The interaction between CEO tenure and conservative accounting

(1)

Amsterdam Business School

Master thesis

The interaction between CEO tenure and

conservative accounting

MSc Accountancy and Control, specialization Accountancy (MSc ACC) Faculty of Economics and Business, University of Amsterdam

Name: Madhvie Tedjai Student number: 10682082 Date: June 20, 2016

Thesis supervisor: Dr. G. Georgakopoulos

(2)

Statement of Originality

This document is written by student Madhvie Tedjai who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This study examines the interaction between CEO tenure and conservative accounting in the period before and after the passage of the Sarbanes-Oxley act (SOX) in 2002. This research is motivated by the lack of studies examining the practice of conservative accounting during a CEOs’ tenure. I predict that CEOs have incentives, due to career concerns, to use more conservative accounting in the beginning of their career compared with the other years of their career. Additionally, I predict that due to the risk of higher litigation costs caused by the passage of SOX, CEOs use more conservative accounting in the post-SOX period. Using both a market-based and an accrual-based measure of conservative accounting, I find no evidence which supports the notion that CEOs use more conservative accounting in the early years of their career compared with the final year of their career. However, the use of negative accruals by CEOs after the passage of SOX did increase. This is consistent with the notion that after the passage of SOX the risk of getting higher litigation costs leads to the use of more conservative accounting. However, no associations are found between the passage of SOX and a lower book-to-market ratio. After conducting several sensitivity analyses the findings remain the same. The findings contribute to the gap in previous literature and complement the findings by Ali and Zhang (2015) that CEOs have incentives to overstate earnings in the beginning of their career.

(4)

1 Introduction

This study examines the interaction between CEO tenure and conservative accounting in the period before and after the passage of the Sarbanes-Oxley act (SOX) in 2002. Conservatism within accounting and financial reporting has been the subject of several prior studies. Looking at the relation between conservative accounting and CEOs, prior studies primarily focus on the interaction between conservative accounting and CEO characteristics such as, the gender of a CEO, whether a CEO is also the Chairman of the Board of Directors

(CEO-duality), managerial ownership, and CEO power. However, the link between CEO tenure and conservative accounting is still missing in prior literature. This research is motivated by the lack of studies examining the practice of conservative accounting during a CEOs’ tenure.

Prior studies mainly focus on the relation between CEO tenure and earnings

management. Ali and Zhang (2015) study the interaction between CEO tenure and earnings management. According to the authors, a CEOs’ ability is linked to higher future

compensation, reappointments, and managerial autonomy. Because of these career concerns, the authors hypothesize that CEOs will engage in more earnings management in the first three years of their career than in the later years of their career. The authors also state that CEOs will overstate their earnings in the final year of their service to boost their final year pay. They also examine the effect of monitoring on the relation between CEO tenure and earnings management, because heavy monitoring will mitigate opportunistic behavior of a CEO. Therefore, they hypothesize that for CEOs who are subject to stronger monitoring, the variance in the use of earnings management between the early and later years of a CEOs’ tenure is smaller. Their findings suggest that the use of earnings management is higher in the first three years and in the final year of a CEOs’ tenure. Another finding of their study is that stronger monitoring of CEOs, mitigates the effect that earnings overstatement is significantly higher in the early years of a CEOs’ service than in the later years of a CEOs’ service.

Furthermore, Hermalin and Weisbach (1998) argue that the ability of a CEO is not clear yet at the beginning of their tenure. Therefore, the CEO is strongly monitored at the beginning of their career. Becoming more competent as a CEO, leads to less monitoring and more bargaining power. Additionally, Garcia Lara et al. (2009) show that CEOs with more bargaining power engage less in accounting conservatism. CEOs are also concerned about their reputation. Francis et al. (2008) argue that CEOs want to establish a good reputation for themselves which can be achieved by engaging in conservative accounting. Subsequently, at the final year of a CEOs’ career the CEO tends to overstate earnings in order to boost their

(6)

6 final year pay (Ali and Zhang, 2015) and to maintain their reputation (Francis et al. 2008).

So, the use of conservative accounting can positively contribute to the desired

perceived ability of a CEO leading to a good reputation. Nowadays conservative accounting is still an important topic as it still increases over the years (Givoly and Hayn, 2000) even

though there is a discussion about whether conservatism jeopardizes the reliability and integrity of information within financial reporting. Under conservative accounting profits are recognized only when it is highly verifiable and losses are recognized when they are

discovered leading to an understatement of net assets and earnings. Conservative accounting can be unconditional (ex ante or news-independent), referring to decreasing earnings and book-value of net assets, which is independent of economic news, through accounting

methods and policies (Pae, 2007). Conservative accounting can also be conditional (ex post or news dependent), referring to accounting methods and policies that verifies bad news as losses faster than verifying good news as gains (Pae, 2007).

According to Watts (2003), firms can benefit from conservative accounting because it mitigates moral hazard arising within contracting, reduces potential litigation costs, reduces the present value of taxes, and reduces political costs. Looking deeper into the contracting theory, agency costs arise because a CEO has private information that is not available to shareholders. CEOs can use this information advantage to manipulate earnings and other financial information leading to a shift of the generated wealth to themselves. Mitigating this opportunistic behavior can be achieved through conservative accounting. Watts (2003) argues that conservative accounting constrains CEOs’ opportunities for opportunistic accounting choices, which are used in contracts. These contracts contain performance measures to have more stricter verification standards for gains than for losses which results in losses being recognized faster compared to gains. This results in avoiding overstated earnings and net assets, reducing the likelihood of opportunistic behavior that violates contracts. This all will lead to an increase of firm value. Furthermore, as said before, a good reputation can be achieved through conservative accounting. This is due to the fact that the users of financial statements and many accountants prefer conservative accounting. The FASB Statement of Concept No. 2 discusses that the users of financial statements desire conservative reporting, because they prefer that possible errors in measurement are caused by understating net income and net assets rather than overstating. They believe that the greater the understatement of assets the greater the margin of safety the assets provide as security for loans or other debts. Purposely understating net assets and earnings is seen as a virtue by many accountants and users of financial statements.

(7)

7 However, there are also concerns regarding the use of conservatism. The FASB

Statement of Concepts No.2 states that conservatism needs to be applied with caution since it jeopardizes the information being faithful, comparable, and neutral. A frequent deliberate understatement of assets may result in overstated income in the future, because when the assets are subsequently realized the earnings can be overstated leading to a surplus.

Conservatism influences the timing of earnings and losses which can mislead investors. There is a lot of discussion whether conservatism yields any benefits for firms (Devine, 1963; Sterling, 1676). According to Hendriksen and Van Breda (1992), conservatism can lead to misstatement of accounting information, and according to Penman and Zhang (2002) conservatism can decrease the quality of earnings.

Despite the controversy surrounding conservative accounting, conservatism increased over the years. According to prior literature this can be a response to the Sarbanes-Oxley act (SOX) which became mandatory in 2002 in the US as a reaction to the major financial scandals (e.g. Enron, Worldcom, Tyco) in the early 2000s. Therefore, in this study I will also test for SOX since it affects conservative accounting during CEO tenure. According to prior literature SOX has significantly influenced financial reporting and disclosures (Gordon et al., 2006; Lobo and Zhou, 2006; Cohen et al., 2008; Zhou, 2008). The increase in conservatism over the years even though there is a discussion about the use of conservatism, provides an interesting topic for further research. It is interesting to see how and why conservatism is still used as a tool by CEOs during their tenure. The research question that will lead through the research is: to what extent do CEOs use conservative accounting during CEO tenure before

and after SOX was conducted? Adopting the statements that CEOs (1) experience career

concerns at the beginning of their career; (2) have a reputation at stake at the end of their career; and (3) opt for a large bonus in the final year of their career, I predict that CEOs are more conservative at the beginning of their career than at the end of their career. Furthermore, I predict that the use of conservative accounting by CEOs will increase after the passage of SOX due to potential higher litigation costs SOX imposes on CEOs.

I test the predictions using a sample of 4,462 U.S. firm-year observations from the 1996-2014 time period. Conservatism is measured using an accrual-based measure and a market-based measure. The accrual-based measure is based on Givoly and Hayn (2000) and describes that conservative accounting leads to negative accruals. The market-based measure is based on Beaver and Ryan (2000) who argue that conservatism leads to an understatement of net assets, which leads to a lower book-to-market ratio.

(8)

8 I find no evidence that supports the notion that CEOs are more conservative at the early years of their career than at the end of their career. Results found using the accrual-based measure show the opposite of the prediction and presents that CEOs use less conservative accounting in the early years of their career and use more conservative accounting in the final year of their career. This finding can be explained by the results of Ali and Zhang (2015) who show that CEOs engage in more earnings management in the early years of their career than at the end of their career due to career concerns. For the market-based measure, I find no relation at all between CEO tenure and a lower book-to-market ratio. Furthermore, the results found regarding the accrual-based measure supports the notion that CEOs are more conservative after the passage of SOX due to higher litigation caused by SOX. Regarding the market-based measure, I find no relation between the passage of SOX and a lower book-to-market ratio. After several sensitivity analyses the findings remain robust.

My study contributes to existing literature in several ways. First, prior studies

primarily focus on the relation between CEO tenure and earnings management and do not address conservative accounting, resulting in a gap in prior literature. This study is to my knowledge the first research that focusses on the relation between CEO tenure and accounting conservatism. With this study I try to fill the gap in existing literature by giving a never before studied analysis of conservative accounting during CEO tenure with an additional focus on the effects of SOX.

As said earlier, conservative accounting increased over time despite the arguments against the use of conservatism in financial reporting. This research can contribute to the ongoing discussion and prior literature by providing insights about the incentives CEOs have to use conservative accounting during their tenure. I examine the use of conservative

accounting in the first three years and the final year of a CEOs’ career and show that CEOs have more incentives to be less conservative in the early years of their career compared to the final year of their career. This evidence complements Ali and Zhang’s (2015) study, in which they find that CEOs overstate earnings in the first three years of their tenure. With this research I also try to expand the research done by Ali and Zhang (2015). My research differs from the research conducted by Ali and Zhang (2015), because I examine the interaction between CEO tenure and conservative accounting. Ali and Zhang (2015) use earnings

management as a characteristic of earnings quality. Instead, I use conservative accounting as a proxy for earnings quality. Furthermore, they do not take the effects of SOX into account.

The remainder of this paper is organized as follows. Section 2 describes the related literature. Derived from the literature review, hypotheses are formulated and presented.

(9)

9 Section 3 discusses the research design. Section 4 presents the data, regression analyses and the subsequent results. Section 5 discusses the additional sensitivity analyses. Finally, section 6 concludes the paper.

2 Literature review and hypotheses

In this chapter literature will be provided concerning conservative accounting, CEO tenure and SOX. Finally, this literature review will be used to formulate hypotheses in this section.

2.1 Conservative accounting

Conservatism is a frequently used qualitative characteristic of financial reporting in previous academic research (Basu, 1997). According to the Financial Accounting Standard Board (FASB), managers, directors, and auditors favor that possible errors in measurement are a consequence of understating net income and net assets rather than an overstatement of net income and net assets and. They also believe that the greater the understatement of assets the greater the margin of safety the assets provide as security for loans or other debts. There are several definitions of conservative accounting defined in previous research. The FASB considers conservatism as a ‘prudent reaction to uncertainty to try to ensure that uncertainties and risks inherent in business situations are adequately considered’ (FASB Statement of Concepts No.2). The general definition of accounting conservatism in previous literature is ‘anticipate no profit, but anticipate all losses’ (Watts, 2003). This means that profits can be recognized only when it is highly verifiable. Losses can be recognized when they are discovered. Basu (1997) introduces the differential verification definition, which means the tendency to require a higher degree of verification to recognize good news as gains than to recognize bad news as losses. The greater the difference in the degree of verification requires for gains versus losses, the greater the conservatism (Watts, 2003; Basu, 1997). Earnings reflects bad news more quickly than good news which leads to systematic differences between bad and good news periods in the timeliness and persistence of earnings (Watts, 2003; Basu, 1997).

There are two types of conservatism. Beaver and Ryan (2005) argue that conservatism can be divided in unconditional conservative accounting and conditional conservative

accounting. Unconditional conservatism is ex ante or news-independent. This refers to decreasing earnings and book-value of net assets, which is independent of economic news, through accounting methods and policies (Pae, 2007). Examples of unconditional

(10)

10 conservative accounting are instantly recognizing the costs of advertising and R&D

expenditures and historical cost accounting for positive net present value projects (Beaver and Ryan, 2005; Pae, 2007). Conditional conservatism is referred to as ex post or news dependent. This means adopting accounting methods and policies that verifies bad news as losses faster than verifying good news as gains (Pae, 2007). The conservatism measure of Basu (1997) can be categorized as a conditional conservative accounting measure.

So, in what way can firms benefit from conservative accounting? According to Watts (2003) the need to use conservative accounting can be explained through four theories: (1) contracting: within contracts, moral hazard problems still exists due to parties having asymmetric information, asymmetric payoffs, limited horizons and limited liability.

Conservative accounting constrains managers’ opportunistic behavior within these contracts. The use of conservative accounting mitigates the risk that these contracts are violated; (2) shareholder litigation: litigation occurs when earnings and net assets are overstated, because overstated earnings and net assets will lead to scrutiny. Conservative reporting leads to an understatement of net assets which reduces potential litigation costs, resulting in incentives for management to disclose conservative values for earnings and net assets; (3) taxation: the asymmetric recognitions of gains and losses, which defines conservatism, helps reducing the present value of taxes. When a firm is profitable, an incentive arises to defer income to reduce the present value of taxes. This leads to an understatement of net assets; (4) accounting

regulation: when firms overstate net assets this can lead to political costs imposed by standard setters like the SEC and the FASB. These political costs can be constrained by firms, through using more conservative accounting, because conservatism leads to an understatement of net assets.

2.2 Agency theory

The agency theory describes the structure of the economic exchange between two parties: the principal and the agent. The principal delegates the decision rights and compensates the agent who performs a task. The agency theory describes a problem of information asymmetry which exists when the agent (e.g., a CEO) has an information advantage over the principal (e.g., shareholders). Fama and Jensen (1983) state that in order for firms to survive it is important that they can control these agency problems. This theory predicts how parties write contracts to protect their own interests. Agency problems exists, because these contracts generates agency costs which contains the costs of structuring, monitoring, and bonding a set of

(11)

11 cost of full enforcement of contracts exceeds the benefits (Fama and Jensen, 1983).

So, contracts cannot fully mitigate agency conflicts. Prior literature shows that accounting conservatism can be a tool in contracting to help reducing agency costs. LaFond and Watts (2008) argue that conservative accounting constrains CEO opportunistic behavior (manipulating and overstating financial performance) and enhances firm cash flow and firm value. The results of their study suggests that an increase in agency costs is associated with an increase in accounting conservatism. They also find that equity investors use conservative accounting to reduce agency costs concluding that conservative accounting assists in reducing agency costs.

Chi et al. (2007) study the interaction between accounting conservatism and corporate governance. The study focusses on whether conservatism is a substitute for, or a complement to corporate governance. If conservatism is a substitute for corporate governance this means that conservative accounting is used to limit information asymmetry, which indicates that firms with more agency problems have a higher demand for conservative accounting. On the contrary, if conservatism is a complement for corporate governance, this means that firms with a high quality of corporate governance leads to better monitoring. The results of their research show that accounting conservatism alternates other corporate governance aspects in reducing agency problems. One of the findings is that in firms that adopt CEO-duality, which is an indication of weak corporate governance, there is a bigger demand for conservative accounting. They also find that when institutional investors own a great amount of shares, the demand for conservative accounting is lower. However, these results are weaker when a change analysis is conducted.

Also, Ahmed and Duellman (2007) find that accounting conservatism can help the board in reducing agency costs. Their research studies the interaction between conservative accounting and board characteristics, such as percentage of inside directors, amount of additional directorships held by directors, CEO-duality, percentage of shares held by outside directors, and board size. In their study, one of their primary findings mentioned is that boards with a higher percentage of inside directors are associated with less conservative accounting than boards with a lower percentage of inside directors. They state that if a high percentage of inside directors takes seat in the board it is a signal of a weak and

non-independent board, which is negatively related to conservative accounting. Another primary finding of their study is that a higher percentage of shares held by outside directors is associated with more conservative accounting. Shares held by outside directors act like an incentive in enhancing monitoring by outside directors. When outside directors have equity

(12)

12 stakes in the firm they will monitor in the best interest of the firm, because they will also benefit from positive firm performance. So, a low percentage of inside directors on the board and a high percentage of shares held by outside directors generates more conservative

accounting which acts as a tool in reducing agency costs.

2.3 The interaction between CEO tenure and conservative accounting

Not much attention is paid to the relationship between conservative accounting and CEO tenure resulting in a gap in prior literature. Ali and Zhang (2015) study the interaction between CEO tenure and earnings management. Several prior studies focus on the first and last year of a CEOs’ tenure. Therefore, the authors’ focus lies on the other years of a CEOs’ tenure. They specifically study the use of earnings management by CEOs during their career, with a focus on the first three years (as a proxy for early years) and the remainder of a CEOs’ years of service. According to the authors, a CEOs’ ability is linked to higher future

compensation, reappointments, and managerial autonomy. This means that the markets’ perception of a CEOs’ ability will influence these long-term benefits for CEOs. Because of these career concerns, the authors hypothesize that CEOs will engage in more earnings management in the first three years of their career than in the later years of their career. By means of measuring the use of discretionary accruals and abnormal discretionary expenses by CEOs during their career, they try to test their hypothesis. They predict that CEOs will

overstate their earnings more in the early years of their service in contrast with the later years of their service, leading to discretionary accruals being greater in the first three years of a CEOs’ service, and abnormal discretionary expenses being smaller in the first three years of a CEOs’ service. The authors also state that CEOs will overstate their earnings in the final year of their service to boost their final year pay, leading to discretionary accruals being large in the final year of a CEOs’ service, and abnormal discretionary expenses being small in the final year of a CEOs’ service. In their study, the authors also consider the effect of

monitoring. They examine the effect of monitoring on the relation between CEO tenure and earnings management, because heavy monitoring will mitigate opportunistic behavior of CEOs. Therefore, they hypothesize that for CEOs who are subject to stronger monitoring, the variance in using earnings management between the early and later years of a CEOs’ tenure is smaller. For their sample period 1992-2010, they find that the use of discretionary accruals is significantly higher and abnormal discretionary expenses are significantly lower in the early years of a CEOs’ tenure as compared to the later years of a CEOs’ service. Another finding of their study is that stronger monitoring of CEOs mitigates the effect that earnings

(13)

13 overstatement is significantly higher in the early years of a CEOs’ service than in the later years of a CEOs’ service. Referring to the horizon problem of the departing CEOs, they initially find no evidence that supports the notion that overstated earnings are higher in the final year of a departing CEO compared to the other years. After controlling for earnings overstatement in the early years of the CEOs’ service, the overstatement becomes

significantly higher in the final year compared to the other years of a CEOs’ tenure supporting the notion that CEOs will overstate the earnings in their final year to boost the final year pay. After certain sensitivity analyses their findings still support their predictions.

Regarding the ability of a CEO, Hermalin and Weisbach (1998) argue that the ability of a newly appointed CEO is not quite clear yet, which leads to heavy monitoring by the board, in contrast with an already serving CEO. A newly appointed CEO has no bargaining power, since the value of the CEO is not known yet. When a CEO proceeds further in his career he gains more competence, which leads to less monitoring by the board, because the ability of the CEO is clear now. The more years a CEO serves, the less scrutiny he receives and the lower the standards are for the CEO. This can lead to the board not detecting

problems timely and not replacing the CEO when necessary. When a CEO after several years of service is not replaced yet, it gives a signal that the board thinks the CEO is of high value. Because of this signal, the CEO attains a stronger bargaining position and can demand a board which he can use for his own benefits, getting more control over the board because now he has more influence on director nomination and election, leading to a board that becomes less independent over the years of the CEOs’ career. Less independent directors are less willing to monitor the CEO. The board has several reasons for not monitoring the CEO intensively. Firstly, the career of an inside director depends on the career of the CEO. Leading to not monitoring the CEO intensively for their own interests. Secondly, the opportunity cost of the directors' time will differ among outside directors. At last, directors who desire to practice interlocking directorate could have incentives to create a reputation for not intensively monitoring a CEO. Because of the lower level of monitoring in the later years of CEOs’ service, the CEO has more room for opportunistic behavior and therefore engages less in conservative accounting.

Looking at the interaction between conservative accounting and CEO power, Garcia Lara et al. (2007) investigate amongst other things the effect of CEO involvement in the board on conservative accounting. The researchers state that CEOs who have substantial bargaining power in the board are associated with a weak governance structure because more power over the board leads to a less independent board and firms with a strong governance structure

(14)

14 engages more in conservative accounting. Their research shows that firms with a CEO that has little control over the board of directors engages more in conservative accounting in contrast with a CEO that has substantial control over the board of directors. Taking the

previous into consideration, I expect that CEOs engage more in conservative accounting at the beginning of their career than at the final years of their service due to more intensively

monitoring in the first years and more generated CEO bargaining power at the final years. Not only the degree of monitoring and bargaining power during the CEO tenure influences the behavior of a CEO, but also the career concerns a CEO has are influential. Gibbons and Murphy (1992) argue that CEOs willingness to engage in conservative accounting depends on their career concerns. As said earlier, the ability of a CEO at the beginning of their service is not known yet. CEOs have career concerns, because the

performance of a CEO is used to determine the ability of a CEO and based on that ability, a firm determines the wages a CEO will receive. Through the years of service the perceived ability of a CEO is frequently updated which affects the wages. Therefore CEOs have incentives at the beginning of their career to take actions which influences their performance in a positive direction, which leads to an increase in the perceived ability, ultimately resulting in higher wages. In the final year of the CEOs’ service the CEO uses less conservative

accounting which overstates assets and earnings, in order to give an extra boost to final years bonus and wage (Ali and Zhang, 2015). Therefore, I state that at the beginning of the CEOs’ service the CEO uses more conservative accounting to positively influence the perceived ability and wages than at the last year of their service.

CEOs are also concerned about their reputation which influences their behavior during their career. CEO reputation is influenced by the performance and perceived ability of a CEO. Newly appointed CEOs want to establish a good reputation for themselves (Francis et al. 2008). Francis et al. (2008) describe the relation between CEO reputation and earnings quality. Their research is based on three theories which are efficient contracting theory, rent extraction theory and the matching theory (the latter will not be discussed in this research). They measure earnings quality with discretionary earnings quality and total earnings quality. They explain that in terms of the “efficient contracting” perspective, CEOs with a high reputation at stake will not indulge in opportunistic behavior for two reasons. The first reason is that CEOs with a high reputation at stake have more to lose in terms of their own human capital. The second reason is that opportunistic behavior will lead to weakened quality of earnings which will result in higher costs of capital. To maintain their level of reputation they want to avoid a higher costs of capital. On the other hand the “rent extraction” perspective

(15)

15 discusses that CEOs with a larger reputation at stake will engage more in opportunistic

behavior leading to higher performance, because frequently underperforming will weaken their reputation. The results of the researchers shows that CEOs with a high reputation are linked to both poorer discretionary earnings quality and poorer total earnings quality, which supports the rent extraction theory. As said earlier, accounting conservatism is also considered as a characteristic of earnings quality. The behavior of the CEO will influence accounting conservatism and will result in lower or higher earnings quality. Because I stated that at the beginning of a CEOs’ career the ability of a CEO is unknown and therefore the CEO is not yet reputed, I will adopt the rent extraction theory which says that at the end of a CEOs career the CEO has built a reputation for themselves which they do not want to weaken. This means that they will overstate earnings at the end of their career leading to less conservative

accounting.

2.4 Sarbanes-Oxley act

The Sarbanes-Oxley act (SOX) became mandatory in 2002 in the US as a reaction to the major financial scandals (e.g. Enron, Worldcom, Tyco) in the early 2000s. Caused by these scandals, investors lost confidence in the financial statements and this led to major changes in regulation and legislation. Nowadays, SOX helps to protect investors against fraudulent accounting activities by companies through employing severe penalties for fraudulent activities, by increasing the independence oversight role of the boards, and by improving the accuracy and reliability of corporate disclosures. A lot of research focusses on the relation between SOX and conservative accounting. The main finding is that conservative accounting has increased after the passage of SOX (Lobo and Zhou, 2006; Cohen et al., 2008; Chang et al., 2012). But contradictory results were also found (Verleun et al., 2011;Jain and Rezaee, 2004). Lobo and Zhou (2006) use discretionary accruals in their research to determine

whether the implementation of SOX caused more conservative accounting and to what extent. They state that a lower use of discretionary accruals would indicate an increase in

conservative accounting. They measure the use of discretionary accruals in the two years preceding and respectively in the two years after the passage of SOX and find that the use of discretionary accruals decreased within firms which indicates that firms are more conservative in the two years after the passage of SOX compared with the two years before the passage of SOX. Referring to what extent conservatism had increased across the two periods, they use the Basu (1997) method and find a significant increase in conservative accounting in the post-SOX period which indicates that losses are recognized faster than profits in the post-post-SOX

(16)

16 period. According to the researchers, the increase in conservative accounting is due to the increase in fines and regulatory scrutiny the passage of SOX imposes for overstating earnings. Specifically SEC. 302 CORPORATE RESPONSIBILITY FOR FINANCIAL REPORTS increases potential litigation against CEOs/CFOs. SEC. 302 mandates CEOs to give a certification about ‘‘material accuracy and completeness of the ﬁnancial statements’’ (Zhou, 2008) and tries to force CEOs in making financial reporting more reliable, more accurate and of high quality to enhance investors’ confidence (Chang et al. 2012). This provides greater incentives to not overstate earnings and to be more conservative when faced uncertainty and increased legal liability.

This is consistent with the research of Watts (2003), which discusses that when firms overstate their earnings and net assets instead of understating earnings and net assets, it will lead to a greater likelihood of litigation and therefore will generate litigation costs.

Conservative accounting leads to an understatement of net assets, which reduces potential litigation costs. Since the expected litigation costs of overstatement are higher than those of understatement, top management and auditors have incentives to report conservative values for earnings and net assets.

However, Verleun et al. (2011) find insignificant results regarding conservatism after the passage of SOX. The authors examine the long-term effects of SOX on accounting quality, including the effect on conservative accounting. Their research showed a minor increase in conservative accounting after the passage of SOX. However, this minor increase was insignificant due to the fact that the increase did not sustain on the long term. In contrast with non-technology firms, the authors found slightly less conservatism within technology firms. However this result is probably a reaction to the dot-com bubble in the year 2000. Furthermore, Jain and Rezaee (2004) find that after the passage of SOX financial reporting did not became more conservative. Their sample contains data from the period 2001-2003, and they use three measures of conservatism (market-value based measure, accrual-based measure, asymmetric timeliness measure), to determine the degree of conservative accounting.

2.5 Hypothesis development

I expect that CEOs engage more in conservative accounting at the beginning of their career than at the final years of their service, because of the following four reasons: (1) CEOs are heavily monitored at the beginning of their career, because the skills and knowledge of the CEO is not clear yet. Throughout their career, their ability becomes clearer and the CEO gets

(17)

17 more bargaining power which leads to a decrease in monitoring. A long established CEO receives less scrutiny. Engaging less in conservative accounting is not quickly detected at the end of a CEOs’ career due to the decrease in monitoring of the CEO. Also Garcia Lara et al. (2007) find that CEOs with more bargaining power engage less in accounting conservatism; (2) CEOs want to create a good reputation for themselves. At the beginning of their career they have to build up their reputation. Engaging in more conservative accounting in the first years will contribute to a good reputation; (3) wages are dependent on the perceived ability. When CEOs’ perceived ability is determined and reputation is established, CEOs have incentives to increase earnings to get higher wages, resulting in a decreased use of

conservative accounting at the final years of their service; (4) as a results of the SOX-act, using less conservative accounting generates higher litigation costs. Therefore, CEOs have incentives to understate assets and earnings. In order to boost final year bonus and wages, the CEO engages in less conservative accounting. In the final year of his service, the CEO does not care anymore about the increased litigation costs, because it will not affect him as he will leave the firm. Based on the above arguments, I propose the following hypothesis:

Hypothesis H1: CEOs are more conservative at the beginning of their career than at the end

of their career.

Hypothesis H1 predicts a positive relationship between the use of conservative accounting and the early years of a CEOs’ career. Furthermore, a negative relationship between the use of conservative accounting and the final years of a CEOs’ career is predicted.

Other factors can also influence the use of conservative accounting by CEOs. The sample used for this research has a time frame from 1996 until 2014. Because SOX was introduced in 2002, SOX forms a big influential factor on conservative accounting in this sample. The passage of SOX results in higher litigation risks and subsequently potential litigation costs when assets and earnings are overstated. SOX forces CEOs to give a

certification about ‘‘material accuracy and completeness of the ﬁnancial statements’’ (Zhou, 2008). Not complying to section 302. generates high fines for the respective CEO.

Conservative accounting can mitigate these risks and costs. Therefore, I propose the following hypothesis:

(18)

18 Hypothesis H2 predicts a positive relationship between the use of conservative accounting and the post-SOX period.

3 Research design

This section discusses the proxies of conservatism in accounting used in this study. Furthermore, an explanation of the independent variables and control variables will be provided. Finally, the empirical model used in this study will be defined.

3.1 Measures of conservatism in accounting

Following Ahmed et al. (2002), I adopt the notion that conservative accounting usually leads to a decrease in book value of equity relative to market value of equity and also leads to a decrease in income relative to operating cash flows. As a consequence of that notion I use two measures of conservative accounting for this study. The two measures are unconditional and firm-specific which are a market-value based measure (Beaver and Ryan, 2000), and an accrual-based measure (Givoly and Hayn, 2000). According to Ahmed et al. (2002) the difference between these two measures is, that the market-value based measure is a stock measure that shows the cumulative effects of conservative accounting on the book value of equity and the accrual-based measure is a flow measure that shows the effects of conservative accounting in the period this measure is calculated.

The market-value based measure based on Beaver and Ryan (2000), is built on the assumption that under conservatism, the tendency exists to require a higher degree of

verification to recognize a gain in asset values than to recognize a loss in asset values (Basu, 1997). This leads to net assets being understated. This understatement can be determined, by comparing the ratio of the firm’s book value of equity to its market value of equity (book-to-market ratio) (Watts, 2003). When firms use conservative accounting, net assets will be understated and lower book-to-market ratios will be recorded (Watts, 2003). The market-value based measure of conservatism will be indicated as CON-BTMit in this study. This

measure will be calculated as the book-to-market ratio multiplied by -1. A positive value of

CON-BTMit signals the use of more conservative accounting.

Following Givoly and Hayn (2000), the accrual-based measure is based on the notion that under conservative accounting gains and losses are not recorded in the same way under similar verifiability. Losses are, in contrast to gains, completely accrued, resulting in understated cumulative accruals which leads to negative accruals (Watts, 2003). The higher

(19)

19 the degree of negativism in accruals, the greater the use of conservative accounting. The accrual-based measure will be indicated as CON-ACCit in this study. This measure will be

calculated as income before extra-ordinary items less cash flows from operations plus depreciations expense deflated by average total assets, multiplied by -1. A positive value of

CON-ACCit signals the use of more conservative accounting.

3.2 Independent variables

To test hypothesis H1, I use two independent variables which are Early Yearsit and Final

Yearit. Early Yearsit is a dummy variable that is defined as one, which captures the first three

years of CEO tenure, and is zero for the remaining years of CEO tenure. Following Ali and Zhang (2015), I chose the first three years of the service of a CEO as a cut-off since it is fifty percent of the median. Figure 1. shows the frequency distribution of the years a CEO served when leaving office. The mean is 7.2 which differs from Ali and Zhang (2015) and is caused by the fact that the statistics of the histogram used by Ali and Zhang is based on 2,278 CEOs, representing 1,688 firms in contrast with the statistics of the histogram used in this study which is based on 3,008 CEOs, representing 2,122 firms. The median of 6 is consistent with the result from Ali and Zhang (2015). A positive coefficient for Early Yearsit is consistent

with hypothesis H1, which says that CEOs engage more in conservative accounting in the early years of their service, than at the final years of their service.

Figure 1. Frequency distribution of years of service as CEO when leaving office.

Histogram is based on 3,008 CEOs, representing 2,122 firms, who leave their firms during the sample period 1996-2014. Mean: 7.2 Median: 6 0 2 00 4 00 6 00 1 00 3 00 5 00 N um b er of C E O s 5 6 10 15 20 4 7 8 9 11 12 13 14 16 17 18 19

(20)

20

Final Yearit is a dummy variable that is defined as one, which captures the final year of a

CEO tenure, and is zero for the remaining years of a CEOs’ years of service. Anegative coefficient for Final Yearit is consistent with hypothesis H1, which says that CEOs engage

less in conservative accounting in the final year of their service, than at the early years of their service.

To test hypothesis H2, I use one independent variable named SOXit, which measures

the influence on conservative accounting for CEOs after the passage of SOX. SOXit is a

dummy variable that is defined as one which captures the post-SOX period, and is defined as

zero which captures the pre-SOX period. A positive coefficient for SOXit is consistent with

hypothesis H2, which says that CEOs are more conservative in the post-SOX period, than in the pre-SOX period due to more litigation the passage of SOX imposes.

3.3 Control variables

In this study, I use seven control variables. According to prior literature there is a numerous range of variables which has an effect on conservative accounting. I make use of particularly these seven control variables, because they are the most adopted in prior studies. The control variables that are used in this research are: CEO Ownershipit, CEO ageit, Leverageit, Firm

Sizeit, Litigationit, Sales Growthit, and CFO/TAit.

The first control variable is CEO Ownershipit, which is equal to the percentage of

outstanding shares of firm i held by the CEO at the beginning of fiscal year t. I control for

CEO Ownershipit, because LaFond and Roychowdhury (2008) found that firms with lower

managerial ownership use more conservatism.

The second control variable is CEO Ageit, which is the CEOs’ age at the beginning of

fiscal year t. I control for CEO Ageit, because Gibbons and Murphy (1992) show that when a

CEO reaches the retirement age, career concerns become weaker, which leads to pay for performance being stronger when a CEO precedes retirement. Therefore, CEOs preceding retirement will overstate earnings to generate higher pay for performance, which influences the degree of conservatism used by the CEO.

The third control variable is Leverageit, which is equal to total liabilities divided by

total assets at the end of the year. I control for Leverageit, because firms with a high level of

leverage likely have greater bond-holder and shareholder conflicts, which influences conservative accounting (Ahmed et al., 2002).

The fourth control variable is Firm Sizeit which is the natural log of total assets. I

(21)

21 conservative accounting in order to avoid such political costs (Watts, 2003). But at the other hand, within large firms the information asymmetry is smaller which leads to a lower requirement for conservative accounting (LaFond and Watts, 2006).

The fifth control variable is Litigationit. I add a dummy variable to control for firms

with high litigation risk. As discussed in section 2.1, when firms overstate their assets and earnings, it generates high litigation costs. With the use of conservative accounting, those litigation costs can be mitigated (Watts, 2003). Litigationit is a dummy variable which equals

one if the firm operates in a high-litigation industry, and is zero otherwise. The SIC codes

corresponding to high-litigation industries are 2833-2836; 3570-3577; 3600-3674; 5200-5691, and 7370-7374 (Ali and Zhang, 2015).

The sixth control variable is Sales Growthit which is equal to the percentage growth in

total sales per year. I control for Sales Growthit, because this can influence the unconditional

measures of conservative accounting, such as CON-BTMit, and CON-ACCit, due to accruals

being affected by sales growth. When accruals such as inventory and receivables are affected by sales growth, this can influence CON-ACCit (Ahmed et al., 2002). Also, a significant

increase in sales can boost market expectations of future cash flows, which affects

CON-BTMit (Ahmed et al., 2002). Following Ahmed and Duellman (2007), I predict a

negative effect of Sales Growthit on CON-ACCit and a positive effect on CON-BTMit.

The final control variable is CFO/TAit, which is equal to cash flow of operations

divided by total assets. I control for CFO/TAit, because firms with higher profits often engage

in more conservative accounting (Ahmed et al., 2002).

3.4 Empirical design

As discussed earlier, the first hypothesis examines the use of conservative accounting by a CEO during its tenure. Based on prior studies of Ahmed and Duellman (2007, 2013) and Ali and Zhang (2015), I estimate the following empirical model using an Ordinary Least Squares (OLS) regression for hypothesis H1:

CONit = a0 + a1 Early Yearsit + a2 Final Yearit + a3 CEO Ownershipit + a4 CEO Ageit + a5

(22)

22 As discussed earlier, the second hypothesis examines the use of conservative accounting by a CEO during its tenure before and after the passage of SOX. Based on prior studies of Ahmed and Duellman (2007, 2013) and Ali and Zhang (2015), I estimate the following empirical model using an Ordinary Least Squares (OLS) regression for hypothesis H2:

CONit = a0 + a1 SOXit +a2 CEO Ownershipit + a3 CEO Ageit + a4 Leverageit + a5 Firm Sizeit +

a6 Litigationit + a7 Sales Growthit + a8 CFO/TAit + εit (2)

CONit =One of the two measures of conservative accounting, which are the

accrual-based measure (CON-ACCit), and the market-value based

measure (CON-BTMit);

CON-ACCit = Income before extra-ordinary items less cash flows from operations

plus depreciations expense deflated by average total assets; multiplied by -1;

CON-BTMit = Book-value of equity divided by the market-value of equity;

multiplied by -1;

Early Yearsit = A dummy variable that is defined as one, which captures the first

three years of CEO tenure, and is zero for the remaining years of CEO tenure;

Final Yearit = A dummy variable that is defined as one, which captures the final

year of CEO tenure, and is zero for the remaining years of CEO tenure;

SOXit = A dummy variable that is defined as one, which captures the

post-SOX period, and is defined as zero which captures the pre-post-SOX period;

CEO Ownershipit = The percentage of outstanding firm stock owned by the CEO;

CEO Ageit = The age of the CEO at the beginning of the year of a firm;

Leverageit = The long-term debt (total liabilities) divided by total assets at the end

of the year of a firm;

(23)

23

Litigationit = A dummy variable which equals one if the firm operates in a

high-litigation industry, and is zero otherwise;

Sales Growthit = The percentage growth of the total sales per year of a firm;

CFO/TAit = Cash flow of operations divided by the average of total assets per year

of a firm.

To summarize, a positive coefficient for Early Yearsit indicates that CEOs engage more in

conservative accounting in the early years of their service, than at the final years of their service. Furthermore, a negative coefficient for Final Yearit indicates that CEOs engage less

in conservative accounting in the final year of their service.

A positive coefficient for SOXit indicates that CEOs are more conservative in the

post-SOX period, than in the pre-post-SOX period.

4 Results

In this chapter the determination of the sample and the sample size will be discussed. Furthermore, the descriptive statistics of the observations will be explained. Also the results of the regressions will be elaborated on. Subsequently will be discussed whether hypothesis H1 and hypothesis H2 are supported or rejected. Finally, several sensitivity analysis will be conducted to test whether the findings are robust.

4.1 Sample selection

The sample selection used for this empirical research consists of data on U.S firms for the years 1996-2014. I have chosen the years 1996-2014 in order to capture complete CEO tenures. Furthermore, I have chosen U.S firms due to the large amount of data available on U.S firms. I obtained data on CEO tenure, CEO age and CEO ownership from Execucomp, and financial statement data from Compustat. TABLE 1 gives an overview of the sample size. I started with 252,912 observations obtained from Compustat. After obtaining data from Execucomp and merging it, the sample contained 110,744 observations. Subsequently, observations with missing values were deleted from the sample resulting in 21,764

observations. Firm years of CEOs who are still employed in the years following 2014 were also eliminated in order to capture the CEOs’ final year of service, resulting in 17,958

observations. Furthermore, I Winsorized the top and bottom 1% of the market-based measure and the accrual-based measure in order to exclude outliers which could have a negative

(24)

24 impact on the results, which yielded in 17,604 observations. After that, incomplete CEO tenures with missing firm years were excluded. This means that when the years of service of a particular CEO (from the year becoming a CEO until the year a CEO resigns from a firm) was missing a firm year, all the years of service of that particular CEO were excluded from the sample which resulted in 4,650 observations. Lastly, firms with SIC codes 6000-6999 were excluded from the sample, because these firms operate in the financial services industries and therefore are subject to regulatory constraints which tends to affect their reporting (Ahmed and Duellman, 2013). Dropping these firms yielded a final sample size of 4,462 observations.

TABLE 2 gives an overview of the firms categorized per SIC industry. Manufacturing has with 51.86 percent the largest observations. Agriculture, Forestry and Fishing has the smallest amount of observations with 0.20 percent and Public Administration has no observations at all.

TABLE 1 Sample size

Observations

Number of observations obtained from Compustat 252,912

Deleted observations after merging with Execucomp (142,168) Deleted observations with missing values (88,980) Deleted observations of CEOs still employed (3,806) Deleted observations after Winsorizing (354) Deleted observations of incomplete CEO tenures (12,954) Deleted observations with SIC code 6000-6999 (188)

Final sample size 4,462

TABLE 2

Observations categorized per SIC industry

SIC code Industry Observations %

0-999 Agriculture, Forestry and Fishing 9 0.20% 1000-1999 Mining & Construction 303 6.79%

2000-3999 Manufacturing 2,314 51.86%

4000-4999 Transportation, Communications, Electric,

Gas and Sanitary service 686 15.37% 5000-5999 Wholesale Trade & Retail Trade 587 13.16% 7000-8999 Services 563 12.62% 9000-9999 Public Administration - 0.00%

(25)

25 4.2 Descriptive statistics

TABLE 3 reports the descriptive statistics of the variables in the regression models. Looking at the measures of conservative accounting, the mean value of the accrual-based conservatism measure CON-ACCit is 0.0132. This is slightly higher than the mean value of the

accrual-based conservatism measure (0.010) reported in Ahmed and Duellman (2007). A positive value of CON-ACCit signals the use of more conservative accounting. The mean value of the

market-based conservatism measure CON-BTMit is -0.4923. This is a larger negative mean

value than the mean value (-0.363) reported in Ahmed and Duellman (2007). This difference is most likely caused by the time periods of the studies. Ahmed and Duellman (2007) use a sample obtained from 1998 to 2002 and the samples used in this study contains data obtained from 1996-2014. The lower the negative amount, the higher the degree of conservatism is found. In conclusion, both the measures of accounting conservatism (CON-ACCit and

CON-BTMit) shows conservatism in the sample.

With respect to the independent variables, approximately 36 percent of the firm-year observations correspond to the early years of the CEOs’ service and approximately 12 percent of the firm-year observations correspond to the final year of the CEOs’ service. These results are in line with the descriptive statistics of Ali and Zhang (2015). About 79 percent of the total firm years corresponds to the post-SOX period (2002 and after). This makes sense considering the sample period which runs from 1996 until 2014. A larger part of CEO tenure takes place in the post-SOX period.

Referring to the control variables, the mean value of CEO Ownershipit is 0.0266 which

indicates that about 2.7 percent of the outstanding firm stock is owned by CEOs. The average age of the CEOs in the sample is 64 year. The mean value of Leverageit is 0.2408 which

indicates that 24 percent of the firm’s assets is the amount of liabilities used to finance the firm’s assets. The mean value of Firm Sizeit is 7.9964, which means that the sample exists of

relatively large firms which is consistent with the mean value reported in Ahmed and Duellman (2007) and Ahmed et al. (2002). The mean value of Litigationit is 0.2331, which

indicates that about 23 percent of the firms in the sample operates in a high litigation industry. This is similar to those of Ali and Zhang (2015). The percentage growth of total sales per year is with 9.6 percent similar to the percentage reported in prior studies. The mean value of

(26)

26 TABLE 3

Descriptive statistics of the variables in the regression models

N Mean STD Median Q1 Q3 CON-ACCit 4,462 0.0132 0.6661 0.0077 -0.0141 0.0335 CON-BTMit 4,462 -0.4923 0.3594 -0.4472 -0.6479 -0.2735 Early Yearsit 4,462 0.3550 0.4786 0.0000 0.0000 1.0000 Final Yearit 4,462 0.1181 0.3228 0.0000 0.0000 0.0000 SOXit 4,462 0.7902 0.4072 1.0000 1.0000 1.0000 CEO Ownershipit 4,462 0.0266 0.0592 0.0025 0.0006 0.0143 CEO Ageit 4,462 64.514 6.4310 65.000 60.000 69.000 Leverageit 4,462 0.2408 0.1787 0.2353 0.1203 0.3417 Firm Sizeit 4,462 7.9964 1.5908 7.9124 6.8653 9.0705 Litigationit 4,462 0.2331 0.2331 0.0000 0.0000 0.0000 Sales Growthit 4,462 0.0960 0.9600 0.0710 -0.0079 0.1573 CFO/TAit 4,462 0.1086 0.1086 0.1014 0.0640 0.1478

TABLE 4 presents the Pearson correlations between the conservatism measures and the independent variables and control variables. The accrual-based measure of conservatism (CON-ACCit) and the market-based measure of conservatism (CON-BTMit) are weakly

negatively correlated but the correlation is not significant. As mentioned earlier, the

difference between the conservatism measures is that CON-ACCit is defined as a flow measure

and CON-BTMit is defined as a stock measure, which causes the weak univariate correlation

between CON-ACCit and CON-BTMit (Ahmed et al., 2002).

Looking at the correlations between the measures of conservative accounting and the independent variables, only weak correlations were found. Early Yearsit is weakly negatively

correlated with CON-ACCit but not at a significant level. This indicates that an increase in

Early Yearsit does not increase the level of negative accruals. Early Yearsit is weakly

positively correlated with CON-BTMit and is significant at a 10% level which means that an

increase in Early Yearsit results in a slight increase in the book-to-market ratio. Final Yearit is

weakly positively correlated with CON-ACCit and is significant at a 5% level which indicates

that an increase in Final Yearit leads to a weak increase in the level of negative accruals. Final

Yearit is weakly negatively correlated to CON- BTMit but not at a significant level. An

increase in Final Yearit does not increase the book-to-market ratio. SOXit is weakly positively

correlated with CON-ACCit and is significant at a 1% level which means that an increase in

(27)

27 correlated with CON-BTMit but not at a significant level. An increase in SOX does not lead to

an increase the book-to-market ratio.

4.3 Accrual-based measure of conservative accounting

Following Ahmed and Duellman (2007), I estimate Eq. (1) and Eq. (2) using an OLS regression. TABLE 5 shows the results of the regression of accrual-based conservatism

(CON-ACCit) on CEO tenure and SOX. Column (1) shows the OLS regression results with the

control variables, excluding the independent variables Early Yearsit, Final Yearit, and SOXit.

Column (2) shows the OLS regression results with both the independent variables Early

Yearsit and Final Yearit included in order to test the first hypothesis, also taken the control

variables into account. Column (3) shows the OLS regression results with the independent variable SOXit included in order to test the second hypothesis, also taken the control variables

into account. Column (4) shows the OLS regression results including all the independent variables and complete set of the control variables.

4.3.1 Results hypothesis H1

To determine the result of hypothesis H1 column (1) and column (2) of TABLE 5 are used. The coefficient on Early Yearsit shown in column (2) is negative and significant at the 10%

level. This finding suggests that losses are, in contrast to gains, not completely accrued which does not result in understated cumulative accruals in the first three years of a CEOs’ years of service. The negative value of the coefficient signals the use of less conservative accounting. The coefficient on Final Yearit shown in column (2) is positive and significant at the 5% level,

which suggests that CEOs make more use of negative accruals in the final year of their service. Both the significant coefficients on Early Yearsit and Final Yearit suggest that the use

of conservative accounting in the first three years of a CEOs’ years of service is significantly smaller than in the final year of a CEOs’ years of service. These findings are not consistent with hypothesis H1, and does not support the notion that CEOs have incentives to engage more in conservative accounting in the first years of their tenure and in less conservative accounting in the last year of their service. This finding supports the results of Ali and Zhang (2015) who conclude that CEOs have incentives to overstate their earnings in the early years of their service compared to the other years due to career concerns.

(28)

28 TABLE 4

Pearson Correlation Matrix: correlation between independent variables, dependent variables, and control variables

CON-ACC CON-BTM Early Years Final Year SOX CEO Ownership CEO Age Leverage Firm Size Litigation Sales Growth CFO/TA

CON-BTM_it -0.0038 1 0.7995 Early Years_it -0.0236 0.0284*** 1 0.1156 0.0577 Final Year_it 0.0306** -0.0094 -0.2715* 1 0.0407 0.5289 0.0000 SOXit 0.059* -0.0179 -0.4575* 0.1886* 1 0.0001 0.233 0.0000 0.0000 CEO Ownershipit 0.005 -0.0164 -0.0298** 0.0084 -0.036** 1 0.7369 0.273 0.0462 0.5741 0.0163 CEO Ageit -0.0429* -0.0245 0.008 0.005 -0.252* -0.0128 1 0.0042 0.1019 0.5939 0.7398 0.0000 0.3937 Leverageit 0.0219 0.1208* 0.0361** 0.0182 -0.073* -0.0317** 0.068* 1 0.1437 0.0000 0.0158 0.2231 0.0000 0.0345 0.0000 Firm Sizeit -0.0399* 0.0778* -0.1084* 0.0547* 0.1443* -0.1248* 0.0242 0.1774* 1 0.0077 0.0000 0.0000 0.0003 0.0000 0.0000 0.1062 0.0000 Litigationit 0.0828* 0.0674* 0.0031 0.0003 0.021 0.0238 -0.1498* -0.2501* -0.1154* 1 0.0000 0.0000 0.8358 0.9854 0.1599 0.1125 0.0000 0.0000 0.0000 Sales Growthit -0.0625* 0.0879* 0.0312** -0.0506* -0.0507* 0.0136 0.0285*** -0.0012 -0.0025 0.0582* 1 0.0000 0.0000 0.0372 0.0007 0.0007 0.3624 0.0572 0.938 0.8648 0.0001 CFO/TAit 0.3027* 0.3216* 0.0119 -0.0557* -0.0137 0.0301** -0.0509* -0.1728* 0.0575* 0.0468* 0.0437* 1 0.0000 0.0000 0.4279 0.0002 0.3618 0.0440 0.0007 0.0000 0.0001 0.0018 0.0035 *= significant at 1% **= significant at 5% ***=significant at 10%

(29)

29 TABLE 5

Regression of Accrual-based conservatism (CON-ACCit) on CEO tenure and SOX

Coefficient P-value Coefficient P-value Coefficient P-value Coefficient P-value

Intercept 0.0078 0.476 0.0097 0.378 -0.0126 0.274 -0.0119 0.320 Early Years_it -0.0038 0.066*** 0.0006 0.795 Final Year_it 0.0082 0.007** 0.0072 0.018** SOXit 0.0129 0.000* 0.0121 0.000* CEO Ownershipit -0.0003 0.435 -0.0003 0.358 -0.0002 0.498 -0.0002 0.475 CEO Ageit -0.0002 0.244 -0.0002 0.245 0.0000 0.818 0.0000 0.893 Leverageit 0.0424 0.000* 0.0430 0.000* 0.0453 0.000* 0.0450 0.000* Firm Sizeit -0.0029 0.000* -0.0032 0.000* -0.0035 0.000* -0.0035 0.000* Litigationit 0.0143 0.000* 0,0142 0.000* 0.0145 0.000* 0.0144 0.000* Sales Growthit -0.0151 0.000* -0,0145 0.000* -0.0145 0.000* -0.0142 0.000* CFO/TA_it 0.2569 0.000* 0.2593 0.000* 0.2600 0.000* 0.2613 0.000* Adjusted R² (1) (2) (3) (4) CON-ACC_it 0.1155 0.1180 0.1210 0.1217

The sample period is from 1996-2014. The sample exists of 4,462 firm year observations. */**/*** represents significance at the 1%/5%/10% level, respectively. CON-ACCit = income before extra-ordinary items less cash

flows from operations plus depreciations expense deflated by average total assets; multiplied by -1. Early Yearsit

= a dummy variable that is defined as one, which captures the first three years of CEO tenure, and is zero for the remaining years of CEO tenure. Final Yearit = a dummy variable that is defined as one, which captures the final

year of CEO tenure, and is zero for the remaining years of CEO tenure. SOXit = a dummy variable that is defined

as one, which captures the post-SOX period, and is defined as zero which captures the pre-SOX period. CEO Ownershipit = the percentage of outstanding firm stock owned by the CEO. CEO Ageit = the age of the CEO at

the beginning of the year of a firm. Leverageit = the long-term debt (total liabilities) divided by total assets at the

end of the year of a firm. Firm Sizeit = natural log of total assets at the end of the year of a firm. Litigationit = a

dummy variable which equals one if the firm operates in a high-litigation industry, and is zero otherwise. Sales Growthit = the percentage growth of the total sales per year of a firm. CFO/TAit = cash flow of operations

divided by the average of total assets per year of a firm.

With respect to the control variables, the coefficient on CEO Ownershipit is in both columns

negative, but not significant. Also the coefficient on CEO ageit is in both columns negative,

but not at a significant level. The coefficient on Leverageit is in both columns positive and

significant at the 1% level which is consistent with the notion that firms with a high level of leverage likely have greater bond-holder and shareholder conflicts, which influences

conservative accounting. The coefficient on Firm Sizeit is in both columns negative and

significant at the 1% level, consistent with the argument that within large firms the information asymmetry is smaller which leads to a lower requirement for conservative accounting (LaFond and Watts, 2006). The coefficient on Litigationit is in both columns

(30)

30 positive and significant at the 1% level which is consistent with the notion that firms that have a high litigation risk engage more in conservative accounting to avoid high litigation costs (Watts, 2003). The coefficient on Sales Growthit is in both columns negative and significant at

the 1% level. These results show that accruals such as inventory and receivables are affected by sales growth which influences conservative accounting measured by accruals, in a negative way, which is consistent with the expectations. The coefficient on CFO/TAit is in both

columns positive and significant at the 1% level which indicates that firms with higher profits often engage in more conservative accounting (Ahmed et al., 2002).

The adjusted R2 of column (1) is 0.1155, which indicates that about 11.55 percent of the variance of the accrual-based measure CON-ACCit is explained by the control variables.

When the independent variables Early Yearsit and Final Yearit are included in the regression

model (see column (2)) the adjusted R2 slightly increases to 0.1180, which indicates that about 11.80 percent of the variance of the accrual-based measure CON-ACCit is explained by

the control variables including the independent variables. So, when the independent variables are included in the regression model, the adjusted R2 only increases with 0.25 percent.

4.3.2 Results hypothesis H2

To determine the result of hypothesis H2 column (1) and column (3) of TABLE 5 are used. The coefficient on SOXit shown in column (3) is positive and significant at the 1% level. This

finding implies that CEOs use more conservative accounting in the post-SOX period in contrast with the pre-SOX period. This is in line with the results of Lobo and Zhou (2006), Cohen et al. (2008), and Chang et al. (2012). This finding is consistent with hypothesis H2 and supports the notion that, because of higher litigation caused by the passage of SOX in 2002, CEOs try to avoid higher litigation costs by the use of conservative accounting.

With respect to the control variables, I find no effect of CEO Ageit on CON-ACCit

since the coefficient of CEO Ageit is negative in column (1) and positive in column (2), but

both not at a significant level. The coefficients of the other control variables are similar to those results reported in 4.3.1 Results hypothesis H1.

The adjusted R2 of Column (1) is 0.1155, which indicates that about 11.55 percent of the variance of the accrual-based measure CON-ACCit is explained by the control variables.

When SOXit is included in the regression model (column (3)), the adjusted R2 slightly

increases and indicates that about 12.10 percent of the variance of the accrual-based measure is explained by the control variables including the independent variable. So, when SOXit is

The interaction between CEO tenure and conservative accounting

Amsterdam Business School

Master thesis