The effect of audit quality on the relationship between short-horizon CEOs and accrual-based earnings management

(1)

The effect of audit quality on the relationship between

short-horizon CEOs and accrual-based earnings management.

Name: Maaike Kistemaker Student number: 10398910 Thesis supervisor: dr. P. Kroos Date: 23 June 2017

Word count: 12.018

MSc Accountancy & Control, specialization Accountancy Amsterdam Business School

(2)

Statement of Originality

This document is written by student Maaike Kistemaker 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

The purpose of this research is to examine the potential relationship between short-horizon CEOs and income-increasing accrual-based earnings management. Based on prior research, I predict that short-horizon CEOs use more earnings management for their personal benefit. Furthermore, this is the first research that investigates the moderating effect of audit quality on this relationship. The monitoring role of high-quality auditors is expected to mitigate the effect of the main relationship. Audit quality is measured as Big Four membership and industry specialism. I also use two methods to measure earnings management: the modified Jones model and performance-matched accruals method. (Kothari et al., 2005).

Contrary to my predictions, I find no clear relationship between short-horizon CEOs and accrual-based earnings management. Also, the mitigating effect of audit quality on this relationship cannot be proved. Nevertheless, there is some indication of an opposite effect of audit quality: short-horizon CEOs use more earnings management when audit quality provided is high. Also, I find prove of a direct negative influence of audit quality on earnings management.

(4)

Contents

1 Introduction ... 6

1.1 Background ... 6

1.2 Research question ... 7

1.3 Relevance of the research question ... 8

1.4 Structure of the thesis ... 9

2 Literature review ... 10

2.1 Agency problem ... 10

2.2 The horizon problem ... 11

2.3 Income-increasing earnings management ... 13

2.4 Auditor quality ... 16 3 Methodology ... 20 3.1 Sample ... 20 3.2 Empirical model ... 21 3.2.1 Auxiliary model ... 22 3.3 Control variables ... 23 4 Results... 25 4.1 Descriptive statistics ... 25

4.2 Results of hypothesis testing ... 28

4.3 Supplemental analysis ... 31 5 Conclusion ... 34 5.1 Summary ... 34 5.2 Conclusion ... 35 5.3 Limitations ... 35 References ... 37

(5)

List of tables

Table 1: Sample………..20

Table 2: Descriptive statistics……….25

Table 3: Descriptive statistics: comparison short horizon vs. non-short horizon………26

Table 4: Pearson correlations……….27

Table 5: OLS regression……….29

Table 6: Interaction effect (Big Four as moderator)………....30

Table 7: Interaction effect (Industry specialism as moderator)………....31

Table 8: OLS regression additional analysis………32

(6)

1 Introduction

1.1 Background

Firms are characterised by separation of ownership and control. Shareholders, as owners of the firm, have different interests than managers, who have control over day-to-day activities. An agency relationship is a contract under which the principal engages the agent to perform some action on his behalf. Some of the decision-making authority is delegated to the agent (Jensen and Meckling, 1976). In the case of a firm, shareholders hire managers to take most of the routine decisions. Assuming that both parties want to maximize their own utility, there is a reason to believe that the agent will not always act in the best interests of the principal (Jensen and Meckling, 1976). Shareholders can take actions to minimize this chance of misalignment, by for instance adjusting incentive plans of managers to more share-based compensation plans. Ownership of equity reduces the agency problem, because the manager’s actions will be reflected in share price, influencing his wealth both in positive and negative ways (Gopalan et al., 2014; McClelland et al., 2012). Taking self-interested actions can reduce share price and has therefore a direct impact on the wealth of the CEO.

However, incentive plans can also trigger the agency problem in itself. An example of this is the horizon problem. A manager facing a short horizon has declining incentives to focus on long-term performance, as he will not be present to claim the proceeds of these benefits in more distant future. Dechow and Sloan (1991) find that CEOs with a short horizon devote less resources to research and development. Investments in R&D mainly exist of future pay-offs, which demonstrates the mechanism of the horizon problem. Besides such real earnings management activities, managers may also engage in accrual-based earnings management. Real earnings management and accrual-based earnings management are different methods for earnings management, but with the same end: adjusting financial reports to mislead stakeholders about the underlying economic performance of the firm, or to influence outcomes that depend on reported accounting results (Healy and Wahlen, 1999). Managers can use accruals to move income across different time periods (Gopalan et al., 2014), for example to achieve their bonus targets. Short-term incentives of CEOs increase the extent to which income-increasing accruals are being used and bonus plans create incentives for managers to select accounting procedures and accruals which maximize the value of their rewards (Healy, 1985).

(7)

Little research up to now examined whether short-horizon employees also use accrual-based earnings management to boost short-term performance. When a CEO knows he will leave the firm, he will have less incentives to act in accordance with the interest of the firm than a manager who has no intention to leave (Kalyta, 2009). McClelland et al. (2012) show that short-horizon CEOs are more likely to use risk-averse strategies when pay-off is after their departure. Also, managers with short-term incentive contracts have a stronger tendency to boost earnings through a higher level of income-increasing accruals (Gopalan et al., 2014). However, firms may anticipate an imminent CEO departure and incentivize short-term CEOs in a way that makes them focus on long-term activities (Cheng, 2004). The focus on accrual-based earnings management instead of real earnings management derives from the property that accruals typically reverse over time. This means that total accruals add up to zero in the long run, so larger accruals in the current period indicate a lower level of accruals in the next period (Gopalan et al., 2014). Short-horizon managers will not be present to experience the adverse consequences of the future accruals reverse.

It should be emphasized that income plans can only trigger adverse side-effects when managers feel they both have incentives and sufficient opportunity to manage earnings. One important mechanism which constrains opportunity is the assurance provided by auditors (Becker et al., 1998). A high-quality auditor is defined as an auditor who delivers credible earnings report. Also, audit quality is determined in its effectiveness in constraining the management of earnings. High audit quality can have a disciplining effect on managers, because the chance that misstatements are discovered is higher (DeAngelo, 1981).

1.2 Research question

This research tests if the relationship between short-horizon CEOs and income-increasing earnings management is influenced by the quality of the auditor. That is, this paper first examines the relation between short-term focused CEOs and income-increasing accrual-based earnings management. Subsequently, this research investigates whether high audit quality negatively moderates the supposed positive relationship between short-term CEOs and income-increasing accrual-based earnings management.

(8)

Audit quality

Short-horizon CEO

Income-increasing accrual-based earnings management Therefore, the research question will be:

Does audit quality have a mitigating effect on the relationship between short-horizon CEOs and income-increasing accrual-based earnings management?

1.3 Relevance of the research question

This research contributes in three different ways to the existing knowledge. First, the relationship between short-term employees and earnings management is re-examined. The prior research on the behaviour of CEOs in their last years of office regarding earnings management provides conflicting findings (Dechow and Sloan, 1991; Gopalan et al., 2014; Kalyta, 2009; Murphy and Zimmerman, 1993; Pourciau, 1993). Pourciau (1993) researches earnings management in situations of routine vs. non-routine executive changes and asks for different classifications of executive changes to provide more insights. She also claims that the causal link between executive changes and discretionary accruals should be investigated while controlling for firm performance. Murphy and Zimmerman (1993) demonstrate that leaving CEOs invest more in their last year of office, in contrast to some prior findings. A potential reason for the ambiguity in the prior research may be that omitted factors (such as audit quality) explain some part of the variation. This lack of consensus asks for more research, which can provide clarification on the topic.

Secondly, most prior research regarding the horizon problem focuses on real earnings management (Cheng, 2004; Dechow and Sloan, 1991; Gibbons and Murphy, 1992a; Matta and Beamish, 2008; McClelland et al., 2012; Murphy and Zimmerman, 1993). Real earnings management has a direct influence on cash flows, by for instance adjusting investments in R&D, PPE and advertising. Accrual-based earnings management does not have a direct impact on cash

(9)

flows, it only transfers earnings between periods by the use of accruals. The focus in this study will be on accrual-based earnings management given that managers may first want to resort to the less costly accrual-based earnings management. When looking at the earnings level over longer windows of time, accruals don’t have an impact, because of their reversal. Real earnings management on the other hand will have an actual impact on firm performance, as investments executed in later periods, or not at all, can have a different yield, due to changed market circumstances and actions of competitors.

Lastly, I will also contribute by specifically looking at the impact of a potential moderating factor: audit quality. Auditors provide value by minimizing the chance of misreporting (Becker et al., 1998), and have a monitoring role in inspecting the data of management before making it publicly available. Audit quality as a monitoring role could influence the tendency of managers to use income increasing earnings management, while auditors are aware of management’s incentives to overstate earnings (Hirst, 1994). Prior research examined the moderating effect of audit quality in general, therefore I will focus on the effect of audit quality in the relationship between short-horizon CEOs and accrual-based earnings management.

1.4 Structure of the thesis

The rest of this paper will be structured in as follows: Section two discusses prior literature and the development of the hypotheses. In section three, the research methodology will be discussed. The results will be presented in chapter four. Finally, chapter five will provide the conclusion and the limitations of the research.

(10)

2 Literature review

2.1 Agency problem

Separation of ownership and control is common in public firms. Ownership of the firm (typically in the hands of shareholders) is separated from control (management). Smaller companies usually have one person fulfilling the role as both owner and manager. However, when companies become larger, more outside capital might be needed, which can be raised through the issuance of shares and disperses ownership. Investors mostly use diversification to rule out firm-specific risks. With diversification, investors invest their money in several companies, because a portfolio consisting of different kind of investments will on average produce higher returns and lower risks than an individual investment. Also, managing a firm requires certain capabilities (for instance regarding leadership) which are not always possessed by owners of a firm. Specialized knowledge is a large benefit of separating ownership and control.

Separation of ownership and control creates an agency relationship between the principal (the shareholder) and the agent (the manager), because the principal engages the agent to perform an action of his behalf. This means that some of the decision-making authority is delegated to the agent (Jensen and Meckling, 1976). Agency theory has three assumptions: humans are self-interested, have bounded rationality and are prone to opportunism. The interests of the agent and principal are not the same and therefore self-interested behavior of the agent is expected (Jensen and Meckling, 1976). Suboptimal decisions taken due to misalignment of interests represent the costs of separating ownership and control.

Controlling the agency problem is important when decision makers (who initiate and implement important decisions) do not bear a major share of the wealth effect of their decisions (Fama and Jensen, 1983, McClelland et al., 2012). In other words, when managers do not have any ownership of equity, their actions (which have an impact on the share price) will not affect their private prosperity. This increases the chance of consuming perks and focusing on private benefits when spending firm’s resources. Therefore, it is important to have effective control procedures preventing this from happening. Both the agent and principal can take measures to reduce the agency problem, by devoting resources to monitoring and bonding costs.

The principal can pay bonding costs to the agent to guarantee that the agent will not take actions which will harm him, or to ensure that he will be compensated if the agent does take harmful actions (Jensen and Meckling, 1976). A common way to realize this is to adjust the agent’s compensation plan, by for example including shares in the plan. Monitoring costs are incurred to

(11)

monitor and restrict the actions of agents. This can include auditor costs and the costs of having an (independent) board of directors. Both are expected to have a monitoring effect on the behavior of the manager and prevent him from taking harmful actions. Despite these expenditures, there will always exist some divergence between the decisions of the agent and the decisions which would maximize the wealth of the principal (Jensen and Meckling, 1976). That is, it is in most cases not cost-effective to perfectly align the interests of agents with the principal’s interests, typically referred to as residual loss.

Both auditors and performance-based income plans are expected to mitigate the effect of the agency problem. On the other hand, while performance-based income plans are intended to align the interests of the principal and agent by motivating productive effort from the agent, at the same time those incentive plans can also be the reason for the agent to take actions that are not in the best interest for the principal. For example, agency theory states that the shorter a relationship between an agent and a principal is, the less efficient it is (Matta and Beamish, 2008), because a short-term relationship increases risks for the agent that external factors will cause fluctuations in firm performance. This creates potential problems for situations where a CEO knows he will leave the company on the short-run.

2.2 The horizon problem

The majority of the literature on CEO compensation argues that the extent to which the compensation of a CEO is short-term vs. long-term will influence the decisions made by the CEO (Gopalan et al., 2014). The horizon problem exists when a CEO is leaving the company in the short term and when he is aware of that. When performance plans are based on current performance, they will only create investment decisions in line with the firm’s best interest when the manager expects to be employed by the company during the entire period of the project (Gibbons and Murphy, 1992a). When a CEO is planning to leave, considerations near the moment of departure will let a CEO focus on the short-term implications of his strategic actions rather than long-term considerations of company growth (Gibbons and Murphy, 1992b), because he will not be able to benefit from long-term growth of the firm.

The horizon problem can have an influence on firms in multiple ways, for example, in triggering earnings management, accounting fraud and suboptimal investment decisions (Kalyta, 2009). It can create situations where a CEO does not act in the interest with the firm’s goals. Prior research has investigated the impact of the horizon problem regarding different factors, for instance investments, mergers and firm performance. Dechow and Sloan (1991) were the first to write about this topic. They discuss the horizon problem by looking at the pattern of R&D

(12)

expenditures of CEOs in their last years of office and find CEOs spend less on R&D in their last years. CEOs also devote less resources to other types of risky investments (Bryan et al., 2000). Matta and Beamish (2008) research international mergers and acquisitions, examples of risky investments, in relation to the time horizon of a CEO. They emphasize that international acquisitions are costly, increase uncertainty about the firm’s short-term income stream and can jeopardize immediate market returns. International acquisitions affect the short-term income of the firm in different ways: through the costs involved in the process of searching an appropriate international target, the problematic process of valuating targets given uncertainty in international contexts, and the information asymmetry between the acquiring firm and the foreign target. These extra costs and uncertainties about future income streams of the firm concern managers that stock prices will fluctuate on the short term, which will impact their wealth (Matta and Beamish, 2008). In line with Bryan et al. (2000) they find that CEOs with a shorter horizon have a lower likelihood of engaging in international acquisitions. A study by McClelland et al. (2012) investigates the effect of the managerial horizon on the overall firm performance and concludes that CEOs with a short horizon are performing worse compared to the ones with long horizons. This is caused by the risk-averseness of CEOs. The higher the percentage of ownership, the more risk-averse CEOs get regarding short-term uncontrollable factors, given that they will not be able to make up for these in the longer term. The risk-aversion may affect their decisions which can have a negative impact on firm performance. Gibbons and Murphy (1992a) find different results. They discover that CEOs invest more in R&D and advertising during their last years. A possible explanation for this unexpected result might be that CEOs do not change the level of their investments, but the type, by for instance investing in projects with immediate pay-offs. Another reason might be the influence of the other decision makers. CEOs are mostly surrounded by a management team who is also involved in the decision-making process and makes it impossible for a CEO to make decisions completely autonomous. Finally, organizational factors such as corporate governance or audit quality which may constrain the horizon problem may not adequately be taken into account. There are also ways to mitigate the effects caused by the horizon problem. The wealth of the CEO can be made dependent on firm value under the assumption that stock price immediately incorporates the value consequences of current actions. Normally, the problem CEOs tend to focus more on the short term is often ‘solved’ by providing stock options and equity as part of the CEO compensation. In this way, the decision horizon of the agent is extended (Dechow and Sloan, 1991, Dikolli, 2001). Investments reduce less in the period before the CEO leaves when he has more stock-based incentives (Gibbons and Murphy, 1992a). Also, in case of a retiring

(13)

CEO, pension plans based on stock incentives generate more motivation to focus on the long term (Kalyta, 2009). These methods are most helpful when the wealth of the CEO is directly connected to stock price, because investment decisions are assumed to be directly reflected in stock price. Long-term payoffs from options and stock are less relevant when a CEO plans to leave a firm soon (Matta and Beamish, 2008) given that the CEO may have to forfeit their equity portfolio when he is bound to leave. Another way to reduce the effect of the horizon problem is to use the relay method of CEO succession (Dechow and Sloan, 1991). This means the CEO is more involved in the transition process and working together with his successor. Also, compensation committees can be introduced. CEOs don’t reduce investments in R&D when their compensation structure is adjusted in the right way by such a committee (Cheng, 2004). That is, besides rewarding a CEO through stock price, another option is to, for example, adjust the CEO’s personal bonus plan for the negative effect of investments when he has a short-expected employment horizon. The assumption here is that firms are aware of the short horizon and adjust the bonus plans of the CEOs accordingly.

2.3 Income-increasing earnings management

Managers can use judgment in several ways when making accounting choices. They can for instance use different depreciation methods and inventory valuation methods. Besides using judgment in accounting decisions, managers also have to exercise judgment in day-to-day business decisions. For example, they can use judgment when altering the level of R&D expenses, advertising costs and other costs. These types of decisions influence a firm’s financial performance (Kalyta, 2009). Earnings management occurs when managers use judgment in financial reporting and are structuring transactions to adjust financial reports with the intention to mislead stakeholders about the underlying economic performance of the firm or to influence outcomes that depend on reported accounting results (Healy and Wahlen, 1999; Leuz et al., 2003). Incentives for earnings management arise from the conflict of interest between the firms’ insiders (managers) and outsiders (stakeholders, board of directors). Insiders can benefit personally at the expense of other stakeholders (Leuz et al., 2003).

There are two main types of earnings management: real earnings management and accrual-based earnings management. Accrual-based earnings management does not have a direct impact on cash flows, while real activities manipulation has (Cohen and Zarowin, 2010). Real earnings management is defined as management actions that deviate from normal business practices, with the objective of meeting certain earnings targets (Roychowdhury 2006). This type of earnings management has an influence by changing the real activity level of the company, instead of only

(14)

adjusting the accounting figure and this can reduce firm value, because actions taken (or disregarded) in the current period to increase earnings can possibly have a negative effect on future cash flows.

Roychowdhury (2006) distinguishes three different types of real earnings management. The first one is sales manipulation, which includes accelerating the timing of sales or generating additional unsustainable sales through for example extra price discounts. When aggressive discounts have been approved earlier to reach a target, customers might expect the same discounts in the future, which has a negative effect on margins. Secondly, by reducing the amount of discretionary expenditures (costs for R&D, advertising, maintenance) firms can increase their earnings. This is most likely to occur when expenditures do not generate quick income. Lastly, overproduction can be used to manipulate earnings. With higher production levels, total costs per unit decline, because fixed costs can be spread out over more units. In his research, Roychowdhury (2006) finds that these types of real earnings management are used to avoid reporting annual losses. Accruals encompass the part of earnings which are not reflected in current cash flows. It is the difference between the reported amount of income and costs, and the moment the actual payment is received or paid. The use of accruals temporarily shifts some of the firm’s earnings between the future and the present (Gopalan et al., 2014). Investors ask for information on periodic basis and in this way financial statements are comparable. Accruals create the possibility for management to influence the timing of transactions (where possible). Also, accruals have the property of reversal, which means that over time the total balance of accruals adds up to zero (Gopalan et al., 2014). A higher amount of income-increasing accruals in the current period typically implies income-decreasing accruals for future periods.

Managers have a certain extent of discretion towards the choices they make in financial reporting. They can exploit this discretion by communicating private information to the market and therefore producing reported accounting performance which is a better representation of the true underlying economic performance. On the other hand, they can use this discretion opportunistically to serve their own interests by enhancing their own wealth, which is at cost of the welfare of the company (Christie and Zimmerman, 1994; Pourciau, 1993). Performance measures are possibly an important motivation, because using accruals to shift earnings across periods can help achieving targets. Therefore, accrual policies of managers are related to the income-reporting incentives of their bonus contracts (Healy, 1985). Bonus plans create incentives for managers to select accounting procedures and accruals to maximize the value of their rewards. Other reasons to use accrual-based earnings management are reporting incentives:

(15)

CEOs want to smoothen income and prefer showing a profit instead of a loss. Smoothening income means that variances of reported earnings between different periods are minimized (Leuz et al., 2003; McNichols and Wilson, 1988). This can be achieved through both income-increasing and decreasing earnings management. McNichols and Wilson (1988) find that provisions for bad debt are used to smoothen earnings. Stakeholders prefer a profit over a loss and therefore accruals are also used to avoid reporting a loss (Dechow et al., 2003). By managing earnings upwards, firms move from reporting a small loss to a small profit. Dechow et al. (2003) find evidence in line with this: small profit firms use more income-increasing earnings management relative to other firms. This suggests that these firms are making an effort to increase earnings for the current year. Rangan (1998) finds that positive abnormal accruals are used in the period before seasoned equity offerings, which implies that these accruals reverse in the period afterwards. In this way, the stock market is misled by the upwardly managed earnings, because they overvalue the firm issuing shares. After that, there is a big chance of disappointment by the declines in predicted earnings. This, however, does not imply that managers generally abuse the accounting discretion to serve their own purposes by inflating their compensation. Financial reporting can also be used to better inform investors about the future performance of the firm (Dechow et al., 2014; Gul et al., 2003). Managers can use their discretion over accounting choices as a signal of their inside information when they perceive current income as an invalid reflection of future income. Signaling can also be used to boost reputation. A CEO can ‘inform’ a potential next employer of his capacities by showing high profits, or making it more difficult for a successor to meet his performance (Pourciau, 1993).

Other research focuses on earnings management in short-horizon situations. CEO pay that is more focused on the short term causes CEOs to use more income-increasing accruals (Gopalan et al., 2014). In line with this, Healy (1985) investigates the effect of different bonus schemes on accounting decisions made. He finds a strong correlation between the use of accruals and the incentives managers have to boost earnings. When their bonus is based on the financial performance of a firm, and has no upper bound, managers are more likely to use income-increasing accruals. This relationship is found in many other studies concerning earnings management, particularly in studies about short-horizon contracts. Managers will for example ignore R&D expenditures because this will have a negative impact on their short-term income (Watts and Zimmerman, 1978). Kalyta (2009) focuses specifically on retiring managers and the choices made in pre-retirement period. He investigates the impact of the type of pension plan on the use of income-increasing earnings management and finds that it happens only when the pension plan is based on the firm performance in the years before retirement, so when the

(16)

manager can benefit from a better firm performance by an increase in personal wealth. Retiring managers have less concerns regarding their reputation (Gibbons and Murphy, 1992b), which makes it more probable that they engage in earnings management. However, for regular CEO changes (non-retirement), evidence on earnings management is mixed, as not all research finds evidence of income-increasing earnings management prior to a CEO change (Murphy and Zimmerman, 1993; Pourciau, 1993, Wells, 2002). In the year prior to a CEO departure, Murphy and Zimmerman (1993) find no evidence of income-increasing earnings management. They argue that declines in growth rate of R&D, advertising and capital expenditures are caused by poor firm performance, which indirectly relates to the chance a manager leaves (Murphy and Zimmerman, 1993) or an increase in monitoring activities due to lower firm performance (Pourciau, 1993). The outcomes were not in line with expectations, as existing literature discussed the incentives for earnings management and concludes that agents, when unrestricted in their behaviour, will take actions to increase their personal wealth as much as possible (Jensen and Meckling, 1976). When CEOs only have a short period ahead to influence their wealth, earnings management of short-term firm performance is the way in which they can use their discretion to create personal benefits (Dechow and Sloan, 1991; Healy, 1985; Watts and Zimmerman, 1978). Accrual-based earnings management does not influence the firm’s activities and the reverse effect is experienced in a later period when the short-horizon CEO has left the firm. Due to this characteristic, CEOs can manage earnings in a way which suits them best and this makes accruals an attractive way of manipulating earnings (Gopalan et al., 2014).

From the prior theory, the first hypothesis can be derived:

Hypothesis 1: CEOs with a short horizon use more income-increasing accrual-based earnings management. 2.4 Auditor quality

Managers want to control their benefits and can use earnings management to conceal firm performance from outsiders. However, there are different methods to mitigate the opportunity for earnings management, for instance a strong internal governance structure, potential high personal costs of revealed earnings management and previously taken accounting decisions, which can limit future choices (Becker et al., 1998). Increased investor protection can limit managers to acquire private benefits (Leuz et al., 2003). Also, when a CEO succession process is implemented, opportunities are constrained for both leaving and incoming managers to undertake earnings management (Dechow and Sloan, 1991; Wells, 2002). Appointing an auditor is another form of monitoring. Auditing is used by firms to reduce agency costs with stakeholders (Jensen and Meckling, 1976) by mitigating the conflicts of interest between owners

(17)

and managers (DeAngelo, 1981). When managers produce financial statements which are verified by an outside party, an auditor reduces the information asymmetry between management and stakeholders (Becker et al., 1998) and the positive bias in management’s valuation of earnings and assets (Kinney and Martin, 1994). Auditors take into account the potential opportunities for earnings management and concentrate on managerial incentives to manipulate earnings (Hirst, 1994). Therefore, auditors are an effective measure to constrain the opportunities for earnings management.

Audit quality is defined as the joint probability of detecting and reporting financial statement errors (DeAngelo, 1981). High-quality auditors can have a disciplining effect on managers, because they are more likely to detect questionable accounting practices, to object to their use and qualify the audit report (Becker et al., 1998). When an auditor detects the need for a restatement, he may force the manager to publish a financial report including the restated numbers, because this reduces the incidence of restatements to be made afterwards. Francis et al. (2013) find that larger Big Four offices (used as a proxy for high audit quality) are associated with fewer client restatements, after controlling for client characteristics. A client restatement provides direct evidence of a low-quality audit, because it indicates that the auditor did not effectively apply the rules and regulations at the time of the audit. This is also the case when industry specialists are used as proxy for high audit quality: the incidence of misstatements is lower for their clients and therefore, these auditors provide higher quality services (Chin and Chi, 2009). A high-quality auditor can also be an effective prevention for earnings management, because management’s reputation is likely to be damaged and firm value to be reduced if misreporting is detected (Becker et al., 1998). If an auditor is known for its high quality, management will anticipate its behavior on this given fact, because chances that earnings management will go unnoticed or will not be disclosed when noticed are reduced. Research concludes that high-quality auditors have a monitoring effect and earnings management will occur less.

There is a debate on how you can measure if an auditor provides high quality. All stakeholders in the financial reporting process may have a different view as to what constitutes audit quality. The user of the financial report may think that high audit quality means the absence of material misstatements, while regulators can view a high-quality audit as one which complies with all professional standards (Knechel et al., 2012). As a basis, every auditor needs to meet certain conditions, on for instance technological and procedural level (DeAngelo, 1981). There is much evidence that larger accounting firms provide higher quality audits. Big Four firms are sued and sanctioned relatively less often by the Security and Exchange Commission, which could stand for

(18)

higher quality and less mistakes. This can be countered by the fact that Big Four firms have more resources to go against lawsuits (Francis, 2004). DeAngelo (1981) agrees that larger firms provide higher audit quality. The reason provided is that auditors with a larger number of clients have more to lose when they are getting caught on not following the rules and regulations. Also, they can lose their client specific quasi-rents and therefore they have more incentives to provide high quality audits.

Moreover, the independence of the auditor can be a sign of audit quality (DeAngelo, 1981). Independence can be influenced by the relative size of a client. When one or a few large clients supply a significant portion of total revenues, the firm will have greater difficulty in maintaining its independence (DeAngelo, 1981). Still, Big Five auditors do not necessarily report more favorably for larger clients (in relation to the total size of the office), but do report more conservatively for larger clients. This is not caused by dependency, but by a higher risk of litigation (Reynolds and Francis, 2000). However, Chung and Kallapur (2003) do not find an association between audit quality and the relative client importance. Lastly, specific knowledge seems to be an indicator for quality, while firms with industry specialized auditors provide on average higher audit quality (Balsam et al., 2003).

Researchers are pointing in the same direction when discussing the effects of audit quality. Higher audit quality reduces the agency problem, because in this way the information asymmetry between management and stakeholders decreases. Both through disciplining managers and the threat of a damaged reputation, high audit quality decreases the chance earnings management is used (Becker et al., 1998). Audit quality also has an effect on the perception of outside parties. A change in auditor quality can evoke reactions from stakeholders. When a firm decides to switch from a Big Four (industry specialist) auditor to a non-Big Four auditor, firms suffer a large negative market reaction (Knechel et al., 2007). Moreover, the market reacts positively from a switch from a non-Big Four auditor to a Big Four auditor. This suggests that the market reacts upon the quality change in auditor. This mechanism also works the other way around. If a Big Six auditor decides to stop auditing a client and resigns, this is received by the market as “bad news”, because in most cases there is no reason disclosed for the auditor’s resignation (Wells and Loudder, 1997). However, the resignation of a Big Six auditor is associated with a greater negative market reaction than the resignation of a non-Big Six auditor (Dunn et al., 1999). This indicates that the loss of a Big Six auditor has a greater impact on the credibility of a firm than the loss of another auditor.

(19)

A high-quality auditor is expected to have more severe impact in restricting management’s actions. The information received from management is inspected more thorough and therefore the expectation is that short-horizon CEOs are less involved in income-increasing earnings management when audit quality is high. Based on existing literature, the second hypothesis can be retrieved:

Hypothesis 2: High audit quality mitigates the positive relationship between short-horizon CEOs and accrual-based earnings management.

(20)

3 Methodology

3.1 Sample

I start with all firms from the Compustat North America database which roughly corresponds with the S&P 1500. Because most data on CEOs became available after implementation of SOX, my data will cover the period from 2004 to 2015. Data on CEOs is retrieved from Execucomp. The data is merged and only the observations concerning CEOs are taken into account. This yields an initial sample of 23.342 firm-year observations, with CEO identifying information present. Missing financial data needed for the calculation of accruals leads to the deletion of 6.678 observations. Because of non-available data on auditors, 1 observation is deleted. Missing data concerning control variables leads to the removal of another 2.142 observations. Based on prior literature (Becker et al., 1998; Kalyta, 2009) firms in the financial industry sector (SIC codes 6000-6999) are deleted from the sample, due to the nature of their operations and the inapplicability of the (modified Jones) model used to estimate expected accruals (Wells, 2002). Therefore, 427 observations are deleted. Lastly, I exclude all firm-year observations where there are fewer than ten observations in any two-digits SIC industry in any year. These observations are excluded because regression-model-based accrual estimates are likely to be imprecise (Kothari et al., 2005). Consequently, 1.358 observations are removed. The final sample consists of 12.736 firm-year observations. Consistent with prior research (Kothari et al., 2005) I winsorize extreme observations of all variables by setting values in the bottom and top one percent equal to the values of the 1st_{and 99}th_percentiles.

Table 1: Sample

Number of observations Initial sample (after merging databases) 23.342

Less: Missing values on financial data 6.678

Less: Missing values on auditor data 1

Less: Missing values on data control variables 2.142

Less: Financial sector observations (SIC 6000-6999) 427

Less: Industries (two-digits SIC codes) with less than 10 observations 1.358

(21)

3.2 Empirical model

To test the first hypothesis the following model will be used: AB-EMit = β0 + β1 ST_CEOit + controls+ ε

Here, accrual-based earnings management denotes signed accrual-based earnings management that is further described in section 3.2.1. For the variable ST_CEO, an indicator variable is used to show whether there is a horizon problem. The variable equals 1 when the year t is the final year prior to a CEO change, and 0 otherwise.1_{Based on my first hypothesis, I expect a positive} relationship between ST_CEO and AB-EM. Therefore, I expect β1 > 0.

To test the second hypothesis the following model will be used:

AB-EM = β0 + β1 ST_CEOit + β2 AQit + β3 ST_CEOit * AQit + controls + ε

The basic model is extended by including auditor quality as moderating variable. A higher audit quality is defined as greater assurance that the financial statements faithfully reflect the firm’s underlying economics (DeFond and Zhang, 2014). Literature uses a lot of different proxies to measure audit quality and there is no consensus on which measure is the best (DeFond and Zhang, 2014) Therefore, in this research AQ is measured in two ways: an indicator variable is used to specify whether an auditor belongs to one of the “Big Four” auditors or not, where a Big Four membership reflects high quality (Becker et al., 1998). Big Four auditors are considered to be more independent, because they have a large client portfolio and are therefore less dependent on the revenues of one client (DeAngelo, 1981). More independence reflects higher audit quality. When a company is audited by a Big Four auditor, the indicator variable for audit quality will be 1, and 0 otherwise.

The second proxy for a high-quality auditor is the extent to which an auditor is a specialist within the industry.2_{Industry-specialized auditors are expected to deliver higher quality, because they} have more expertise in a specific field (Balsam et al., 2003; DeFond and Zhang, 2014). For each year separately, industry specialization is measured using the market share approach, based on total sales. Industries are based on two-digits SIC codes. Market share is calculated by computing the total sales per industry. After that, all sales within an industry are assigned to an auditor. By calculating which audit firm has clients with the highest sum of sales within an industry, the

1_{My assumption here is that people at that period know that their horizon is short. This is an assumption, because}

there might be instances where a CEO is dismissed at the end of the year, while at the beginning of his final year he was not knowledgeable that he would be dismissed at the end of the year. This however, biases against my hypothesis.

2_{In the United States, there is no mandatory audit firm rotation in the time of my sample (Cameran et al., 2015).}

This means that auditing firms can still be industry specialists, which would be difficult to observe when auditing firms would have to rotate mandatory.

(22)

industry specialist per year is determined (Balsam et al., 2003). For firms audited by an industry specialist, the indicator variable is 1 and for all other firms, the variable is 0.

The interpretation of the coefficients is as follows. When audit quality is low, the indicator variable for audit quality becomes 0 and the relationship between ST_CEO and AB-EM will be represented by β1. When audit quality is high, the relationship between ST_CEO and AB-EM will be equal to β1+ β3. My expectation is that the positive relationship between ST_CEO and AB-EM is weaker when AQ is high, because high audit quality will constrain the opportunities of earnings management when CEOs know they will leave soon. Therefore, β1+ β3 < β1 and this means that for the second hypothesis, I expect that β3 < 0.

3.2.1 Auxiliary model

In this research, the focus will be on signed accrual-based earnings management. Earnings management can be used to decrease or increase income. I will only be looking at income-increasing earnings management,3_{so I will not look at the absolute value of earnings} management, but at the signed value. To measure earnings management, the level of discretionary accruals is used. McNichols and Wilson (1988) describe a framework for classifying accounting accruals into non-discretionary and discretionary accruals.

TA = NA + DA Where:

TA = total accruals

NA = non-discretionary accruals DA = discretionary accruals

The nondiscretionary part reflects accruals used for the specific business conditions, such as growth and operating cycle. The discretionary component reflects the choices made by management. Because the level of discretionary accruals is unobservable, previous researchers have used different proxies. Since it is a noisy proxy, I will use two different methods to measure the level of discretionary accruals. The first proxy is the modified Jones model (Kothari et al., 2005). This model is commonly used to estimate the level of discretionary accruals.

3_{My assumption is that the CEOs are rewarded based on their financial performance. This means that influencing}

(23)

To estimate discretionary accruals, I use the following formula: TAit = α0 + α1 (1/ASSETSit-1) + α2 (∆REV- ∆REC)it + α3PPEit + εit

(Kothari et et al., 2005, p. 173) Where, for fiscal year t and firm i : TA it = total accruals

ASSETS_it,-1 = total assets at year t-1

REV it = revenue (scaled by lagged total assets)

REC it = receivables (scaled by lagged total assets)

PPE it = Property, plant and equipment (scaled by lagged total assets)

ε = residuals

The residuals in the model are used as the measure of discretionary accruals.

An alternative measurement of accrual-based earnings management is the method of Kothari et al. (2005) using performance-matched discretionary accruals. With this method, firms are matched based on their performance (ROA) and industry (two-digits SIC code) within the same year. For every observation, the match is made with an observation which has the closest ROA (within the same two-digits SIC code and year). Controlled for the effect of performance on discretionary accruals, a firm’s estimated amount of discretionary accrual is adjusted by subtracting the corresponding discretionary accrual of the matched firm (Kothari et al., 2005). I define the performance-matched discretionary accrual for firm i in year t as the modified Jones-model discretionary accrual (Dechow et al., 1995) in year t minus the matched firm’s modified Jones model discretionary accrual for year t. (Kothari et al., 2005). ROA is computed by dividing the net income by the total assets.

3.3 Control variables

I include several control variables in my study. First, performance and growth opportunity are argued to affect discretionary accounting choices (Kalyta, 2009). Therefore, I control for change in market value (PERF) and the book-to-market ratio (BTM) in my analysis. Second, firm size, measured as the logarithm of total assets (LSIZE) and operating cash flows (OCF) are being controlled for (Becker et al., 1998; Kalyta, 2009). Thirdly, managers of highly leveraged firms may have incentives to use income increasing accruals (to avoid violations of debt covenants) and therefore leverage (LEV) is introduced as control variable, measured by the debt-assets ratio (Becker et al., 1998; Kalyta, 2009). Accrual management can be used to smoothen earnings and

(24)

decrease volatility in earnings (Leuz et al., 2003; McNichols and Wilson, 1988). Consequently, volatility measured by the standard deviations of ROA over the previous five years is added as a control variable (VOL). Moreover, accrual policies can be influenced by loss (avoidance) (Burgstahler and Dichev, 1997). I create an indicator variable to control for losses (LOSS), where I take net income and create an indicator variable equal to one if the net income < 0, zero otherwise. Also, indicator variables are introduced to control for industry-specific effects. Lastly, I control for the effects of time-series trends by including year dummies.

(25)

25

4 Results

4.1 Descriptive statistics

Table 2 provides the descriptive statistics for the dependent and independent variables of interest, as well as the control variables. When accruals are measured using the modified Jones model, DA (MJM) is on average positive (0.0003), as well as the median (0.0045), which is in line with prior research (Kalyta, 2009; Kothari et al., 2005). However, when discretionary accruals are measured using the performance-matched approach, the mean (-0.0014) and median (-0.0009) are both negative. This is similar to the findings of Kothari et al. (2005). From all observations, 9.85% are classified as short horizon (final year before the CEO leaves office). On average, 89.93% of the firms is audited by a Big Four auditor and 29.17% by an industry specialist. The average increase in market value is $0.19 million and 19.34% of the firm-year observations incurred a loss.

Table 2: Descriptive statistics

Variables Obs. Mean St. Dev. 10% 25% 50% 75% 90%

DA (MJM) 12,736 0.0003 0.0656 -0.0691 -0.0266 0.0045 0.0344 0.0664 DA(PM) 12,736 -0.0014 0.0804 -0.0935 -0.0443 -0.0009 0.0422 0.0901 ST_CEO 12,736 0.0985 0.2979 0 0 0 0 0 Big Four 12,736 0.8993 0.3009 0 1 1 1 1 Ind. Spec 12,736 0.2917 0.4546 0 0 0 1 1 PERF 12,736 0.1900 1.7301 -0.3853 -0.1561 0.0683 0.3286 0.6979 BTM 12,736 0.7856 7.251 0.1012 0.2451 0.4603 0.8417 1.4321 LSIZE 12,736 7.4186 1.6313 5.4616 6.2978 7.3301 8.4625 9.6157 OCF 12,736 687.19 1,722.15 8.07 44.50 145.04 479.02 1,551.00 LEV 12,736 0.2237 1.2885 0.0000 0.0159 0.1755 0.3095 0.4533 VOL 12,736 0.0663 0.0917 0.0090 0.0168 0.0333 0.0750 0.1554 LOSS 12,736 0.1934 0.3950 0 0 0 0 1

In table 3, descriptives are compared between the observations with and without short-horizon CEOs. For both the modified Jones model and the method using performance-matched accruals, discretionary accruals are lower when the CEO has a short horizon. Specifically, both DA (MJM) (-0.0076) and DA (PM) (-0.0054) show a negative result for the short horizon CEOs, which indicates that CEOs use income-decreasing discretionary accruals. For non-short horizon CEOs, DA (MJM) is positive (0.0012) and DA (PM) is negative (-0.0010). Both differences in the mean discretionary accruals are significant at a 5% level.

(26)

Observations with short-horizon CEOs more frequently report losses (28.39% vs. 18.35%). This coincides with the intuition that those CEOs are more likely to be replaced in the near future. Surprisingly, PERF (change in market value) is higher for firms with a short-horizon CEO (0.2827) than the control group (0.1799). Furthermore, the volatility in accounting performance is greater for those observations with short-horizon CEOs (0.0725 compared to 0.0656). This finding indicates that the ROA for short-horizon CEOs fluctuates relatively more.

Table 3: Descriptive statistics: comparison short horizon vs. non-short horizon

Short horizon Non-short horizon Significance

Mean Median Mean Median T-test Ranksum

DA (MJM) -0.0076 0.0019 0.0012 0.0048 ** ** DA (PM) -0.0054 -0.0018 -0.0010 -0.0008 ** Big Four 0.9059 1 0.8986 1 ** Ind. Spec 0.2815 0 0.2928 0 ** PERF 0.2827 0 0.1799 0.0077 ** ** BTM 0.9602 0.5423 0.7665 0.4546 ** ** LSIZE 7.41 7.32 7.42 7.33 OCF 713.37 127.19 684.33 147.97 ** LEV 0.2069 0.1705 0.2255 0.1761 VOL 0.0725 0.0375 0.0656 0.0329 ** ** LOSS 0.2839 0 0.1835 0 ** ** ** = 5% significance level

Table 4 provides data on the correlations between different variables. The different measures of discretionary accruals are highly correlated and significant (0.64), while the different measures of audit quality have a lower but also significant correlation (0.21). This suggests that the different proxies to a reasonable extent represent the underlying theoretical construct. As already indicated by table 3, both proxies of accrual-based earnings management are negatively correlated with short-horizon CEOs. When we examine the correlation between accrual-based earnings management and the two proxies for audit quality, the correlations suggest that in three out of the four cases (with one of these significant), higher audit quality is associated with less income- increasing accrual-based earnings management. The correlation table also shows that larger firms more often hire audit firms that are Big-Four auditors and industry specialists.

(27)

Pe ar son c or re lati ons V ar iab les D A (M JM ) D A (P M ) ST_ CE O Bi g Fo ur Ind. _Spec PERF BTM LS IZ E O CF LE V VO L LO SS D A (M JM ) 1 D A (P M ) 0. 64 1 ST_ CE O -0 .0 4 -0 .0 2 1 Big F our -0 .0 1 -0 .0 3 0. 01 1 Ind. S pe c 0. 02 -0 .0 1 -0 .0 1 0. 21 1 PERF 0. 06 0. 02 0. 02 0. 00 -0 .0 1 1 BTM -0 .0 6 0. 05 0. 01 0. 00 0. 00 -0 .0 2 1 LS IZ E 0. 04 -0 .0 4 0. 00 0. 37 0. 19 0. 01 0. 02 1 O CF -0 .0 2 -0 .0 7 0. 01 0. 12 0. 13 -0 .0 2 -0 .0 2 0. 64 1 LE V -0 .0 6 -0 .0 5 -0 .0 1 -0 .0 2 -0 .0 1 0. 00 0. 01 -0 .0 4 0. 00 1 VO L -0 .0 6 0. 02 0. 02 -0 .15 -0 .0 6 0. 06 0. 02 -0 .2 9 -0 .11 0. 08 1 LO SS -0 .3 4 -0 .0 3 0. 08 -0 .11 -0 .0 6 -0 .0 4 0. 09 -0 .2 0 -0 .13 0. 05 0. 02 5 1 Re su lts in b old in dic at e a s ig nif ic an ce le ve l of 5% Re su lts in it alic s in dic at e a s ig nif ic an ce le ve l o f 10% Ta ble 4 : Pe ar son co rre lat ion s

(28)

In general, the data provide no concerns about multicollinearity issues between the control variables. Only the correlation between size and operating cash flows exceeds 0.6. As a robustness analysis, I repeated my main analyses while excluding operating cash flows. My findings remained unaltered.

4.2 Results of hypothesis testing

In the following section, the first hypothesis is tested. Table 5 reports the results of an OLS regression on the dependent, independent and control variables. The results reported in the left column are for discretionary accruals as the dependent variable measured via the modified Jones model. As table 5 shows, the regression coefficient is negative (-0.0027). However, the result is not significantly different from zero. So, against my expectations, I find no significant relationship between the presence of short-horizon CEOs and accrual-based earnings management. The results are similar when the hypothesis is tested using discretionary accruals measured according to the method of performance-matched accruals. This is reported in the right column of table 5. Again here, the coefficient on ST_CEO is negative but insignificant. Thus, based on these findings, I have to reject my first hypothesis. Several control variables are significantly related to the dependent variable. PERF, BTM, LSIZE and VOL have a positive and significant coefficient in at least one of the two models. Becker et al. (1998) find similar results as they also use LSIZE as control variable in their model. In addition, OCF, LEV and LOSS have a significant negative coefficient in both models. OCF is both used by Becker et al. (1998) and McNichols and Wilson (1988) and these papers provide comparable coefficients. However, the negative coefficient of LEV is not in line with prior research (Kalyta, 2009). That is, where higher leverage would imply a higher likelihood of debt covenant violation and therefore a higher likelihood of income-increasing accrual-based earnings management, my findings suggest a greater use of income-decreasing discretionary accruals. The R-squared of both models have low explanatory power (12.9% and 1.29%). Overall, the regression results do not support my hypothesis that short-horizon CEOs make greater use of income-increasing accrual-based earnings management.

(29)

Table 5: OLS regression

Variables Predicted sign DA (MJM) DA (PM)

Intercept β0 0.0129*** 0.0043 (0.0049) (0.0065) ST_CEO β1

+

-0.0027 -0.0032 (0.0018) (0.0024) PERF 0.0018*** 0.0015*** (0.0019) (0.0004) BTM 0.0001* 0.0003** (0.0001) (0.0001) LSIZE 0.0010** 0.0002 (0.0005) (0.0006) OCF -0.0000*** -0.0000*** (0.0000) (0.000) LEV -0.0022*** -0.0032*** (0.0004) (0.0006) VOL 0.0252*** 0.0338*** (0.0066) (0.0086) LOSS -0.0593*** -0.0119*** (0.0015) (0.0020)

Industry effects Yes Yes

Year effects Yes Yes

F-statistics 75.63*** 6.67***

R-squared 0.1290 0.0129

N 12,736 12,736

T-statistics are based on OLS and standard errors are reported in parentheses. ***,**,* corresponds to 1%, 5% and 10% significance levels.

In the following analysis, the moderating effect of audit quality will be discussed. Table 6 reports the results of including the moderator variable audit quality (AQ), measured as Big Four membership. When audit quality is low, the relationship between ST_CEO and AB-EM is represented by β1. As table 6 shows, the regression coefficient β1 is positive when discretionary accruals are measured via the modified Jones model (0.0036), while the coefficient is negative when the performance-matched method is used (-0.0082). However, both results are not significant.

The coefficient of ST_CEO*AQ is expected to be negative. This coefficient shows the difference in the relationship between ST_CEO and AB-EM between firms with a low vs. high audit quality. When DA is measured via the modified Jones model, the coefficient has the expected sign, but is not significant (-0.0069). For the performance-matched method, the coefficient is positive and insignificant (0.0056). Based on these findings, I have to reject my

(30)

hypothesis. So, against expectations, audit quality (measured as Big Four membership) does not influence the relationship between short horizon CEOs and discretionary accruals. However, the coefficient for AQ is negative and significant for both measurements of discretionary accruals. This suggests that audit quality has a direct mitigating effect on income-increasing accrual-based earnings management. This is consistent with prior research (Becker et al., 1998).

Table 6: Interaction effect (Big Four as moderator)

Intercept β0 0.0043 0.0070 (0.0027) (0.0065) ST_CEO β1

+

0.0036 -0.0082 (0.0059) (0.0077) AQ β2 -0.0114*** -0.0104*** (0.0021) (0.0028) ST_CEO * AQ β3

-

-0.0069 0.0056 (0.0062) (0.0081)

Controls Yes Yes

F-statistics 71.54*** 6.73***

R-squared 0.1319 0.0141

N 12,736 12,736

I replicate the analysis reported in table 6 with industry specialization as the proxy for audit quality. The results are shown in table 7. In general, the results substantiate my prior results. Again, the coefficient on ST_CEO is not positive and significant in these analyses. On the contrary, for discretionary accruals measured according to the modified Jones model, I find a negative and significant relation (-0.0039). This suggests that when audit quality is low, short-horizon CEOs use less income-increasing discretionary accruals. Furthermore, AQ is not significantly related to discretionary accruals (for both different proxies of earnings management). This contrasts with prior research (Balsam et al., 2003) where a negative relationship between industry specialization and discretionary accruals is found. In addition, the coefficient on ST_CEO*AQ is again insignificant. So, also this analysis does not provide support for my hypothesis.

(31)

Table 7: Interaction effect (Industry specialist as moderator)

Intercept β0 0.0132*** 0.0045 (0.0049) (0.0065) ST_CEO β1

+

-0.0039* -0.0043 (0.0022) (0.0029) AQ β2 0.0001 0.0002 (0.0013) (0.0017) ST_CEO * AQ β3

-

0.0045 0.0038 (0.0041) (0.0053)

Controls Yes Yes

F-statistics 70.00*** 6.20***

R-squared 0.1276 0.0130

N 12,736 12,736

For my first hypothesis, I expected short-horizon CEOs to have incentives to manipulate earnings in their own interest. This could not be confirmed from my analysis, as the results of the regression where not significant. Therefore, I reject my first hypothesis.

Secondly, I predicted that due to the increased control from auditors, managers will use less discretionary accruals when being audited by a high-quality auditor. While the regression (using Big Four as AQ) showed that high quality auditors have a negative impact on the use of discretionary accruals, I am unable to provide evidence for the moderating effect. Measuring AQ as industry specialism does not provide any additional support. Therefore, I also reject my second hypothesis.

4.3 Supplemental analysis

In this research, short horizon is measured as the year prior to CEO leave. To assess whether my results are sensitive to this specific definition of short horizon CEOs, I repeat the analysis with an alternative definition of short horizon.

As CEOs might be aware of their departure more than one year in advance (for example when CEOs almost reach the age of retirement), I will create a new the indicator variable measuring the short horizon of a CEO as the two years prior to CEO leave. My new indicator variable (SH2_CEO) will have the value of 1 in the two years before departure and 0 otherwise. Table 8

(32)

and table 9 show the results of the additional analysis. The regression in table 8 shows similar results to the regular regression in table 5. The coefficient of ST2_CEO is negative, but insignificant in both cases.

Table 8: OLS regression additional analysis

Intercept β0 0.0131*** 0.0040

(0.0049) (0.0064)

ST2_CEO β1

+

-0.0011 -0.0002

(0.0015) (0.0019)

Controls Yes Yes

F-statistics 75.70*** 6.62***

R-squared 0.1292 0.0128

N 12,736 12,736

Table 9 reports the results of including moderator variable audit quality. The coefficients of ST2_CEO are negative in three out of the four cases. However, none of the results is significant, while a negative significant effect was found in the previous analysis (table 7). Moreover, there is one significant result for the interaction effect, while there were none before. When AQ is measured as industry specialism and discretionary accruals via the modified Jones model, the interaction is positively significant at a 10% level. This suggests that higher audit quality has a positive effect on the relationship between short-term CEOs and accrual-based earnings management. The increased use of accruals when audit quality is high and CEOs leave within a short timeframe is not in line with expectations based on previous research. Concerning the direct effect of AQ, the results are similar to earlier analyses. When AQ is measured as Big Four, the coefficients are both negative and significant at a 1% level (-0.0113 and -0.0101). When AQ is measured as industry specialism, the coefficients are both not significant.

Overall, the additional analyses provide one new insight, which is the significant interaction effect when AQ is measured as industry specialism. Still, both hypotheses are rejected based on this, because the significant positive effect is against the expectation of the second hypothesis.

(33)

Table 9: Interaction effects additional analysis

Variables Predicted sign AQ as Big Four AQ as Ind. Spec

DA(MJM) DA(PM) DA (MJM) DA (PM) Intercept β0 0.0149*** 0.0063 0.0134*** 0.0042 (0.0050) (0.0065) (0.0049) (0.0065) ST2_CEO β1

+

0.0039 -0.0013 -0.0028 -0.0011 (0.0046) (0.0061) (0.0017) (0.0023) AQ β2 -0.0113*** -0.0101*** -0.0003 0.0001 (0.0021) (0.0028) (0.0013) (0.0017) ST_CEO2 * AQ β3

-

-0.0055 0.0012 0.0056* 0.0033 (0.0049) (0.0064) (0.0032) (0.0042)

Controls Yes Yes Yes Yes

Industry effects Yes Yes Yes Yes

Year effects Yes Yes Yes Yes

F-statistics 71.48*** 6.65*** 70.02*** 6.12***

R-squared 0.1318 0.0139 0.1295 0.0129

N 12,736 12,736 12,736 12,736

(34)

5 Conclusion

5.1 Summary

This paper examines if short-horizon CEOs are more inclined to use accrual-based earnings management. Additionally, the role of audit quality is examined, where audit quality could possibly influence the behavior of short-term CEOs concerning accrual-based earnings management. Therefore, the following research question is investigated:

Does audit quality have a mitigating effect on the relationship between short-horizon CEOs and income-increasing accrual-based earnings management?

Accruals can be used to move income across different time periods. CEOs can use this opportunity to manage earnings in their own interests (Gopalan et al., 2014). The reverse property of accruals causes that leaving CEOs do not experience the opposite effect (income-decreasing accruals). Based on this, I tested the following hypothesis:

H1: CEOs with a short horizon use more income-increasing accrual-based earnings management.

An important mechanism to constrain opportunities to manage earnings is the assurance provided by auditors. (Becker et al., 1998). High audit quality can have a disciplining effect on managers, because the chance that misstatements are discovered is higher (DeAngelo, 1981). Therefore, the second hypothesis was as follows:

H2: High audit quality mitigates the positive relationship between short-horizon CEOs and accrual-based earnings management.

I use two methods to estimate accrual-based earnings management: the modified Jones model (Dechow et al., 1995; Kothari et al., 2005) and the method of performance-matched discretionary accruals by Kothari et al. (2005). Audit quality is measured in two different ways. First, Big Four auditors are considered to be of higher quality. Their large portfolio makes them less dependent on the revenues of one client (Becker et al., 1998; DeAngelo, 1981). The second proxy for audit quality is industry specialism, as more expertise leads to increased quality (Balsam et al., 2003; DeFond and Zhang, 2014).

Using a sample of 12,736 observations over a period from 2004 until 2015, the mean of discretionary accruals differs significantly between the short-horizon and non-short-horizon CEOs. Short-horizon CEOs use significantly less income-increasing discretionary accruals, under both proxies for earnings management. I expected short-horizon CEOs to have incentives to manipulate earnings upwardly, so these results are against my expectations. The analysis of the

(35)

first hypothesis shows no significant positive relationship for both earnings management proxies. Therefore, I reject my first hypothesis.

For my second hypothesis, I predicted that due to increased control from auditors, managers will use less discretionary accruals when being audited by a high-quality auditor. While the regression (using Big Four as AQ) shows that high quality auditors have a negative impact on the use of discretionary accruals, I am unable to provide evidence for the moderating effect. Measuring AQ as industry specialism does not provide any additional support. Therefore, I also reject my second hypothesis. So, against expectations, audit quality does not have a mitigating effect on the relationship between short-horizon CEOs and discretionary accruals. However, my research also provides different insights. Audit quality (Big Four membership) has a direct moderating effect on earnings management, which is consistent with prior research and confirms the theory that auditors perform a monitoring role (Becker et al., 1998). Lastly, when industry specialism is used as proxy, results suggest that when audit quality is low, short-horizon CEOs use less income-increasing discretionary accruals, which is against expectations.

I conducted a supplemental analysis where the measure for short-horizon CEOs was adjusted. Opposite of what I expected, I found evidence for a positive significant interaction effect, which means higher audit quality strengthens the positive relationship between short-term CEOs and accrual-based earnings management. This is not in line with prior research.

5.2 Conclusion

This research provides no evidence that CEOs with a short horizon use more income-increasing accruals. Also, the mitigated effect of audit quality on this relationship cannot be proved. Nevertheless, there is some indication of an opposite effect of audit quality: a strengthening effect on the relationship between short-horizon CEOs and accrual-based earnings management, which is in contrast with what I predicted based on prior literature. Also, I do find prove of a direct influence of audit quality on earnings management.

5.3 Limitations

There are a few limitations within this study. Firstly, I had to assume that CEOs are aware of the fact that they are going to leave the company, as it would otherwise be impossible to measure conscious earnings management. In reality, it may be the case that part of the short-horizon CEOs are unaware of their upcoming departure. It is really difficult to indicate which of the