The effect of CFO power on earnings management following CEO
turnover
Name: Rik Hunse
Student number: 11377763 Thesis supervisor: R.S. Ghita Date: June 22, 2018
Word count: 17.327 (12.788 core text, excluding tables) MSc Accountancy & Control, specialization Control
Statement of Originality
This document is written by student Rik Hunse who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
I test for the effect of CEO turnover on earnings management and the extent to which this effect is moderated by CFO power. I test for accruals-based earnings management using the Modified Jones Model and use Roychowdhury’s (2006) model to test for real earnings management. I use CFO tenure as a proxy for CFO power. After controlling for, amongst others, firm performance and governance strength, I find only weak evidence for accruals-based earnings smoothing surrounding routine CEO transitions. CFO power has a slight but nearly negligible mitigating effect on this form of earnings smoothing. I find weak evidence for an accruals-based big bath in CEO transitions where departure was forced, and succession was internal. More powerful CFOs are not found to be better able to constrain such big bath behavior. Surprisingly, I find that CEO turnover positively affects income-decreasing abnormal discretionary expenditure in each year surrounding a CEO transition for nonroutine CEO transitions and more powerful CFOs seem better able to limit this downward form of earnings management.
Contents
1 Introduction... 5
2 Literature review & Hypotheses development ... 8
2.1 General earnings management incentives... 8
2.2 Earnings management incentives specific to new CEOs ... 9
2.3 CFO power’s potential as a remedy to earnings management ... 12
2.4 Hypotheses Development ... 13
3 Research design ... 19
3.1 Identification of CEO transitions ... 19
3.2 Earnings management proxies... 19
3.3 CEO Turnover and CFO power proxy ... 21
3.4 Control Variables ... 22
3.5 Empirical method ... 22
4 Empirical results ... 25
4.1 Sample selection and descriptive statistics ... 25
4.2 OLS-Regression Results ... 29
4.2.1 Type 1 CEO transitions: peaceful departure with internal succession ... 31
4.2.2 Type 2 CEO transitions: peaceful departure with external transition ... 34
4.2.3 Type 3 CEO transitions: forced departure with internal succession ... 36
4.2.4 Type 4 CEO transitions: forced departure with external succession... 39
5 Discussion and Conclusion... 42
5.1 Discussion of main findings ... 42
5.1.1 Discretionary accruals-based earnings management ... 42
5.1.2 Real earnings management ... 44
5.2 Conclusion ... 46
6 References ... 48
1 Introduction
The role of the Chief Financial Officer (CFO) is a complex one. On the one hand, the CFO is charged with a fiduciary responsibility to safeguard financial statement integrity, on the other hand the CFO fulfills an important role as strategic partner to the Chief Executive Officer (CEO) (Kroos, Schabus & Verbeeten, 2017); Friedman, 2014). This conflict of interest is an interesting factor when studying the earnings management efforts of CEOs. Earnings management, especially accruals-based earnings management, poses a threat to the quality of reported earnings and therefore also undermines the fiduciary duty of CFOs. Prior studies have mainly focused on the extent to which more powerful CFOs are better able to constrain income-increasing forms of earnings management (Feng, Ge, Luo and Shevlin, 2011; Baker, Lopez, Reitenga and Ruch, 2018). A far less explored area is the question whether more powerful CFOs can and do constrain income-decreasing forms of earnings management as well. A so-called earnings bath or big bath and earnings smoothing may not be as harmful in the eyes of investors, regulators and creditors as income-increasing earnings management, I consider downward earnings management and earnings smoothing nevertheless as being threats to the quality of financial statements as well, and therefore indirectly undermine the duties of the CFO.
CEO transitions are a setting in which earlier studies identified increased likelihood of big baths and earnings smoothing to occur (e.g., Choi, Kwak & Choe, 2014; Wells, 2002). Therefore, CEO turnover cases form a suitable setting in which to address the gap in the literature, which I attempt to fill by answering the research question:
“How does the power that CFOs have in organizations help in combatting earnings management behavior by CEOs who are engaged in a CEO transition?”
I draw on several streams of extant literature. First, there is a large base of literature on general forms of earnings management, mostly income-increasing forms of earnings management (e.g., Cohen & Zarowin, 2010; Guidry, 1999; McAnally, Srivastava & Weaver, 2007). Second, I discuss a part of the literature that concentrates on earnings management surrounding CEO turnover. I discuss how the forms of earnings management and the incentives for those forms of earnings management are different in times of CEO turnover than they would be in normal course of business (Choi et al., 2014; Ali & Zhang, 2015; Wells, 2002; Bornemann, Kick,
Pfingsten & Schertler, 2015). Third, I describe a stream of literature that focuses on the power relation between CEO and CFO, and how more powerful CFOs are more able to achieve their goals (Simsek, 2007). I attempt to bring these areas of research together and as such create a new starting point for study.
I test for the effect of CFO power on two forms of earnings management: accruals-based earnings management (which involves accounting choices) and real earnings management (which involves deviations from normal actual business decisions). I hypothesize that CFOs have more interest in fighting accruals-based earnings management than in fighting real earnings management, as accruals-based earnings management is more in conflict with the fiduciary duties of the CFO, and real earnings management is not. I use the Modified Jones Model to estimate discretionary accruals and Roychowdhury’s (2006) model to estimate discretionary expenditure, a proxy for real earnings management. I first test for indications of both accruals-based and real activity-based earnings management across four types of CEO transitions, where type 1 transitions involve peaceful departures and internal successions, type 2 transitions are characterized by peaceful departure and external succession, type 3 transitions are forced departures followed by internal succession and type 4 transitions involve forced departures and external succession. Type 1 transitions are also referred to as routine transitions, whereas type 2-4 transitions are also referred to as nonroutine transitions. I then regress the earnings management indicators on a dummy variable representing CEO turnover and subsequently test whether the effect of CEO turnover on earnings management is moderated by CFO power.
I find weak evidence for accruals-based earnings baths subsequent to type 3 CEO transitions. I also find weak evidence for earnings smoothing surrounding routine CEO transitions. I find that CEO turnover is associated with income-decreasing abnormal discretionary expenditure in all years surrounding a CEO turnover, and that this effect is weakly negatively moderated by CFO tenure. I find weak evidence for a mitigating effect of CFO tenure on accruals-based earnings smoothing in the transition year for routine CEO transitions. Contrary to expectations, I do, however, not find evidence for a mitigating effect of CFO tenure on accruals-based big baths surrounding nonroutine CEO transitions.
Together, my findings suggest that accruals-based earnings management patterns in my sample are consistent with those found in prior literature, although effect sizes are rather small.
Surprisingly, I do not find convincing evidence that more powerful CFOs are better able to constrain accruals-based earnings management. The results on real earnings management are surprising as well: the pattern of income-decreasing abnormal discretionary expenditure is not consistent with what should be expected in an earnings bath setting. The results indicate that more powerful CFOs can and do have a small mitigating effect on income-decreasing abnormal discretionary expenditure, even though this is not their core responsibility.
My contribution to extant literature is twofold. first, I add to the literature on earnings management surrounding CEO transitions. This stream of literature is severely divided as to which forms of earnings management occur under which circumstances. The results of my study show an overlap with some of those found in existing literature and contradict others, another indication that it is not straightforward to associate specific forms of earnings management with specific turnover circumstances. I argue, moreover, that within CEO transition types, circumstances and earnings management incentives may vary greatly. The second contribution is a nuance to the fiduciary duty of the CFO. I argue that CFOs do work to protect the quality of financial statement, but that the extent to which they succeed in doing so is not per se dependent on the power they have in their organization, or more specifically: on their tenure. In the next section I present an overview of existing literature. In the third section I develop my research method. In section four I discuss my sample and present empirical results of the study. In the fifth section I discuss my results and draw conclusions. 1
1 In the remainder of this text, when I refer to the CEO or CFO using words such as ‘he’, ‘his’ and ‘him’, this
should be read as ‘he/she’, ‘him/her’ and ‘his/her. For readability concerns, however, I decide not to make this distinction throughout the entire text.
2 Literature review & Hypotheses development
2.1 General earnings management incentives
Earnings management in general is an intensively discussed phenomenon in existing literature. The concept of earnings management is somewhat intangible as annual reports do not explicitly report on earnings management: it is hidden in the actions and reporting decisions. Earnings management can take place through accounting choices (e.g. by means of discretionary accruals) and/or through real earnings management, which involves managers influencing earnings through real actions, when those deviate from normal business activity (Ugrin & Emley, 2017); Lo Ramos & Rogo, 2017; Ali & Zhang, 2015; Cohen & Zarowin, 2010). Real earnings management is performed through, for example, abnormal R&D expenditure, or asset write-downs.
A study on the extant literature on earnings management learns that earnings management is mostly performed to overstate earnings. Burgstahler and Dichev (1997) argue that organizations engage in earnings management when it would help the company present an increasing earnings pattern or to avoid losses. According to the prospecting theory, presented by Kahneman and Tversky (1979), the incentives for earnings management are the largest when it leads the company to report a gain instead of a loss or when earnings management leads the company to report an increase rather than a decrease in earnings. In support of this theory, Hayn (1995) demonstrates how significantly more companies report a small profit than a small loss, implying that organizations manage earnings to show a positive earnings figure if not doing so would lead them to report a small loss.
There is a good reason why organizations are inclined to present earnings that look better than they actually are. Higher earnings can reduce transaction costs with stakeholders. For example, suppliers and lenders offer better terms, customers pay higher prices for continued service and employees are more inclined to stay with the organization (Bowen, DuCharme & Shores, 1995; Zhao, 2017). Managers can also manipulate earnings to meet performance thresholds for their bonus awards (Guidry, Leone & Rock, 1999) or stock market expectations (Dichev, Graham, Harvey & Rajgopal, 2013). However, there are also reasons to manipulate earnings downwards. Managers can understate earnings when the strike price of their equity option grants is related to the company’s stock price (which is influenced by earnings) to make personal profit
subsequently (McAnally, Srivastava & Weaver, 2007). A combination of upward and downward earnings management can occur when managers perceive that the stock market values earnings stability Dichev et al., 2013).
2.2 Earnings management incentives specific to new CEOs
In the previous section I presented a selection of the literature on earnings management in general. However, this study is concentrated toward earnings management behavior among new CEOs, that is, earnings management subsequent to CEO transitions. One reason why specifically CEO turnover is an interesting context to study earnings management can be found in the fact that CEO turnover brings forth different opportunities and incentives to engage in different forms of earnings management. The types of earnings management observed may be dependent on the conditions under which the transition takes place (Choi et al., 2014). I do, however, not intend to claim that the general incentives for earnings management are absent among new CEOs. What I demonstrate in this section is how the specific earnings management incentives for new CEOs differ from general incentives and why CEO turnover is an interesting setting for studying earnings management (Choi et al., 2014). The literature on earnings management in general is rather extensive and well-developed. In this literature, I observe how earnings management behavior surrounding CEO turnover is different from general, more common earnings management behavior, and it is this type of earnings management that has received relatively less attention from researchers to date. It is these contrasts why I consider it relevant to also shed some light on general earnings management: to better understand the context of this study, in which I view earnings management following CEO turnover as a subset of earnings management in general.
Few authors found evidence for upward earnings management subsequent to CEO turnover in their studies. One incentive to engage in upward earnings management may be considered to be not much different from general upward earnings management incentives: Ali and Zhang (2015) find that CEOs who are new to their role are likely to report overstated earnings. These authors argue that those CEOs are subject to severe scrutiny: the new CEO still has to prove his ability to the market. Other incentives to manipulate earnings upward are presumably the same as the general incentives as discussed earlier.
A more dominant place in extant research is reserved for so-called big bath or earnings bath behavior. Such an earnings bath allows the new CEO to take as much losses as possible in his
first year and blame the departing CEO for the poor performance in his first year. This lowers expectations for subsequent years and at the same time reduces the likelihood that losses need to be taken in subsequent years, making it easier for the incoming CEO to meet or beat expectations in subsequent years (Choi et al., 2014; Bornemann et al., 2015). Choi et al. (2014) make a classification within their sample of Korean firms, that distinguishes four types of CEO turnovers. The first distinction is made based on whether the departing CEO was either forced to leave or left peacefully. Next, CEO transitions were divided based on whether the new CEO was recruited from outside the organization or promoted from within the company. In this way, a 2x2 setting was created, resulting in four subsamples, each representing a different combination of circumstances to the CEO transition (Choi et al, 2014). Similar distinctions can be found in the work of Wells (2002) and Bornemann et al. (2015).
Consensus as to which forms of earnings management to expect in which CEO turnover circumstances has not been reached. Choi et al. (2014) find that new CEOs who are promoted from inside are more likely to take an earnings bath when they succeed a CEO that was forced to leave, and that they do so through both discretionary accruals and discretionary expenditure, often referred to as real earnings management. Moreover, according to Choi et al. (2014), externally recruited CEOs in a peaceful transition are inclined to use discretionary accruals to manage earnings upward in their first year. The latter finding is consistent with the findings of Ali and Zhang (2015), but inconsistent with Bornemann et al. (2015), who find evidence for big baths among externally sourced incoming CEOs. Bornemann et al. (2015) do not find evidence that suggests that there is a difference in big bath behavior between peaceful and forced CEO departures, contrary to Wells (2002) and Choi et al. (2014). The findings and theories by Choi et al. (2014) and Bornemann et al. (2015) are rather contradictory. On the one hand, Bornemann et al. (2015) claims that externally recruited CEOs take earnings baths and that it is less relevant whether the departing CEO left peacefully or was forced to leave, while on the other hand Choi et al. (2014) find no support for big baths among externally recruited CEOs. Instead, they only find evidence for big bath earnings management among internally sourced CEOs who succeed CEOs who were forced out.
Bornemann et al. (2015) explain their findings through two concurring theories: externally sourced CEOs may have incentives to lower benchmarks for future performance, and externally recruited CEOs may be specifically hired with the intention to have them ‘clean up’ the balance sheet (Bornemann et al., 2015, p. 200). I argue, however, that both of these are somewhat
general reasons for taking a big bath, and that this does not convincingly explain why externally recruited CEOs are more likely to take an earnings bath than internally promoted CEOs. Choi et al. (2014) theorize that big bath earnings management by internally promoted CEOs following a forced departure of the former CEO may be induced by a sense of rivalry between the departing and incoming CEO. Wells (2002) also finds evidence for downward earnings management following a CEO turnover and finds that this effect is stronger when the CEO transition was a non-routine event, which is partly consistent with Choi et al. (2014), but inconsistent with Bornemann et al. (2014). Wells (2002) argues that externally recruited CEOs are more likely to take an earnings bath as they do not carry responsibility for past results. I consider this, moreover, a better explanation for the results of Bornemann et al. (2015), albeit that Bornemann does not find any differences between forced and peaceful departure, and their results would thus not implicitly suggest that an earnings bath is taken to blame the predecessor.
It should be noted, however, that Wells (2002) only finds weak evidence for real earnings management, and no evidence for accruals-based earnings management. Although there is some overlap, especially between the studies by Wells (2002) and Bornemann et al. (2015), it may be considered surprising that evidence on big bath earnings management among externally sourced CEOs who succeed ousted CEOs is not more obviously apparent in the literature, as the incentives for a big bath seem largest in such a situation: a CEO who was forced out is an easy scapegoat, whereas the externally recruited CEO does not feel any responsibility for decisions made prior to the transition.
Less surprising is that, to my knowledge, no author claims to find evidence for big baths subsequent to routine turnover, i.e., transitions where internally promoted CEOs succeed peacefully departed CEOs. Instead, Lam Detzler and Machuga (2002) present the coaching theory. According to the coaching theory, in a routine turnover, both departing CEO and incoming CEO may have incentives to give the impression of a smooth transition. In other words, the departing CEO does not have incentives to overstate earnings for short-term personal gain and the incoming CEO does not have incentives to take an earnings bath (Lam Detzler & Machuga, 2002). Presenting a smooth transition sounds as something that everybody would favor, however, if the actual performance does not follow such a smooth pattern, presenting smoothed earnings may be an unfaithful representation of reality.
In summary, there are three types of earnings management that new CEOs engage in: upward earnings management, downward earnings management and earnings smoothing. This is as such not that different from earnings management in general, but the incentives to engage in the different forms of earnings management are different in times of CEO turnover than in normal circumstances. From the literature can be derived that it depends on the circumstances under which a CEO turnover occurs, what kinds of earnings management can be expected. There is no clear consensus regarding which forms of earnings management should be expected under which circumstances. Researchers have made attempts to find explanations for these differences. For instance, Choi et al. (2014) suggest that differences may arise due to differences in management contracts. It could, for example, be that CEO transition circumstances have influence on (variable) compensation contracts and thus earnings management incentives. Wells (2002) however, conducted his study in an Australian context, where variable executive compensation related to firm performance has no prominent role in management contracts. He argues that earnings management takes place despite the absence of such compensation factors.
2.3 CFO power’s potential as a remedy to earnings management
The role of the modern CFO is an interesting one. On the one hand, the increasing uncertainty in the economic environment has made the tasks of executives more demanding, which led the role of CFO to transform from mainly performing financial reporting tasks to more of a strategic partner within the organization (Ilie, 2015), on the other hand, traditionally, the CFO’s most important responsibility has been to prepare the company’s financial statements and safeguarding the integrity of financial reporting (Kroos et al., 2017). Kroos et al. (2017) suggest that there is a relation between CFO bonus incentives and earnings management behavior by CFOs. That is, the more the CFO’s compensation is linked to company performance, the more that CFO is inclined to manipulate the financial reports.
Organizations attempt to manage this interest conflict for CFOs, for example through the voluntarily adoption of bonus clawbacks (Kroos et al., 2017). This dual role of CFOs could mean that CFOs could not only be seen as the preventers of earnings management, but also as the instigators of earnings management at the same time. Feng et al. (2011) find that CFOs are indeed involved in earnings management practices. However, in these cases, CFOs involved in earnings manipulations were not found to have significantly different equity incentives than matched CFOs who did not engage in earnings management. Instead, the authors find that these CFOs were mostly pressured by the CEO, who in his turn, was found to be sensitive to equity
incentives. The greater the difference between equity incentives for both CEO and CFO, the more likely the CEO would successfully involve the CFO in earnings manipulations (Feng et al., 2011). Personal gain was not found to be a motivator for CFOs to engage in earnings management, but CEO pressure was, which is consistent with Friedman (2014), who also suggests that the quality of reported earnings tends to decrease as the CFO’s power relative to the CEO decreases.
An interesting nuance is proviced by Baker et al., (2018). In their recent study on CEO-CFO power relations and earnings management efforts, they find that CEOs are more likely to manage earnings through accruals (accounting choices), whereas CFOs are more likely to manage earnings through discretionary expenditure (real business choices). The CEO and CFO attempt to prevent each other from successfully doing so, and the more powerful, the better they can prevent each other’s’ earnings management efforts. This suggests that both CEO and CFO have their own areas of responsibility which they protect, while they try to manage earnings through the area of responsibility of the other (Baker et al., 2018).
What the study by Baker et al (2018) has in common with the majority of studies, is a focus on upward earnings management. The effect of CFO power on other forms of earnings management remains a somewhat unexplored matter. Prior literature implies that financial reporting integrity is enhanced by a somewhat powerful CFO. Increased CFO power leads to better knowledge of the firm’s operations, risks and resources (Beck & Mauldin, 2014) and powerful CFOs also have more established relationships with key stakeholders on which they can rely in achieving their goals (Simsek, 2007). However, ultimately, CFOs are responsible to their CEOs, which gives the latter power over the former, while, contrary to the CFO, the main task of the CEO is to maximize shareholder value (Friedman, 2014).
2.4 Hypotheses Development
CFOs may be argued to have incentives to engage in earnings management themselves, a case in which it would be contradictory to expect that CFOs play a role in reducing or preventing earnings management. However, Feng et al. (2011) find that CFOs do only actively engage in earnings management when they are pressured by their CEOs. Baker et al. (2018) find that CFOs may also engage in earnings management themselves, but that it is more likely that this occurs through influence on real activity, rather than through financial statement manipulations. These are initial indications that CEO-CFO power relations are an interesting factor to involve
in this study. Kroos et al. (2017) emphasize the CFOs’ personal cost of earnings restatement following earnings management practices. As the financial gatekeeper, the negative consequences of such earnings restatements are especially significant to the CFO (Kroos et al., 2017). Furthermore, Friedman (2013) demonstrated that the quality of financial reports diminished as the CFO’s power relative to the CEO decreased. Summarized and simplified: CFOs have no strong incentives to initiate earnings management themselves, but rather to protect financial statements instead. They are, however, less likely to succeed in doing so when they have relatively low power compared to their CEO.
That is, however, in the normal course of business, where earnings management usually involves upward earnings management, for example to improve credit ratings, to be eligible for bonus payments, or to keep up with analyst expectations (Dichev et al., 2013; Guidry et al., 1999). Subsequent to CEO transitions on the contrary, the literature is especially focused on downward earnings management (Choi et al., 2014; Bornemann et al., 2015; Wells, 2002) and earnings smoothing (Lam Detzler & Machuga, 2002). Not only the different incentives for and types of earnings management observed differ from normal course of business. Subsequent to a CEO transition, the CEO-CFO power relations may also be reconfigured as there simply is a new CEO. First, I test whether the patterns observed by previous authors are also observable across my sample of US-based firms with a stock exchange listing. Choi et al. (2014), Bornemann et al. (2015) and Wells (2002) all observed evidence for downward earnings management subsequent to a CEO transition in their samples. Also, all three researches made a subdivision in their samples based on different scenarios based on peaceful versus forced departure and internal versus external CEO succession. The existing literature is not entirely unanimous as to in which scenarios to expect which forms of earnings management, but none of the aforementioned authors found evidence for a big bath subsequent to a peaceful CEO transition involving internal succession.
In such routine CEO transitions there is practically no reason to take the bath: there is no scapegoat to blame, the departing CEO might take a supervisory role (which increases the chance of confrontation) and the incoming CEO might even be partly responsible for prior performance if promoted from inside, hence it does not make sense to take an earnings bath. The aforementioned authors each found evidence for earnings management in some of the three remaining scenarios, but although there is an overlap in findings, there are no two researchers who find exactly the same form of earnings management in the same scenario. In line with prior
literature, I also facilitate for a similar subdivision based on CEO transition circumstances and hypothesize that downward earnings management takes place subsequent to a CEO transition involved either forced CEO departure, external CEO succession, or both. Building on the coaching theory (Lam Detzler & Machuga, 2002), I expect to find evidence for accruals-based earnings smoothing surrounding a peaceful CEO transition with internal succession. Big bath earnings management can take place through either discretionary accruals and discretionary expenditure:
H1a: Incoming CEOs take an earnings bath through discretionary accruals in those cases where the CEO-transition was not a peaceful event with internal succession H1b: Incoming CEOs take an earnings bath through discretionary expenditure in those cases where the CEO-transition was not a peaceful event with internal succession
H1c: Following coaching theory, accruals-based earnings smoothing can be observed surrounding a peaceful CEO-transition with internal succession
If it is already established that CFOs can use their power to prevent earnings management from occurring, which is basically reasoning the other way around from Friedman’s (2013) finding that CEOs use their power over CFOs to make earnings management happen, and that the CFO’s personal cost of earnings restatements are an incentive in itself to combat earnings management (Kroos et al., 2017), why would it not automatically be the same in the case of earnings management surrounding CEO turnover?
The answer would lie in the nature of the type of earnings management that may be expected to occur surrounding a CEO transition. The existing literature has, when discussing the role of CFOs in earnings management, so far predominantly (if not entirely) focused on upward earnings management. As upward earnings management basically involves creating a better picture than justified by reality, this poses a potential danger to investors, lenders and other stakeholders. The responsibility of the CFO is really to protect these stakeholders against misleading information. I consider downward earnings management far less risky, I argue that it could almost be seen as ‘overly prudent behavior’. Of course, there are downsides for stakeholders, an investor might argue that an earnings bath might perhaps unnecessarily reduce share prices and/or dividends, but in the long run these negative effects will be offset by the significantly better future results that are inherent to earnings baths. To a lesser extent, I argue that the same could be said about earnings smoothing, in which better and worse results are
smoothened against each other. In other words, I argue that a big bath and earnings smoothing are less dangerous for stakeholders than earnings overstatements, which leads me to suggest that results of prior research are not readily transferrable to this context without further examination.
It remains, however, the CFO’s duty to protect financial statement integrity and the quality of reported earnings. Furthermore, as an executive, the CFO’s compensation is usually partly variable and based on the company’s performance, which means that the CFO himself has no incentives to facilitate an earnings bath. I argue that the only one taking benefit from an earnings bath would be the incoming CEO. Therefore, I theorize that the CFO intends to fulfill his fiduciary duty and attempts to protect the financial statements’ integrity. I find support for this assumption in recent work by Baker et al. (2018), who find that CFOs and CEOs attempt to protect their own fields of responsibility. More powerful CFOs can be expected to experience greater success in inhibiting accruals-based earnings management efforts by CEOs, but not in preventing real activity-based earnings management (Baker et al., 2018). I consider this plausible, as accruals-based earnings management is much more of a threat to the CFO’s fiduciary responsibilities than real activity-based earnings management would be and therefore I hypothesize that:
H2a: CFO power is negatively related with accruals-based earnings management indicators
H2b: CFO power is not significantly related with real earnings management indicators
In summary, I combine two lines of executive research. First, I build on literature that suggests that incoming CEOs have, under specific circumstances, incentives to engage in specific forms of earnings management. In my first set of hypotheses I test whether these assumptions hold throughout my sample of US companies. I facilitate a subdivision of my sample into subsamples based on CEO transition circumstances, just as prior researchers have done, to evaluate whether indications of earnings management can be observed in each of four scenarios. It could, for example, be that I do not find significant evidence for abnormal discretionary accruals in my whole sample, but I might find significant effects for isolated subsamples categorized by scenario. Second, I build on theory on CEO-CFO power relations and how CFOs attempt to fulfill their fiduciary duty and protect financial statement integrity, but may be less interested in restraining a CEO’s earnings management endeavors through real strategic decisions. Surrounding a CEO transition, the interests of the incoming CEO and the CFO may clash, and
I hypothesize that more powerful CFOs are more likely to successfully combat earnings management efforts by incoming CEOs.
In my first set of hypotheses, I expect to find indications of earnings management. Finding evidence in support of coaching theory (Lam Detzler & Machuga, 2002) would indicate that, in a peaceful CEO transition with internal succession, the departing CEO and incoming CEO successfully attempt to smoothen net earnings using discretionary accruals. This would, for instance, imply that the CEOs want to show a smooth transition with low noise to the outside world. Furthermore, in such a peaceful routine transition, departing and incoming CEO have usually worked together for a longer period leading up to the transition, and afterwards the departing CEO often remains active in the organization in some sort of supervisory role, another set of reasons for both CEOs to show a smooth transition (Lam Detzler & Machuga, 2002).
I expect to find indicators of downward earnings management (‘big bath’ or ‘earnings bath’) in those cases where the departing CEO was either forced to leave, the incoming CEO was recruited externally, or both. Forced departure of the former CEO provides the incoming CEO with an incentive to take an earnings bath and blame the former CEO, and to claim the credits for subsequent, better, results. An externally recruited CEO carries no responsibility for decisions made prior to his assignment, which also gives him increased possibility to take the bath. In those cases where either one of these circumstances is present, I expect to find indications of a positive causal relation between downward earnings management and CEO turnover. If evidence is indeed obtained for the first set of hypotheses, the assumptions based on extant literature can be reinforced. If not, I might need to reconsider these assumptions and perhaps conclude that earnings management surrounding CEO transitions might not be as significant across US-based firms as in Australian and South-Korean firms, for example.
In my second set of hypotheses, I expect to find a moderating effect of CFO power on the relation between CEO turnover and abnormal discretionary accruals. If this effect is indeed observed and considered significant, I theorize that more powerful CFOs are indeed more likely to be able to constrain new CEOs’ earnings smoothing and downward earnings management through discretionary accruals. If that would be the case, then CFO power might be considered a useful indicator for accruals-based earnings management risk in the wake of a CEO transition. If no significant effect will be obtained, there could be different possible reasons: it could be that CFO power is simply unrelated to earnings management, or that CFOs do not deter
accruals-based earnings management, but facilitate it instead (which would both contradict my theoretical framework), or that my proxies of CFO power do not successfully capture the relationship.
I theorize that CFOs use their power to counter accruals-based earnings management, but not real activity-based earnings management. A non-significant relationship between CFO power and discretionary expenditure would support that idea. It would indicate that CFOs award no priority to the deterrence of real activity-based earnings management as it would not directly compromise their own responsibilities: real earnings management poses no threat to financial statement integrity in itself. If, contrary to my expectations, a significant negative relationship between CFO power and discretionary expenditure differentials would be obtained, this would indicate that CFOs also use their power to counter real activity-based earnings management. In that case I might conclude that CFOs (also) operate against earnings management from their role as a strategic partner to the firm.
3 Research design
3.1 Identification of CEO transitions
Prior literature (Choi et al., 2014; Wells, 2002; Bornemann et al., 2015) indicate that (type of) earnings management is different across CEO transition scenarios. I therefore attempt to compose a balanced sample containing companies that underwent a CEO transition of either type. I distinguish four CEO transition types: type 1 CEO transitions (peaceful departure with internal succession), type 2 CEO transitions (peaceful departure with external succession), type 3 CEO transitions (forced departure with internal succession) and type 4 CEO transitions (forced departure with external succession). On request I obtain annual S&P500 CEO transition overviews for the years 2004-2016 from Spencer Stuart, a large executive search company in the United States. I consult Spencer Stuart on the interpretation of these reports to minimize the risk of interpretation errors.2 By far most of CEO transitions in North-America were routine events, that is, they would qualify as type-1 CEO transitions. It is a common problem in comparable studies that samples tend to be small. However, to have a balanced sample with sufficient of each CEO transition types I desire to have at least 30 companies in each subsample. In addition to the CEO transitions that I identify in the reports by Spencer Stuart (2004-2016), I conduct an extensive search on the internet for CEO transition events. In line with extant literature, I classify each CEO transition into one of the four types based on information in press releases and articles in business press. I acknowledge that there is a risk of misinterpretation due to potential window-dressing by companies or simply due to lack of experience with interpreting such communications for the author. To minimize this risk, I consult multiple communications before classifying each CEO transition.
3.2 Earnings management proxies
Two dominant earnings management mechanisms can be identified in the literature: earnings management through abnormal discretionary accruals – which involves accounting choices – and real earnings management, which means that managers actually make business decisions that deviate from normal practice (Choi et al., 2014; Wells, 2002; Bornemann et al., 2015). Testing for two different forms of earnings management allows me to evaluate whether or not different forms of CEO turnover give rise to different forms of earnings management and
2 For example, on the difference between ‘resignation’ and ‘stepping down’: after consultation I learn that
‘stepping down’ refers to a CEO who retires or leaves in good standing, whereas ‘resignation’ refers to a CEO who leaves in poor standing.
whether or not CFO power has a mitigating effect on either form of earnings management (hereafter interchangeably used with ‘EM’). Abnormal discretionary accruals are that part of total accruals that is estimated not to be explained by normal course of business. I use the Modified Jones Model as presented by Dechow, Sloan and Sweeney (1995) to estimate abnormal discretionary accruals. An important input to the Modified Jones Model is total accruals (TA). I identify two different ways to calculate TA. A commonly used method, amongst others by Dechow et al. (1995) to identify total accruals is to subtract operating activities net cash flow (OANCF) from net income (NI):
𝑇𝐴 = 𝑁𝐼 − 𝑂𝐴𝑁𝐶𝐹 (1)
An alternative way to calculate TA is proposed by Jones (1991):
𝑇𝐴𝑡= (∆𝐶𝐴𝑡− ∆𝐶𝐿𝑡− ∆𝐶𝑎𝑠ℎ𝑡+ ∆𝑆𝑇𝐷𝑡− 𝐷𝑒𝑝𝑡)/𝐴𝑡−1 (2)
Where ∆𝐶𝐴𝑡 and ∆𝐶𝐿𝑡 denote the increase of current assets and current liabilities in year t, ∆𝐶𝑎𝑠ℎ𝑡 is the increase in cash and cash equivalents in year t, ∆𝑆𝑇𝐷𝑡 denotes incremental debt in current liabilities in year t, 𝐷𝑒𝑝𝑡 is depreciation expenses in year t and 𝐴𝑡−1 denotes lagged total assets. The Modified Jones Model then requires that I regress total assets on a set of variables and then use the residuals of that regression as DADE (residuals of regression with TA obtained from equation 1) and DAJO (residuals of regression with TA obtained from equation 2). This regression formula reads as follows:
𝐷𝐴𝐷𝐸𝑡 = 𝛽1(1 𝐴⁄ 𝑡−1) + 𝛽2(∆𝑅𝐸𝑉𝑡− ∆𝑅𝐸𝐶𝑡) + 𝛽3(𝑃𝑃𝐸𝑡) (3)
𝐷𝐴𝐽𝑂𝑡 = 𝛽1(1 𝐴⁄ 𝑡−1) + 𝛽2(∆𝑅𝐸𝑉𝑡− ∆𝑅𝐸𝐶𝑡) + 𝛽3(𝑃𝑃𝐸𝑡) (4)
Where 𝐴𝑡−1 denotes lagged total assets, ∆𝑅𝐸𝑉𝑡 and ∆𝑅𝐸𝐶𝑡 denote incremental revenue and receivables in year t, respectively, and 𝑃𝑃𝐸𝑡 denotes gross property, plant and equipment in year t.
I test for real earnings management through abnormal levels of discretionary expenditure (DISEX). I employ a widely used model as presented by Roychowdhury (2006). According to
this model, I first calculate total discretionary expenditure for each company year as the sum of advertising expenses, research and development expenses and selling, general and administrative expenses. I then standardize total discretionary expenditure by dividing the sum by lagged total assets. To arrive at DISEX, I estimate the difference between total discretionary expenditure and normal discretionary expenditure by taking the residuals of the formula:
𝐷𝐼𝑆𝐸𝑋𝑃𝑡 𝐴𝑡−1 = 𝛼0+ 𝛼1( 1 𝐴𝑡−1) + 𝛽 ( 𝑅𝐸𝑉𝑡 𝐴𝑡−1) + 𝜀𝑡 (5)
For conducting additional tests, I also calculate the absolute values of DADE, DAJO and DISEX:
𝐴𝐵𝑆𝐷𝐴𝐷𝐸 = √𝐷𝐴𝐷𝐸2 (6) 𝐴𝐵𝑆𝐷𝐴𝐽𝑂 = √𝐷𝐴𝐽𝑂2 (7) 𝐴𝐵𝑆𝐷𝐼𝑆𝐸𝑋 = √𝐷𝐼𝑆𝐸𝑋2 (8)
I argue that, as these variables can assume values both above and below zero, there could be information in their absolute values. I prevent from concluding that no earnings management takes place if effect sizes on DADE, DAJO and DISEX are not significantly different from zero when in fact, for example, large positive effects on DADE offset large negative effects. Through absolute values, I attempt to eliminate this risk. Data required to calculate earnings management proxies is obtained from the COMPUSTAT North-America database.
3.3 CEO Turnover and CFO power proxy
CEO Transitions are denoted as dummy variable CEOTOV, which assumes value 1 for companies that are classified in each of four CEO transition types, and 0 if otherwise.
Extant literature provides several suggestions for CFO power proxies. I decide to use CFO tenure (CFOTNR) as my proxy for CFO power. As CFOTNR increases, so does his company-specific expertise, which, according to Beck and Mauldin (2014), provides him with a source of power. Furthermore, the longer a CFOs tenure, the more likely he has developed connections with important stakeholders. Important stakeholders can help an executive exercising influence and achieving his goals (Simsek, 2007). CFOs can usually be identified using the EXECUCOMP database, however, tenure is in the vast majority of cases not provided. The dataset on Pension Benefits in some cases provides some information, but in the majority of
cases CFOTNR values are hand-collected, mostly from Bloomberg’s executive search function, company websites (if CFO is currently employed) or CFOs’ personal LinkedIn profiles.
3.4 Control Variables
I control for firm performance using an adjusted measure of return on assets (adjROA). I calculate the return on assets by dividing net earnings by average total assets, and then subtract from that ratio the company’s average ROA for the preceding five years. I use an adjusted form of performance to control for differences between industries (Choi et al., 2014). For example, labor-intensive firms may typically have larger ROAs than capital-intensive firms. I control for corporate governance strength using a measure for board size (BOARD) and the percentage of independent directors on the board (OUTSIDE). Corporate governance data is obtained from the ISS Directors and ISS Directors Legacy database. In 24 cases, the data needed to be hand-collected from companies’ proxy statements filed with the Securities and Exchange Committee, because of missing data. I control for firm growth using sales growth (GROWREV) and total asset growth (GROWTA), calculated as the percentage difference between revenue and total assets, respectively. Finally, I control for firm size using the natural logarithm of total assets (SIZE). All control variables positively contribute to the fit of my model: the inclusion of each control variable leads to an increased overall F-statistic and adjusted R-squared.
3.5 Empirical method
First, I test whether my earnings management measures DADE, DAJO and DISEX have means significantly different from zero across subsamples, from year t= -1 to year t= +1, where the year of CEO transition is denoted with year t= 0. I then conduct a one-way ANOVA to test whether mean earning management indicators (EM) are different across subsamples. This series of tests should give a first impression whether EM occurs across subsamples and whether the subsamples are indeed different from each other, which I assume based on prior literature. Following H1a, I expect to find evidence for accruals-based big baths surrounding nonroutine CEO transitions.3 This would be indicated by significant negative DADE and DAJO in year t= 0 and significant positive DADE and DAJO in year t= +1. In year t= -1 one might expect significant positive DADE and DAJO, which would suggest that the departing CEO manages earnings upward to cover up poor performance. Following H1b, I expect to find evidence for real activity-based big baths surrounding nonroutine CEO transitions. This would be indicated
by significant positive DISEX in year t=0 and significant negative DISEX in year t= +1. Significant negative DISEX may be expected in year t= -1. Following H1c, I expect to find evidence for accruals-based earnings smoothing surrounding routine CEO transitions. Ideally, this would be indicated by nonsignificant DADE and DAJO, and highly significant ABSDADE and ABSDAJO. This would mean that abnormal discretionary accruals may be used to manage earnings, but that the direction in which this happens is not universal, consistent with what should be expected in an earnings smoothing setting: the one company manages earnings upward in the one year to offset a loss and downward in the next year to smooth an increase in profitability, whereas another company might do so in exact opposite direction. This alone, however, would be insufficient to conclude that earnings smoothing is occurring in type 1 CEO transitions, therefore I conduct an additional test if mean DADE and DAJO are not significantly different from zero.
To identify accruals-based earnings smoothing, Lam Detzler and Machuga (2002) compare net earnings in the year prior to a CEO transition (t= +1) with net earnings in the two years subsequent to CEO turnover (t= 0 and t= -1). Coaching theory then suggests that net income before and after the CEO turnover will be equal due to earnings smoothing (Lam Detzler & Machuga, 2002). I consider the coaching theory as developed by Lam Detzler and Machuga (2002) interesting, although I consider Bouwman’s (2014) method to test for earnings smoothing to better capture the concept of earnings smoothing: Bouwman (2014) suggests looking at the correlation between discretionary accruals and incremental operating activities net cash flow (dOANCF) instead. The more negative this correlation, the stronger the indication of earnings smoothing:
𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑆𝑚𝑜𝑜𝑡ℎ𝑖𝑛𝑔 = 𝐶𝑜𝑟𝑟(𝐷𝐴𝐷𝐸, 𝑑𝑂𝐴𝑁𝐶𝐹) (9) 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 𝑆𝑚𝑜𝑜𝑡ℎ𝑖𝑛𝑔 = 𝐶𝑜𝑟𝑟(𝐷𝐴𝐽𝑂, 𝑑𝑂𝐴𝑁𝐶𝐹) (10)
After determining the forms of EM and the extent to which this is visible across subsamples, I proceed with testing whether, in line with H1a-c, these forms of EM are indeed driven by CEO turnover. Following H2a-b, I test whether there is a mitigating effect of CFO power on these forms of earnings management. The OLS-regression formula for this set of tests is as follows:
𝐸𝑀𝑡= 𝛽1𝐶𝐸𝑂𝑇𝑂𝑉𝑡+ 𝛽2𝐶𝐹𝑂𝑇𝑁𝑅𝑡+ 𝛽3𝐶𝐸𝑂𝑇𝑂𝑉 ∗ 𝐶𝐹𝑂𝑇𝑁𝑅𝑡+ 𝛽4𝑎𝑑𝑗𝑃𝐸𝑅𝐹𝑡+ 𝛽5𝑆𝐼𝑍𝐸𝑡 + 𝛽6𝐺𝑅𝑂𝑊𝑇𝐴𝑡+ 𝛽7𝐺𝑅𝑂𝑊𝑅𝐸𝑉𝑡+ 𝛽8𝐵𝑂𝐴𝑅𝐷𝑡+ 𝛽9𝑂𝑈𝑇𝑆𝐼𝐷𝐸𝑡+ 𝜀
Where 𝐸𝑀𝑡 represents the outcomes of formulae 3-8. For H2a-b, I am mainly interested in the interaction effect CEOTOV*CFOTNR. If a mitigating effect of CFO power is present, then this should be indicated by a significant interaction term with an effect sign opposite to that of the coefficient of CEOTOV.
4 Empirical results
4.1 Sample selection and descriptive statistics
My sample is a carefully composed set of companies that fit a specific set of selection requirements. First, I need to be sufficiently confident about the circumstances under which a CEO transition takes place to be able to classify each transition in a turnover type. Previously, I discussed how I conduct this search and the limitations to this practice. CEO transitions of which it was overly ambiguous to determine the circumstances were not included in the sample. The second requirement is that companies continue to exist post-transition. In many cases where the CEO transition was a nonroutine event, the company either merged with another company or either went bankrupt shortly after the CEO was fired, or discontinued its stock listing, which impacts data availability. The third requirement is that sufficient data needs to be available. I use 15 company years of data to estimate DADE, DAJO and DISEX. All input should be available in the COMPUSTAT, EXECUCOMP and ISS Directors databases, or should be somewhat easily hand-collected. A majority of nonroutine CEO transitions fail to meet all these criteria. This results in a somewhat small turnover sample, with a total of 131 cases (393 firm year observations). All subsamples do contain at least 30 cases. The small sample size is unfortunate, but not extremely deviant from those in extant literature: Choi et al. (2014) for example, identify a total of 142 non-routine turnovers in Korea, of which most occurred among the smaller companies. Wells (2002) identified only 26 of such cases in an Australian setting. The descriptive statistics of the CEO turnover sample are presented in table 1, panel A, and its distribution by transition type is presented in panel C. In panel B I report the descriptive statistics for the control group, consisting of 63 companies4 (198 firm year observations) that did not experience CEO transitions. These observations were collected from the same period as those of the turnover sample, divided over that period as equally as possible, in order to minimize year-fixed effects. I acknowledge, however, that there may be differences between the control sample and the treatment sample, possibly due to selection bias. To illustrate, many data required hand-collection for the turnover sample, whereas the control sample was composed of companies for which sufficient data was available. This could indicate that not all companies may not have been subject to the same reporting requirements during the
4 The control sample is smaller than the total turnover sample, however, as I conduct the regression analysis on a
subsample-by subsample basis the control group is in each case larger than the treatment group, by a factor of at least 1,5
Table 1: Descriptive Statistics t= + 1 Max P an el A : C E O T u rn ove r s am p le E ar ni gns m anage m ent pr o xie s ,288 ,364 ,608 ,629 ,364 ,608 Inde pe nde nt v ar ia bl es 55 ,205 12,977 2,503 3,3264 20 ,929 Median ,003 ,002 ,005 ,023 ,023 ,014 8 -,002 9,534 ,018 ,0093 11 ,875 Min -,629 -,308 -,173 ,000 ,001 ,000 0 -,266 4,909 -,360 -,8143 6 ,500 std. Dev ,079 ,073 ,078 ,068 ,059 ,071 11 ,070 1,468 ,267 ,3236 2,108 ,551 Mean ,000 ,005 ,020 ,040 ,043 ,037 11 -,007 9,567 ,050 ,0408 10,608 ,873 Obs. 130 130 122 130 130 122 131 130 130 130 130 130 130 t= 0 Max ,382 ,123 ,556 ,626 ,266 ,556 36 1,024 12,935 1,011 ,421 20 1,000 Median -,002 -,002 ,003 ,024 ,017 ,018 8 -,007 9,427 ,028 ,020 11 ,833 Min -,626 -,266 -,109 ,001 ,000 ,000 0 -,486 5,022 -,339 -,872 6 ,333 std. Dev ,085 ,048 ,077 ,073 ,038 ,072 9,462 ,132 1,517 ,149 ,140 2,256 ,810 Mean -,006 -,011 ,023 ,044 ,031 ,037 10,802 -,007 9,502 ,034 ,017 10,603 ,890 Obs. 131 131 123 131 131 123 131 131 131 131 131 131 131 t= -1 Max ,428 ,394 ,398 ,428 ,394 ,398 35 1,024 12,960 ,858 ,677 20 ,941 Median ,003 ,002 ,005 ,020 ,020 ,013 8 -,010 9,380 ,019 ,019 11 ,846 Min -,221 -,285 -,111 ,000 ,000 ,000 0 -,530 4,462 -,453 -,337 5 ,100 std. Dev ,070 ,071 ,060 ,058 ,060 ,054 9,029 ,130 1,511 ,172 ,134 2,307 ,134 Mean ,002 ,008 ,018 ,038 ,038 ,030 10,351 -,005 9,478 ,029 ,039 10,725 ,804 Obs. 131 131 123 131 131 123 131 131 131 131 131 131 131 DADE DA JO D IS E X A B S D A D E A B S D A JO A B S D IS E X C F O T N R adj P E R F S IZ E G R O W T A G R O W R E V B O A R D O U T S ID E
Table 1, continued t= + 1 Max P an el B : C on tr ol s am p le E ar ni gns m anage m ent pr o xie s ,144 ,139 ,093 ,199 ,421 ,155 Inde pe nde nt v ar ia bl es 36 ,370 11,018 1,075 ,473 20 ,929 P an el C : D is tr ib u tio n of C E O t ran sit io n s Median ,007 -,002 -,004 ,022 ,023 ,012 9 ,002 8,127 ,071 ,050 9 ,818 Min -,199 -,421 -,155 ,002 ,000 ,002 1 -,358 4,535 -,351 -,518 4 ,385 std. Dev ,051 ,094 ,039 ,039 ,081 ,030 8 ,089 1,763 ,239 ,166 2,471 ,137 Mean -,001 -,021 -,003 ,033 ,051 ,024 11 -,001 8,104 ,124 ,080 9,635 ,789 Obs. 63 63 59 63 63 59 63 63 63 63 63 63 63 t= 0 Max ,297 ,235 ,111 ,297 ,284 ,116 36 ,203 10,943 1,601 ,793 19 ,947 Median ,005 ,004 ,001 ,020 ,027 ,011 8 ,019 7,998 ,071 ,060 9 ,833 Min -,087 -,284 -,116 ,000 ,003 ,001 0 -,083 4,051 -,356 -,409 4 ,385 std. Dev ,053 ,070 ,036 ,043 ,055 ,029 8,656 ,055 1,788 ,236 ,163 2,550 ,137 40 31 30 30 Mean ,010 -,007 -,004 ,031 ,043 ,022 10,603 ,024 8,006 ,115 ,094 9,603 ,790 Obs. 63 63 59 63 63 59 63 63 63 63 63 63 63 P ea ce ful de pa rtu re + I nt er na l suc ce ss io n P ea ce ful de pa rtu re + E xt er na l suc ce ss io n F or ce d de pa rtu re + I nt er na l suc ce ss io n F or ce d de pa rtu re + E xt er na l de pa rtu re t= -1 Max ,081 ,266 ,126 ,146 ,651 ,126 35 ,260 10,894 ,528 ,412 14 ,923 Median ,000 ,003 ,000 ,021 ,031 ,012 7 ,008 7,875 ,074 ,090 9 ,857 Min -,146 -,651 -,097 ,000 ,001 ,000 0 -,158 4,355 -,214 -,281 4 ,100 std. Dev ,046 ,102 ,033 ,032 ,087 ,026 8,394 ,069 1,770 ,149 ,136 2,264 ,166 Mean -,005 ,000 -,001 ,033 ,052 ,020 9,635 ,011 7,913 ,091 ,085 9,524 ,775 Obs. 63 63 59 63 63 59 63 63 63 63 63 63 63 T ype 1 T ra ns it io ns T ype 2 T ra ns it io ns T ype 3 T ra ns it io ns T ype 4 T ra ns it io ns DADE DA JO D IS E X A B S D A D E A B S D A JO A B S D IS E X C F O T N R adj P E R F S IZ E G R O W T A G R O W R E V B O A R D O U T S ID E
data collection period, and therefore may be different from each other. I attempt to reduce this risk as much as possible by controlling for size, performance and corporate governance metrics. However, there may be differences that cannot be filtered out by my model.
In table 2 I report the Spearman correlation coefficients of all variables. Spearman correlation coefficients are calculated after assigning ranks to each observation, which I consider desirable, as I have an ordinal value (CEOTOV) in my model. I find significant correlations between DADE, DAJO and DISEX, however, they all measure EM and will only be used as dependent variables. DADE, DAJO and DISEX significantly correlate with adjPERF, suggesting that there is a link between performance and abnormal discretionary accruals and abnormal discretionary expenditure, which I do not consider surprising. Also, there is a weak but significant relation between CEOTOV and adjPERF, which indicates that CEO transitions are more likely observed among weak performers and vice versa. Apart from the correlation between board size and firm size, there are no other correlations greater than ,300.
Table 2: Spearman correlation matrix
DADE DA JO D IS E X A B S D A D E A B S D A JO A B S D IS E X C E O T O V adj P E R F C F O T N R B O A R D O U T S ID E G R O W T A G R O W R E V S IZ E DADE ,425** -,138** 0,04 ,033 ,035 -,009 ,440** -0,08 -,038 0,06 ,103* ,003 -,021 DAJO ,425** -,08 -,006 -,036 ,001 -,003 ,175** -,043 -,018 ,016 ,133** ,027 -,06 DISEX -,138** -,08 ,04 -,004 ,265** ,156** -,212** -,099* -,111** -,028 -,126** -,167** -,042 ABSDADE ,04 -,006 0,04 ,397** ,201** ,026 0 -,197** -,160** -,121** -,125** -,036 -,172** ABSDAJO ,033 -,036 -,004 ,397** ,284** -,113** -,011 -,201** -,251** -,145** -,002 ,146** -,265** ABSDISEX ,035 ,001 ,265** ,201** ,284** 0,06 -,034 -,259** -,264** -,141** -,045 -,023 -,238** CEOTOV -,009 -,003 ,156** ,026 -,113** 0,06 -,150** -,032 ,234** ,102* -,244** -,223** ,385** adjPERF ,440** ,175** -,212** 0 -,011 -,034 -,150** -,025 -,033 ,017 ,161** ,172** -,122** CFOTNR -,08 -,043 -,099* -,197** -,201** -,259** -,032 -,025 ,224** 0,03 ,104* ,038 ,218** BOARD -,038 -,018 -,111** -,160** -,251** -,264** ,234** -,033 ,224** ,257** -,066 -,105* ,604** OUTSIDE ,06 ,016 -,028 -,121** -,145** -,141** ,102* ,017 0,03 ,257** -,068 -,084* ,237** GROWTA ,103* ,133** -,126** -,125** -,002 -,045 -,244** ,161** ,104* -,066 -,068 ,511** -,028 GROWREV ,003 ,027 -,167** -,036 ,146** -,023 -,223** ,172** ,038 -,105* -,084* ,511** -,069 SIZE -,021 -,06 -,042 -,172** -,265** -,238** ,385** -,122** ,218** ,604** ,237** -,028 -,069
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
Nevertheless, I run VIF tests to make sure that multicollinearity is not a problem in my sample. In all cases, VIF scores are far below 10 for each variable, indicating that there is no problem with multicollinearity in my data.
4.2 OLS-Regression Results
Following Choi et al. (2014), I first test whether there are differences in earnings management indicators across subsamples. In panel A-D of Table 3 I report the mean earnings management indicators for each CEO transition type. On the left side I present the mean actual values of estimated DADE, DAJO and DISEX, on the right side I present the mean absolute values of estimated DADE, DAJO and DISEX. In panel E I report the results of a one-way ANOVA test for differences between the transition types. It should not be considered overly informative that all absolute values are significantly different from zero, as absolute values cannot be negative. However, in a later stage, where I conduct the regression analyses, the absolute values may help providing interesting insights.
The results in panel A of table 1 indicate that neither of the abnormal discretionary accruals estimates has a mean that is significantly different from zero. This is consistent with both prior literature and my expectation, as I do not expect accruals-based EM in any specific direction surrounding a routine CEO transition. At the same time, this leaves the possibility open that I find support for accruals-based earnings smoothing, for which I conduct an additional test later on.5 I find some indication for income-decreasing abnormal discretionary expenditure in year t= +1, significant at the .10 level. This may be considered somewhat surprising. Subsequent tests will show whether or not this is related to a CEO transition.
I find weak evidence for income-increasing DAJO in year t= +1 following type 2 transitions, see panel B. Although mean DADE and DAJO are negative in the transition year, they are not significantly different from zero. I find strong evidence for income-decreasing DISEX in both year t= -1 and year t= 0.
5 As both ABSDADE and ABSDAJO are significantly different from zero, there is reason to assume that
abnormal discretionary accruals are used in type 1 transitions, but that the positive ones cancel out the negative ones in my sample, resulting in actual mean estimates of abnormal discretionary accruals being not significantly different from zero.
Table 3: Earnings management indicators by CEO transition type
EM Proxy DADE DAJO DISEX ABSDADE ABSDAJO ABSDISEX
Panel A: Type 1 CEO transitions: Peaceful departure & internal succession
year t= -1 (n= 40) -,002 ,002 ,005 ,023 ,026 ,016 p-val. ,623 ,747 ,318 ,000 ,000 ,001 year t= 0 (n= 40) -,004 -,002 ,008 ,027 ,021 ,018 ,452 ,635 ,150 ,000 ,000 ,000 year t= +1 (n= 40) -,001 -,002 ,011 ,021 ,025 ,020 ,803 ,679 ,055 ,000 ,000 ,000
Panel B: Type 2 CEO transitions: Peaceful departure & external succession
year t= -1 (n= 31) ,006 -,008 ,020 ,025 ,024 ,026 p-val. ,429 ,168 ,011 ,000 ,000 ,001 year t= 0 (n= 31) -,002 -,007 ,023 ,041 ,029 ,030 ,837 ,384 ,014 ,000 ,000 ,001 year t= +1 (n= 31) ,011 ,021 ,012 ,038 ,038 ,030 ,248 ,084 ,140 ,000 ,001 ,000
Panel C: Type 3 CEO transitions: Forced departure & internal succession
year t= -1 (n= 30) ,012 ,034 ,033 ,039 ,048 ,040 p-val. ,339 ,065 ,050 ,001 ,008 ,014 year t= 0 (n= 30) -,029 -,026 ,047 ,061 ,043 ,054 ,235 ,044 ,059 ,007 ,000 ,028 year t= +1 (n= 30) ,026 ,023 ,034 ,041 ,047 ,054 ,027 ,135 ,059 ,000 ,000 ,026
Panel D: Type 4 CEO transitions: Forced departure & external succession
year t= -1 (n= 30) -,005 ,005 ,015 ,073 ,059 ,038 p-val. ,814 ,750 ,220 ,000 ,000 ,001 year t= 0 (n= 30) ,011 -,011 ,016 ,051 ,031 ,043 ,517 ,181 ,191 ,001 ,000 ,001 year t= +1 (n= 30) -,038 -,018 ,020 ,067 ,067 ,043 ,144 ,351 ,130 ,006 ,000 ,000
Panel E: Difference tests for each EM proxy by CEO transition type
F-stat. year t= -1 ,391 2,094 1,323 5,601 2,748 1,569 p-val. ,759 ,104 ,270 ,001 ,046 ,200 F-stat. year t= 0 1,138 1,563 1,627 1,456 2,004 1,725 ,336 ,202 ,187 ,230 ,117 ,165 F-stat. year t= +1 3,708 2,196 ,586 2,666 3,020 1,576 ,013 ,092 ,625 ,051 ,032 ,199
