Amsterdam Business School
Does manager change affect the bottom line?
The effect of CFO replacement on earnings management
Name: Tanja Langendijk Student number: 10331506
Thesis supervisor: dhr. prof. dr. J.F.M.G. (Jan) Bouwens Date: 16 June 2016
Word count: 15.358
Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Tanja Langendijk 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.
2
Abstract
The purpose of this research is to examine the potential relation between replacement of Chief Financial Officers (CFOs) and earnings management. I argue that CFOs have both the opportunity and incentives to manage earnings. Whereas opportunity arises from their responsibility with financial reporting, this research identifies several incentives for CFOs in a certain phase of CFO replacement to manage earnings. Whereas I predict that both leaving CFOs and CFOs in their first full year manage earnings upwards, I predict CFOs in the year of appointment to manage earnings downwards.
With this research, I relay on the distributional approach and use two methods for testing the hypotheses: Murphy (2000) and Bouwens and Kroos (2011). Contrary to my predictions, I found no clear separate relation between CFO replacement and earnings management. However, results do provide some indication for a relation between both. A potential explanation for not finding the predicted relations is the introduction of the Sarbanes-Oxley Act in 2002 as this increased cost of misreporting for both CEOs and CFOs. Nevertheless, findings of the Murphy (2000) models suggest managers to be still willing to manage earnings.
Contents
1 Introduction ... 5
2 Literature Review ... 7
2.1 Why can CFOs manage earnings? ... 7
2.2 CFO replacement ... 8
2.2.1 Leaving CFO ... 8
2.2.2 New CFO in year of appointment ... 13
2.2.3 New CFO in first full year ... 16
2.2.4 Sarbanes-Oxley Act ... 18 3 Research methodology ... 21 3.1 Earnings management ... 21 3.2 Variables ... 22 3.2.1 Undesired behavior ... 22 3.2.2 Year-to-date performance ... 24 3.2.3 CFO replacement ... 24 3.2.4 Control variables ... 25 3.3 Samples ... 26 3.4 Models ... 28
4 Analyses and results ... 30
4.1 Descriptive statistics ... 30
4.2 Main analyses ... 32
4.3 Additional analysis ... 38
5 Summary and conclusion ... 40
5.1 Summary ... 40 5.2 Conclusion ... 43 5.3 Limitations ... 43 5.4 Further research ... 43 6 References ... 45 7 Appendix ... 48
4
Figures and Tables
Figure 3.1 Predictions on the effect of CFO replacement on earnings distributions 23
Table 2.1 Hypotheses and supporting literature ... 20
Table 3.1 Samples ... 27
Table 3.2 Predicted signs ... 29
Table 4.1 Descriptive statistics Murphy (2000) approach ... 30
Table 4.2 Descriptive statistics Bouwens and Kroos (2011) approach ... 31
Table 4.3 Regression Murphy (2000) approach ... 34
Table 4.4 Regression Murphy (2000) approach: subsamples ... 35
Table 4.5 Regression Bouwens and Kroos (2011) approach ... 37
Table 4.6 Regression Murphy (2000) approach: new CEO has assumed office ... 39
Table 7.1 Data collection Murphy (2000) approach ... 48
Table 7.2 Data collection Bouwens and Kroos (2011) approach ... 49
Table 7.3 Correlation matrix Murphy (2000) approach... 50
1
Introduction
In this paper, I examine the extent to which the replacements of Chief Financial Officers (CFOs) are associated with reported earnings. I examine specifically whether CFOs are inclined to use income-increasing earnings management both in the year they leave office and in the first full year after appointment, and to use income-decreasing earnings management in the year of appointment.
Zorn (2004) provides evidence on the increased importance of and reliance on Chief Financial Officers (CFOs) by firms. His sample period ran from 1963 to 2000 and during this period, the percentage of firms having a CFO in place rose from zero to eighty percent. At first, corporate finance was more of a back-office function but as its role developed, it became a key function within firms.
This research follows up on the study of Geiger and North (2006), which studies the influence of the replacement of a CFO on the amount of earnings management within a firm. The primary responsibility of CFOs is with financial reporting (Jiang, Petroni, & Wang, 2010). Therefore, managers in such positions are expected and found to have the opportunity to manage earnings. Ge, Matsumoto, and Zhang (2011) provide evidence of personal characteristics to have an influence on accounting decisions being made. CFO replacement may have an influence on earnings management since different CFOs may choose differently. Finally, Geiger and North (2006) found a significant negative relation between CFO replacement and discretionary accruals, which led them to conclude that the amount of earnings management is significantly decreased after a CFO is replaced.
With this research, I try to answer the following question: Is the appointment of
a new CFO associated with changes in earnings distributions? For the purpose of this
study, I use the distributional approach by relying on two different methods: Murphy (2000) and Bouwens and Kroos (2011). I have identified three thresholds around which predictions have been tested. First, I predict firms featuring a leaving CFO to show better performance as such a CFO faces the horizon-problem (Dechow & Sloan, 1991), CFO compensation is often tied to earnings and a leaving CFO seems to have an incentive to perform well if he plans to stay attached to the firm after retirement. The second threshold is the year during which a new CFO is appointed. Incentives for such a CFO to use income-decreasing earnings management include both contracting
6
motivations, as earnings-based compensation schemes are often fully or partly deferred to the first full year (Wilson & Wang, 2010), and expectations with respect to performance to be managed downwards (O'Glove, 1987). Finally, I expect firms featuring a new CFO in his first full year to show better performance as CFO compensation is again related to earnings. However, a new CFO may face adverse incentives as his perspective is more long-term and the reversing nature of earnings management may lead future earnings, and along with them compensation, to be decreased.
Findings do provide some indication for a relation between CFO replacement and earnings management as I found a joint effect of year-to-date income and CFO replacement with the Murphy (2000) models. However, contrary to my predictions, I do not find a clear separate relation between both, which might be explained as high compliance towards the Sarbanes-Oxley Act (SOX). Nevertheless, I do manage to find evidence that managers are still willing to manage earnings regardless of their expected tenure and I do this with an undisputed proxy for earnings management.
This research contributes to prior literature in several ways. First, as the amount of accrual-based earnings management seems to have decreased after the implementation of SOX, the question arises whether CFO replacement still has an influence on earnings management. However, no pre- and post-SOX comparison is being made, further research into this area might therefore be interesting. Second, whereas prior literature tends to measure earnings management based on the criticized modified cross-sectional Jones’ model (McNichols, 2000), this study uses the distributional approach to operationalize earnings management. Finally, this research is of practical relevance as it indicates what consequences CFO replacement might have on earnings management.
The remainder of this paper is organized as follows. In the second section I discuss why CFOs might have both the opportunity and incentives to manage earnings based upon prior literature. In this section I identify three phases during the CFO replacement process. In the third section, I discuss the research methodology. Subsequently, I discuss the analyses that have been performed based on both the Murphy (2000) and the Bouwens and Kroos (2011) approach. This section includes the descriptive statistics, the results and additional research that has been performed. Finally, in the fifth section I provide a summary and a conclusion.
2
Literature Review
In this section I will discuss why CFOs may have the opportunity and incentives to manage earnings. I understand earnings management as the act of using the discretion in accounting standards in order to present managers’ performance favorably. I will argue that CFOs have both the opportunity and incentives to manage earnings given that they are responsible for the design and implementation of financial systems, and often their personal wealth depends on financial results.
2.1 Why can CFOs manage earnings?
This research focuses on the effect of replacing CFOs on earnings management. The role of finance seems to have gained in importance over the past decades as evidenced by Zorn (2004). His sample period ran from 1963 to 2000 and during this period, he found that the percentage of firms having a CFO in place rose from zero to eighty. Corporate finance seems to have developed from a back-office function to a key function within firms. Zorn (2004) provides some explanations for the role change of the CFO, of which the most important one seems to be the passage of FASB 33 in 1978 that changed and extended some reporting requirements. It has been suggested that those changes resulted in a development in the role of finance. The author expected and found a significant increase in the number of CFO positions after 1978, in response to the passage of FASB 33. This increase was enhanced by the mimetic behavior of firms resulting from the popularity of this managerial position.
The primary responsibility of the CFO is with the design and the implementation of the financial system (Mian, 2001). This responsibility suggests such managers to have an opportunity to influence both the extent to which a company uses earnings management and how it is being used. This relation has been evidenced by the research of Jiang, Petroni, and Wang (2010). Their study shows that CFOs both have an indirect effect on earnings management through the CEO, who may “pressure” the CFO into using earnings management for his personal gain, and a direct effect as the CFO may also benefit from earnings management himself. Evidenced by the increased focus on CFO responsibility for financial reporting of SOX, regulators also seem to acknowledge this influence.
I argue that managers in the role of CFO have the opportunity to personally influence the extent and type of earnings management that is used within a firm. Ge,
8
Matsumoto, and Zhang (2011) studied whether personal characteristics of CFOs have an influence on the accounting choices being made. In the traditional view, personal characteristics wouldn’t influence the firm’s accounting choices. Contrary, upper echelons theory suggests that personal characteristics do affect decision making. This theory suggests that CFOs do not necessarily make the same accounting choices under similar circumstances, rather, their choices are influenced by their personal characteristics. However, it is questionable whether this influence has economic significance such that it can be supported with significant evidence. Besides, personal characteristics may be dominated by the CEO as he might set most of the one at the top such that these are not influential. In support of the upper echelons theory, the paper, yet, finds that CFO characteristics do have an influence on accounting choices. Therefore, I assume that replacing the CFO affects decision making with respect to earnings management.
Some studies, such as that of Arthaud-Day, Certo, Dalton, and Dalton (2006), found that restating firms have a higher likelihood of CFO termination, both in the pre-SOX period and in the post-pre-SOX period. This might indicate some influence of CFOs on earnings management as they are replaced more often if earnings are restated thereby indicating some relation to be existent between both.
2.2 CFO replacement
This section is divided into three subsections that discuss different phases in the CFO replacement process: the year of leave, the year of appointment and the first full year of the new CFO. I argue CFOs in different phases of the CFO replacement process to have different incentives which might trigger them into using earnings management. Each subsection describes the incentives identified for CFOs under such circumstances to manage earnings, substantiated with prior literature.
2.2.1 Leaving CFO
When a new CEO is selected, the prior CEO usually continues to serve as chairman of the board of directors for a few years.
(Vancil, 1987, p. 57) This subsection discusses the incentives identified for a CFO upon the end of his tenure. I have identified two main incentives for a CFO in this phase to manage
earnings: (1) personal wealth and (2) career opportunities. The latter can be divided in the last resort for maintaining the position, cross-industry career opportunities and staying attached to the firm in another position which is related to the opening citation of this subsection. As argued previously, I understand earnings management as the act of using the discretion in accounting standards in order to present managers performance favorably.
First, personal wealth relates to the contracting motivations as described by Healy and Wahlen (1999). Motivations for managing earnings can be categorized into contracting, capital market and regulatory motivations. Contracting motivations includes both lending contracts and management compensation contracts. Lending contract motivation indicates that firms manage earnings in order to avoid debt covenants whereas the management compensation contract motivation indicates that managers manage earnings in order to increase personal wealth as their executive compensation is often earnings-based. Capital market motivations is about using earnings management in order to influence short-term stock price performance. Evidence indicates that this might be done by firms prior to IPOs, in order to meet expectations of analysts and shareholders, and to influence these expectations. This category is also described by Dechow and Skinner (2000). Finally, firms may also have regulatory motivations to manage earnings, these motivations might be motivated by industry-specific, anti-trust, or other regulations. The latter two categories seem less relevant for the personal incentives of a leaving CFO to use earnings management.
The management compensation contract seems to be the primary motivation for CFOs as earnings-based compensation schemes makes earnings management advantageous to managers, it provides them with incentives to show better performance and thus to behave in a way shareholders would not want them to as this might destroy value. In contrast, Graham, Harvey, and Rajgopal (2006) find that maintaining bonuses were considered unimportant motivations for meeting or beating earnings benchmarks. Managing earnings upwards now should result in lower earnings during future periods due to the reversing nature of earnings management. This would be disadvantageous to the manager as future personal wealth would be affected negatively. However, this disadvantage is of less relevance at the end of tenure as future periods are of less interest to leaving managers.
10
Related to the latter consideration, prior literature (Dechow & Sloan, 1991; Jain, Jiang, & Mekhaimer, 2015) acknowledges that there exists a horizon problem with persons in managerial positions. This horizon problem clearly relates to the agency theory and the theory of the firm which are based on the separation between ownership and control in corporations. Ownership is usually with shareholders whereas the control on daily operations is with management, among whom the CEO and the CFO. Management and shareholders often encounter conflicts of interest: whereas shareholders care about the firm and often have a long-term perspective, management is generally viewed as self-interested, cares for the short-term and therefore tends to show undesired behavior that destroys rather than creates value (Murphy & Jensen, 2011). Due to these conflicting interests, shareholders demand information in return for the capital and decision rights that have been provided to the management. Management has an informational advantage compared to the shareholders of the firm as the latter is usually not aware of what really happens within the firm.
Conflicts of interest seem to be even more present at the end of management’s tenure as I have argued previously. This would explain the horizon problem, which assumes that if the CEO is about to leave the firm, he has an incentive to increase short-term performance as future financial performance is irrelevant to him. Management compensation is often based on some incentive contract which often contains an earnings-related part. Murphy and Jensen (2011) show that bonus plans can provide incentives to manipulate earnings and show how these counterproductive incentives might be reduced. One of the solutions mentioned by them is using externally based performance measures such as relative performance against a peer group. Whereas internally based performance measures can be manipulated by either influencing performance or influencing the benchmark, externally based performance measures can be influenced less easily. This is evidenced by Murphy (2000) as this author argues and finds that internally determined performance standards are more easily influenced and are more often found to be related to income smoothing. Moreover, he indicates seventy-seven percent of the sample firms relied on such internal standards and only twelve percent on a mixture of internal and external standards. This indicates that earnings-based compensation schemes are generally not designed to avoid undesired behavior.
A long horizon deters the CEO from using income-increasing earnings management as the reversing nature of earnings management would result in loss of earnings in future periods (Dechow & Sloan, 1991). Future performance is of no relevance when the CEO is leaving as he won’t be attached to the firm anymore in those future periods and compensation is often primarily related to short-term rather than long-term performance. I argue that the same logic applies to CFOs as their compensation is often tied to earnings in a similar fashion. Therefore, I expect that leaving CFOs will show undesired behavior as shareholders do care for the future performance and want management to behave in such a fashion. Dechow and Sloan (1991) found evidence of this horizon problem: they found that leaving CEOs tend to manage discretionary investment expenditure to improve short-term earnings performance. Jain, Jiang, and Mekhaimer (2015) also provide evidence on this horizon problem and on how this might be mediated by internal governance. They found that internal governance, operationalized as the difference between the horizon of the CEO and of his subordinates, is positively related to liquidity when the CEO is close to retirement. A great difference in horizon suggests that the subordinates are more incentivized to monitor the actions of the CEO as CEOs facing a shorter horizon have a short-term perspective whereas the subordinates face a longer horizon and, therefore, seem to care more for the longer term. It is found that better internal governance improves the transparency of information, thereby reduces the information asymmetry and increases the liquidity of the firm. Internal governance may therefore be regarded as a mediating factor in the horizon problem of CEOs. I argue that this horizon problem is also present for leaving CFOs as they face similar earnings-based compensation schemes which may trigger them into using earnings management in a similarly.
The second incentive I have identified, career opportunities, is to be divided into three situations. First, the situation in which showing good performance is beneficial is the last resort to maintain the current position, this incentive is also acknowledged by Geiger and North (2006). CFO replacement may be either voluntary or involuntary. Voluntary replacement can be either due to routine replacements or to recruitment by another firm such that the CFO faces new job opportunities whereas involuntary replacement is often disciplinary (Mian, 2001). Whereas replacement creates incentives to use income-increasing earnings-management, CFOs facing involuntary
12
replacement and under recruitment by another firm might face another incentive to show strong financial performance at the end of their term. The involuntarily leaving CFO may want to use this last year to show his capabilities in order to maintain his current position. Second, the situation may arise that the manager is to be recruited by other companies and is therefore incentivized to show good performance. Good performance poses managers to better outside career opportunities than bad performance as competitors will use such information in their selection of candidates. CFOs therefore believe repeatedly failing to meet earnings benchmarks may affect intra-industry mobility: their ability to find a new job after replacement (Graham, Harvey, & Rajgopal, 2006). Both situations, voluntary and involuntary, create an additional incentive to show better performance, either through actually improving performance or through managing earnings. Although the motivations behind showing good performance are different in both situations, both provide incentives to show better performance. Therefore, it may not be necessary to differentiate between either types of replacement and this might explain why Geiger and North (2006) do not differentiate between these either.
Finally, the last situation related to the career opportunity incentive assumes that the CFO will stay attached to the firm by taking on another position within the firm, and relates to the opening citation of this subsection. This is related to the incentive to maintain the current position or to improve outside career opportunities as the current manager might show his capabilities by showing good performance and benefit from this in his further career. Upon the end of his tenure, the CFO may be willing to stay attached to the firm such that he benefits from showing good performance, also at the end of his term, as the company is probably more eager to continue the relationship if the leaving CFO performs well. Vancil (1987) discusses the type of succession in which the CEO “passes the baton”. In this case, the person currently covering the position of CEO will retain another position within the company, such as chairman.
The fact that the leaving CEO resigns from his current position but stays attached to the firm, incentivizes that person to show good performance. Again, I argue CFOs to have similar incentives. Although the CFO loses his current managerial position within the firm, this does not necessarily mean that he’ll immediately leave the firm. Another position may be occupied by him in the near future such that the CFO stays closely attached to the firm. It may therefore be beneficial to the person to show
good performance. For CEOs, the succession process is often designed such that the CEO will only “pass the baton”, transfer his managerial title, if he’s confident that the successor is capable of taking the position. If such a succession process would also exist for CFOs, this would create another incentive for performing well at the end of the tenure. However, existence of a similar succession process for CFOs is less likely as the CEO position is more senior and a good succession process is therefore more important. Yet, it is to be expected that not all CFOs leave immediately after resigning from their current position.
Substantiated by prior literature, I predict that firms will show better performance through undesired behavior such as earnings management if their CFO resigns from his current position. This leads to the following hypothesis:
H1a: In the year that the CFO leaves office, the firm reports abnormally high earnings.
2.2.2 New CFO in year of appointment
This subsection discusses the incentives a newly appointed CFO may face in the year of appointment. I have identified two main incentives in this phase of CFO succession: (1) facilitating to show improved performance in future periods and (2) earnings-based compensation schemes are often deferred to the first full year. This latter incentive relates to the previously described personal wealth incentive and management compensation contracts motivation. These incentives relate to artificial changes in reported earnings, referred to as Big Bath accounting. However, as CFO retirement is often disciplinary, it should be acknowledged that some real changes will also take place upon appointment of a new CFO.
As argued previously in relation to the incentives for a leaving CFO, retirement may be either voluntary or involuntary. Mian (2001) provides evidence that CFO turnovers are most often disciplinary, as the preceding three-year period financial performance seems to be declining. Substantiated with this evidence, it can be expected that performance will change after the appointment of a new manager as change is clearly demanded by the company. Poor performance by a predecessor might require more “cleaning up the mess” than actually being able to enhance performance. Really changing something takes time such that I expect the performance in the first year of appointment not to increase significantly, it might even
14
decrease. Financial performance will likely decrease as previous performance was abnormally high, based on the incentives described in relation to the leaving CFO which led me to predict that the firm would show abnormal good performance.
However, part of the performance of a new CFO might not be explained by the real change that is required. Big bath accounting is the situation in which management takes great expenses and defers revenue to future periods in order to show worse performance in the current period. Earnings are managed downwards in order to exaggerate how bad the situation is. This type of undesired behavior is often incentivized by the existence of bonus thresholds: if management expects it can never attain the threshold for a given year, they might benefit from deferring current earnings to future periods. In the case of CFO appointment, big bath accounting may benefit the new CFO such that it may lower expectations with respect to performance and that it can be used to lower targets for future periods. This relates to the first incentive I have identified as artificially low performance in the current period can be used to show improved performance in the future. Poor performance can be blamed on the predecessor CFO, his short-term based decisions may have led to reversals of earnings management and this might explain unsatisfying long-term performance. Whereas the predecessor CFO can be blamed for current performance due to both the reversing nature of earnings management and the “mess” he made, the successor CFO can take credit for improved performance in future periods (O'Glove, 1987). Besides for the purpose of being able to show performance improvements, big bath accounting is used to attain lower targets upon which managers will be evaluated and compensated. Lower targets are easier to attain, consequently it will be easier to obtain compensation tied to financial performance.
Whereas earnings-based compensation plans seem to provide new CFOs with a similar incentive to show good performance as I have identified for leaving CFOs, in practice this seems to work out differently. Prior literature states that earnings-based compensation is often deferred to the first full year of a CFO’s tenure (Wilson & Wang, 2010). This suggests that less negative personal wealth consequences are attached to the decrease in earnings and therefore provide another incentive to show worse performance. However, after studying the 2014 proxy statements of both Apple Inc. and Facebook Inc., it seems that these corporations compensate their new coming CFOs on a pro-rata basis calculated from their promotion date.
Ms. Ahrendts’ and Mr. Maestri’s bonus opportunities were pro-rated based on the time they served on the executive team during the year.
(Apple Inc., 2014, p. 34)
Mr. Wehner's target bonus was increased in the third quarter of 2014 from 50% of his base salary to 75% of his base salary in connection with his appointment as our Chief Financial Officer.
(Facebook Inc., 2014, p. 21) This suggests that compensation of newly appointed CFOs is also related to performance in the year of appointment, as for the leaving CFO. Yet, although decreasing earnings may have some effect on income as a proportion of their income may still be earnings-based, the absolute amount of income affected by this is less than it would be for a complete year as it is calculated on a pro-rata basis. The management compensation motivation, therefore, seems to be of less importance to a CFO in his first full year of tenure.
Vancil (1987) describes three tasks a new CEO faces in the first few months of tenure. These tasks include the following: managing the expectations, taking ownership of the strategic trust of the corporation, and building confidence by achieving realistic first year performance. On first sight, this does not seem to rhyme with the incentives described previously for new CFOs to manage earnings downwards. However, this income-decreasing earnings management is actually used to manage expectations as the expectations as well as targets are adjusted downwards. This does not seem to build confidence and trust, however, due to the reversing nature of earnings management, the decrease in performance can be blamed upon the predecessor CFO as was previously argued such that confidence and trust in the new CFO are less affected.
However, as Vancil (1987) describes several types of CEO succession, it can be concluded that not all new CEOs face the same incentives. Whereas CEOs replacing a retired CEO have a long tenure ahead, replacement CEOs may only face a short tenure as they are only appointed “interim”. Due to resignation of the previous CEO, the succession process with respect to selecting a successor may not be completed such that a replacement is appointed as long as no permanent successor has been found. This short tenure may result in the same horizon-problem as
16
mentioned for a leaving CEO (Dechow & Sloan, 1991). However, figures provided by the author indicate that such departures are only nine percent of the cases. Therefore, this group of managers facing adverse incentives seems not to be representative.
The following prediction takes both performance increasing and performance decreasing incentives for the CFO in the year of appointment into account. Substantiated by prior literature, I conclude that performance decreasing incentives weigh more heavily as the financial consequences related to this will be greater in the future than those of increasing short-term performance. This leads to the following hypothesis:
H1b: In the year a new CFO assumes office, the firm reports abnormally low earnings.
2.2.3 New CFO in first full year
The last phase I have identified in the CFO succession process is the newly appointed CFO in his first full year. Incentives faced by these CFOs are similar to those faced by the leaving CFO with respect to (1) personal wealth, except for the horizon problem that is present for leaving CFOs. An additional incentive for new CFOs facing a longer horizon is related to the (2) capital market motivations as described by Healy and Wahlen (1999).
After the year of appointment, the CFO faces similar incentives as any other CFO. Motivations for earnings management as described previously (Healy & Wahlen, 1999) may all influence the manager as he may be incentivized to enhance some party’s interests. As described previously, the most likely motivation to affect the decision making behavior of a CFO is the contracting motivation. Management compensation contracts may incentivize managers to manage earnings since compensation schemes are often tied to earnings. The CFO might increase earnings to show better performance, this relates to the previously motivated prediction about the behavior of a leaving CFO. However, the new CFO in his first full year of appointment does have less incentives to increase earnings due to the reversing nature of earnings management. Whereas leaving CFOs face the horizon problem, new CFOs will be personally affected by future performance as they, most probably, don’t plan on leaving the firm soon after their appointment as CFO. This is evidence by Vancil (1987) as he shows “interim” succession only to be the case in nine percent
of his observations. Newly appointed CFOs generally face a longer horizon such that future compensation is affected negatively if earnings are increased in the current period. Current income-increasing earnings management should lead to a decrease in earnings at some point in the future at which the new CFO most probably still occupies his current position. The focus of the new CFO will therefore be more long-term than was the case for a leaving CFO.
Besides the incentive related to the personal wealth of a new CFO, incentives arise to show good performance related to capital market motivations. Degeorge, Patel, and Zeckhauser (1999) show that earnings management is used around certain performance thresholds to exceed these thresholds. These relate to reporting positive profits, sustaining recent performance and meeting analyst’ expectations. According to Graham, Harvey, and Rajgopal (2006), earnings are assumed to be the most important performance measure as their informational content seems to be greater than that of any other measure. Short-term benchmarks are often related to earnings and CFOs seem to be willing to use earnings management in order to meet those benchmarks, even without worrying about the reversing nature of the management of earnings. It is therefore expected that a manager will try to show better performance in this phase of the CFO replacement process. This incentive might also be related to the personal wealth incentive as Matsunaga and Park (2001) found that CEO compensation is negatively related to failing to meet analyst forecast for at least two quarters such that an incentive arises to meet earnings forecasts. Applying this finding to CFOs, I would suggest that personal wealth may also depend on meeting or failing to meet earnings benchmarks.
Personal wealth incentives along with capital market incentives lead to the prediction that newly appointed CFOs show better performance in their first full year. However, the lack of the horizon problem may also make the CFO taking the longer term into account. Yet, this leads to the following hypothesis:
H1c: In the year next to the year a new CFO has assumed office, the firm reports abnormally high earnings.
18
2.2.4 Sarbanes-Oxley Act
Although I have identified several incentives for managing earnings, new regulation may serve as an explanation if I do not manage to find any relation between CFO replacement and earnings management. SOX, which was introduced in 2002, was implemented with one of the main objectives to restore the integrity of financial statements by restraining earnings management and accounting fraud (Cohen, Dey, & Lys, 2008). This regulation was intended to hold both CEOs and CFOs more accountable for the financial statements (Collins, Masli, Reitenga, & Sanchez, 2009). CEOs and CFOs now have to certify personally to the material accuracy and completeness of the financial information and disclosures released by the company (Geiger & North, 2006). Collins et al. (2009) summarized the main changes of SOX with respect to CFOs as follows:
(1) Section 302 requires CFOs to certify, among other things, the fairness of the presentation of the financial statements; (2) Section 906 imposes on CFOs potential criminal sanctions for misstated financial statements; and (3) Section 304 requires that CFOs disgorge any bonuses or stock market gains that result from misstated earnings.
(Collins, Masli, Reitenga, & Sanchez, 2009, p. 6) CEOs and CFOs are thus now both recognized to be significantly influential in what is reported by the company, as those can now be held personally responsible for it.
SOX increases the cost of misreporting for CFOs individually. The likelihood of firms misreporting is a function of the likelihood of getting caught and the cost of misreporting, both for the firm and for the individual itself (Indjejikian & Matejka, 2009). By increasing the cost of misreporting for the CFO, the likelihood of firms misreporting is thus decreased. Therefore, it was expected that accrual-based earnings management would decrease after SOX was implemented. Cohen, Dey, and Lys (2008) provide evidence for this as they found that the amount of accrual-based earnings management has decreased significantly in the post-SOX period whereas the use of real earnings management has increased. Accrual-based earnings management is the use of discretionary accruals, the discretion in accounting standards, to influence what is reported and is obviously related to what I understand to be earnings management: the act of using the discretion in accounting standards in
order to present managers performance favorably. Real earnings management is about actually changing what happens within the firm, the operating decisions, this type of earnings management actually changes cash flows instead of only what is being reported. I mainly expect CFOs to have an influence on discretionary accruals since their main responsibility is with financial reporting.
Geiger and North (2006) performed their research on the effect of CFO replacement on earnings management on a pre-SOX sample. Their sample covered the period from 1994 to 2000, therefore, their study may have been recognized as outdated as it did not provide evidence on the current situation at publishing date. Since SOX seems to ban out some amount of accruals-based earnings management, it is conceivable that the CFO, as well as other members of management, have less opportunity to manage earnings than before the introduction of this regulation. Therefore, I acknowledge that no relationship between CFO replacement and earnings management may be present due to the introduction of SOX. Table 2.1 includes the hypotheses regarding CFO replacement supported with prior literature.
20
Table 2.1 Hypotheses and supporting literature
Hypothesis Literature Argument
H1a: Leaving CFOs show better performance.
Healy and Wahlen (1999) Contracting motivations: management compensation
contracts.
Dechow and Sloan (1991) Contracting motivations: Horizon problem due to short
tenure.
Jain, Jiang and Mekhaimer (2015) Contracting motivations: Horizon problem due to short
tenure and internal governance may mediate this.
Murphy and Jensen (2011) Contracting motivations: Conflicts of interest: bonus
plans create incentives for earnings management which can be solved by using external standards.
Murphy (2000) Contracting motivations: Firms more often use internal
standards.
Mian (2001) Career opportunities: Voluntary turnover vs.
involuntary turnover.
Graham, Harvey and Rajgopal (2006) Career opportunities: Failing to meet earnings
benchmarks may affect intra-industry mobility. Also
provides a counterargument for contracting
motivations as maintaining bonuses is found not to be seen as a valid motivation for earnings management.
Vancil (1987) Career opportunities: Passing the baton.
H1b: Newly appointed CFOs show worse performance in the year of appointment.
Mian (2001) Replacements are often disciplinary such that real
change is expected. Cleaning up the “mess” of predecessor might actually lower performance before it can improve.
Healy and Wahlen (1999) Contracting motivations: management compensation
contracts.
O’Glove (1987) Big bath accounting can make it easier to show
improved future performance and to lower targets.
Wilson and Wang (2010) Earnings-based bonus scheme often deferred to first
full year.
Vancil (1987) Small amount of replacements are “interim”, relatively
long tenure ahead such that no horizon problem exists.
H1c: Newly appointed CFOs show better performance in the first full year.
Vancil (1987) Small amount of replacements are “interim”, relatively
long tenure ahead such that no horizon problem exists.
Healy and Wahlen (1999) Contracting motivations: management compensation
contracts, along with capital market motivations.
Degeorge, Patel and Zeckhauser (1999) Capital market motivations: performance thresholds:
report positive profits, sustain recent performance, and meet analyst forecasts, seem to lead to earnings management.
Matsunaga and Park (2001) Capital market motivations: CEO annual cash bonuses
are negatively related to failing to meet quarterly analyst’ forecasts.
Graham, Harvey and Rajgopal (2006) Capital market motivations: CFOs seem willing to use
earnings management in order to meet short-term earnings benchmarks and are not afraid of the reversing nature of it.
3
Research methodology
In this section I will define earnings management and discuss how I operationalize and measure earnings management for the purpose of this study. Subsequently I will discuss the remaining variables of interest along with the sample and models I use to test the hypotheses formulated in the previous section.
3.1 Earnings management
The definitions for earnings management that are used in the literature suggest that managers make use of judgment in order to manage earnings:
Earnings management is a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain.
(Schipper, 1989, p. 92)
Earnings management occurs when managers use judgment in financial reporting and in structuring transactions to alter financial reports to either mislead some stakeholders about the underlying economic performance of the company or to influence contractual outcomes that depends on reported accounting numbers.
(Healy & Wahlen, 1999, p. 368) These definitions explain that, and how, managers have the opportunity to exercise judgment as accounting regulation provides some room to decide on how standards apply to a certain financial event. In this paper I understand earnings management as the act of using the discretion in accounting standards in order to present managers’ performance favorably.
22
3.2 Variables
3.2.1 Undesired behavior
Following both Burgstahler and Dichev (1997) and Degeorge, Patel, and Zeckhauser (1999) I use the distributional approach for measuring earnings management1. This method aims at identifying specific distributions that are in line with predictions around some identified thresholds. This research focuses on whether CFO replacement leads to a change in the extent and type of earnings management, the distributional approach seems appropriate for this purpose. Whereas the usual distributional approach relays on distributions around a certain interval, this study uses OLS regression to see whether the different stages of CFO replacement are related to earnings management as operationalized through earnings behavior.
For the purpose of this study, I have identified three different thresholds that are used to test the hypotheses of this study. These include: the year in which the predecessor leaves, the year in which his successor is appointed, and the first full year of the new CFO. Based upon prior literature and the forthcoming hypotheses, I have made predictions about the behavior of earnings in narrow intervals around these thresholds. The first hypothesis states that firms that feature a leaving CFO show abnormally high performance, I expect this to be done by the use of income-increasing earnings management. Second is hypothesized that firms featuring a new CFO will show abnormally low performance, this might be both through real changes and through the use of income-decreasing earnings management in the year of appointment. Finally, the third hypothesis states that firms in the first full year of the new CFO will again show abnormally high performance, presumably by the use of income-increasing earnings management. These predictions are visualized in Figure 3.1.
Prior research (Das, Shroff, & Zhang, 2009) found that fourth-quarter reversals reflect earnings management efforts. This finding is to be expected as most compensation schemes are based on audited annual earnings instead of unaudited quarterly earnings. Management is incentivized to manage fourth quarter
1 Beneish (2001) and McNichols (2000) describe three methods for measuring earnings management:
(1) the discretionary accruals method, (2) the specific accruals method, and (3) the distributional approach. Whereas Geiger and North (2006) use the first, this method has been heavily criticized. The second method does not seem appropriate for the purpose of this research.
earnings as it is then estimable whether they will meet or fail to meet certain targets. According to O’Glove (1987), previously discussed big bath accounting tends to take place during this fourth quarter, this again indicates that earnings management is often present during the fourth quarter.
For the purpose of this study I follow two distinct methods to test my hypotheses. First, I follow the approach of Murphy (2000) which uses the fourth quarter share of net income as a proxy for earnings management. Second, I follow the approach of Bouwens and Kroos (2011). This approach uses the ratio of net income and the target of the last quarter as a proxy for earnings management. For the purpose of this study, quarterly earnings forecasts are used as the target since these provide a benchmark for comparing actual net income to what has been expected.
Figure 3.1 Predictions on the effect of CFO replacement on earnings distributions H1a: Leaving CFO
Q1 Q2 Q3 Q4
Earnings managed upwards in Q4
H1b: New CFO in year of appointment
Q1 Q2 Q3 Q4
Earnings managed downwards in Q4
H1c: New CFO in first full year
Q1 Q2 Q3 Q4
24
3.2.2 Year-to-date performance
Both methods include a year-to-date variable to account for the performance during the first three quarters since I expect earnings management to take place mainly during the fourth quarter, as visualized in Figure 3.1. Generally, managers are expected to smoothen income (Murphy, 2000). Therefore, I expect managers to manage earnings downwards if year-to-date performance is good, independent of the incentives identified during the phases of CFO replacement. Similarly, I expect managers facing bad year-to-date performance to manage earnings upwards in order to smoothen income. Das, Shroff, and Zhang (2009) refer to this as the reversal phenomenon. Consistent with Murphy (2000) and Bouwens and Kroos (2011), I predict the year-to-date variable to be negatively related to my proxies for earnings management.
This year-to-date variable thus indicates how good performance was during the first three quarters. Both methods, however, use a different year-to-date variable. For the Murphy (2000) approach I use a variable that indicates whether current income over the first three quarters was running ahead of prior year’s income (YTD). YTD is for this method equal to one if year-to-date income is good, such that current income is higher than prior year’s. The year-to-date variable included in the Bouwens and Kroos (2011) model provides a comparison between current net income over the first three quarters and the consensus forecast of that period, which is the average of all forecasts per company per period. For this method, YTD is equal to one if current income exceeds the consensus forecast.
3.2.3 CFO replacement
In addition to the incentives for a manager to smoothen income based on year-to-date performance, I expect earnings to be managed in some predicted direction given the current CFO replacement phase. Whereas I expect year-to-date performance to be negatively related to the proxies for earnings management, I expect the relation to be even more negative in the year in which a new CFO is appointed. Contrary, for leaving CFOs and new CFOs in their first full year, I expect the relation to be less negative or even positive. In order to test for this expected effect, I include three CFO replacement related dummy variables. CFOLEAVE is equal to one if the CFO is in his last year of tenure. CFONEW is equal to one if this is the year in which a new CFO gets appointed, the year of replacement. Finally, CFOFULL is equal to one if this is the first full year of
the new CFO in his current position. Additionally, I include interaction variables between the year-to-date variable and each of the CFO replacement variables. In line with my hypotheses, I expect the combined coefficient of the CFO replacement variable and the interaction variable for the CFOLEAVE, CFONEW, and CFOFULL models to be positive, negative, and positive, respectively.
3.2.4 Control variables
Following the research of Geiger and North (2006), I have included several control variables in the model based on Murphy (2000). The reason for Geiger and North (2006) to include such control variables is mainly based upon prior literature. In order to control for firm size, operating cash flow, and sales growth, I include the following control variables: Firstly, the log of market value of equity (MVE) is included in order to control for firm size. Secondly, I include the book-to-market ratio of equity (BM) to control for the growth-opportunities and this is expected to be negatively related to discretionary accruals. Thirdly, cash flows from operations are scaled by total assets (CFFO) and this is expected to have a negative relation with abnormal accruals. Fourthly, I include a variable to control for sales growth (GROWTH) which is expected to be positively related to discretionary accruals. Finally, the return of assets from the prior period (ROA) is included to control for prior performance which is expected to be positively related to the earnings management proxy.
These control variables, however, are based on a research (Geiger & North, 2006) that uses the Jones’ model and therefore discretionary accruals as its proxy for earnings management. Contrary, I use the distributional approach. Although the methods differ, both use proxies for earnings management. Therefore, I expect similar relations to be present in my sample. Differences between both methods explain my decision for excluding some of the control variables which were used by Geiger and North (2006), among which a variable to control for financial condition. Their reason for including that variable was directly related to a possible weakness of the Jones’ model such that I should not control for this. The other two control variables which I have excluded from the model are related to the financing structure and the acquisitions done by the company. Both FINANCE and ACQ were included in the research of Geiger and North (2006) to account for concerns about significant changes in the capital structure and the effect this might have on accruals. However, after
26
running some of the regressions, these did not seem to be significant in my model such that I chose to exclude the variables from my model.
As the Bouwens and Kroos (2011) approach compares performance within one year, it compares to a target rather than prior year’s income, rather than between years, there is no need to control for the abovementioned factors.
3.3 Samples
The samples initially consisted of US companies after 2002 as I have mentioned SOX as an alternative explanation and SOX is only applicable to US firms as off 2002. Besides, the availability of quarterly data is greater for US corporations as the SEC mandates quarterly reports, Form 10Q, for publicly trading corporations. Following Geiger and North (2006) as well as other prior discretionary accruals studies, financial service and utilities industry firms are excluded due to idiosyncratic and industry-specific financial reporting issues.
To retrieve the appropriate data, I have addressed four databases, namely: Compustat, ExecuComp, CRSP, and I/B/E/S. Respectively these provide financial, statistical and market information, information on executive compensation, information on security prices, and consensus and detail forecasts from security analysts. The main variables of interest, the variables with respect to CFO replacement, are retrieved from ExecuComp. ExecuComp mainly provides information as off 2006, this is related to the implementation of FASB 158 which improved guidance on reporting over related aspects. Therefore, the sample periods run from 2006 to 2014 and initially included 17,492 firm-year observations after which the ExecuComp sample was merged with the other datasets. Merging lead to final samples for the Murphy (2000) and the Bouwens and Kroos (2011) approach containing respectively 8,919 and 11,762 observations as shown in Table 3.1 as well as in Appendix Tables 7.1 and 7.2.
Table 3.1 Samples
Number of observations
ExecuComp sample 131,476
Less: not attributable to CFO 113,972
Less: years included twice2 12
Final ExecuComp sample 17,492
Less: observations lost due to merging datasets 8,573
Final sample Murphy (2000) approach3 8,919
Final ExecuComp sample 17,492
Less: observations lost due to merging datasets 5,730
Final sample Bouwens and Kroos (2011) approach4 11,762
This study relies upon calendar year data rather than fiscal year data, that is, I only include observations with data dates on the 31st of December such that fiscal years follow calendar year. Although this reduces the number of observations included in my samples, by only including such observations I avoid several difficulties. First, by using fiscal year data, economic shocks have a different effect on the results as these may impact a different period for different firms dependent on the fiscal year periods used by the firm. Secondly, due to the fact that I use data from different databases, it is very difficult to retrieve all data based on different fiscal years, i.e. stock prices for the market value of equity should be gathered at the end of the fiscal year, which would be different for each firm following a different fiscal year.
2 ExecuComp dataset included 12 double firm-year observations. In order to identify the right data I
used EDGAR Company Search and manually searched through the corresponding annual reports.
3 ExecuComp dataset is merged with Compustat and CRSP datasets such that observations are lost in
merging. The full development of final sample Murphy (2000) approach is included in Appendix Table 7.1.
4 ExecuComp dataset is merged with Compustat and I/B/E/S datasets such that observations are lost
in merging. The full development of final sample Bouwens and Kroos (2011) approach is included in Appendix Table 7.2.
28
3.4 Models
I use the following equations to test hypothesis 1a: Murphy (2000) approach:
𝑄4 𝑆𝐻𝐴𝑅𝐸𝑖 = 𝛼𝑖+ 𝛽1𝑌𝑇𝐷 + 𝛽2𝐶𝐹𝑂𝐿𝐸𝐴𝑉𝐸 + 𝛽3𝑌𝑇𝐷𝐿𝐸𝐴𝑉𝐸 + 𝛽4𝑀𝑉𝐸 + 𝛽5𝐵𝑀 + 𝛽6𝐷𝐼𝑆𝑇𝑅𝐸𝑆𝑆 + 𝛽7𝐶𝐹𝐹𝑂 + 𝛽8𝐺𝑅𝑂𝑊𝑇𝐻 + 𝛽9𝑅𝑂𝐴 + 𝜀𝑖
Bouwens and Kroos (2011) approach: 𝑆𝐴𝐿𝐸𝑆 𝑄4
𝑇𝐴𝑅𝐺𝐸𝑇 𝑄4𝑖 = 𝛼𝑖 + 𝛽1𝑌𝑇𝐷 + 𝛽2𝐶𝐹𝑂𝐿𝐸𝐴𝑉𝐸 + 𝛽3𝑌𝑇𝐷𝐿𝐸𝐴𝑉𝐸 + 𝜀𝑖
Both hypotheses 1b and 1c are tested using similar equations as presented above, however, the circumstances under which earnings management is expected are different such that the variables related to CFO replacement are replaced by either CFONEW and YTDNEW or CFOFULL and YTDFULL. As argued previously, I expect the relation between YTD and my proxies for earnings management to be negative, I therefore predict 𝛽1 to be negative. To test my hypotheses, regression coefficients 𝛽2 and 𝛽3 are evaluated such that their combined coefficient should be in line with my predictions. Therefore, I expect the combined coefficient for the year in which the CFO leaves to be positive, for the year in which the new CFO assumes office to be negative and for the first full year of tenure of the new CFO to be again positive. Predicted signs for the regression coefficients are reported in Table 3.2.
Table 3.2 Predicted signs
This table presents the predicted signs for the test and control variables. YTD is a dummy variable equal to one if current income over the first three quarters is good, either compared to prior year’s income over that period (Murphy (2000) approach) or compared to the consensus forecast of that period (Bouwens and Kroos (2011) approach); CFOLEAVE is a dummy variable equal to one if a CFO left this year; CFONEW is equal to one if a new CFO assumed office this year; CFOFULL is equal to one if a new CFO assumed office last year; YTDLEAVE is an interaction variable between YTD and CFOLEAVE; YTDNEW between YTD and CFONEW; and YTDFULL between YTD and CFOFULL; MVE is log of market value of equity; BM is book-to-market ratio of equity; CFFO is cash flows from operations scaled by total assets; GROWTH is sales growth over the last period; ROA is prior year’s return on assets.
Variable CFO replacement variable
Interaction variable Hypothesis Predicted sign Combined effect
YTD -
CFOLEAVE YTDLEAVE H1a +
CFONEW YTDNEW H1b - CFOFULL YTDFULL H1c + MVE - BM - CFFO - GROWTH - ROA +
30
4
Analyses and results
In this section I will discuss the analyses that I have performed along with their results. To be able to compare the descriptive statistics and the results of both approaches, I will discuss both methods before turning to the next part of the analyses.
4.1 Descriptive statistics
Table 4.1 Descriptive statistics Murphy (2000) approach
This table presents descriptive statistics on the dependent, test, and control variables. Q4SHARE is Q4’s share of net income; YTD
is a dummy variable equal to one if current income over the first three quarters exceeds prior year’s income over that period;
CFOLEAVE is a dummy variable equal to one if a CFO left this year; CFONEW is equal to one if a new CFO assumed office this year; CFOFULL is equal to one if a new CFO assumed office last year; YTDLEAVE is an interaction variable between YTD and CFOLEAVE; YTDNEW between YTD and CFONEW; and YTDFULL between YTD and CFOFULL; MVE is log of market value of equity; BM is book-to-market ratio of equity; CFFO is cash flows from operations scaled by total assets; GROWTH is sales growth over the last period; ROA is prior year’s return on assets.
Variables Obs. Mean Std. Dev. 10% 25% 50% 75% 95%
Q4SHARE 8,877 .2741 .1403 .1220 .2012 .2579 .3161 .5435 YTD 8,877 .6538 .4758 0 0 1 1 1 CFOLEAVE 8,877 .1127 .3162 0 0 0 0 1 CFONEW 8,877 .1001 .3002 0 0 0 0 1 CFOFULL 8,877 .0768 .2663 0 0 0 0 1 YTDLEAVE 8,877 .0725 .2594 0 0 0 0 1 YTDNEW 8,877 .0612 .2397 0 0 0 0 1 YTDFULL 8,877 .0499 .2178 0 0 0 0 0 MVE 8,877 14.8446 1.5014 13.0867 13.7447 14.6822 15.8072 17.5331 BM 8,877 .5086 .5879 .1501 .2767 .4497 .6758 1.1076 CFFO 8,877 .1036 .0812 .0178 .0531 .0919 .1414 .2402 GROWTH 8,877 .1174 .2743 -.0662 .0081 .0774 .1724 .4588 ROA 8,877 .1006 .0924 .0206 .0471 .0836 .1336 .2558
Table 4.1 presents descriptive statistics on the dependent, test, and control variables included in the Murphy-based (2000) model. With this approach I use fourth quarter share of net income as a proxy for earnings management. The mean of this earnings management proxy is positive, in line with prior literature (Cohen, Dey, & Lys, 2008). The year-to-date variable has a mean of 0.6538, which means that for 65.38% of my observations, current year’s income is running ahead of prior year’s income. The means for the three variables related to the CFO replacements – CFOLEAVE,
CFONEW, and CFOFULL – represent the percentage of CFOs under such
7.68% (682) of the firm-year observations represent respectively leaving CFOs, newly appointed CFOs, and CFOs in their first full year of tenure. The means of the interaction variables between CFO replacement and year-to-date performance represent the percentage of observations in which the firm encounters one of the phases of CFO replacement along with a good year-to-date performance. To control for the influence of outliers, I have winsorized the continuous variables. With winsorization, outliers are replaced by a less extreme value since those extreme observations can significantly affect the final results. In order to check whether the independent variables are highly correlated, Appendix Table 7.3 includes a correlation matrix. Except for the correlation between the interaction variables and the CFO replacement variables, no high correlation seems to be present between the variables of interest.
Table 4.2 Descriptive statistics Bouwens and Kroos (2011) approach
This table presents descriptive statistics on the dependent, test, and control variables. SALESTARGETQ4 is Q4’s ratio of net income and target; YTD is a dummy variable equal to one if current income over the first three quarters exceeds the consensus forecast of that period; CFOLEAVE is a dummy variable equal to one if a CFO left this year; CFONEW is equal to one if a new CFO assumed office this year; CFOFULL is equal to one if a new CFO assumed office last year; YTDLEAVE is an interaction variable between YTD and CFOLEAVE; YTDNEW between YTD and CFONEW; and YTDFULL between YTD and CFOFULL.
Variables Obs. Mean Std. Dev. 10% 25% 50% 75% 95%
SALESTARGETQ4 11,762 1.5081 71.3159 -0.4269 0.5323 0.9133 1.1238 2.6628 YTD 11,762 0.4052 0.4910 0 0 0 1 1 CFOLEAVE 11,762 0.1172 0.3217 0 0 0 0 1 CFONEW 11,762 0.1087 0.3112 0 0 0 0 1 CFOFULL 11,762 0.0823 0.2748 0 0 0 0 1 YTDLEAVE 11,762 0.0411 0.1986 0 0 0 0 0 YTDNEW 11,762 0.0381 0.1914 0 0 0 0 0 YTDFULL 11,762 0.0307 0.1725 0 0 0 0 0
Descriptive statistics for the dependent, test, and control variables included in the Bouwens and Kroos-based (2011) model are presented in Table 4.2. With this approach, I use the ratio of fourth quarter net income and the target as a proxy for earnings management. For the purpose of this study, I have measured the target as the consensus forecast for the appropriate period. The mean of this earnings management proxy is again positive, in line with prior literature (Cohen, Dey, & Lys, 2008). The year-to-date variable has a mean of 0.4052 which means that for 40.52% of my observations, actual performance exceeds the forecast. Approximately 11.72%
32
(1,379), 10.87% (1,278), and 8.23% (968) of the observations represent respectively leaving CFOs, newly appointed CFOs, and CFOs in their first full year of tenure. In all circumstances, this percentage is slightly higher than for the previously discussed Murphy-approach (2000) which might be explained by the fact that this sample includes more firm-year observations: 11,762 compared to 8,877. Appendix Table 7.2 includes the correlation matrix for the Bouwens and Kroos (2011) approach. As with the prior method, no high correlation is present except for the interaction variables with the CFO replacement variables.
4.2 Main analyses
In order to test the hypotheses of this research, I have ran several regressions. The results of those regressions are included in this subsection, results are used to either reject or accept the hypotheses.
The first approach I have used to test the hypotheses is the previously described approach of Murphy (2000), which includes the fourth quarter share of net income as dependent variable. The regression results for this approach have been included in Table 4.3. In order to correct for heteroscedasticity and to solve for cross-observation correlation in the variables, I have used robust and clustered standard errors in running the regression. Consistent with the findings of Murphy (2000), I find a negative and significant coefficient for the year-to-date variable which indicates earnings management to be present.
I have used the first model to test my prediction on incentives for a CFO to use income-increasing earnings management in his last year of tenure. As Table 4.3 shows, the regression coefficients of the replacement and the interaction variable are -0.0177 and 0.0109, respectively. However, only the first coefficient is significant such that I find no significant evidence for their combined effect on earnings management. The second model is used to test my prediction that income-decreasing earnings management to be used by a new CFO. Both regression coefficients for the CFO related variables are insignificant such that I find no evidence to support the second hypothesis. The third model provides significant evidence with respect to my prediction that income-increasing earnings management to be present in the first full year in office. However, as the coefficients of the replacement and the interaction variable are respectively 0.0221 and -0.0237, the combined effect is approximately zero (-0.0016).
This is not in line with my prediction, the slightly negative effect even provides some evidence for a contradictory relation as earnings seem to be managed slightly downwards in this phase of CFO replacement. Overall, the results in Table 4.3 are not in support of my hypotheses. Whereas I do not manage to find evidence for the CFO replacement to be related to earnings management, the joint effect of both year-to-date performance and the interaction variable between year-to-year-to-date performance and the CFO replacement seems to be significant as indicated by the F-score. This significance is present for all models which allows me to conclude that CFO replacement has some effect on earnings management.
In order to distinguish between situations in which managers may have different incentives to manage earnings, I have performed the same regression on two subsamples. The first subsample contains 5,804 observations in which year-to-date income is good, YTD is equal to one. The second subsample contains 3,073 observations for which YTD is zero, year-to-date income is negative. The results of the regression models are presented in Table 4.4. Contrary to what my predictions, I only manage to find significant coefficients with respect to the main variables of interest for the second subsample. Whereas I find a negative and significant coefficient (-0.0172) for the CFOLEAVE variable, I find a positive and significant coefficient (0.0232) for the CFOFULL variable. These results assume earnings to be managed slightly downwards if a CFO will leave the firm and current income is running behind and earnings to be managed upwards if a CFO is in his first full year of tenure. These findings provide contradictory evidence for hypothesis 1a such that I would have to reject this and significant evidence to support hypothesis 1c. However, this is only in the case if current income is running behind and it should be noticed that this coefficient is again only slightly positive.
