EARNINGS MANAGEMENT BEFORE AND AFTER CEO RESIGNATION ZHI QIANG NATIONAL UNIVERSITY OF SINGAPORE 2002 - I- EARNINGS MANAGEMENT BEFORE AND AFTER CEO RESIGNATION ZHI QIANG (Bachelor of Economics, Peking University) A THESIS SUBMITTED FOR THE DEGREE OF MASTER OF SCIENCE IN MANAGEMENT DEPARTMENT OF FINANCE & ACCOUNTING NATIONAL UNIVERSITY OF SINGAPORE 2002 - II - Abstract This study examines the earnings management behavior of outgoing and incoming CEOs. The Jones Model is used to hypotheses that outgoing CEOs do not engage in significant earnings management before resignation, and incoming CEOs manage earnings downwards in the quarter in which they are appointed as CEO and manipulate earnings upwards in the following quarter. The empirical evidence is consistent with my hypotheses. My findings on outgoing CEOs are not consistent with Pourciau (1992) and further tests suggest that the contradictory results are due mainly to the inherent bias in Pourciau’s research methodology. Key words: Earnings Management, Earnings Manipulation, CEO turnover, CEO Resignation, Total Accruals, Discretionary Accruals - III - ACKNOWLEDGEMENTS This academic thesis was completed with the help and support of the following people. I am thankful to my girlfriend, who is my dearest soul mate and provide me with the constant stream of motivation, encouragement and love. All these are the most intangible, but nevertheless the most needed and important element to give me the reason to keep going. I am grateful to my family, who lend me their support and encouragement unselfishly, and who love and care for me endlessly. Special thanks to my supervisor, A/P Michael Shih. The completion of this thesis could not have been possible if not for the expertise and guidance of him. I also appreciate his consideration and patience in guiding my life in many aspects. Last, but not least, I want to give my thanks to Mr. Lin Zhixing as well as other my fellow classmates and friends in NUS, who brought endless pleasure and support to me. Zhi Qiang Dec. 2002 - IV - TABLE OF CONTENT ABSTRACT III ACKNOWLEDGEMENTS IV TABLE OF CONTENT V LIST OF TABLES AND FIGURES VIII CHAPTER ONE: INTRODUCTION 1 CHAPTER TWO: LITERAT URE REVIEW 5 2.1 Introduction 5 2.2 Definitions of Earnings Management 5 2.3 Motivations for Earnings Management 6 2.4 Development of Research Models 9 2.5 Earnings Management Associated with CEO Turnover 11 2.5.1 CEOs’Incentives and Methods to Manipulate Earnings 11 2.5.2 Earnings Management Associated with CEO Turnover 12 Explanations of Earnings Management Associated with CEO Turnover 12 Classification of CEO Tu rnover and Earnings Management 14 Earning Management Or Poor Performances 15 2.6 Concluding Remarks 17 -V - CHAPTER THREE: HYPOT HESIS DEVELOPMENT 18 3.1 Limitations of Prior Research 18 3.2 Hypotheses for Ea rnings Management Before CEO Resignation 21 3.3 Hypotheses for Earnings Management After CEO Resignation 21 CHAPTER FOUR: SAMPLE SELECTION AND RESEARCH MODELS 23 4.1 Sample Selection 23 4.1.1 Timeline Surrounding CEO Resignation and Sample Selection 23 4.1.2 Descriptive Statistics for The Sample 24 4.2 Research Methodology 27 4.2.1 Total Accruals 27 4.2.2 Estimation of Discretionary Accruals 27 Estimation of Discretionary Accruals for Individual Sample Firm 27 Estimation of Discretionary Accruals by Pooled Sample 29 Testing Earnings Management by Pooled Sample on Aggregate Basis 30 Testing Earnings Management for Pooled Sample on a Quarterly Basis 31 CHAPTER FIVE: EMPIRICAL RESULTS AND DISCUSSION 32 5.1 Estimation of Discretionary Accruals by Individual Sample Firm 32 5.2 Estimation of Discretionary Accruals by Pooled Sample 37 5.2.1 Testing Earnings Management by Pooled Sample on Aggregate Basis 39 5.2.2 Testing Earnings Management by Pooled Sample on a Quarterly Basis 41 5.3 Additional Evidence 43 - VI - 5.3.1 The Healy Model 43 5.3.2 The DeAngelo Model 45 5.3.3 The Refined Jones Model 47 5.3.4 The Kang and Sivaramakrishnan Model 48 CHAPTER SIX: SUMMARY AND CONCLUSIONS 50 6.1 Summary of Findings 50 6.2 Implication of Findings 51 6.3 Contribution of Study 51 6.4 Limitations and Suggestion for Future Research 52 6.5 Conclusions 53 APPENDICES 55 - VII - LIST OF TABLES AND FIGURES List of Tables Table 1 Frequency distribution of published papers exploring different motivations of Earnings Management 7 Table 2 Selection of sample of CEO resignation, 1998-1999 24 Table 3 Descriptive statistics for the sample 25 Table 4 Mean discretionary accruals for each quarter surrounding resignation date 33 Table 5 Averages of coefficients in firm-specific regressions 35 Table 6 Estimation of discretionary accruals by pooled data 38 Table 7 Estimation of discretionary accruals by pooled data on aggregate basis 40 Table 8 Estimation of discretionary accruals by pooled data on quarterly basis 42 Table 9 The empirical results of Healy Model 44 Table 10 The empirical results for DeAngelo Model 46 List of Figures Figure 1 Timeline surrounding CEO resignation 23 Figure 2 Mean Discretionary Accruals by individual firm 32 Figure 3 Median Unexpected Accruals using my model and Pourciau’s model (firm-by-firm regressions) 36 Figure 4 Mean Discretionary Accruals by pooled data 38 Figure 5 Median Unexpected Accruals using my model and Pourciau’s model (pooled data regressions) 41 Figure 6 Mean Discretionary Accruals by Refined Jones Model 48 Figure 7 Mean Discretionary Accruals by Kang and Sivaramakrishnan Model 49 - VIII - Chapter One Introduction Chapter One: Introduction There has been intense interest in the management of earnings by managers/firms, especially after the accounting frauds committed by Enron and WorldCom were uncovered. Academic research exploring the issue has examined the circumstances under which managers/firms are expected to manipulate earnings and the direction of the manipulation under each circumstance, while employing increasingly reliable earnings management detection models. One of the circumstances under which earnings management is expected to occur is when there is a change in the CEO. Earnings management largely involves shifting earnings in future periods to the current period (borrowing earnings from the future) or deferring current earnings to future periods. Earnings management upward (downward) in one period, therefore, eventually will be offset by earnings decreases (increases) in the future. Thus, not every CEO has incentives to manage earnings in a particular direction in every quarter. Departing CEOs are a special type of CEOs, however. They are not expected to continue to hold their position long into the future, and therefore have nothing to lose in the future if they borrow earnings from the future by accelerating the recognition of earnings. Newly appointed CEOs may have incentives to manage earnings in a different direction. Since they are “new kids in town”, they probably will not be blamed for any short-term earnings weakness (“it’s the old CEO’ s fault”). New CEOs therefore are expected to manipulate earnings downward right after their appointment, and manipulate earnings upward later to claim credit for an earnings turnaround. - 1- Chapter One Introduction CEO turnover can be classified either as routine (retirement) or non-routine (resignation). This paper examines earnings management behavior before and after non-routine CEO turnover. While many studies have investigated earnings management associated with CEO turnover in general, very little is known about how firms manage earnings in periods before and after a CEO resignation. The only study that we are aware of is one by Pourciau (1992), which was published more than a decade ago. Her test results suggest that resigning CEOs take less accounting accruals and manage earnings downward before their departures. Pourciau herself finds the results surprising and perplexing, and surmises that they may be attributed to resigning CEOs having to report lower earnings in the year before their departure as a result of their manipulating income upward (in such a subtle way to avoid detection) in the preceding years. While that explanation is not entirely impla usible, one must question why the outgoing CEOs are so good at manipulating earnings upward in prior years only to suddenly lose their skill at doing so in the year preceding their resignation. Since Pourciau’ s study was conducted a long time ago and she could not avail herself of more sophisticated methodologies used in today’s detecting earnings management, her surprising test results could have been explained by flaws in her test design. Specifically, Pourciau (1992) detected earning management using the amount of total accruals (earnings minus operating cash flow). There are several problems with such approach: (1) failure to control for non-discretionary accruals; (2) failure to control for poor corporate performance, which usually precedes CEO resignations; (3) failure to control for assets write-offs/ write -downs, which are likely to be associated with poor performance but may be not discretionary. In addition, the sample size in her study was also relatively small. -2 - Chapter One Introduction In my study, I look for signs of earnings management in each quarter, employing a variety of models commonly used to estimate discretionary accruals, including the Jones model (1991) and refined Jones model. I find evidence that incoming CEOs manipulate earnings downward in the quarter in which they are appointed and manage earnings upward in the following quarter. Unlike Pourciau (1992), however, I find no evidence that resigning CEOs manipulate earnings in the four-quarter period before the date of resignation, either downward or upward. More importantly, I am able to replicate Pouciau’s surprising results by adopting her test design for my study. This is evidence that her results were primarily caused by flaws in her test design. My study contributes to the literature in several ways. First, Pourciau’ s (1992) finding that resigning CEOs manipulate earnings downward in the periods just before their departure is puzzling. My study reexamines this issue using more sophisticated methodology. The evidence from my study shows that these CEOs have no reasons to manipulate earnings downward in those periods to their own detriment. My study shows that it is very important for researchers to employ sophisticated models to detect earning management. Second, my study also examines the earnings management of the incoming CEO and finds that incoming CEOs manipulate earnings downward and upward in different periods after the departure of the old CEOs. This suggests that boards of directors should be alert to earnings manipulation in those periods. The remainder of the study is organized as follows: Chapter two presents an overview of the issue of earnings management and reviews the literature. Chapter three identifies several methodological deficiencies in prior research on earnings management by -3 - Chapter One Introduction resig ning and incoming CEOs, and develops the research hypotheses. Details of the empirical test, such as the data sources, sample selection procedure and research methodology, are presented in Chapter four. Chapter five reports and interprets the empirical test results. Finally, I summarize the findings of the study, discuss the implications of the results and suggest areas for future research in Chapter six. -4 - Chapter Two Literature Review Chapter Two: Literature Review 2.1 Introduction What is earnings management? What are the motivations for earnings management? What are the most developed research models in this field? This part of the thesis attempts to provide the answers. 2.2 Definitions of Earnings Management Although much research has been done in the area of earnings management, researchers find it is difficult to give a very clear definition of earnings management. The followings are several definitions given in the academic literature: “… a purposeful intervention in the external financial reporting process, with the intent of obtaining some private gain (as opposed to, say, merely facilitating the neutral operation of the process)…” Schipper (1989) (Page 95) “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 depend on reported accounting numbers.”Healy and Wahlen (1999) (Page 104) - 5- Chapter Two Literature Review “These statements imply that within -GAAP choices can be considered to be earnings management if they are used to “obscure” or “mask” true economic performance, bringing us again back to managerial intent.”Dechow and Skinner (2000) (Page 67) Although researchers have different definitions of earnings management, they are consistent in at least two aspects. First, the objective of earnings management is to mislead the stakeholders and to “obscure”the true economic performance of the firm. Second, managers can get benefits from earnings management. For example, managers can boost stock price, increase earnings-based bonus awards and avoid regulation by means of earnings management. I will discuss these motivations of earnings management in detail below. 2.3 Motivations for Earnings Management The motivations for earnings management are very important to researchers in this field. Only with a good understanding of the motivations for earnings management, can researchers hypothesize circumstances under which managers are most likely to manipulate accounting numbers, thus do their studies accordingly. According to prior research, there are at least three kinds of motivations for earnings management, including: (1) capital market motivation; (2) compensation (bonus) motivation; and (3) anti-regulation motivation. To characterize the recent earnings management research, a search of the academy journals in the 1993-2000 period was conducted in The Accounting Review, Contemporary Accounting Research, Journal of Accounting and Economics, Journal of Accounting, Auditing and Finance, Journal of Accounting and Public Policy, Journal of Accounting Research, and Journal of -6 - Chapter Two Literature Review Business, Finance and Accounting. The search identified 47 articles examining earnings management based on one of the three motivations. The statistics are reported in Table 1. Table 1: Frequency distribution of published papers exploring different motivations of Earnings Management Motivations for earnings management Number of articles % Capital market motivation 16 34.0 Personal bonus award motivation 14 29.8 Anti-regulation motivation 17 36.2 Total 47 100.0 Notes: A search of The Accounting Review, Contemporary Accounting Research, Journal of Accounting and Economics, Journal of Accounting, Auditing and Finance, Journal of Accounting and Public Policy, Journal of Accounting Research, and Journal of Business, Finance and Accounting for the period 1993-2000 was conducted. The search identified 47 articles examining earnings management based on one of the three motivat ions. As Table 1 shows, 34 percent of these research studies investigated the capital market motivation for earnings management. For example, Erickson and Wang (1999) investigated whether acquiring firms attempt to increase their stock price prior to a stock for stock merger in order to reduce the cost of buying the target and found that acquiring firms did manage earnings upward in the periods prior to the merger agreement. Teoh, Wong and Rao (1998) found evidence that initial public offering (IPO) firms, on average, have high positive issue-year earnings and abnormal accruals, followed by poor long-run earnings and negative abnormal accruals. They believed that the incentives to manage earnings might be especially strong when the firm is planning to sell shares to the market. Tech, Welch and Wong (1998) provided evidence -7 - Chapter Two Literature Review that seasoned equity issuers raise reported earnings, by altering discretionary accounting accruals, to mislead investors. 29.8 percent of these prior research studies have examined the compensation motivation for earnings management. For instance, Holthausen, Larcker and Sloan (1995) investigated the extent to which executives manipulate earnings to maximize the present value of bonus plan payments and found evidence consistent with the hypothesis that managers manipulate earnings downwards when their bonuses are at their maximum. In addition to the earnings management by top managers, Guidry, Leone and Rock (1999) examined bonus incentives and unexpected accruals at the business unit level and found that earnings management is used to increase business unit managers’earnings-based bonus awards. Finally, 36.2 percent of previous studies focused on the anti-regulation motivation of earnings management. Beatty, Chamberlain and Magliolo (1995) investigated how banks altered the timing and magnitude of transactions and accruals to achieve primary capital, tax, and earnings goals and satisfy the bank-industry regulatory constraints. Key (1997) examined unexpected accruals for firms in the cable television industry at the time of Congressional hearings on whether to deregulate the industry. Her evidence is consistent with firms in the industry deferring earnings during the period of Congressional scrutiny. Han and Wang (1998) tested whe ther oil firms that expected to profit from the 1990 Persian Gulf Crisis used accruals to reduce their reported quarterly earnings and, thus, political exposure. Their results are consistent with their hypothesis. -8 - Chapter Two Literature Review 2.4 Development of Research Models Besides exploring the motivations for earnings management, another fundamental task facing researchers is to build up a reliable model to measure management’ s discretion over earnings and to detect earnings management. Beginning with Healy (1985), many contemporary studies in this field have focused on accounting accruals. Accounting accruals are the “summary measure of the timing differences that result from all accounting choices” (Watts and Zimmerman, 1990). Total accruals, which are the difference between net income and cash flow from operations, can be divided into the discretionary accruals and nondiscretionary accruals. Total accruals are observable, while discretionary accruals and nondiscretionary accruals are not. So accounting researchers believe tha t the vital issue for testing earnings management is to find a way to measure nondiscretionary and discretionary accruals. Three main research methods have been developed to measure nondiscretionary and discretionary accruals. They are: (1) aggregate accruals method; (2) specific accruals method; and (3) frequency distribution method. The aggregate accrual method identifies discretionary accruals based on the relation between total accruals and hypothesized explanatory factors. For example, Healy (1985) and DeAngelo (1986) used total accruals and change in total accruals respectively, as measures of management’s discretion over earnings. Jones (1991) introduced a regression approach to control for nondiscretionary accruals, specifying a linear relation betw een total accruals and change in sales and property, plant and equipment. According to the variables used to estimate nondiscretionary accruals, this method can be further divided into: (1) the Healy Model; (2) the DeAngelo Model; (3) the Jones Model; (4) the Refined Jones Model; and (5) the Kang and Sivaramakrishnan -9 - Chapter Two Literature Review Model. The Jones Model is the most frequently used model by researchers, which suggests that the Jones Model is widely accepted as providing an adequate proxy for earnings management. The specific accrual method is another popular approach. It was first developed by McNichols and Wilson (1988). Unlike the aggregate accrual method, the specific accrual method often focuses on a certain industry in which a specific accrual or a set of accruals can reliably reflect discretionary and nondiscretionary accruals. For example, McNichols and Wilson (1988) used residual provision for bad debt as the discretionary accrual proxy. They estimated the residual provision for bad debt as the residual from a regression of the provision for bad debts on the beginning balance of the provision, and current & future write -offs. Another example is Beaver and Engel (1996), who used the residual allowance for loan losses as the discretionary accrual proxy. They estimated the residual allowance for loan losses as the residual from a regression of the allowance for loan losses on net charge -offs, loan outstanding, nonperforming assets and one -year ahead change in nonperforming assets. Other studies using specific accruals are Moyer (1990), Petroni (1992), Beaver and McNichols (1998), Penalva (1998), Nelson (2000) and Petroni, Ryan and Wahlen (1999). The frequency distribution method was developed more recently. It assumes that nondiscretionary accruals and reported earnings in the absence of earnings management should be distributed evenly around a specified benchmark, such as zero, prior quarter’ s earnings or analysts’forecast. The frequency distribution method then examines the statistical properties of reported earnings and identifies whether - 10 - Chapter Two Literature Review discontinuities around the benchmark, which suggest the exercise of discretion, exist. Studies using the frequency distribution method are by Burgstahler and Dichev (1997) and Degeorge, Patel and Zeckhauser (1999). 2.5 Earnings Management Associated with CEO Turnover 2.5.1 CEOs’Incentives and Methods to Manipulate Earnings CEOs’compensation contracts normally contain incentive provisions that link CEOs’ compensation to firms’ accounting-earnings performance. Therefore, researchers predict that the usage of these compensation contracts will induce CEOs to engage in earnings management to boost their salary and bonus. Prior studies have found empirical evidences consistent with these predictions. For example, Healy, Kang and Palepu (1987) examined the effect of accounting procedure changes on cash salary and bonus compensation to CEOs and obtained some indirect evidence. They found the salary and bonus payments to CEOs were dependent on reported earnings, rather than on those “true”earnings that are “uncontaminated”. According to prior studies, CEOs generally manipulate accounting earnings by discretionary accruals. They normally shift earnings in future periods to the current period (borrowing earnings from the future) or defer current earnings to future periods. To maximize the present value of their cumulative salaries and bonuses, CEOs may choose accounting discretionary accruals to balance the short-term and long-term benefits. Previous studies have provided evidence consistent with this. For instance, Holthausen, Larcker and Sloan (1995) investigated the extent to which executives manipulate earnings to maximize the present value of bonus plan payments and found - 11 - Chapter Two Literature Review evidences consistent with the hypothesis that executives manipulate earnings downwards when their bonuses are at their maximum. 2.5.2 Earnings Management Associated with CEO Turnover Since CEOs generally manage earnings by moving accounting numbers from one period to another period, researchers are very interested in how CEOs behave just before their departing the post of CEO and just after their assuming the post of CEO. The short horizon of CEOs’departure provides researchers a good opportunity to examine CEOs’earning management incentives. Explanations of Earnings Management Associated with CEO Turnover Three main explanations of earnings management associated with CEO turnover that have been examined in prior studies are: (1) Horizon problem. The horizon problem suggests that outgoing CEOs approaching a known departure date use accounting discretion to increase earnings and earningsbased compensation in their final years, at the expense of future earnings. Since maximizing the present value of their cumulative salaries and bonuses is the goal of the CEOs, researchers expect executives who put less value on future earnings than current earnings to have stronger incentives to improve short -term earnings performance. Typical executives who may place little value on future earnings are those who are expecting to leave their positions in the near future. Therefore, CEOs are most likely to manage earnings for short-term gains before they depart. Several - 12 - Chapter Two Literature Review previous studies have focused on earnings management associated with CEO turnover. Dechow and Sloan (1991) investigated the hypothesis that CEOs in their final years of office manage discretionary investment expenditures to improve short-term earnings performance. They found evidence that CEOs spent less on R&D during their final years in office. (2) Cover-up. Outgoing CEOs in firms with poor performance are threatened by termination, thus use accounting discretion to cover up the firm’s deteriorating economic performance. Weisbach (1988) suggested that the dominance of the board of directors by insiders might reduce the threat of CEO dismissal associated with poor reported earnings. Therefore, the composition of the board may influence the CEO’s incentives to manage earnings. (3) Big-bath. Incoming CEOs use accounting discretion to boost future earnings at the expense of transition-year earnings by writing off unwanted operations and unprofitable divisions. Weisbach (1992) examined the relation between management turnover and divestitures of recently acquired divisions. The empirical results indicated that the move to incomereducing accounting methods, the write-off of unwanted operations and the write-off of unprofitable divisions can be attributed to incoming CEOs who implicitly blame their predecessors for past performance. - 13 - Chapter Two Literature Review Classification of CEO Turnover and Earnings Management While many studies have investigated earnings management associated with CEO turnover in general, very little is known about how firms manage earnings in periods before and after a CEO resignation. The only study that we are aware of, which focuses on CEO resignation, is done by Pourciau (1992). It was published more than a decade ago. Pourciau (1992) classified CEO turnover as routine and nonroutine and examined evidence of earnings management associated with “nonroutine” executive changes. Her results are consistent with the hypothesis that incoming executives manage accruals in a way that decreases earnings in the year of the executive change and increases earnings in the following years. However, she failed to find evidence to support her hypothesis that outgoing executives manage accruals upward before their departure. On the contrary, her evidence indicated that outgoing CEOs manage earnings downward. One suggested reason is that the executive had successfully managed earnings, avoid ing termination for a number of years. Besides Pourciau’ s paper, Murphy and Zimmerman (1993) also examined earnings management associated with routine and nonroutine CEO turnover. However, they found no obvious earnings management. They documented the behavior of a variety of financial variables surrounding CEO departures and concluded that turnover-related changes in R&D, advertising, capital expenditures, and accounting accruals are mostly due to poor performance. They also found the managerial discretion appears to be limited to firms whose poor performance precedes the CEO’s departure. They found - 14 - Chapter Two Literature Review no evidence of managerial discretion in strongly performing firms where the CEO retires as part of the normal succession process. Earning Management Or Poor Performances Murphy and Zimmerman’s (1993) paper raised another issue regarding earnings management associated with CEO turnover: Are the unexpected accruals documented in prior studies indicative of CEOs’earnings management, or just the results of poor corporate performance? Prior researchers found CEO turnover often coincides with poor performance of firms. For example, Weisbach (1988) and Warner et al. (1988) documented that CEO turnover is preceded by adverse share-price and earnings performance. These studies indicate that CEO turnover is associated with poor performance of firms. The systematic poor performance preceding CEO changes confounds the interpretation of tests of earnings management in two respects: (1) poor performance preceding CEO replacement is likely to disguise attempts by the outgoing CEOs to boost earnings; and (2) the “big bath”associated with incoming CEO may represent a correction for the earnings boost by the former CEO or reflect a further deterioration in firm performance, rather than opportunistic behavior of the new CEO. These possibilities make it difficult to interpret the results as evidence of earnings management by outgoing and incoming CEOs. Therefore, to investigate earnings management associated with CEO turnover, one fundamental task is to control for performance preceding CEOs’ departure. Prior - 15 - Chapter Two Literature Review researchers have acknowledged this point and tried to control for poor performance. One well-known study that did so is Murphy and Zimmerman’s (1993). In their paper, Murphy and Zimmerman (1993) tried to control for the poor performance of firms in two ways: (1) Unlike prior studies, which typically focused on a single financial variable, Murphy and Zimmerman’s (1993) study examined eight financial variables jointly (research and development, advertising, capital expenditures, accounting accruals, earnings, sales, assets, and stock prices). Some of the variables (such as, R&D, advertising, capital expenditures, and accounting accruals) were assumed to be subject to considera ble managerial discretion, while others (such as sales, assets and stock-price performance) were assumed to be less discretionary and to reflect largely the performance of the firm. They then documented the behavior of these financial variables and considered the implications of simultaneous changes among the variables. Their findings indicate that the changes in R&D, advertising, capital expenditures, and accounting accruals surrounding CEO turnover are due mostly to poor performance, not management’s discretion over accruals. (2) Contrary to the prior studies’ assumption regarding the exogeneity of CEO turnover, this study allowed for endogenous CEO departures by using a system of simultaneous equations. To reduce the heteroscedasticity in the simultaneous system, ordinary least squares (OLS) and two-stage least squares (2SLS) were used. The empirical results suggest that, after controlling for firm performance and the endogeneity of CEO turnover, there is little evidence that CEOs use accruals to - 16 - Chapter Two Literature Review manage earnings around their departure date. In addition, the authors segmented the sample into subsamples in which departures were unrelated to performance, and concluded that managerial discretion is limited to performance-related CEO departures. 2.6 Concluding Remarks In summary, prior research studies have devoted considerable efforts to earnings management. Their research has defined the concepts of earnings management, explored the different motivations of earnings management and developed models to detect earnings management. Among all the available models, the Jones Model has been the most frequently used. Very few studies have investigated earnings management associated with CEO turnover, and even fewer have classified CEO turnover and conducted the fur ther study. This relative blank field give us incentives to do some research in deep. - 17 - Chapter Three Hypothesis Development Chapter Three: Hypothesis Development 3.1 Limitations of Prior Research While many studies have investigated earnings management associated with CEO turnover, few have examined specifically earnings management before and after CEO resignations. In a study by Pourciau (1992), CEO resignations are divided into two groups: forced resignations and voluntary resignations. The author hypothesizes that both groups of outgoing CEOs will manipulate reported earnings upwards before their resignations to increase their compensation prior to departure. She argued that while CEOs who resign voluntarily are in full control of the timing of their resignations, CEOs who are forced to resign are aware of their departure, and therefore manage earnings upwards in a bid to change the probability or timing of the forced resignations. The empirical results, however, seem to suggest that what transpires before CEO resignations is the exact opposite of what is expected. Not only does Pourciau (1992) fail to find evidence that more accounting accruals are taken in the year preceding CEO resignations, but accounting accruals in that period seem to be manipulated downward. Pourciau (1992) herself finds the results surprising and perplexing. She surmises that they may be caused by outgoing CEOs having to take less accounting accruals in the year before their departure as a result of their manipulating income upward (in such a subtle way to avoid detection) in the preceding years. While that explanation is not entirely implausible, it is unclear why the outgoing CEOs are so good at manipulating earnings upward in prior years only to suddenly lose their skill at doing so in the year preceding their resignations. The explanations for the surprising results, I believe, lie - 18- Chapter Three Hypothesis Development elsewhere. Specifically, I have identified several methodological deficiencies in the Pourciau (1992) study and they are as follows: (1) Failure to control for non-discretionary accruals. In her paper, Pourciau simply assumed a “random walk process”for earnings, accruals and cash flows to control for nondiscretionary accruals. As previously mentioned, this is the same methodology used by DeAngelo (1986). Dechow, Sloan and Sweeney (1995) have demonstrated the low earnings management detecting power of the DeAngelo model, compared with other more complicated models (such as, the Jones model). When Pourciau’s study was conducted, there was no well-specified model that can control accurately for nondiscretionary accruals. (2) Failure to control for poor corporate performance. Acknowledging the importance of controlling for corporate performance, Pourciau (1993) attempts to control for poor performance prior to CEO resignations by deflating earnings and accruals by contemporaneous sales. However this procedure is likely to fail because accruals will not change proportionately with sales. When sales go down by x%, accruals are likely go down by more than x%. For example, if sales deteriorate by 10%, working capital accruals would go down by roughly 10% as well, but depreciation and amortization expenses would largely remain the same. As a result, total accruals would decline more than 10%. Interestingly, before CEO resignations, sales are usually on the decline. As a result, total accruals deflated by sales is likely to show a downward trend before CEO resignations. This is exactly what Pourciau (1992) found. The downward trend exhibited by accruals deflated by sales would offset the CEO’ s manip ulation of - 19 - Chapter Three Hypothesis Development earnings by taking more accruals, and make such earnings management more difficult to detect. (3) Pourciau (1992) also examined write -offs/ write-downs recorded by firms before CEO resignations. However, write -offs/ write-downs are likely to be associated with poor performance and have little to do with earnings management. (4) Pourciau included special items in her study, which may not be the discretionary part of accruals. Bernard and Skinner (1996) argued that special items should not necessarily be viewed as discretionary. “For example, it is much less likely that gain on the sale of a subsidiary or gains and losses from lawsuits are discretionary.”(Page 319) Pourciau did mention that she also computed the total accruals excluding special items. However, choosing sales as the deflator makes her test results difficult to interpret. (5) Small sample size. Pourciau’s total final sample size was only 73. In addition to the inherent methodological deficiency, an increase in monitoring activities before CEO resignation may make it difficult for CEOs to engage in earnings management before they are forced to resign or voluntarily do so. Previous studies have documented a significant relation between firm performance and the probability of executive changes. For instance, Weisbach (1988) found that the probability of executive change in a firm that is in the lowest performance decile may range between 6 and 13 percent, while the probability of an executive change in a top-decile company is between 3 and 9 percent. This relation implies that firm performance is a signal for a - 20 - Chapter Three Hypothesis Development company to monitor the performance of the CEO and the poorly performing companies can expect that they are more likely to have CEO turnover and therefore increase their monitoring activities. 3.2 Hypotheses for Earnings Management Before CEO Resignation Given the controversial nature of outgoing CEOs’incentives to manipulate earnings, I propose my hypothesis in null form: H1 (Null Hypothesis): The CEOs who resign do not mana ge earnings upward before their resignations. 3.3 Hypotheses for Earnings Management After CEO Resignation While it is debatable whether outgoing CEOs have incentives to manipulate earnings upward or downward before their resignation, there is no significant disagreement over the prediction that the incoming CEOs have incentives to manage earnings downward. Vancil (1987) summed up the three critical tasks a new CEO must face: (1) managing the expectations; (2) taking ownership of the strategy of the corporation; and (3) achieving performance goals to build up confidence. Manage expectations and achieve performance goals are therefore crucial to incoming CEOs. It is suggested that the new CEOs could try to blame the outgoing CEO for poor performance and manage earnings upwards to meet performance goals later. Prior research studies provide support for this argument. Strong and Meyer (1987) found that new CEOs are more likely to make large discretionary write-offs to draw attention to the inferior decisions of prior CEOs. Pourciau (1992) found that new CEOs manage accruals downwards in the year of the - 21 - Chapter Three Hypothesis Development executive change and upwards in the following years. I incorporate prior studies’ insights and propose the second hypothesis (in alternative form): H2: Incoming CEOs manage earnings downward immediately after they are appointed and manage earnings upward in the following periods. - 22 - Chapter Four Sample Selection and Research Models Chapter Four: Sample Selection and Research Models 4.1 Sample Selection 4.1.1 Timeline Surrounding CEO Resignation and Sample Selection For all sample firms, the quarter during which a CEO resigns is defined as quarter 0 (Q0). The first quarter preceding quarter 0 is defined as quarter –1 (Q-1) and the first quarter after quarter 0 is defined as quarter 1 (Q1), etc. Figure 1 demonstrates these designations. FIGURE 1 Timeline Surrounding CEO Resignation Q-20 Q-4 Q-3 Q-2 Q-1 Estimation Period Q0 Q1 Outgoing CEO Q2 Q3 Q4 Incoming CEO Resignation Date A keyword search of the Compustat Executive Compensation database was conducted. This search revealed 454 companies that had a CEO retiring, resigning, or passing away over the sample period 1998-1999. Of these 454 companies, I identified 179 companies that had CEO a resigning during the sample period. I then deleted 37 companies because they were financial institutions and regulated companies. (Prior - 23- Chapter Four Sample Selection and Research Models studies suggest deleting financial institutions and regulated companies from data sample.) I obtain the quarterly accounting data from the Compustat Industrial Quarterly database. To be included in my sample, companies must have complete accounting data from quarter –20 (5 years before resignation date) to quarter 4 (1 year after resignation date). Since some firms failed to meet data availability requirements, the selection process resulted in a final sample of 126 firms. Table 2 summarizes the sample selection process. Table 2: Selection of sample of CEO resignation, 1998 -1999 Criteria Number of Firms Total CEO turnover in 1998-1999 454 Less: CEO retiring, passing away and other reasons (275) Financial institutions or regulated companies (37) Not complete accounting data (16) Final sample for regression analysis 126 Note: A keyword search of Compustat Executive Compensation database was conducted. This search revealed 454 companies that had a CEO retiring, resigning, or passing away over the sample period 1998-1999. Of these 454 companies, I identified 179 companies that had CEO resigning during the sample period. Then I deleted 53 companies because they were financial institutions, regulated companies or failed to meet data availability requirements. The selection process resulted in a final sample of 126 firms. 4.1.2 Descriptive Statistics for The Sample Descriptive statistics for the 126 sample firms are presented in Table 3. Table 3 shows that sample companies have a fairly wide range of size and performance (panel A) and - 24 - Chapter Four Sample Selection and Research Models CEO resignations are roughly evenly divide d between Year 1998 and Year 1999 (panel B). Panel C shows that the sample firms represent 36 industries, with the highest concentration of firms (13 firms) in the Transportation Equipment Industry (SIC 37), followed by the Electronics & Other Electrical Equipment Industry (SIC 36), Wholesale Trade -durable Goods Industry (SIC 50) and Chemicals & Allied Products Industry (SIC 28). The remaining firms in the sample are relatively evenly distributed among the other 32 industries. Table 3: Descriptive statistics for the sample Panel A: Descriptive statistics for revenue, total assets, net income and ROA Mean Median Mode Revenue 668.7 170.1 168.5 1257.2 1.2 11816.0 Total assets 3122 697 348 8734 3 213277 Net income 35 7 36 188 -1521 197441 0.008 0.013 0.037 0.048 -0.825 0.490 ROA Std. dev. Minimum Maximum Note: Total assets and net income are all reported in millions of US dollars. ROA is return on assets, which is the ratio of net income to total assets. Panel B: Sample distribution by year Year of resignation Number of companies 1998 61 1999 65 - 25 - Chapter Four Sample Selection and Research Models Panel C: Sample distribution by industry SIC code Industry Frequency 13 Oil And Gas Extraction 4 14 Mining And Quarrying 5 15 Building Construction 1 20 Food And Kindred Products 5 21 Tobacco Products 1 23 Apparel 4 25 Furniture And Fixtures 2 26 Paper And Allied Products 1 27 Printing, Publishing, And Allied Industries 1 28 Chemicals And Allied Products 9 29 Petroleum Refining And Related Industries 1 31 Leather And Leat her Products 5 32 Stone, Clay, Glass, And Concrete Products 2 33 Primary Metal Industries 1 36 Electronic And Other Electrical Equipment 11 37 Transportation Equipment 13 38 Instruments and related products 4 39 Miscellaneous Manufacturing Industries 5 40 Railroad Transportation 3 44 Water Transportation 2 45 Transportation By Air 1 48 Communications 2 50 Wholesale Trade-durable Goods 10 51 Wholesale Trade-non -durable Goods 4 52 Building Materials, Hardware, Garden Supply 3 55 Automotive Dealers And Gasoline Service Stations 1 56 Apparel And Accessory Stores 1 57 Home Furniture, Furnishings, And Equipment Stores 2 58 Eating And Drinking Places 5 59 Miscellaneous Retail 3 70 Hotels, Rooming Houses, Camps, And Other Lodging Places 2 73 Business Services 2 79 Amusement And Recreation Services 1 80 Health Services 4 81 Legal Services 1 87 Engineering, Accounting, Research, Management 4 Total 126 - 26 - Chapter Four Sample Selection and Research Models 4.2 Research Methodology 4.2.1 Total Accruals In my study, I calculate total accruals as income before extraordinary items minus cash flow from operating activities. As mentioned earlier, whether special items are discretionary or not is subject to debate. Therefore, I subtract the special items (on an after -tax basis) fr om the total accruals. To adjust it to after-tax basis, I multiply special items by 0.6. (Assuming tax rate equal to 0.4) Total Accruals = Income before Extraordinary Items (Item 8) – Cash Flow From Operating Activities (Item 108) –0.6 Special Items (Ite m 32) 4.2.2 Estimation of Discretionary Accruals Estimation of Discretionary Accruals for Individual Sample Firm Prior research by Han and Wang (1998) and Erickson and Wang (1999) examined discretionary accruals as the residuals of regressions using pooled data. However, this procedure ignores the fact that different firms normally have different accrual policies and accrual patterns, leading to biased coefficient estimates and standard error estimates. To solve this problem, I use a variation of the Jones Model: TTAC it / ASTit = β 0 (1 / ASTit ) + β 1 (∆REV it / ASTit ) + β 2 (PPE it / ASTit ) + β 3 Q1 β 4 Q2 + β 5 Q3 + β 6 Q4 + ε it where TTAC it is total accruals for firm i in quarter t; - 27 - (1) Chapter Four Sample Selection and Research Models ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to 1 for the quarter j (j=1, ... ,4) of the fiscal year and 0 otherwise; ε it is the error term for firm i in quarter t. Like prior studies, changes in revenues and in gross property, plant and equipment are used to control for the nondiscretionary components in total accruals. The coefficient of ∆REVit is expected to be positive because changes in working capital accounts (e.g., changes in accounts receivable, changes in inventory, etc.) are part of total accruals and are positively related to changes in revenues. The expected sign for PPE it is negative because higher fixed assets are expected to lead to higher depreciation and deferred taxes. The quarter indicator variables are used to account for variation in accruals across quarters. In this model, I assume that firms do not change their accruals policy significantly among different years. Therefore, I do not use year indicator variables in this model. I estimate Equation (1) for each firm in the sample individually, using time-series quarterly data from the 25 quarters in six years (from 20 quarters before the quarter with CEO resignation to 4 quarters after that quarter). Discretionary accruals for each firm-quarter observation are estimated as the difference between reported total accruals for the quarter and the fitted va lue of total accruals using coefficients from the Equation (1). I then calculate the mean discretionary accruals across firms in every - 28 - Chapter Four Sample Selection and Research Models quarter from quarter -4 to quarter 4 to judge whether earnings management has occurred. Estimation of Discretionary Accruals by Pooled Sample Following Han and Wang (1998) and Erickson and Wang (1999), I also estimate discretionary accruals as the residuals,ε it , from the following model, using data pooled across all quarters (1993-2000) and across firms: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + ε it (2) where Yk is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise. Similar to prior studies, the coefficient for ∆ REVit is expected to be positive and the coefficient for PPEit is expected to be negative. The quarter (year) indicator variables here are used to account for variation in accruals across quarters (years). Prior researchers believe accounting accruals are influenced by the economic conditions at different points in time. Therefore, adding the quarter and year indicators to the equation can control for the variation across periods and improve the accuracy of the estimation. To make my study comparable to prior studies, I incorporate yearly dummy variables in the pooled data regressions. I used three quarterly dummy variables to account for the variation across four calendar quarters and seven yearly dummy variables to account for the variation across eight calendar years (1993-2000). - 29 - Chapter Four Sample Selection and Research Models The base quarter and base year are the first calendar quarter and the year 1993, respectively. Testing Earnings Management by Pooled Sample on Aggregate Basis To capture outgoing and incoming CEOs’earnings management, two dummy variables are added into the equation (2). TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + β 13T1 + β 14T2 + ε it (3) where T1 is an indicator variable equal to 1 if quarter t is one of the four quarters immediately before the resignation quarter, and 0 otherwise. T2 is an indicator variable equal to 1 if quarter t is one of the four quarters immediately after the resignation quarter, and 0 otherwise. If the outgoing CEOs engage in earnings management before their resignations, the coefficient of T1 is expected to be significantly different from zero. However, if the incoming CEOs manage earnings downward immediately after they are appointed and manage earnings upward in the following periods, the coefficients of T2 may not be significantly different from zero. (The upward effect and downward effects may cancel each other off.) Thus, we test Hypotheses H1 and H2 on a quarterly basis. - 30 - Chapter Four Sample Selection and Research Models Testing Earnings Management for Pooled Sample on a Quarterly Basis Instead of estimating earnings management on an aggregate basis, the following equation is estimated to analyze outgoing and incoming CEO’ s earnings management for every quarter (from quarter –4 to quarter 4) before and after resignation date. TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + β 13 D−4 + β 14 D−3 + β 15 D− 2 + β16 D−1 + β17 D0 (4) + β 18 D1 + β 19 D2 + β 20 D3 + β 21D4 + ε it Where D j (j= -4 to 4) is an indicator variable equal to 1 if quarter t is one of the four quarters immediately before and after the resignation quarter and 0 otherwise. If the CEOs that resign engage in earnings management upward before their resignations, some of the coefficients for D− 4 to D−1 are expected to be significantly higher than zero. If the incoming CEOs manage earnings downward immediately after they are appointed and manage earnings upward in the following periods, some of the coefficients for D0 through D4 are expected to be significantly lower than zero and some higher than zero. - 31 - Chapter Five Empirical Results and Discussion Chapter Five: Empirical Results and Discussion 5.1 Estimation of Discretionary Accruals by Individual Sample Firm I estimate Equation (1) for each firm individually and this produces a time series of estimated quarterly discretionary accruals for each firm. I then compute the mean of estimated discretionary accruals across firms in each quarter. Figure 2 plots the mean discretionary accruals from Equation (1) for the quarters before and after a CEO resignation, including quarters from Q-4 to Q4. The mean residuals are -0.0168, 0.0602, -0.0409, 0.0052, -0.1387, 0.0889, -0.0796, -0.0943, -0.0194, for quarters from Q-4 to Q4, respectively. The mean residuals suggest that the sample firms do not have abnormally high or low unexpected accruals for the quarters before Q0. They also suggest that the sample firms had abnormally low discretionary accruals in the quarter where resignations occurred and abnormally high discretionary accruals in the quarter immediately after the resignation quarter. The statistical significance of the mean discretionary accruals in these quarters is presente d in Table 4. FIGURE 2 Mean Discretionary Accruals by individual firm Mean Discretionary Accruals 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 -0.2 Q-4 Q-3 Q-2 Q-1 Q0 Q4 - 32- Q1 Q2 Q3 Chapter Five Empirical Results and Discussion Table 4: Mean discretionary accruals for each quarter surrounding resignation date Quarter　 MDA Standard Error t-value Q-4 -0.0168 0.0209 -0.8044 Q-3 0.0602 0.0428 1.4055 Q-2 -0.0409 0.0724 -0.5650 Q-1 0.0052 0.0087 0.5929 Q0 -0.1387 0.0837 -1.6569** Q1 0.0890 0.0423 2.1010* Q2 -0.0796 0.1041 -0.7648 Q3 -0.0943 0.1229 -0.7674 Q4 -0.0194 0.0315 -0.6150 Notes: MDA refers to Mean Discretionary Accruals. MDA are the average across firms of the error term for each quarter of the following model estimated for each firm individually: TTAC it / ASTit = β 0 (1 / ASTit ) + β 1 (∆REV it / ASTit ) + β 2 (PPE it / ASTit ) + β 3 Q1 β 4 Q2 + β 5 Q3 + β 6 Q4 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. ** Statistically significant at the 10% level. - 33 - Chapter Five Empirical Results and Discussion Table 4 shows that the mean discretionary accruals for Q0 and Q1 are significantly different from zero at the 10% and 5% level, respectively. These empirical results are consistent with my Hypothesis H1 that outgoing CEOs do not engage in significant earnings management. They are also consistent with Hypothesis H2 that the incoming CEOs manage earnings downward immediately after they are appointed and manage earnings upward in the following periods. As well, I compute the average of each regression coefficient across firms. The results are reported in Table 5. The mean coefficient of change in revenues is 1.1490 (t statistic = 2.6683) and the mean coefficient for property, plant and equipment is 0.6673 (t-statistic = -2.5005). The signs for both coefficients are in the predicted direction and statistically significant at the 5% level. The quarter indicator variables are all negative and statistically significant at the 5% level. The results are not consistent with Pourciau’s findings, since she found negative unexpected total accruals before CEO resignation in her paper. Acknowledging the importance of controlling for corporate performance, Pourciau attempts to control for poor performance prior to CEO resignations by deflating earnings and accruals by contemporaneous sales. However, the procedure is likely to fail because accruals will not change proportionately with sales. Since sales are likely to go down prior to CEO resignations, total accruals deflated by sales are likely to show a downward trend before CEO resignations. - 34 - Chapter Five Empirical Results and Discussion Table 5: Averages of coefficients in firm-specific regressions 　 Mean Coefficients 1/ ASTit Standard Error t-value 110.3444 50.5784 2.1816* ∆REV it / ASTit 1.1490 0.4306 2.6683* PPEit / ASTit -0.6673 0.2669 -2.5005* Q1 -0.1993 0.0816 -2.4430* Q2 -0.4132 0.0470 -8.7873* Q3 -0.2955 0.0449 -6.5762* Q4 -0.3674 0.0496 -7.4101* Notes: Mean coefficients are the average coefficients of each term from the following model estimated for each firm individually: TTAC it / ASTit = β 0 (1 / ASTit ) + β 1 (∆REV it / ASTit ) + β 2 (PPE it / ASTit ) + β 3 Q1 β 4 Q2 + β 5 Q3 + β 6 Q4 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. To show the flaw of Pourciau’s research methodology, I calculate the unexpected accruals for my sample firms in Q–3, Q–2 and Q-1, following Pourciau’s methodology. In her study, Pourciau defined unexpected accruals as the first difference of the total accruals, after scaling by contemporary sales. Since Pourciau used median of unexpected accruals as the indicator of unexpected accruals in her study, I did the same. - 35 - Chapter Five Empirical Results and Discussion The results are reported in Figure 3. The median unexpected accruals using Pourciau’ s model (i.e., deflated by current sales) are –0.0484, -0.0861 and -0.0465 for Q-3, Q-2, and Q-1, respectively. Similar to Pourciau’ s study, all the variables are negative. The median unexpected accruals using my model (i.e., deflated by current total assets) are –0.0026, 0.0007 and 0.0017 for the Q-3, Q-2, Q-1, respectively. They are all around zero, and two of them are positive. This comparison indicates that Pourciau’ s perplexing finding in her paper that outgoing CEOs managed earning downwards before their nonroutine turnover may be mainly due to an inherent bias in her methodology. FIGURE 3 Median Unexpected Accruals using my model and Pourciau's model (firm-by-firm regressions) Mean Unexpected Accruals 0.02 0 -0.02 -0.04 -0.06 -0.08 -0.1 Q-3 Q-2 Q-1 Median unexpected accruals by my model Median unexpected accruals by Pourciau's model - 36 - Chapter Five Empirical Results and Discussion 5.2 Estimation of Discretionary Accruals by Pooled Sample My regression using data pooled across firms and quarters generates similar results. I estimated Equation (2) by the ordinary least squares (OLS) method and the results are reported in Table 6. The coefficient for the change in revenues is 0.0419 (t -statistic = 2.8293) and the coefficient for the gross book value of property, plant and equipment is –0.0349 (t-statistic = -10.3034). The signs of both coefficients are in the predicted direction and statistically significant at the 5% level. The quarter indicator for Q4 is negative and all other quarter indicators are positive. None of these quarter indicator variables is statistically significant at the 5% level. All the year indicator variables are positive and statistically significant at the 5% level except for the indicator variables for year 1994. Since my purpose for using quarter and year dummy variables is to control for yearly and quarterly economic variation caused by different business cycles and economic conditions in different years and quarters, I do not report the coefficients and results of the quarter and year indicator variables in the tables and discussion in the remainder of the paper. ( Results for these indicator variables are reported in the Appendices.) Figure 4 plots the mean residuals from Equation (2) for the quarters before and after a CEO resignation, including quarters from Q-4 to Q4. The mean residuals are -0.0037, 0.0030, 0.0002, 0.0034, -0.0101, 0.0114, -0.0009, -0.0001, -0.0028 for quarters from Q-4 to Q4, respectively. - 37 - Chapter Five Empirical Results and Discussion Table 6: Estimation of discretionary accruals by pooled data 　 Coefficients Standard Error t-value Intercept -0.0579 0.0071 -8.1625 1/ ASTit 0.2284 0.1394 1.6380 ∆REV it / ASTit 0.0419 0.0148 2.8293* PPEit / ASTit -0.0349 0.0034 -10.3034* Notes: The results are estimated from the following model by pooled data across firms and quarters: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenu es for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; Yk is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise; ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. FIGURE 4 Mean Discretionary Accruals by pooled data Mean Discretionary Accruals 0.015 0.01 0.005 0 -0.005 -0.01 -0.015 Q-4 Q-3 Q-2 Q-1 Q0 Q4 - 38 - Q1 Q2 Q3 Chapter Five Empirical Results and Discussion The mean of the residuals from the regression also suggests that there are no abnormally high or low unexpected accruals for the quarters before Q0. They also suggest that the sample firms have low unexpected accruals in the quarter where resignation happens and high unexpected accruals in the quarter immediately after Q0. I assess the statistical significance of the unexpected accrual surrounding the resignation date on both aggregate and quarterly basis in the following section. 5.2.1 Testing Earnings Management by Pooled Sample on Aggregate Basis To test for outgoing and incoming CEOs’earnings management on an aggregate basis, I estimate Equation (3). The results are reported in Table 7. The coefficient for the change in revenues is positive and the coefficient for the gross book value of property, plant and equipment is negative. They are both significantly different from zero at the 5% level. The coefficient for T1 is positive and that for T2 is negative. Neither is statistically significant. These results are consistent with my Hypothesis H1 that predicts no earnings management before CEO resignation. The evidence is inconsistent with Pourciau’s finding, since she found negative unexpected total accruals before CEO resignation. As I have done with the firm-by-firm regressions, I also compare the median discretionary accruals for my pooled sample model and using Pourciau’ s model. The results are reported in Figure 5. The median unexpected accruals using Pourciau’ s model are –0.0484, -0.0861 and -0.0465 for Q-3, Q-2 and Q-1, respectively. Noticeably, they are all negative, which is similar to what Pourciau found in her paper. In contrast, the median unexpected accruals for my pooled sample model are all - 39 - Chapter Five Empirical Results and Discussion positive. The difference in results is consistent with the argument that Pourciau’ s finding is mainly due to limitation in her methodology. Table 7: Estimatio n of discretionary accruals by pooled data on aggregate basis Coefficients Standard Error t-value Intercept -0.0580 0.0071 -8.1442 1/ ASTit 0.2294 0.1396 1.6439 ∆REV it / ASTit 0.0419 0.0149 2.8100* PPEit / ASTit -0.0349 0.0034 -10.3084* T1 0.0008 0.0046 0.1818 T2 -0.0024 0.0044 -0.5379 Notes: The results are estimated from the following model by pooled data across firms and quarters: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + β 13T1 + β 14T2 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; Yk is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise; T1 is an indicator variable with 1 if in the year (namely, form quarter -4 to quarter -1) immediately preceding to resignation date and 0 otherwise; T2 is an indicator variable with 1 if in the year (namely, form quarter 0 to quarter 4) immediately after resignation date and 0 otherwise. ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. - 40 - Chapter Five Empirical Results and Discussion Mean Unexpected Accruals FIGURE 5 Median Unexpected Accruals using my model and Pourciau's model (pooled data regression) 0.01 0 -0.01 -0.02 -0.03 -0.04 -0.05 -0.06 -0.07 -0.08 -0.09 -0.1 Q-3 Q-2 Q-1 Median unexpected accruals by my model Median unexpected accruals by Pourciau's model The fact that I am unable to find earnings management by new CEOs during the one year after resignation date may be due to two reasons: (1) there is no earnings management occurring during this period; and (2) there is earnings management for this period, but the pattern of earnings management for very quarter is different, and the upward and downward effects of earnings management for different quarters may offset each other. I will next examine earnings management on a quarterly basis. 5.2.2 Testing Earnings Management by Pooled Sample on a Quarterly Basis To test earning management on a quarterly basis for the pooled sample, I estimated Equation (4). The results are reported in Table 8. The coefficients for the change in revenues and the gross book value of property, plant and equipment both have the correct sign and are significant at the 5% level. Only the coefficients for D0 and D1 - 41 - Chapter Five Empirical Results and Discussion are statistically significant at the 5% level. Since quarter 0 is the first quarter in which the incoming CEOs have the power to manage the financial reports, the significance of the coefficients for D0 and D1 indicate that the incoming CEOs manipulate earnings downward for the first quarter (Q0) in which they have power to manage financial reports and manage earnings upwards in the following quarter (Q1). As for Q2, Q3 and Q4, I cannot detect significant earnings management in these quarters. Generally, this result is consistent with Hypothesis H2 that new CEOs manage earnings downward immediately after they are appointed and manage earnings upward in the following periods. Table 8: Estimation of discretionary accruals by pooled data on quarterly basis Coefficients Standard Error t Stat Intercept -0.0580 0.0071 -8.1510 1 / ASTit 0.2302 0.1394 1.6514 ∆REV it / ASTit PPEit / ASTit 0.0423 0.0149 2.8424* -0.0349 -0.0019 0.0034 0.0068 -10.3260* -0.2839 0.0045 -0.0045 0.0068 0.0068 0.6637 -0.6555 0.0048 0.0071 0.6843 -0.0149 0.0138 0.0071 0.0068 -2.1040* 2.0176* -0.0021 0.0068 -0.3147 -0.0046 0.0069 -0.6740 -0.0037 0.0070 -0.5250 D−4 D −3 D−2 D−1 D0 D1 D2 D3 D4 Notes: The results are estimated from the following model by pooled data across firms and quarters: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + β 13 D−4 + β 14 D−3 + β 15 D− 2 + β16 D−1 + β17 D0 + β 18 D1 + β 19 D2 + β 20 D3 + β 21D4 + ε it - 42 - Chapter Five Empirical Results and Discussion where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; Yk is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise; D j (j=-4, -3, -2, -1, 0, 1, 2, 3, 4) is an indicator variable with 1 for the quarter j and 0 otherwise; ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. 5.3 Additional Evidence I also use other models of detecting earnings management to examine whether the results are robust to the choice of model. Since my sample firms cover a large number of industries, it is not practical to use the Dechow and Sloan Model (Industry Model) to calculate average total accruals for different industries and compare them to total accruals. Therefore, the alternative testing models that used are: the Healy Model (1985), the DeAngelo Model (1986), the Refined Jones Model (1995) and the Kang & Sivaramakrishnan Model (1995). 5.3.1 The Healy Model Using the Healy Model, I calculate the mean total accruals (deflated by the book value of total assets) for the estimation period (quarter –20 to quarter –5) and for the period from quarter –4 to quarter 4. The results are shown in Table 9. - 43 - Chapter Five Empirical Results and Discussion Table 9: The empirical results of Healy Model Panel A: The mean total accruals for the estimation period and event period Period Mean total accruals Period Mean total accruals Estimation period -0.0351 Q0 -0.0481 Q-4 -0.0398 Q1 -0.0232 Q-3 -0.0334 Q2 -0.0363 Q-2 -0.0403 Q3 -0.0352 Q-1 -0.0318 Q4 -0.0403 Notes: The mean total accruals are deflated by the book value of total assets. Estimation period is the period from quarter –20 to quarter –5. Panel B: The tests for differences in means (compared to estimation period) Quarter t-value ANOVA F-value Probability Q-4 0.7857 0.6173 0.4321 Q-3 0.2887 0.0833 0.7728 Q-2 0.8639 0.7463 0.3877 Q-1 0.5695 0.3243 0.5690 Q0 2.1398* 4.5788 0.0324 Q1 1.9809* 3.9239 0.0476 Q2 0.1904 0.0363 0.8490 Q3 0.0045 0.0000 0.9964 Q4 0.8343 0.6960 0.4042 Note: * Statistically significant at the 5% level. - 44 - Chapter Five Empirical Results and Discussion The mean total accruals in the estimation period and in every quarter from Q–4 to Q4 are all negative (Panel A). To investigate whether mean total accruals in the event quarters are significantly different from that for the estimation period, I apply the t-test and ANOVA F-statistic (Panel B). The results in the Panel B indicate that the mean total accruals for Q–4, Q–3, Q–2, Q–1, Q2, Q3, and Q4 are not significantly different from the mean total accruals of estimation period. On the other hand, the mean total accruals in Q0 and Q1 are both significantly different from the mean total accruals in the estimation period at the 5% level. The mean total accruals in Q0 is –0.0481, which is significantly lower than mean total accruals in estimation period; the mean total accruals for Q1 is –0.0232, which is significantly higher than mean total accruals in the estimation period. Therefore, similar to Jones’model, the Healy Model indicates that the new CEOs will manage earnings downward immediately after they take the position of CEO and manipulate earnings upward in the following quarter. The Healy Model also fails to detect any significant earnings management in Q2, Q3 and Q4. 5.3.2 The DeAngelo Model The DeAngelo Model assumes that the average change in nondiscretionary accruals from two adjacent periods is approximately zero and tests whether the average value of the change in total accruals is significantly negative before an event date. Using this model, I calculated the average value of the change of total accruals for the period from quarter –4 to quarter 4. The empirical results are shown in Table 10. As expected, the difference in the mean total accruals between Q-1 and Q0 is negative and statistically significant at the 10% level. The difference in the mean total accruals between Q0 and Q1 is positive and - 45 - Chapter Five Empirical Results and Discussion statistically significant at the 5% level. These results indicate that the sample firms have abnormally low total accruals for the quarter in which the outgoing CEOs resign and abnormally high total accruals for the second quarter after a CEO resignation. The differences in the mean total accruals between other quarters are all insignificantly different from zero at the 10% level. These results confirm the findings in prior models that there is no earnings management in other quarters except Q0 and Q1. Table 10: The empirical results for DeAngelo Model Period Mean of changes in total accruals t-value Probability Q-3 to Q-4 0.0037 0.5440 0.5872 Q-2 to Q-3 -0.0046 -0.7345 0.4636 Q-1 to Q-2 0.0115 1.5206 0.1271 Q0 to Q-1 -0.0132 -1.8630** 0.0645 Q1 to Q0 0.0178 2.1638* 0.0320 Q2 to Q1 -0.0144 -1.5212 0.1302 Q3 to Q2 0.0041 0.5756 0.5657 Q4 to Q3 0.0039 0.5310 0.5962 Notes: The mean total accruals are deflated by the book value of total assets. * Statistically significant at the 5% level. ** Statistically significant at the 10% level. - 46 - Chapter Five Empirical Results and Discussion 5.3.3 The Refined Jones Model In the Jones Model, revenues are assumed to be an objective measure of a firm’ s operations before managers’manipulations. However, Dechow, Sloan and Sweeney (1995) argued that the reported revenues may be not completely exogenous and could be affected to some extent by managers. To overcome the limitation of the Jones model, Dechow, Sloan and Sweeney postulated a refined version of the Jones model. The nondiscretionary accruals are estimated as follows. TTAC it / ASTit = β 0 (1 / ASTit ) + β 1 (∆REV it / ASTit − ∆REC it / ASTit ) + β 2 ( PPEit / ASTit ) + ε it (5) where ∆REC it is the change in net receivables for firm i in period t. In their modified Jones model, the authors assume that it is easier to manipulate earnings by using discretion based on credit sales than on cash sales. They also remove the change in net receivables from the change of revenues and obtained the net change of revenues based on cash sales. They believe the net change in revenues base d on cash sales can accurately control for nondiscretionary accruals and “then the estimate of earnings management should no longer by biased toward zero in samples where earnings management has taken place through the management of revenues”. (Page 201) I estimate the Refined Jones Model by individual sample firm and report the mean discretionary accruals for these quarters in Figure 6. Figure 6 plots the mean discretionary accruals from the Refined Jones Model. The mean discretionary accruals are -0.0005, 0.0021, -0.0062, 0.0091, -0.0128, 0.0154, - - 47 - Chapter Five Empirical Results and Discussion 0.0055, -0.0045, 0.0019, in quarters Q-4 through Q4, respectively. The mean of the discretionary accrual in Q0 is significantly lower than zero at the 10% level and that in Q1 is significantly higher than zero at the 5% level. This result is consistent with what I find using the Jones Model. It also does not indicate that the outgoing CEOs manage earnings. FIGURE 6 Mean Discretionary Accruals by Refined Jones Model Mean Discretionary Accruals 0.02 0.015 0.01 0.005 0 -0.005 -0.01 -0.015 Q-4 Q-3 Q-2 Q-1 Q0 Q1 Q2 Q3 Q4 5.3.4 The Kang and Sivaramakrishnan Model In the Kang and Sivaramakrishnan Model, net revenues, operating expenses and gross book value of property, plant and equipment are used as the control variables for the nondiscretionary accruals. The Model is as follows: TTAC it / ASTit = β 0 + β 1 ( REVit / ASTit ) + β 2 ( EXPit / ASTit ) + β 3 (PPE it / ASTit ) + ε it where - 48 - (6) Chapter Five Empirical Results and Discussion REVit is the net revenue for firm i in period t; EXPit is the operating expenses (cost of good sold, selling and administrative expenses before depreciation, etc) for firm i in period t. I use the Kang and Sivaramakrishnan model to estimate discretionary accruals by individual sample firm and the mean discretionary accruals are shown in Figure 7. The mean discretionary accruals are -0.0012, -0.0005, -0.0076, 0.0081, -0.0164, 0.0142, -0.0071, -0.0066, -0.0006, for quarters from Q-4 to Q4, respectively. The mean of the discretionary accrual in Q0 is significantly lower than zero at the 5% level and that in Q1 is significantly higher than zero at the 10% level. Similar to the other previous models, this model also shows evidence consistent with Hypothesis H1 and H2. FIGURE 7 Mean Discretionary Accruals by Kang and Sivaramakrishnan Model 0.02 Mean Discretionary Accruals 0.015 0.01 0.005 0 -0.005 -0.01 -0.015 -0.02 Q-4 Q-3 Q-2 Q-1 Q0 Q4 - 49 - Q1 Q2 Q3 Chapter Six Summary and Conclusions Chapter Six: Summary and Conclusions 6.1 Summary of Findings This study investigates the earnings management behavior of outgoing CEOs and incoming CEOs. A sample of 126 CEO resignations during the period 1998-1999 is used. For each firm, I examine discretionary accruals in the four quarters immediately before and after the CEO resignation. The results suggest that outgoing CEOs use accruals to increase earnings one year before resignation date on the aggregate basis. The quarterly analysis for the individual firms sample shows that outgoing CEOs record accruals to increase earnings in quarter -4, quarter –3 and quarter-1. However, all these discretionary accruals are not significantly different from zero even at the 10% significance level. Therefore, my empirical results support my Hypothesis H1 (Null Hypothesis), which predicts no earnings management by outgoing CEOs. Since my results are not consistent with Pourciau’s findings, I also calc ulate the unexpected accruals using Pourciau’ s methodology. My results indicate Pourciau’ s findings are due to an inherent bias in her methodology. My empirical results also indicate that incoming CEOs record accruals to decrease earnings in quarter 0 and increase earnings in quarter 1. The significance tests show that the discretionary accruals for these two quarters are significantly different from zero either at the 5% level or the 10% level. These results support my Hypothesis H2 that incoming CEOs manage earnings downward immediately after they are appointed and manage earnings upward in the following periods. - 50- Chapter Six Summary and Conclusions 6.2 Implication of Findings The evidence for earnings management behavior before CEOs’ departure from previous studies is controversial. My study shows that outgoing CEOs do not engage in significant earnings management one year before resignation. One possible explanation is that the increase in monitoring activities before CEO resignation may make it difficult for outgoing CEOs to engage in earnings management before they are forced to resign or voluntarily do so. While prior evidence on outgoing CEOs’ earnings management behavior iscontroversial, there is no significant disagreement in the prediction of incoming CEOs’ earnings management. My study provides additional empirical evidence to support the “Big-bath” hypothesis. My results indicate that incoming CEOs will use accounting discretion to blame the outgoing CEOs for poor performance and manage earnings upwards to meet performance target later. 6.3 Contribution of Study Compared with other issues in earnings management, relatively few researchers have investigated earnings management associated with CEO turnover. Even fewer researchers have focused on earnings management associated with CEO resignation. Only Pourciau has investigated earnings management associated with nonroutine CEO turnover. However, due to several methodological deficiencies, Pourciau herself found some of her results surprising and perplexing. I applied a more reliable methodology in my study. My methodological improvements include: (1) using quarterly data instead - 51 - Chapter Six Summary and Conclusions of yearly data; (2) using the Jones Model to control for nondiscretionary accruals and poor corporation performance before CEO resignation; (3) deleting special items from total accruals; and (4) increasing total sample size to 126 and a fair variety of industries. My study contributes to the literature in several ways. First, Pourciau’ s (1992) finding that resigning CEOs manipulate earnings downward in the periods right before their departure is puzzling. My study shows that they appear no reason for manipulating earnings downward in those periods to their own detriment. Second, I employ more sophisticated models to detect earning management. Results from les s sophisticated models may be likely to be unreliable and mislead researchers. 6.4 Limitations and Suggestion for Future Research One limitation of this study is that I only studied CEOs earnings management behavior during the period from four quarters be fore the resignation date to four quarters after resignation date. Future research can prolong the time horizon to better understand CEOs’earnings smoothing behaviors. The other limitation is that I can only detect new CEOs’earnings management in the resignation quarter and the quarter following resignation quarter. I find no significant earnings management behavior in other quarters within one year after resignation date. I cannot conclude whether this is due to the fact that new CEOs stop manipulating earnings during this period or the inability of my model to detect earnings management, if any. Future research can focus on these quarters for fully understanding the pattern of new CEOs’earnings management - 52 - Chapter Six Summary and Conclusions 6.5 Conclusions This study examines the ear nings management behavior of outgoing and incoming CEOs surrounding CEO resignation. The Jones Model is used to test two hypotheses. The empirical evidence is consistent with my hypotheses, indicating that outgoing CEOs do not engage in significant earnings management before resignation and incoming CEOs manage earnings downwards in the quarter in which they are appointed as CEO and manipulate earnings upwards in the following quarter. My empirical results regarding outgoing CEOs are not consistent with Pourciau’s findings. Further analysis indicates that this is mainly due to the inherent bias in Pourciau’ s research methodology. - 53 - Appendices Appendix 1: Full table of estimation of discretionary accruals by pooled data Coefficients Standard Error t Stat Intercept 1/ ASTit -0.0579 0.0071 -8.1625 0.2284 0.1394 1.6380 ∆REV it / ASTit 0.0419 0.0148 2.8293* PPEit / ASTit -0.0349 0.0034 -10.3034* Y00 ０．０４８７　 ０．００７５ 　 ６．５１９７＊　 Y99 Y98 ０．０３４６　 ０．０３９２　 ０．００７３ 　 ０．００７３ 　 ４．７３９０＊　 ５．４０８８＊　 Y97 ０．０３６７　 ０．００７２ 　 ５．０９１１＊　 Y96 ０．０３６９　 ０．００７２ 　 ５．０９９６＊　 Y95 ０．０３５４　 ０．００７５ 　 ４．７４６２＊　 Y94 ０．０１３６　 ０．００７８ 　 １．７５１４＊＊　 Q4 －０．０００６　 ０．００３８ 　 －０．１５１２　 Q3 ０．００２３　 ０．００３３ 　 ０．６８５９ 　 Q2 ０．０００８　 ０．００３６ 　 ０．２２７６ 　 Notes: The results are estimated from the following model by pooled data across firms and quarters: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for f irm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; Yk is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise; ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. - 55 - Appendix 2: Full table of testing Earnings Management by pooled data on aggregate basis Coefficients Standard Error t Stat Intercept -0.0580 0.0071 -8.1442 1 / ASTit ∆REV it / ASTit 0.2294 0.0419 0.1396 0.0149 1.6439 2.8100* PPEit / ASTit -0.0349 0.0008 0.0034 0.0046 -10.3084* 0.1818 -0.0024 0.0044 -0.5379 ０．０４９８　 ０．００７７ 　 ６．４６３８＊　 Y99 Y98 ０．０３５９　 ０．０３９６　 ０．００７９ 　 ０．００７７ 　 ４．５６６０＊　 ５．１０５５＊　 Y97 ０．０３６４　 ０．００７３ 　 ４．９６４５＊　 Y96 ０．０３６９　 ０．００７２ 　 ５．０９８４＊　 Y95 ０．０３５４　 ０．００７５ 　 ４．７４７１＊　 Y94 ０．０１３６　 ０．００７８ 　 １．７５４７＊＊　 Q4 －０．０００６　 ０．００４３ 　 －０．１３６４　 Q3 ０．００２６　 ０．００１１　 ０．００３５ 　 ０．００３８ 　 ０．７４７７ 　 ０．２８５５ 　 T1 T2 Y00 Q2 Notes: The results are estimated from the following model by pooled data across firms and quarters: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + β 13T1 + β 14T2 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; Yk T1 is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise; is an indicator variable with 1 if in the year (namely, form quarter -4 to quarter -1) immediately preceding to resignation date and 0 otherwise; T2 is an indicator variable with 1 if in the year (namely, form quarter 0 to quarter 4) immediately after resignation date and 0 otherwise. ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. ** Statistically significant at the 10% level. - 56 - Appendix 3: Full table of testing Earnings Management by pooled data on quarterly basis Coefficients Standard Error t Stat Intercept -0.0580 0.0071 -8.1510 1 / ASTit ∆REV it / ASTit 0.2302 0.0423 0.1394 0.0149 1.6514 2.8424* PPEit / ASTit -0.0349 -0.0019 0.0034 0.0068 -10.3260* -0.2839 0.0045 -0.0045 0.0068 0.0068 0.6637 -0.6555 0.0048 0.0071 0.6843 -0.0149 0.0071 -2.1040* D1 0.0138 0.0068 2.0176* D2 -0.0021 0.0068 -0.3147 D3 -0.0046 -0.0037 0.0069 0.0070 -0.6740 -0.5250 ０．０４９４　 ０．００７８ 　 ６．３７０９＊　 ０．０３５６　 ０．０３９６　 ０．００７９ 　 ０．００７８ 　 ４．５１８５＊　 ５．０８２７＊　 Y97 Y96 ０．０３６７　 ０．０３６８　 ０．００７３ 　 ０．００７２ 　 ５．０００９＊　 ５．１０３８＊　 Y95 Y94 ０．０３５４　 ０．０１３６　 ０．００７４ 　 ０．００７７ 　 ４．７５１１＊　 １．７５２７＊＊　 Q4 Q3 －０．０００６　 ０．００２２　 ０．００４３ 　 ０．００３５ 　 －０．１３０６　 ０．６３７４ 　 Q2 ０．００１６　 ０．００３８ 　 ０．４２９３ 　 D−4 D −3 D−2 D−1 D0 D4 Y00 Y99 Y98 Notes: The results are estimated from the following model by pooled data across firms and quarters: TTAC it / ASTit = α + β 0 (1 / ASTit ) + β 1 (∆ REVit / ASTit ) + β 2 ( PPEit / ASTit ) + β 3 Q2 + ... + β 5 Q4 + β 6 Y94 + ... + β 12Y00 + β 13 D−4 + β 14 D−3 + β 15 D− 2 + β16 D−1 + β17 D0 + β 18 D1 + β 19 D2 + β 20 D3 + β 21D4 + ε it where TTAC it is total accruals for firm i in quarter t; ∆ REVit is the change in revenues for firm i in quarter t; ASTit is total assets for firm i in quarter t; PPEit is the gross book value of property, plant and equipment for firm i in quarter t; - 57 - Q j is a quarter indicator variable set equal to one for the quarter j (j=1, ... ,4) of the fiscal year and zero otherwise; Yk is a year indicator variable taking the value one for the year k (k=1994, ... ,2000) and zero otherwise; D j (j=-4, -3, -2, -1, 0, 1, 2, 3, 4) is an indicator variable with 1 for the quarter j and 0 otherwise; ε it is the error term for firm i in quarter t. * Statistically significant at the 5% level. ** Statistically significant at the 10% level. - 58 - Bibliography Aharoney, J., Lin, C. and Loeb, M., 1993. Initial public offerings, accounting choices and earnings management. Contemporary Accounting Research 10, 61¯81. Ali, A. and Kumar, k., 1993. Earnings management under pension accounting standards: SFAS 87 versus APB 8. Journal of Accounting, Auditing and Finance 8, 427-446. Arya. A., Glover, J. and Sunder, S., 1998. Earnings management and the revelation principle. Review of Accounting Studies 3, 7-34. Baber, W., Kang, S. and Krishna, K., 1998. Accounting earnings and executive compensation: The role of earnings persistence. Journal of Accounting and Economics 25, 169-193. Ball, R., and Brown, P., 1968. An empirical evaluation of accounting income numbers. Journal of Accounting Research 6, 159-178. Balsam, S., 1998. Discretionary accounting choices and CEO compensation. Contemporary Accounting Research 15, 229-252. Balsam, S., Haw, I. And Lilien, S., 1995. Mandated accounting changes and managerial discretion. Journal of Accounting and Economics 20, 3-29. Barth, M., Elliot, J. and Finn, M., 1999. Market rewards associated with patterns of increasing earnings. Journal of Accounting Research 37, 387-413. Bartov, E., 1993. The timing of asset sales and earnings manipulation. The Accounting Review 68, 840-855. Beatty, A., Chamberlain, S. and Magliolo, J., 1995. Managing financial reports of commercial banks: the influence of taxes, regulatory capital and earnings. Journal of Accounting Research 33, 195-212. Beaver, W. and Engel, E., 1996. Discretionary behavior with respect to allowances for loan losses and the behavior of security prices. Journal of Accounting and Economics 22, 177-206. Beneish, M., 1997. Detecting GAAP violation: implications for assessing earnings management among firms with extreme financial performance. Journal of Accounting and Public Policy 16, 271-309. Bernard, V. and Skinner, D., 1996. What motivates managers’choice of discretionary accruals? Journal of Accounting and Economics 22, 313-325. Bizjak, J., Brickley, J. and Coles, J., 1993. Stock-based incentive compensation and investment. Journal of Accounting and Economics 16, 349-372. - 59 - Black, E., Sellers, K. and Manly, T., 1998. Earnings management using asset sa les: an international study of countries allowing noncurrent asset revaluation. Journal of Business Finance and Accounting 25, 1287-1317. Bowen, R., Burgstahler, D. and Daley, L., 1986. Evidence on the relationships between earnings and various measures of cash flow. The Accounting Review 61, 713¯725. Burgstahler, D. and Dichev, I., 1997. Earnings management to avoid earnings decreases and losses. Journal of Accounting and Economics 24, 99-126. Cahan, S., 1992. The effect of antitrust investigations on discretionary accruals: A refined test of the political cost hypothesis. The Accounting Review 65, 77-95. Cahan, S., Chavis, B. and Elmendorf, R., 1997. Earnings management of chemical firms in response to political costs from environmental legislation. Journal of Accounting, Auditing and Finance 12, 37-65. Collins, J., Shackelford, D. and Wahlen, J., 1995. Bank differences in the coordination of regulatory capital, earnings and taxes. Journal of Accounting Research 33, 263-291. Coughlan, A. and Schmidt, R., 1985. Executive compensation, management turnover, and firm performance. Journal of Accounting and Economics 7, 43-66. DeAngelo, H., DeAngelo, L. and Skinner, D., 1994. Accounting choice in troubled companies. Journal of Accounting and Economics 17, 113-144. DeAngelo, L., 1986. Accounting numbers as market valuation substitutes: A study of management buyouts of public stockholders. The Accounting Review 61, 400-420. Dechow, P. and Sloan, R., 1991. Executive incentives and the horizon problem: An empirical investigation. Journal of Accounting and Economics 14, 51-89. Dechow, P., Sloan, R. and Sweeney, A., 1995. Detecting earnings management. The Accounting Review 70, 193¯225. Dechow, P., Sloan, R. and Sweeney, A., 1996. Causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research 13, 1¯36. DeFond, M. and Jiambalvo, J., 1994. Debt covenant violation and manipulation of accruals. Journal of Accounting and Economics 17, 145¯176. DeFond, M. and Park, C., 1997. Smoothing income in anticipation of future earnings. Journal of Accounting and Economics 23, 115-139 Degeorge, F., Patel, J. and Zeckhauser, R., 1999. Earnings management to exceed thresholds. Journey of Business 72, 1-33. - 60 - Demski, J., 1998. Performance measure manipulation. Contemporary Accounting Research 15, 261-285. Eldenburg, L. and Soderstrom, N., 1996. Accounting system management by hospital operating in a changing regulatory environment. The Accounting Review 71, 23-42. Erickson, E. and Wang, S., 1999. Earnings management by acquiring firms in stock for stock mergers. Journal of Accounting and Economics 27, 149-176. Elliott, J. and Hanna, J., 1996. Repeated accounting write -offs and the information content of earnings. Journal of Accounting Research 34 (Supplement), 135-156. Evans, J. and Sridhar, S., 1996. Multiple control systems, accrual accounting, and earnings management. Journal of Accounting Research 34, 45-65. Francis, J., Hanna, J. and Vincent, L., 1996. Causes and effects of discretionary asset write-offs. Journal of Accounting Research 34 (Supplement), 117-134. Friedlan, J., 1994. Accounting choices by issuers of initial public offerings. Contemporary Accounting Research 11, 1¯31. Gaver, J., Gaver, K. and Austin, J., 1995. Additional evidence on bonus plans and income management. Journal of Accounting and Economics 19, 3-28. Gaver, J. and Gaver, K., 1998. The relation between nonrecurring accounting transactions and CEO cash compensation. The Accounting Review 73, 235-253. Guay, W., Kothari, S. and Watts, R., 1996. A market-based evaluation of discretionary accrual models. Journal of Accounting Research 34 (Supplement), 83-105. Guenther, D., 1993. Earnings management in response to corporate tax rate changes: evidence from the 1986 tax reform act. The Accounting Review 69, 230-243. Guidry, F., Leone, A. and Rock, S., 1999. Earnings-based bonus plans and earnings management by business-unit managers. Journal of Accounting and Economics 26, 113-142. Hall, S., 1993. Political scrutiny and earnings management in the oil refining industry. Journal of Accounting and Public Policy 12, 325-351. Hall, S. and Stammerjohan, W., 1997. Damage awards and earnings manageemnet in the oil industry. The A ccounting Review 72, 47-65. Han, J. and Wang, S., 1998. Political costs and earnings management of oil companies during the 1990 Persian Gulf crisis. The Accounting Review 73, 103¯117. Healy, P., 1985. The impact of bonus schemes on the selection of accounting principles. Journal of Accounting and Economics 7, 85¯107. - 61 - Holthausen, R., Larcker, D. and Slaon, R., 1995. Annual bonus schemes and the manipulation of earnings. Journal of Accounting and Economics 19, 29-74. Jiambalvo, J., 1996. Discussion of ca uses and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research 13, 37¯47. Jones, J., 1991. Earnings management during import relief investigations. Journal of Accounting Research 29, 193¯228. Kang, S. and Sivaramakrishnan, K., 1995. Issues in testing earnings management and an instrumental variable approach. Journal of Accounting Research 33, 355-367. Kaplan, P., 1985. Comments on Paul Healy: Evidence on the effect of bonus schemes on accounting procedure and accrual decisions. Journal of Accounting and Economics 7, 109-113. Kasanen, E., Kinnunen, J. and Miskanen, J., 1996. Divident-based earnings management: empirical evidence from Finland. Journal of Accounting and Economics 22, 283-312. Kasznik, R., 1999. On the association between voluntary disclosure and earnings management. Journal of Accounting Research 37, 57-81. Key, K., 1997. Political cost incentives for earnings management in the cable television industry. Journal of Accounting and Economics 23, 309-337. Kumar, K., Ghicas, D. and Pastena, V., 1993.Earnings, cash flows and executive compensation: An exploratory analysis. Managerial Finance 19, 55-74. Lambert, R., 1993. The use of accounting and security price measures of performance in managerial compensation contracts. Journal of Accounting and Economics 16, 101123. Lambert, R. and Larcker, D., 1987. An analysis of the use of accounting and market measures of performance in executive compensation contracts. Journal of Accounting Research 25, 85-126. LaSalle, R., Jones, S. and Jain, R., 1993. The association between executive succession and discretionary accounting changes: earnings management or different perspectives? Journal of Business Finance & Accounting 22, 497-520. Lewellen, W., Loderer, C. and Martin, K., 1987. Executive compensation and executive incentive problems. Journal of Accounting and Economics 9, 287-310. Liberty, S. and Zimmerman, J., 1986. Labor union contract negotiations and accounting choices. The Accounting Review 61, 692 ¯712. Maydew, E., 1997. Tax-induced earnings management by firms with net operating losses. Journal of Accounting Research 35, 83-96. - 62 - Mensah, Y., Considine, J. and Oakes, L., 1994. Statutory insolvency regulations and earnings management in the prepaid health-care industry. The Accounting Review 69, 70-95. McNichols, M., 2000. Research design issues in earnings management studies. Journal of Accounting and Public Policy 19, 313-345. McNichols, M. and Wilson, G., 1998. Evidence of earnings management from the provisions for bad debts. Journal of Accounting Research , Supplement 26, 1-31. Miller, G. and Skinner, D., 1998. Determinants of the valuation allowance for deferred tax assets under SFAS No. 109. The Accounting Review 73, 213-233. Murphy, K. and Zimmerman, J., 1993. Financial performance surrounding CEO turnover. Journal of Accounting and Economics 16, 273-315. Natarajan, R., 1996. Stewardship value of earnings components: additional evidence on the determinants of executive compensation. The Accounting Review 71, 1-22. Peasnell, K., Pope, P. and Young, S., 2000. Detecting earnings management using cross-sectional abnormal accrual models. Accounting and Business Research 30, 313326. Perry, S. and Williams, T., 1994. Earnings management preceding management buyout offers. Journal of Accounting and Economics 18, 157¯179. Petroni, K., Ryan, S. and Wahlen, J., 1999. Discretionary and non-discretionary revisions of loss reserves by property-casualty insurers: Differential implications for future profitability, risk, and market value. Working paper, Indiana University. Pourciau, S., 1993. Earnings management and nonroutine executive changes. Journal of Accounting and Economics 16, 317-336. Rees, L., Gill, S. and Gore, R., 1996. An investigation of asset write-downs and concurrent abnormal accruals. Journal of Accounting Research 34 (Supplement), 157170. Schipper, K., 1989. Commentary on earnings management. Accounting Horizons 3, 91-102. Shivakumar, L., 2000. Do firms mislead investors by overstating earnings before seasoned equity offerings? Journal of Accounting and Economics 29, 339-371. Soo. B., 1999. Accrual response to mandated accounting principles: the case of SFAS No. 2 and 34. Journal of Accounting and Public Policy 18, 59-84. Subramanyam, K., 1996. The pricing of discretionary accruals. Journal of Accounting and Economics 22, 249-281. - 63 - Sweeney, P., 1994. Debt-covenant violations and managers’accounting responses. Journal of Accounting and Economics 17, 281-308. Teoh, S., Welch, I. and Wong, T., 1998. Earnings management and the long run underperformance of seasoned equity offerings. Journal of Financial Economics 50, 53¯100. Teoh, S., Welch I., Wong, T., 1998. Earnings management and the long run market performance of initial public offerings. Journal of Finance 53, 1935-1974. Trueman, B. and Titman, S., 1988. An explanation for accounting income smoothing. Journal of Accounting Research , Supplement 26, 127-143. Wahlen, J. 1994. The nature of inf ormation in commercial bank loan loss disclosures. The Accounting Review 69, 455-478. Watts, R. and Zimmerman, J., 1986. . Positive accounting theory Prentice-Hall, Englewood Cliffs, NJ. Watts, R. and Zimmerman, J., 1990. Positive accounting theory: A ten-year perspective. The Accounting Review 65, 131-156. Weisbach, M., 1988. Outside directors and CEO turnover. Journal of Financial Economics 20, 431-460. Wu, Y., 1997. Management buyouts and earnings management. Journal of Accounting, Auditing and Finan ce 12, 373-390. - 64 - [...]... are very interested in how CEOs behave just before their departing the post of CEO and just after their assuming the post of CEO The short horizon of CEOs’departure provides researchers a good opportunity to examine CEOs’earning management incentives Explanations of Earnings Management Associated with CEO Turnover Three main explanations of earnings management associated with CEO turnover that have been... controversial nature of outgoing CEOs’incentives to manipulate earnings, I propose my hypothesis in null form: H1 (Null Hypothesis): The CEOs who resign do not mana ge earnings upward before their resignations 3.3 Hypotheses for Earnings Management After CEO Resignation While it is debatable whether outgoing CEOs have incentives to manipulate earnings upward or downward before their resignation, there is no... efforts to earnings management Their research has defined the concepts of earnings management, explored the different motivations of earnings management and developed models to detect earnings management Among all the available models, the Jones Model has been the most frequently used Very few studies have investigated earnings management associated with CEO turnover, and even fewer have classified CEO turnover... examines the earnings management of the incoming CEO and finds that incoming CEOs manipulate earnings downward and upward in different periods after the departure of the old CEOs This suggests that boards of directors should be alert to earnings manipulation in those periods The remainder of the study is organized as follows: Chapter two presents an overview of the issue of earnings management and reviews... benefits from earnings management For example, managers can boost stock price, increase earnings- based bonus awards and avoid regulation by means of earnings management I will discuss these motivations of earnings management in detail below 2.3 Motivations for Earnings Management The motivations for earnings management are very important to researchers in this field Only with a good understanding of the... 2.5 Earnings Management Associated with CEO Turnover 2.5.1 CEOs’Incentives and Methods to Manipulate Earnings CEOs’compensation contracts normally contain incentive provisions that link CEOs’ compensation to firms’ accounting -earnings performance Therefore, researchers predict that the usage of these compensation contracts will induce CEOs to engage in earnings management to boost their salary and. .. Pourciau (1992), CEO resignations are divided into two groups: forced resignations and voluntary resignations The author hypothesizes that both groups of outgoing CEOs will manipulate reported earnings upwards before their resignations to increase their compensation prior to departure She argued that while CEOs who resign voluntarily are in full control of the timing of their resignations, CEOs who are... CEO turnover and conducted the fur ther study This relative blank field give us incentives to do some research in deep - 17 - Chapter Three Hypothesis Development Chapter Three: Hypothesis Development 3.1 Limitations of Prior Research While many studies have investigated earnings management associated with CEO turnover, few have examined specifically earnings management before and after CEO resignations... very little is known about how firms manage earnings in periods before and after a CEO resignation The only study that we are aware of, which focuses on CEO resignation, is done by Pourciau (1992) It was published more than a decade ago Pourciau (1992) classified CEO turnover as routine and nonroutine and examined evidence of earnings management associated with “nonroutine” executive changes Her results... operations and the write-off of unprofitable divisions can be attributed to incoming CEOs who implicitly blame their predecessors for past performance - 13 - Chapter Two Literature Review Classification of CEO Turnover and Earnings Management While many studies have investigated earnings management associated with CEO turnover in general, very little is known about how firms manage earnings in periods before ... investigated earnings management associated with CEO turnover, few have examined specifically earnings management before and after CEO resignations In a study by Pourciau (1992), CEO resignations... Manipulate Earnings 11 2.5.2 Earnings Management Associated with CEO Turnover 12 Explanations of Earnings Management Associated with CEO Turnover 12 Classification of CEO Tu rnover and Earnings Management. .. Burgstahler and Dichev (1997) and Degeorge, Patel and Zeckhauser (1999) 2.5 Earnings Management Associated with CEO Turnover 2.5.1 CEOs’Incentives and Methods to Manipulate Earnings CEOs’compensation