation is consistent with hypothesis 1a, and in line with prior literature (f.e Becker et. Al, 1998). The coefficient on HQAHLIT in model 2 is negative and significant on the 99% confidence interval. This implies that management of firms that operate in a high-litigation risk industry and have a high-quality auditor, are constrained in the use of
discretionary accruals. Furthermore the coefficient on HQAHLIT in model 2 is -,405. This is more negative than the coefficient on HQA (-,123) in model 1. This implies that a high-quality auditor is decreasing the ability of managers to use discretionary accruals to manage earnings when the firm operates in a high-litigation risk industry. This is consistent with hypotheses 2a, which states that litigation risk increases the negative association between the
26
presence of a high-quality auditor and accrual based earnings management. Furthermore the coefficient on HQA in model 2 is negative but insignificant. This can be explained because HQAHLIT dominates HQA. The coefficient on HLIT is positive and significant. This implies that managers of firms that operate in a high-litigation risk industry make more use of
discretionary accruals. A possible explanation of this observation can be found in that the motivation to engage in earnings management can come from pressure from affiliated parties, as explained by Chen & Tsai (2010). When the firm is not audited by a high-quality auditor, management engages in accrual based earnings management to show a (small) profit instead of a loss to keep for example shareholders satisfied. Management then hopes that the auditor of lower quality (and also the shareholders) won’t recognize that the result is due to earnings management.
Turning to the control variables the coefficient on SIZE is negative and significant in model 1 and model 2. This implies that smaller firms engage more in accrual based earnings management. A possible explanation is that most of the (smaller) larger firms are audited by a (non) high-quality auditor, and thus management of (smaller) larger firms are constrained (less) more in the use of accrual based earnings management. The coefficient on ROAt-1 is
positive and significant in model 1 and model 2. This implies that firms that have a higher return on assets in year t-1 engage more in earnings management in the next year, compared to firms with a lower return on assets. A possible explanation is that there is pressure from affiliated parties (Chen & Tsai, 2010) to reach a certain earnings benchmark. And for
example this affiliated parties expect that the firm will make more profit than the year before. The coefficient on MTBt-1 is negative but insignificant. I expected the sign on MTBt-1 to be
positive because of the explanation of Skinner and Sloan (2002; in: Sun & Liu, 2012), that firms with high MTB ratios are suggested to have high growth opportunities, and are more likely to engage in earnings management. I can’t find a logical explanation for this. The coefficient on LEVERAGEt-1 is positive but insignificant. This means that there is a weak
positive association between leverage and accrual based earnings management, which is in line with prior literature (f.e. Dechow et al, 1996). Lastly the coefficient on Tenure is positive and significant. This implies that longer auditor tenure is associated with higher levels of accrual based earnings management. This observation implies that the auditor is becoming too friendly with management and thus allows management to use discretionary accruals to manage earnings, as is suggested for example by Davis et. al (2009)
27
TABLE 6
OLS regression of accrual based earnings management measure discretionary accruals on the presence of high quality auditors (I) and on the presence of high quality auditors
where firms operate in a high litigation risk industry (II)
I II Intercept 0,059** 0,030 (,023) (,260) HQA -,123** -,035 (,012) (,505) HLIT ,393*** (,000) HQAHLIT -,405*** (,000) control variables SIZE -,198*** -,203*** (,000) (,000) ROAt-1 ,120*** ,123*** (,006) (,004) MTBt-1 -,054 -,053 (,170) (,178) LEVERAGEt-1 ,043 ,047 (,287) (,261) TENURE ,108*** ,106*** (,007) (,007) n 617 617
Data are for the years 2004 – 2013. Data are on 617 firm-year observations. Significance is based on a two-tailed
test.
* significant at 90% level
** significant at 95% level
28
In table 7 the results of the OLS regression on real earnings management are presented. In column 1 and 3 the results are presented with respect to the regression including HQA as variable of interest. In column 2 and 4 next to HQA also HLIT and HQAHLIT are included. The R-squared of models 2 and 4 are 0.063 and 0,067 respectively. This means 6,3% (6,7%) of the variation in ABN_CFO (ABN_Disexp) (the measures for real earnings management) is explained by the variables of interest (HQA, HLIT, HQAHLIT) and the 5 control variables. The Anova F-value of ABN_CFO (ABN_Disexp) is 5,065 (5,350). This means that the models are significant and fitting the data.
The coefficient on HQA in model 1 is negative but insignificant. A negative sign on HQA in the model ABN_CFO means that firms engage more in this type of real earnings management. This is because for example when managers of firms try to boost sales at year-end, earnings can be taken on the sales made but no cash is received yet, and thus earnings are increasing, but cash flow from operations are not. The coefficient on HQA in model 3 is negative but insignificant. This implies that management of firms that are audited by a high-quality auditor have cut discretionary expenses, thereby increasing earnings, and thus engage more in real earnings management. This observations, in addition to the observations from table 6 (that high-quality auditors constrain the use of accrual based earnings management), imply that managers of firms with a high-quality auditor resort to the use of real earnings management methods (at least cutting discretionary expenses, which is significant) to manage earnings. This is consistent with hypotheses 1b, and in line with prior literature (f.e. Chi et. al, 2011). Furthermore the coefficient on HQAHLIT in model 2 is negative and significant. This implies that managers of firms which operate in a high-litigation risk industry and that are audited by a high-quality auditor engage in this type of real earnings management
(ABN_CFO). Furthermore the coefficient on HQAHLIT in model 4 is negative but
insignificant. This implies a weak association between the presence of a high-quality auditor for firms in a high-litigation risk industry and the use of this type of real earnings
management (ABN_Disexp). When I compare the coefficient on HQAHLIT in model 2 with the coefficient on HQA in model 1, I find that the coefficient on HQAHLIT (-,257) is more negative than the coefficient on HQA (-,018). This implies that managers of firms who operate in a high-litigation risk industry and employ a high-quality auditor, resort to this type of real earnings management (ABN_CFO) more compared to other firms. This is consistent with hypotheses 2b, which states that litigation risk increases the positive association between the presence of a high-quality auditor and real earnings management. When I compare the coefficient on HQAHLIT in model 4 with the coefficient on HQA in model 3, I find that the
29
coefficient on HQAHLIT (-,038) is less negative than the coefficient on HQA (-,079). This is not consistent with hypotheses 2b, because this implies that managers of firms who operate in a high-litigation risk industry and employ a high-quality auditor, resort to this type of real earnings management (ABN_Disexp) less compared to other firms. But because of the significance of the coefficients the results on ABN_CFO are more reliable than the results on ABN_Disexp. Furthermore the coefficients on HLIT are positive and significant for models 2 and 4. This implies that managers of firms that operate in a high-litigation risk industry make less use of real earnings management. A possible explanation of this observations can be found in the observation I did with respect to accrual based earnings management. I there found that managers of firms that operate in a high-litigation risk industry make more use of accrual based earnings management (because of pressure from affiliated parties). And thus if they make more use of accrual based earnings management there is less need to use real earnings management to manage earnings.
Turning to the control variables to coefficient on SIZE is negative but insignificant for models 1 and 2 (ABN_CFO). This implies that managers of smaller firms engage more in this type of real earnings management. A possible explanation is that smaller firms can more easily boost sales at year-end because operations are less structured then for larger firms. For models 3 and 4 (ABN_Disexp) the coefficients on SIZE are positive and significant at 90% and 95% respectively. This implies that management of larger firms make more use of this type of earnings management. A possible explanation is that managers of larger firms have more discretionary expenses they can cut, because probably larger firms have more
advertising expenses and invest more in R&D expenses, compared to smaller firms. The coefficients on ROAt-1 are positive and significant for models 1 and 2 (ABN_CFO), This
implies that firms that have a higher return on assets in year t-1 engage more in earnings management in the next year, compared to firms with a lower return on assets. A possible explanation is the same as described above in the discussion of control variables of table 6, that there is pressure from affiliated parties (Chen & Tsai, 2010) to reach a certain earnings benchmark. The coefficients on ROAt-1 are negative and insignificant for models 3 and 4
(ABN_Disexp). I have no logical explanation of this observations. The coefficients on MTBt-1
are positive but insignificant for models 1 through 4. This implies that managers of firms with higher market-to-book ratios engage more in both types of real earnings management. This is consistent with the explanation of Skinner and Sloan (2002; in: Sun & Liu, 2012). They argued that firms with high MTB ratios are suggested to have high growth opportunities, and
30
TABLE 7
OLS regression of real earnings management measures ABN_CFO and ABN_Disexp on the presence of high quality auditors (I+III) and on the presence of high quality auditors
where firms operate in a high litigation risk industry (II+IIII)
ABN_CFO ABN_Disexp I II III IIII Intercept 0,017 0,044 -,206*** -,247*** (,695) (,388) (,000) (,000) HQA -,018 ,032 -,079 -,080 (,716) (,553) (,111) (,133) HLIT ,369*** ,209** (,000) (,019) HQAHLIT -,257*** -,038 (,005) (,676) control variables SIZE -,046 -,057 ,130* ,118** (,398) (,285) (,016) (,027) ROAt-1 ,129*** ,139*** -0,044 -,033 (,004) (,002) (,321) (,458) MTBt-1 ,049 ,037 ,054 ,036 (,228) (,363) (,180) (,377) LEVERAGEt-1 -,078* -,040 -,177*** -,126*** (,061) (,350) (,000) (,004) TENURE ,063 ,069* -,015 -,004 (,125) (,088) (,715) ,928 n 617 617 617 617
Data are for the years 2004 – 2013. Data are on 617 firm-year observations. Significance is based on a two-tailed
test.
* significant at 90% level
** significant at 95% level
31
are more likely to engage in earnings management. The coefficients on LEVERAGEt-1 are
negative for models 1, 2, 3 and 4 (90%, insignificant, 99%, 99% respectively). This means that there is a negative association between leverage and real earnings management, which is in line with prior literature (Chi et. al, 2011). The coefficients on TENURE are positive for models 1 and 2 (insignificant and 90% respectively). This implies that longer auditor tenure is weakly associated with higher levels of this type of real earnings management. Lastly the coefficients on TENURE are slightly negative and insignificant for models 3 and 4. No reasonable conclusion can be drawn from this observations.
Below I will summarize the findings of table 6 on accrual based earnings management and table 7 on real earnings management. In model 1 of table 6 I find that the coefficient on HQA is negative and significant. This implies that when the financial statements of a firm are audited by an auditor of high quality, management of these firms are constrained in the use of discretionary accruals to manage earnings. This result is consistent with hypotheses 1a. In model 1 and 3 of table 7 I find that the coefficients on HQA are negative but insignificant. Negative coefficients on HQA in the models 1 (ABN_CFO) and 3 (ABN_Disexp) imply that management of firms engage more in this types of real earnings management. Although not statistically significant, there is a positive association between the presence of a high-quality auditor and the use of real earnings management. This results are consistent with hypotheses 1b. Furthermore these results, in addition to the results in model 1 of table 6 (management of firms that employ a high-quality auditor use less accrual based earnings management), imply that managers of firms that are audited by a high-quality auditor and thus are constrained in using accrual based earnings management to manage earnings, resort to the use of real earnings management methods.
The coefficient on HQAHLIT in model 1 of table 6 is negative and significant. This implies that managers of firms which operate in a high-litigation risk industry and employ a high-quality auditor are constrained in the use of discretionary accruals to manage earnings. In addition I find that the coefficient on HQAHLIT (-,405) in model 2 of table 6 is more negative than the coefficient on HQA (-,123) in model 1 of table 6. This results suggest that litigation risk increases the negative association between the presence of a high-quality auditor and the use of accrual based earnings management. This is consistent with hypotheses 2a. The coefficient on HQAHLIT in model 2 of table 7 is negative and significant. This implies that managers of firms which operate in a high-litigation risk industry and that are audited by a high-quality auditor engage in this type of real earnings management
32
is more negative compared to the coefficient on HQA (-,018) in model 1 of table 7. This results suggests that litigation risk increases the positive association between the presence of a high-quality auditor and the use of real earnings management. This is consistent with
hypotheses 2b. The coefficients on HQAHLIT in model 4 of table 7 is negative but insignificant. There is a weak association between the use of this earnings management method (ABN_Disexp) for firms operating in a high-litigation risk industry and the presence of a high-quality auditor. Contrary to the results on ABN_CFO, I find that the coefficient on HQA (-,079) in model 3 of table 7 is actually more negative than the coefficient on
HQAHLIT (-,038). This result suggests that litigation risk decreases the positive association between the presence of a high-quality auditor and the use of real earnings management. This is not consistent with hypotheses 2b.
Overall, in the results of the OLS regressions I found support for hypotheses 1a that the presence of a high-quality auditor is negatively associated with accrual based earnings management. In addition I found evidence that supports hypotheses 2a that litigation risk increases the negative association between the presence of a high-quality auditor and accrual based earnings management. Furthermore I found weak support for hypotheses 1b that the presence of a high-quality auditor is positively associated with real earnings management. Both real earnings management methods investigated are positively associated the presence of a high-quality auditor but the results were insignificant. Lastly I found mixed evidence with respect to hypotheses 2b. The results on the real earnings management method ABN_CFO are significant and support hypotheses 2b that litigation risk increases the positive association between the presence of a high-quality auditor and real earnings management. Contrary to the results on ABN_CFO are the results on ABN_Disexp. These results contradict hypotheses 2b. But the coefficient on HQAHLIT (ABN_Disexp) is insignificant, where the coefficient on HQAHLIT (ABN_CFO) is significant.
5 Conclusion
In a more general setting managers can have an incentive to manage earnings for example to avoid reporting a loss, or reach a certain earnings benchmark. They can engage in accrual based earnings management and/or real earnings management. Prior literature has shown that managers use the two earnings management methods as substitutes. In addition prior literature has shown that firms resort to real earnings management when their ability to use accrual based earnings management is constrained.
33
auditor in a high-litigation risk industry has an influence on firms earnings management choice. I therefore investigated the relationship between one of the earnings management methods and whether the firm employs a high-quality auditor. Furthermore I investigated whether there was a difference if the firm operates in a high-litigation risk industry and employs a high-quality auditor.
I found empirical evidence that there is a negative relationship between the use of accrual based earnings management by management of firms who have an incentive to manage earnings and the presence of a high-quality auditor. So this suggests that auditors of high quality constrain managers of firms in the use of accrual based earnings management. In addition I find evidence that management of firms which are operating in a high-litigation risk industry are more constrained in the use of discretionary accruals by the presence of a high-quality auditor. Furthermore I found weak evidence that management of firms use more real earnings management when their firm is audited by a high-quality auditor. Lastly I found mixed evidence regarding the use of real earnings management by managers of firms that operate in a high-litigation risk industry and are audited by a high-quality auditor.
To summarize and to answer the research question: The presence of a high-quality auditor in a high-litigation risk industry has an influence on the earnings management choice that managers make. It is safe to say that managers make less use of accrual based earnings management, and thus that the presence of a high-quality auditor (in a high-litigation risk industry) constrains the use of accrual based earnings management. The weak evidence on the use of real earnings management when a high-quality auditor is present suggests that firms that employ a high-quality auditor resort to real earnings management, but this is just a weak association. So the presence of a high-quality auditor in a high-litigation risk industry results in less use of accrual based earnings management, but the evidence is not strong enough to conclude that these firms resort to the use of real earnings management.
The weak associations found regarding real earnings management are a limitation of my study. A recommendation for future research is to investigate the use of real earnings management more extensively. It is still especially interesting to investigate the tradeoff decision between accrual based earnings management and real earnings management in a more general setting. Possibly future research can focus more on specific industries where the tradeoff decision is more likely to happen. For example in industries where firms rely heavily on R&D investments, so where managers can cut discretionary expenses easily.
