The adoption of a clawback provision and its association with errors and irregularities

(1)

MSc Accountancy & Control, Accountancy track

University of Amsterdam, Faculty of Economics and Business

Master thesis accountancy:

The Adoption of a Clawback Provision and its

Association with Errors and Irregularities

Name Nienke Doorenspleet

Student nummer 10076646

Course Master’s Thesis Accountancy

University University of Amsterdam

Supervisor Mario Schabus

Date 22th, June 2015

(2)

2

STATEMENT OF ORIGINALITY

This document is written by Nienke Doorenspleet who declares 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)

3

ABSTRACT

I investigate the impact of clawback adoption on accounting quality in terms of intentional and unintentional error restatements. I predict that robust clawback adoption is associated with a decrease in unintentional errors and that this decrease will be more significant than the decrease in intentional errors. Contrary to my expectations, I find that clawback adoption is not significantly related to a reduction in unintentional errors. Additionally I examine whether a firm takes into consideration its experience with past restatements when adopting a clawback provision. The findings imply that firms are more likely to opt for a robust clawback when they have had restatements in the past that were caused by unintentional errors. On the contrary I expect that firms adopt a misconduct clawback when they didn’t have experience with unintentional errors, but I don’t find significant evidence for this relationship. This study highlights the importance of recognizing the differences between intentional and unintentional

(4)

4

1 INTRODUCTION

Executive compensation is a central issue of corporate governance in aligning shareholders’ interest with managers’ interest. Famously is the statement of Jensen and Murphy (1990) who argued that what matters the most is not how much CEOs are getting paid, but how CEOs got paid. Within this ongoing debate about pay structures a new segment of interest arose, the so called ‘clawback provisions’, also known as ‘compensation recovery provisions’. These provisions are introduced by section 304 of the Sarbanes Oxley Act in 2002 (SOX 304). SOX 304 authorizes the recovery of any compensation gained by CEOs and CFOs of public firms in case of material noncompliance or misconduct. However despite the efforts of the SEC, eventually it only led to a small number of claims (Sidel, 2010). This was mainly due to the fact that only the SEC could enforce the recoup of compensations and due to the fact that it was quite difficult to assess and prove managerial misconduct (Chan, Chen, Chen and Yu, 2012; Fried and Shilon, 2011). The Dodd-Frank act however, provided a new generation of clawback provisions in 2010. This new act, referred to as the DFA 954, entitles the board of directors to enforce clawback provisions. DFA 954 also requires the SEC to direct national stock exchanges to prohibit the listing of firms that did not implement such compensation clawback policies. Therefore the Dodd-Frank Act seeks to make the clawback of ‘ill-gotten’ performance-based compensation mandatory.

However, as of the start of 2015 the SEC still didn’t implement DFA 954, and therefore the adoption of a clawback provision by listed firms is still not mandatory. This delay might imply that there are difficulties related to the design of an acceptable clawback policy. Since it is therefore not possible to conduct studies by means of mandatory clawbacks, studies use the adoption of voluntary clawback provisions as a derivative of mandatory clawbacks.

According to Yu-Chun Lin (2013) there are four research areas with regard to the adoption of clawback provisions. The first research area is focusing on the specific determinants of firms’ adopting clawback provisions, related to the size of firms, corporate governance structures etc. (Brown, Davis-Friday and Guler 2011; Addy and Yoder 2011). The second focus is based on whether clawbacks effectively improve financial reporting quality and whether they affect auditor behavior (Chan et al. 2012). The third research area is focusing on the market reaction and signalling effects (Brown et al. 2011; Gao, Iskandar-Datta and Jia 2010) and the fourth research area is more directed to changes in CEO compensation (Babenko, Bennett and Bizjak 2012; Dehaan, Hodge and Shevlin 2013; Chan et al. 2012).

(6)

6 Referring to the second strand of literature, Dehaan et al. (2013) suggest that the main reason for the introduction of the clawback provision would be the improvement of the financial reporting quality. Previous studies have shown that managers with stock based and performance based compensation plans manipulate accounting information in order to maximize their personal gain. These manipulations could eventually lead to restatements which strongly suggest that the financial statement was of poor quality (Palmrose and Scholz, 2004). In other words there exists a positive relationship between these compensation plans and accounting restatements (Burns and Kedia, 2006). Much research has been done on moderating factors that would lead to less earnings management. For example there is a large research area focusing on the effects of IFRS adoption on earning management activities. Tightened accounting standards may put pressure on management to comply with IFRS standards resulting in less earnings management (Ewert and Wagenhofer, 2005). The introduction of a clawback provision should, according to Dehaan et al. (2013), also create incentives to reduce earnings management and to produce high-quality information. After all, the introduction of a clawback provision increases the cost of being caught. If caught, ‘ill-gotten’ performance-based compensation should be handed in by the manager responsible for these misstatements.

Based on the above one could expect a negative relationship between the adoption of a clawback policy and incentives to manage earnings, which will in turn lead to less restatements. However this point of view is mainly based on the assumption that restatements are the result of intentional manipulations, therefore ignoring restatements resulting from unintentional errors. The adoption of a clawback provision might as well lead to increased incentives to avoid unintentional errors. For that reason some clawback provisions will also be triggered in case unintentional errors are made. These so-called “robust clawbacks” will trigger the provision regardless whether the error was made intentional or unintentional. It basically refers to the setting in which clawback provisions require repayment irrespective of the reason of restatements and it should in theory lead to the reduction of both intentional errors and unintentional errors.

“Misconduct clawbacks” on the other hand, will only be triggered in case of fraud, when intentional errors are made. This trigger event only requires repayment when the restatement is a consequence of intentional errors (Dehaan et al., 2013). It is up to the company to decide whether it will adopt a robust clawback or a misconduct clawback.

Despite of having different trigger events and therefore acknowledging the difference between intentional errors and unintentional errors, most current studies fail to recognize this distinction. In case of robust clawbacks, it would be interesting to have insight in whether they

(7)

7 reduce both intentional and unintentional errors. Most studies are just focusing on the total reduction of restatements and therefore don’t have insight in whether adoption has led to less fraud or to less mistakes, as they are focused on the total sum of restatements.

By using a more detailed measurement technique I will expand existing literature regarding clawback policies and will fill the gap in literature by focusing on both intentional and unintentional errors and whether clawback adoption affect both kind of restatements.

In their additional analysis Dehaan et al. (2013) similarly attempt to make a distinction between intentional and unintentional errors. Their cross-sectional analysis shows some evidence that the adoption of a robust clawback lead to incremental benefits in comparison with a misconduct clawback. This incremental effect is suggesting that the adoption of a robust clawback will also lead to a reduction in unintentional errors, but it doesn’t give insight into how these intentional errors and unintentional errors are related to each other; how large is this reduction of unintentional errors relative to the intentional errors? Therefore their research does not fully reflect the direct link between a certain kind of clawback and a certain kind of restatement. To get more insight in how these two relate and referring to the incremental effect of Dehaan et al. (2013), I predict that robust clawback adoption is associated with a decrease in unintentional errors and that this decrease will be more significant than the decrease in intentional errors.

In addition I will provide insight in the relationship between the different kind of restatements and the choice between the adoption of either a robust clawback or a misconduct clawback. That is, for example, whether firms experiencing fraudulent restatements make different choices than firms with restatements due to errors, in their selection of clawback type. I make the assumption that firms are more likely to adopt a robust clawback when they experienced a restatement resulting from unintentional error prior to adoption and that firms are more likely to adopt a misconduct clawback when they haven’t experienced a restatement resulting from unintentional error prior to adoption.

Unfortunately I do not find any significant association between the adoption of a robust clawback and a decrease in unintentional errors. Additionally the relationship between the robust clawback and intentional errors can’t be tested, due to missing data regarding intentional errors. The results of the second part support my assumption that firms are indeed more likely to adopt a robust clawback when they experienced error restatements in the past. However I did not find evidence for the assumption that firms will be more likely to adopt a misconduct clawback when they haven’t experienced error restatements in the past. Research could be improved by using a different dataset.

(8)

8 Besides filling a gap in literature, this research also contributes to the literature as the SEC is considering to implement mandatory clawback policies. However these plans are established in 2010 and yet they are still not implemented. This might be due to possible concerns. More research done in this area could contribute to the decision whether or not to introduce mandatory clawback policies. Therefore this study is of particular interest for regulators, but could also be of interest for stakeholders. Stakeholders are concerned with the quality of the financial statement, upon they make decisions, therefore it is in their interest to know whether clawbacks will lead to better financial statement quality or not.

2 LITERATURE REVIEW, THEORY AND HYPOTHESES

2.1 Historical Background

Through the alignment of interests between principal and agent, equity-based manager compensation has been seen as a potential solution to the agency problem between managers and stakeholders (Jensen and Meckling, 1976). Yet, since equity-based compensation plans don’t impose a downward risk on management it creates incentives for managers to take excessive risks and to engage in a more aggressive form of earnings management. Shareholders have little resources and power to monitor managers and to prevent them from pursuing their own needs. Additionally, according to Bebchuk, Fried and Walker (2002), shareholders find it hard to seek legal remedy in a derivative suit. The clawback provision has since found its way as a new corporate governance tool (Gao et al. 2010). Although the clawback provision is not truly a new tool, Gao et al. (2010) speak of a “newly-found value” referring to its increased use in recent years.

A clawback provision entrust the board of directors to request CEOs and CFOs to pay back compensation, received in case of material noncompliance or misconduct. In absence of a clawback provision, firms still have the possibility to recover previously paid bonuses by means of a lawsuit, however the burden of proof is significant larger than in the case of a clawback provision. As such, a clawback provision provides an important tool to mitigate earnings management behaviour. After all, clawback provisions could impose an ex-ante risk for executives when it turns out that compensation is gained at the expense of shareholders. In case the provision will be triggered regardless the nature of the restatement (robust clawback), the clawback provision might also be a tool to reduce the number of unintentional errors.

(9)

9 Companies started to implement clawback provisions into their firm policies in mid-2000 after big scandals like Enron and WorldCom (Schwartz, 2008), well before the introduction of the Sarbanes Oxley Act in 2002. After 2002 there have been several of restatements, but the SEC didn’t use its exercise power until 2007 (Salehi and Marino, 2008). In 2010 the introduction of the Dodd-Frank Act seeks to make the clawback provision mandatory. The main purpose of both regulations however has remain the same: reducing the agency problem. Tying payment to performance encourages managers to engage in earnings management that lead to less optimal compensation contracts. Implementing a clawback makes engaging in earnings management more costly and has therefore a deterrent effect on managers. Managers are subjected to the risk of repayment. (Iskandar-Datta and Jia, 2013).

2.2 The Adoption of Clawback Provisions

Upon adoption of a clawback provision, there are a number of issues to consider. Gibson, Dunn and Crutcher (2008) define a couple of matters to take into consideration before adopting a clawback provision. In the first place the board needs to consider to whom the clawback provisions should apply. Clawback provisions could apply to the CEO, the CFO, to all executives or even to all employees. Secondly the question should be asked to which awards the clawback provisions should apply. The single most popular approach that boards choose to adopt includes all performance-based awards, and excludes base salary as base salary is not linked to a specific performance. A distinction between short-term and long-term incentives could be made. It is worth noting that not only cash bonuses are covered. A substantial part of executives’ pay is received in incentives compensation that covers both bonuses and equity (Gao, Lemmon and Li, 2011). According to Equilar (2011), 88 percent of firms have a clawback that covers equity incentives.

Another decision that should be taken into consideration is related to the circumstances that will trigger the clawback provision. A clawback provision could for example be applicable in the event of all restatements, with exception of changes in accounting policy. Earlier I referred to this kind of clawback as the robust clawback.

According to Plumplee and Yohn (2010) a restatement could occur for many reasons. 57 percent occurs as a result of internal company errors, only 3 percent occurs due to fraud and 3 percent as a result of complexity. The remaining 37 percent is explained by characteristics of accounting standards. However, in the case that the clawback provision covers all restatements it will also apply when a restatement should be made due to an unintentional error. Since 57

(10)

10 percent occurs due to an error and only 3 percent occurs as a result of fraud, such a provision could influence a firm’s ability to retain or attract employees.

A fourth issue to consider would be the duration of the “look-back period”. To which period should the clawback be linked? Consistent with section 304 of SOX, some organizations introduced a reach back period of 12 months, which means that the organization is allowed to recoup received compensation upon 12 months following the publication of misstated financial information. However according to Gibson et al. (2008) most firms chose to adopt a longer reach back period, varying between three till five years and even unlimited periods. Although these unlimited periods are quite uncommon, as it will be perceived as being unfair and it might impact a firm’s ability to attract or retain new staff. Section 954 of Dodd-Frank provides more guidance as it sets a standard of three years. In addition, the Dodd-Frank act also states that the clawback will be “calculated as the excess amount paid on the basis of the restated results” suggesting that one is not reliable for the full amount of bonus received.

2.3 SOX 304 Versus DFA 954

The aim of the Dodd Franc act is to make clawback provision mandatory for listed firms. However since by the beginning of 2015 the act still didn’t lead to mandatory clawback adoption, up until now DFA 954 and SOX 304 are both used as guidelines during the adoption of a clawback. This indiscriminate use of both standards might create confusion. The following table will provide a short overview of the main differences between both laws, therefore trying to enlarge understanding.

SOX 304 DFA 954

Only enforceable by SEC Enforceable by board of directors

Only triggered in case of a restatement based on misconduct

Triggered in case of any restatement

Recovery of all incentive-based pay and equity-based compensation

Recovery of only incentive-based pay

Applies to CEO’s and CFO’s only Applies to current and former executive

officers*

(11)

11 * The definition of executive officer is not defined, and therefore doesn’t provide enough guidance.

2.4 Clawback Adoption, Determinants and Consequences

Increased interest for clawback provisions in recent years has led to a new stream of academic literature. Brown et al. (2011) investigated the characteristics of firms that have chosen to voluntary adopt a clawback provision. They find that the size of the firm, goodwill impairment charges and extraordinary M&A bonuses are positively associated with the decision to voluntarily adopt a clawback policy. Furthermore they find a negative relationship with the level of CEO influence over the Board of Directors. In addition to the research on the determinants of voluntary clawback adoption, Addy, Chu and Yoder (2011) have found that adopters of voluntary clawback provisions have lower management entrenchment scores. On the contrary the likelihood of a clawback provision increases with the existence of director interlocks with other adopting companies and with the existence of recent restatements. They also find that when the CEO of the firm likewise serves as the chairman of the board, the adoption of a clawback provision will increase.

Another set of academic papers is focusing more on the consequences of clawback initiation rather than on the determinants. In order to get more knowledge with regard to clawback effectiveness, Chan et al. (2012) compare results for voluntary adopters versus non-adopters. They examine whether the adoption of a clawback provision lead to a lower likelihood that auditors identify and report material internal control weaknesses. By looking at the number of material internal control weaknesses between firms with and without clawback adoption, they conclude that auditors are indeed less likely to find and report material weaknesses in the internal control of clawback adopters. In addition adoption is linked to lower audit fees and to a higher earnings response coefficient. This earnings response coefficient could be seen as a part of a research stream trying to examine market reaction and signaling effects. By examining the earnings response coefficient one can examine whether the adoption of a clawback clause improves perceived financial reporting quality. Dehaan et al. (2012) show that voluntary adopters have a higher earnings response coefficient and a lower analyst forecast dispersion. In their research they try to examine a wide range of research questions regarding clawback adoption. Therefore they also find an increase in CEO compensation in case a clawback provision is adopted.

(12)

12

2.5 Quality Improving Mechanisms

A quick look in the academic auditing literature provides every reader with a staggering amount of studies, trying to examine the effects of various matters on a firms’ financial reporting quality. One of these literature streams provides evidence that managers with stock based and performance based compensation plans manipulate accounting information in order to maximize their personal gain. According to Palmrose and Scholz (2004) these manipulations could lead to restatements that give an indication that the financial reports are of poor quality. Policymakers are concerned with reducing the number of restatements by implementing mechanisms that should reduce managers’ incentives to ‘cook the books’. These new audit mechanisms or standards will be followed by academic research investigating whether or not they have the intended effect. For example there is a large research area focusing on the effects of IFRS adoption on earnings management activities. Tightened accounting standards may put pressure on management to comply with IFRS standards resulting in less earnings management (Ewert and Wagenhofer, 2005). Besides trying to reduce fraud by tightening accounting standards, Rockness and Rockness (2005) raise the question as to whether it is possible to legislate ethical behaviour. Providing ethical training could make managers (and other employees) more aware of what they are actually doing. More stringent regulation and the provision of ethical training are just two ways of effectively trying to reduce managers’ incentives to commit fraud, most regulation however, is focused on punishment of the wrong behaviour. The introduction of a clawback provision should, according to Dehaan et al. (2013), create incentives to reduce earnings management and to produce high-quality information. After all, the introduction of a clawback provision increases the cost of being caught. If caught, ‘ill-gotten’ performance-based compensation should be handed in by the manager responsible for these misstatements. Therefore clawback provisions fall within this category of punishment.

In addition to the reduction of earnings management, in case the firm has in place a robust clawback, it might as well lead to increased efforts to avoid unintentional errors (Dehaan et al., 2013). This increased effort will be shown in terms of a better focus on the underlying subject. Besides increasing one’s own effort it could also lead to a firm’s effort to increase training and therefore increase the technical skills of its employees. Better skilled employees will in theory make less unintentional mistakes.

(13)

13

2.6 Clawbacks and Financial Reporting Quality

As the main aim of introducing the clawback provision is to reduce the number of errors in the financial statements, and therefore to improve the quality of the financial statements, most of the research regarding clawbacks is focused on this relationship. Measuring financial reporting quality has been done in numerous ways. Chan et al. (2012) examines whether the clawback provision lead to a reducing number of financial restatements. They mention that the outcome of their research could have two possible outcomes, either a positive or negative relationship exists. The most obvious prediction would be that clawbacks lead to a change in managerial’ behaviour. Managers have less incentives to commit fraud by means of manipulating earnings if they are aware that this behaviour will be punished (Desai, Hogan and Wilkins, 2006). However, clawback adoption might not lead to a reduction in financial restatements as these clawbacks are mostly linked to the existence of misconduct. Brown et al. (2010) show that 47 percent of their sample choose to adopt a misconduct clawback, followed by 34 percent choosing a robust clawback. According to Lublin (2010) and Weiss (2009) the definition of misconduct is somewhat vague, a company has no guaranty of winning a lawsuit against a case of misconduct. In addition, even if the clawback is triggered, it is up to management to make the decision whether or not compensation will be recouped. Therefore, adoption of a clawback provision increases the discretion of the board. Because of this uncertainty, the financial media put question marks regarding the effectiveness of clawback provisions. It is also a possibility that clawback provisions are only used to signal high quality and integrity, instead of actual change in managerial behaviour (Spence, 1973).

Whereas the study of Chan et al. (2012) examines whether the clawback provision lead to a reducing number of financial restatements, a later study of Chan, Chen, Chen and Yu (2014) measures audit quality as the total amount of earnings management. A lower degree of earnings management is a signal of higher audit quality. They found that clawbacks lead to a reduction in accrual management, but will instead lead to an increase in real transactions management. This increase in real transactions management boosts short-term profitability. As such, the total amount of earnings management, combination of accrual management and real transaction management, does not decrease. So in contrast with earlier researches these results imply that a reduction in restatements does not automatically suggest an improvement in earnings quality.

(14)

14

2.7 Intentional Errors Versus Unintentional Errors

Restatements are commonly used as a proxy for measurement of financial reporting quality. According to Dechow, Ge and Schrand (2010) a major benefit of using restatement data is the lower ‘type 1 error’. In addition, restatement samples are mostly large in sample size. In comparison they state that researchers who are using accruals as a proxy for financial reporting quality, largely depend on a firms’ fundamental performance when using accrual-based measurements. Therefore there is a higher risk on distortions if not properly modelled.

Most studies assume that misreporting is mainly caused by intentional manipulation, nevertheless restatement samples contain both intentional errors and unintentional errors. A wide range of previous research has shown that this assumption, that misreporting is mainly caused by intentional manipulation, is not correct. Several studies illustrate that unintentional errors occur more often than intentional errors (Hayes, 2014). For instance, Scott A. Taub (2006), Acting Chief Accountant of the SEC, observed that more than half of all restatements were triggered by “ordinary books and records deficiencies or by simple misapplications of the accounting standards”.

Hennes Leone and Miller (2008) mention how important it is to make a distinction between restatements driven by intentional errors and restatements driven by unintentional errors. They claim that research related to restatement data could be enhanced by focusing more on the distinction between irregularities and errors, especially since the relative occurrence of errors in restatements is increasing. They stress that research based on the conventional assumption that restatements are mainly due to fraudulent behaviour rather than the occurrence of simple errors, are likely lead to the wrong conclusions as the frequency of unintentional error-driven restatements have increased substantially. They also demonstrate that restatements caused by intentional errors could have more consequences to a mangers’ career, due to job loss and/ or reputation loss, relative to restatements caused by unintentional errors.

According to Palmrose, Richardson and Scholz (2004), research using restatements as a proxy for financial reporting quality could be enhanced by distinguishing between the different types of restatements, as restatements caused by different types of misstatements can have different consequences. Fraud driven restatements tend to be more damaging to the integrity and reputation of a firm, which in turn lead to a higher litigation risk and a higher CEO turnover rate. These findings implies that in general fraud-driven restatements have a stronger impact on dependent variables.

(15)

15 Doyle, Ge and McVay (2007) also made this distinction by mentioning that a weak control environment allows both intentionally biased accruals and unintentional errors. With intentionally biased accruals one could think about a lack of segregation of duties and with unintentional errors one could think about a lack of experience, which could lead to wrong estimations regarding certain provisions. Subsequently Yu-Chun Lin (2013) used the concept of trigger events. Comparing ‘unintentional error triggers’ with ‘fraud triggers’ they found that different triggers could lead to different effects on financial reporting quality. Although the empirical value of this study can’t be considered as being of very high quality, the main message is that one can’t automatically assume that a drop in restatements is an indication of less fraud, it might as well be the result of a reduction in unintentional errors, due to example extended knowledge.

As I aim to do in this research I try to fill this gap in literature by focusing on both irregularities and errors and whether clawback adoption affect both kind of misstatements.

Chan et al. (2012) demonstrate that the occurrence of accounting restatements significantly declines after adoption of a clawback provision. Their explanation is that managers have lower incentives to conduct fraud and engage in earnings manipulation. However they don’t mention the possible occurrence of unintentional errors, they just assume that a drop in fraud cases is a logical explanation for the reduction in restatements. Based on the prior, suggesting that a small part of all restatements is linked to fraud makes me wonder if this reduction in restatements could be assigned to the ‘fraud part’. Unintentional errors happen in ratio more often, due to different causes. Restatements due to intentional errors might be attributed to management incentives, whereas restatements as a consequence of unintentional errors could be an indication of either lower competence (Demerjian, Lev, Lewis and McVay, 2013) or an indication of lower incentives to conduct effective controls over financial reporting (Hoitash, Hoitash and Johnstone, 2012). The adoption of a clawback provision may lead to increased efforts to avoid such unintentional errors (Dehaan et al., 2013). Besides arguing that clawback initiation could lead to increased incentives from employees it might also affect a firms’ effort to improve control. Chan et al. (2012) found that for firms adopting a clawback provision, the auditor is less likely to report a material internal control weakness. These results are consistent with management trying to improve the internal control to prevent restatements. In addition, under the signalling argument of Spence (1973), auditors may perceive clawback adopters as having a better internal control system, therefore reducing their own efforts to uncover misstatements, which also might lead to less restatements.

(16)

16 I build my first hypothesis based on the above-discussed views and upon the study of Dehaan et al. (2013). By making a distinction between robust clawbacks and misconduct clawbacks they provide some evidence that robust clawbacks have a larger effect on financial reporting quality compared to misconduct clawbacks. Suggesting that the combination of intentional errors and unintentional errors lead to a stronger reduction in restatements than the effect of intentional errors on its own. Therefore I see an incremental effect of unintentional errors on the reduction of restatements. I state my first hypothesis as follows:

H1: Robust clawback adoption is associated with a decrease in unintentional errors

The study of Dehaan et al. (2013) indicates that introduction of a robust clawback provision has led to a reduced number of fraud based restatements. However based on the first hypothesis in combination with the argument that the occurrence of fraud is relatively small relative to simple errors I expect that the decrease in unintentional errors will be more significant than the decrease in intentional errors.

H2: Robust clawback adoption is associated with a higher decrease in unintentional errors

as compared with intentional errors.

Whereas I have mainly focused on the post-adoption period, the second part of this thesis aims at providing insight in the relationship between the different kind of restatements and the choice between the adoption of either a robust clawback or a misconduct clawback. This way I am trying to increase understanding in the decision to adopt a particular clawback provision

The type of clawback provision that a firm has in place is possibly related to its previous experience with restatements. Addy and Yoder (2011) found that misstatements that are the result from irregularities have a significant relationship with the adoption of a clawback adoption. Gao et al. (2011) stated that restatements are the most essential factor in the decision to adopt a provision for clawbacks. Both these studies do not pay attention to the different types of clawbacks, Brown et al. (2011) do. Taking into consideration the different types of clawback, they don’t find an incremental effect of restatements on the likelihood to adopt a clawback provision. They only found that restatements resulting from irregularities are related to the adoption of a misconduct clawback provision.

Whether the previous history of misstatements of a firm is decisive in the choice to adopt a clawback adoption or that this choice is made on the basis of other factors, isn’t really

(17)

17 of relevance for what I aim to investigate in this analysis. Once the decision is made to adopt a clawback provision, in order to determine which kind of clawback provision will be adopted I still expect firms to take into consideration a firm’s past experience with restatements. To illustrate, a firm might adopt a clawback provision with the purpose of signalling high quality and integrity or because adoption is made mandatory. Once the decision is made to introduce a clawback provision, step two is to decide which kind of clawback to adopt, hence I expect that the previous history of restatements is relevant in making this decision.

If a firm has previously experienced a restatement resulting from an unintentional error I expect that a firm is more likely to adopt a robust clawback aimed at reducing both unintentional errors and intentional errors. On the contrary if the firm hasn’t been experiencing a restatement as a result of an unintentional error I expect that the particular firm will adopt a misconduct clawback, as they will be less concerned with preventing unintentional errors. After all, they haven’t had experience with unintentional errors in the last couple of years, which might be for example an indication of good qualified, dedicated staff. Hence, they don’t have a reason to focus on reducing these type of errors.

H3: Firms are more likely to adopt a robust clawback when they experienced a restatement resulting from unintentional error prior to adoption.

H4: Firms are more likely to adopt a misconduct clawback when they haven’t experienced

a restatement resulting from unintentional error prior to adoption.

Instead of making assumptions about the relationship between unintentional errors and the initiation of a misconduct clawback, one could argue that instead a positive relationship between intentional errors and misconduct clawbacks is expected. However, as I mentioned before, I expect that the occurrence of intentional errors is rare, therefore the focus of these hypotheses is on unintentional errors.

3 METHODOLOGY

3.1 Sample Selection 1

To test for hypothesis one and two I use a different dataset than for testing hypothesis three and four.

(18)

18 As I aim to do in this research I try to fill the gap in literature by focusing on the intentional and unintentional dimensions of restatement data. Restatement data is a commonly used proxy for measurement of financial reporting quality. Basically it means that companies declare that they have made mistakes, misstatements in the past, which is a direct signal that they were in violation with accounting rules. Also like mentioned earlier a major benefit of using restatement data is the lower type 1 error.

Clawback data will be retrieved from the Corporate Library database, consisting of North America firm year observations from 2007-2012. This data is classified according type of clawback provision. Therefore it is possible to investigate which trigger event the firm has in place and to find a possible relationship with future reductions in restatements. The first category is the fraudulent clawback which generally applies only in case of a fraudulent act leading to a restatement. The second type is the performance-based clawback that will be triggered when compensation has been distributed in case of incorrect financials. To stick with the terms used in this paper, the former could be seen as a misconduct clawback, whereas the performance-based clawback is similar to the robust clawback. The third category of clawback is the non-compete clawback, meaning that the clawback will be triggered in case of violating a restrictive covenant, for example violating a non-compete clause. The rest category, other clawbacks, is used to indicate that the firm has adopted a clawback provision other than a robust, misconduct or non-compete clawback.

The initial sample consists of 18,616 observations, dropping missing data results in a sample size of 18,413.

Audit Analytics provides restatement data, which covers the reason for a firm’s restatement. Errors due to ‘financial fraud, irregularities and misrepresentations’ are indicated as intentional errors whereas errors due to ‘accounting and clerical applications’ and errors due to ‘accounting rule (GAAP/FASB) and application failures’ will be classified as unintentional errors, leaving out the errors due to other significant issues. It is important to consider that Audit Analytics reports on the year in which the restatement is made and not on the year to which this restatement is related.

The number of observations remains intact after merging with the restatement data, retrieved from Audit Analytics, as I make the assumption that there hasn’t been a restatement if the field is empty.

The control variables, as described in section 3.3, will be mainly retrieved from the Compustat database, with exception from the corporate governance variables (BOARDSIZE,

(19)

19 control variables with the clawback and restatement data and dropping missing values the number of observations is reduced to 4988. After correcting for outliers, by means of dropping

variables equal to the 1st and 99th percentile, the final number of observations is 4697, consisting

of 1119 unique firms.

3.2 Sample Selection 2

The sample needed to test for the second and fourth hypothesis deviates somewhat from the first sample, as I use some other control variables and time variables.

I initially start with the same dataset, therefore the original number of observations remains the same. That is, after dropping missing values the number of firms that adopted a clawback provision is 18,413. Also the same procedure has been followed in merging the clawback data with the restatement data. Control variables are gathered from the Compustat database and the ISS database. Besides using partly the same control variables as needed for testing hypothesis one, new control variables are included and different time variables are set. After having done that, the sample consists of 3142 observations. Correcting for outliers, the same way I did in the first sample, the final number of observations is 2964, consisting of 1003 unique firms.

3.3 Control Variables - H1 + H2

To control for other factors that could possibly impact the relationship between misstatements and clawback provisions, some variables to control for this influence should be included in the model.

According to Woodland and Reynolds (2003) it is more likely that larger firms will have restatements than that their smaller counterparts will. Since Brown et al. (2011) argue that the size of a firm is one of the most important determinants for adoption of a clawback provision, I’ll include SIZE as a control variable in the model to reduce the possibility of bias. As a measure for SIZE the natural log of total assets will be used.

Also as larger audit firms tend to supply higher audit quality than smaller auditors (e.g. Lennox, 1999; Francis, 2004), the dummy variable BIG4 will be added, indicated as one when being audited by a Big-4 audit firm and zero otherwise.

Burgstahler and Dichev (1997) imply that firms with negative returns have an incentive to hide these losses from credit markets because this might negatively impact the cost of debt.

(20)

20 Consequently, this may lead to the increase of earnings management. LOSS is a dummy variable that equals one if a firm reports a net loss in one of the three years subsequent to the introduction of a clawback and zero if the firm didn’t report a loss in those years. A loss in one of these three years could be an indication of earnings management in year t or in the years before t. Accruals might have been used to hide losses, subsequently leading to a reversal of accruals in the following years and hence resulting in a loss. Wayne, Guay, Kothari and Watts (1996) state that a manager has limited opportunities to prevent a so-called reversal.

The Market-to-book ratio is a good indication of performance as it incorporates both a forward-looking perspective and a historical perspective. Prior literature demonstrates that managers of firms with high growth opportunities, shown by high market-to-book ratios, are more likely to receive more stock options than those working for a firm with low growth opportunities (Murphy, 2003). As a consequence, managers of firms with growth opportunities might have a higher incentive to achieve set performance targets and therefore the occurrence of manipulated financial numbers might increase. MTB will be included to control for this performance indication, and will be measured as the total market value divided by common/ordinary equity.

I expect that Return on assets might negatively influence earnings, as firms with a low

ROA ratio might have incentives to boost the results, eventually resulting in a better ROA ratio. ROA is measured as net income divided by total assets.

An important area of research is concerned with the role of corporate governance on earnings management. Therefore it is important to control for the effect of corporate governance variables on the occurrence of restatements. Consequently BOARDSIZE,

CEO-TENURE and CEO-DUALITY are added as control variables in the empirical model. The role

of the size of a board of directors on earnings management is inconclusive. A smaller board might have better financial oversight, accordingly leading to a better firm performance. Interchangeably, larger boards might in turn have a broader range of experience, which could also lead to a better firm performance (Xie, Davidson and DaDalt, 2003).

According to Pfeffer (1983) the length of tenure increases job-specific knowledge. In addition Hermalin and Weisbach (1991) provide evidence that a longer tenure enables a director to work more effectively due to increased knowledge and experience. Based on these results one would expect that tenure might influence the degree of earnings management. To control for this measure I include the tenure of the most important director in the board, the CEO. CEO-TENURE is measured by subtracting the year the current CEO started in the board of directors by the year the current CEO ended it service. A critical note is that this doesn’t

(21)

21 automatically indicate that the current CEO has fulfilled the role of CEO within the board during his or her total time of service.

Another subject of academic research is the practice in which a CEO also serves as the chairman of the board of directors. Davidson, Jiraporn, Kim and Nemec (2004) provide evidence that CEO’s who are also fulfilling the role of chairman are operating under greater pressure to show positive results. Whereas a lot of studies indicate that duality doesn’t harm overall firm performance, Davidson et al. (2004) claim that it may lead to greater agency problems under certain conditions. This discrepancy is just one of the examples in the literature that indicates that there hasn’t been reached consensus about the real effects of CEO-duality.

CEO-DUALITY is a dummy variable that is equal to one if the CEO is also the chairman of the

board and zero in absence of this duality. This control variable is derived by looking at “employment title CEO”, “employment title chairman” and “board affiliation”. If a director has been assigned the title of CEO, it doesn’t automatically indicate that the director is the CEO of that particular company. Whether the director is CEO of the particular firm can be determined by looking at board affiliation. If the variable indicates that the director is employed, one could draw the conclusion that the director is CEO of the firm. However if the variable indicates that the director is independent, one could conclude that the director is CEO of another firm; this data has been removed from the sample. Next it will be checked whether an employed CEO also has the title of chairman.

TABLE 1: VARIABLES + DESCRIPTION – SAMPLE 1

ERRORRESt Variable equal to 1 if the financial statement has been

restated in year t due to unintentional error, 0 otherwise

FRAUDRESt Variable equal to 1 if the financial statement has been

restated in year t due to intentional error, 0 otherwise

CB(ROBUST)t Variable equal to 1 if the firm has a robust clawback

provision in year t, 0 otherwise

CB(MISCONDUCT)t Variable equal to 1 if the firm has a misconduct

clawback provision in year t, 0 otherwise

CB(NONCOMPETE)t Variable equal to 1 if the firm has a non-compete

CB(OTHER)t Variable equal to 1 if the firm has a clawback

provision other than a robust, misconduct or non-compete in year t, 0 otherwise

(22)

22

SIZEt Variable measured as the natural log of total assets in

year t

BIG4t Variable equal to 1 if the firm is audited by a BIG4, 0

otherwise

LOSSt+1;t+2;t+3 Variable equal to 1 if the firm has reported a net loss

in the years t+1, t+2 or t+3

MTBt Market to book ratio in year t, measured as total

market value divided by common/ordinary equity

ROAt Net income divided by total assets in year t

BOARDSIZEt Number of directors on the board in year t, derived by

counting the number of individual directors for each firm and year

CEO-DUALITYt Dummy variable equal to 1 if the CEO is also the

chairman of the board in year t, 0 otherwise

CEO-TENUREt Years the current CEO has been on the board of

directors in year t, measured by subtracting the year the current CEO started in the board by the year the current CEO ended

3.4 Control Variables - H3 + H4

The control variables SIZE and MTB are also used to control for the adoption of a clawback provision. The inclusion of the variable SIZE is based on the political cost theory of Watts and Zimmerman (1986). According to this theory, firms are motivated to engage in activities that reduce political intervention. The initiation of a clawback provision could be one of those activities adopted with the intention to reduce political costs and since exposure increases with the size of a firm, SIZE is included in the empirical model. The same argumentation could be used for the control variable MTB.

To test for the third and fourth hypothesis, corporate governance control variables are added to the model based on the theory of Adams et al. (2005). Adams et al. (2005) primarily discuss a CEO’s ability to affect firm’s performance. Their overall conclusion is that an increase in CEO power lead to an increase in performance variability. Translated to the setting of clawback adoption, meaning that higher CEO power could affect the adoption of a clawback provision. Arguing that the adoption of a clawback could negatively influence a manager’s bonus pay-outs, I would say that CEO’s power might negatively influence clawback adoption, as they are using their power to stop the firm from adopting a clawback policy. To control for CEO power, the following two control variables are added to the model: CEO-TENURE and

(23)

23

CEO-DUALITY. Respectively referring to the years the current CEO has been on the board of

directors and to the event that the CEO is chairman of the board at the same time.

Following the same reasoning as behind the corporate governance variables as described above, I included the variable BOARDSIZE into the model; suggesting that besides the power of the CEO, the power of the board also influences firm’s performance. To control for this possible effect, BOARDSIZE represents the number of directors on the board. A large number of directors on the board will increase managerial power (Yermack, 1996). Data for this variable is obtained by counting the number of individual directors for each firm and year.

To control for positive influences on the adoption of a clawback, the control variables

PROFIT, ISSUE and IMPAIRMENT are included. Firms that have recently issued equity are

likely to benefit more from the positive signalling effects that comes with the adoption of a clawback provision (Brown et al., 2011). Signalling high quality and integrity might help a firm to get access to the capital markets, therefore the issuance of equity could positively influence the adoption of a clawback provision.

The recognition of a goodwill impairment involves a lot of judgement. Errors could arise in determining the amount of goodwill (wrong estimation of future cash flows) and errors could arise during determination of the amount of impairment. These errors could be the result of a wrong estimation, but could also occur as a result of fraud. To hedge themselves against these risks, firms that have had goodwill impairments in the past might sooner adopt a clawback provision relative to firms that didn’t experience a goodwill impairment. The last control variable to discuss is the variable PROFIT, based on the argument of Brown et al. (2011) that profitable firms are more likely to be subject to scrutiny if they don’t adopt a clawback provision.

TABLE 2: VARIABLES + DESCRIPTION – SAMPLE 2

CB(ROBUST)t Variable equal to 1 if the firm has a robust clawback

provision in year t, 0 otherwise

CB(MISCONDUCT)t Variable equal to 1 if the firm has a misconduct

ERRORRESt-1;t-2;t-3 Variable equal to 1 if the financial statement has been

restated in the years t-1, t-2 or t-3 due to unintentional error, 0 otherwise

(24)

24

FRAUDRESt-1;t-2;t-3 Variable equal to 1 if the financial statement has been

restated in the years t-1, t-2 or t-3 due to intentional error, 0 otherwise

SIZEt-1 Variable measured as the natural log of total assets in

year t-1

PROFITt Profit in year t measured as net income divided by

total market value

ISSUEt-1;t-2;t-3 Dummy variable equal to 1 if shares has been issued

in the years t-1, t-2 or t-3, 0 otherwise

IMPAIRMENTt-1 ;t-2 ;t-3 Dummy variable equal to 1 if there has been a

goodwill impairment (pretax) in the years t-1, t-2 or t-3, 0 otherwise

MTBt-1 Market to book ratio in year t-1, measured as total

market value divided by common/ordinary equity

BOARDSIZEt-1 Number of directors on the board in year t-1, derived

by counting the number of individual directors for each firm and year

CEO-DUALITYt-1 Dummy variable equal to 1 if the CEO is also the

chairman of the board in year t-1, 0 otherwise*

CEO-TENUREt-1 Years the current CEO has been on the board of

directors in year t-1, measured by subtracting the year the current CEO started in the board by the year the current CEO ended*

3.5 Empirical Model

The first regression model will test whether clawback adoption is associated with a decrease in unintentional errors and can be stated as followed:

H1

ERRORRESt = α + β1*CB(ROBUST)t + β2*CB(MISCONDUCT)t +

β3*CB(NONCOMPETE)t + β4*CB(OTHER)t + β5*SIZEt + β6*BIG4t +

β7*LOSSt+1;t+2;t+3 + β8*ROAt + β9*MTBt + β10*BOARDSIZEt +

(25)

25

Where CB(ROBUST)t is a dummy variable that will equal one if the firm adopted a robust

clawback and zero otherwise. The coefficient β1shows the variance in the dependent variable

CB(ROBUST)t and is expected to have a negative coefficient, implying that the adoption of a

robust clawback has led to a reduction in unintentional errors.

To test the second hypothesis, whether robust clawbacks are associated with a higher decrease in unintentional errors as compared with intentional errors, I conduct a similar

regression, but will replace ERRORRESt as the dependent variable for FRAUDRESt:

H2

FRAUDRESt= α + β1*CB(ROBUST)t + β2*CB(MISCONDUCT)t +

β3*CB(NONCOMPETE)t + β4*CB(OTHER)t + β5*SIZEt + β6*BIG4t +

β7*LOSSt+1;t+2;t+3 + β8*ROAt + β9*MTBt + β10*BOARDSIZEt +

β11*CEO-DUALITYt + β12*CEO-TENUREt + Residual

CB(ROBUST)t will equal one if the earnings has in place a robust clawback and zero otherwise.

I expect that robust clawbacks will lead to a higher decline in restatements due to unintentional

errors than due to intentional errors, therefore I expect coefficient β1in regression two to be

less significant than coefficient β1in regression one (but still be negatively associated with the

dependent variable). Instead I expect a negative significant relation on coefficient β2, implying that restatements due to intentional errors are mainly reduced through the adoption of a misconduct clawback.

Hypothesis three has been stated as follows: “Firms are more likely to adopt a robust clawback when they experienced a restatement resulting from unintentional error prior to adoption”. To test for this hypothesis, the following regression model will be used:

H3

CB(ROBUST)t= α + β1*ERRORRESt-1;t-2;t-3+ β2*FRAUDRESt-1;t-2;t-3 + β3*SIZEt-1 +

β4*PROFITt + β5*ISSUEt-1;t-2;t-3 + β6*IMPAIRMENTt-1 ;t-2 ;t-3 + β7*MTBt-1 +

β8*BOARDSIZEt-1 + β9*CEO-DUALITYt-1 + β10*CEO-TENUREt-1 +

Residual

ERRORRESt-1;t-2;t-3 is a dummy variable that will equal one if the firm experienced an

(26)

26 difference in the dependent variable robust clawback for firms with and without unintentional errors and is expected to have a positive coefficient.

Hypothesis four has been stated as follows: “Firms are more likely to adopt a misconduct clawback when they haven’t experienced a restatement resulting from unintentional error prior to adoption”. To test for this hypothesis, the following regression model will be used:

H4

CB(MISCONDUCT)t= α + β1*ERRORRESt-1;t-2;t-3+ β2*FRAUDRESt-1;t-2;t-3 + β3*SIZEt-1 +

β4*PROFITt + β5*ISSUEt-1;t-2;t-3 + β6*IMPAIRMENTt-1 ;t-2 ;t-3 + β7*MTBt-1 +

β8*BOARDSIZEt-1 + β9*CEO-DUALITYt-1 + β10*CEO-TENUREt-1 +

Residual

Instead of expecting a positive coefficient for β1, I now expect to find a negative relationship

between ERRORRESt-1;t-2;t-3 and CB(MISCONDUCT)t. Meaning that firms won’t be interested

in adopting a misconduct clawback when they have experienced restatements due to unintentional errors in the past three years.

4 DESCRIPTIVE STATISTICS AND EMPIRICAL

RESULTS

4.1 Descriptive Statistics - Sample 1

Table 3 gives an oversight of the distribution, mean and standard deviation from the different types of clawback provisions.

TABLE 3: CLAWBACK DESCRIPTIVES - SAMPLE 1

N Mean Std. Dev. Min. Max.

CB(MISCONDUCT)t 4697 0.199 0.400 0 1

CB(ROBUST)t 4697 0.170 0,376 0 1

CB(NONCOMPETE)t 4697 0.015 0.122 0 1

CB(OTHER)t 4697 0.009 0.093 0 1

Firms that have a robust clawback or a misconduct clawback in place are almost equally distributed over the sample; 20 percent of the observations are registered as having a

(27)

27 misconduct clawback, whereas 17 percent have a robust clawback in place. Only a very small portion of the sample could be assigned to the non-compete clawback and other clawback, the remainder of the observations don’t have a clawback provision in place.

To correct for duplicate firm observations, table 3 will provide an overview of the descriptive statistics for the most recent year of a unique firm value.

TABLE 3: UNIQUE FIRM CLAWBACK DESCRIPTIVES - SAMPLE 1

CB(MISCONDUCT)t 1119 0.252 0.434 0 1

CB(ROBUST)t 1119 0.224 0,417 0 1

CB(NONCOMPETE)t 1119 0.012 0.107 0 1

CB(OTHER)t 1119 0.007 0.084 0 1

As shown above 25 percent of the firms have adopted a misconduct clawback and 22 percent a robust clawback. These percentages are slightly higher than for the total observations, but most importantly is that the distribution remains equal. Most firms adopted a misconduct clawback, followed by firms adopting a robust clawback. This distribution is similar to that of Brown et al. (2011), who show that of their sample 47% percent choose to adopt a misconduct clawback followed by 34% choosing a robust clawback. On the one hand, the higher level of clawback adopters could be explained by the difference in sample, their sample covers the period 2005-2009, and on the other hand by the lower level of observations of Brown et al.’s study. Although a robust clawback might be more effective in that it might reduce both intentional and unintentional errors, this result could be explained by the fact that punishing unintentional errors is not effective as it could be perceived as being unfair.

TABLE 4: RESTATEMENT DESCRIPTIVES - SAMPLE 1

ERRORRESt 4697 0.047 0.212 0 1

FRAUDRESt 4697 0.001 0.025 0 1

Based on table 4 I can tell that 4.7 percent of the observations have experienced a restatement as a consequence of an unintentional error. Only 0.2 percent of the observations have had a restatement resulting from fraud. This last observation is in line with for example the study of Hayes et al. (2014), illustrating that unintentional errors occur more often than intentional errors. The same is supported by Scott A. Taub (2006), stating that more than half of all

(28)

28 restatements were triggered by “ordinary books and records deficiencies or by simple misapplications of the accounting standards”.

After correcting for missing values there were only three observations left that reported a restatement due to fraud. Due to this extreme low level of fraud observations it is not possible to perform reliant testing, therefore hypothesis two can’t be tested. I’ll perform an additional analysis that restores the data to its original state (before removing data because of missing value), including just the clawback data and the restatement data. The restatement data has been retrieved from the Audit Analytics database, in total, this database has 5345 observations, of which only 64 observations have a fraud restatement, while the remainder is assigned to unintentional errors. Therefore, even though I had to drop observations because of missing values, I could draw the conclusion that only a small portion of all restatement have had a

restatement resulting from fraud.

With respect to the control variables, table 5 gives an overview of the descriptive statistics of these variables, subdivided in firms that have a clawback and firms that don’t have any form of clawback.

(29)

29 The average size for all firms in the Compustat database (not restricted by this sample size), measured as the natural log of total assets, is 20.031. The average firm size for clawback firms

TABLE 5 : CONTROL VARIABLES DESCRIPTIVES – SAMPLE 1

TOTAL SAMPLE N Mean Std. Dev. 25% Median 75% Min. Max.

SIZEt 4697 21.832 1.541 20.658 21.742 22.875 18.776 25.886 BIG4t 4697 0.933 0.249 1.000 1.000 1.000 0.000 1.000 LOSSt+1;t+2;t+3 4697 0.014 0.116 0.000 0.000 0.000 0.000 1.000 ROAt 4697 0.050 0.065 0.018 0.050 0.086 -0.293 0.243 MTBt 4697 2.582 2.042 1.334 1.993 3.110 -0.702 17.126 BOARDSIZEt 4697 9.407 2.257 8.000 9.000 11.000 4.000 34.000 CEO-DUALITYt 4697 0.515 0.500 0.000 1.000 1.000 0.000 1.000 CEO-TENUREt 4697 11.786 8.822 5.000 9.000 15.000 1.000 62.000

CLAWBACK FIRMS N Mean Std. Dev. 25% Median 75% Min. Max.

NONCLAWBACK FIRMS N Mean Std. Dev. 25% Median 75% Min. Max.

(30)

30 in the sample of this research is 22.460. This is in line with the research of Brown et al. (2011) who show that clawback firms are on average larger than the control group. In addition 93 percent of the firms having a clawback provision are audited by a BIG 4 auditor.

Brown et al. (2011) provide a good oversight of the difference in statistics between firms adopted a clawback provision and firms that didn’t adopt a clawback provision. As they stated, the average market-to-book ratio for clawback-firms is higher than for the non clawback firms, which is consistent with the results shown in table 5. This slightly higher average, that is more distinct in the sample of Brown et al. (2011), could indicate that firms with growth opportunities might indeed earlier adopt a clawback provision than firms with lower growth opportunities.

The mean for return on assets for clawback firms is 0.050, which is slightly higher than the mean in the study of Chan et al. (2014), which is 0.037. This could possibly be explained by the larger sample size of 4704 observations versus a sample size of 343 clawback adopters. Only 1.4 percent of the total observations have reported a loss in the three following years.

With respect to the corporate governance control variables, on average half of the firms, 52 percent, have a CEO who is also the chairman of the board. This result is about similar with the results of Brown et al. (2011) who report a percentage of 59 percent. In addition the average tenure of a CEO in the board of directors is 12 years and the average number of directors in the board of a firm is 9 directors, similar to the results of Addy et al. (2011) who report an average board size of 10 directors.

The following table (table 6) represents the Pearson correlation table. Coefficients marked by one star indicates that the relevant coefficient has a significance level lower than 10 percent. Two stars are an indication of a significance level lower than 5 percent and a significance level lower than 1 percent are represented by three stars.

(31)

31

TABLE 6: PEARSON CORRELATION – SAMPLE 1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 1. CB(ROBUST)t 1 2. CB(MISCONDUCT)t -0.23*** 1 3. CB(NONCOMPETE)t -0.06*** -0.06*** 1 4. CB(OTHER)t -0.04*** -0.05*** -0.01 1 5. ERRORRESt -0.01 -0.01 -0.01 0.03** 1 6. FRAUDRESt -0.02 -0.01 -0.01 -0.00 0.06*** 1 7. SIZEt 0.17*** 0.23*** 0.01 0.01 -0.02 -0.02 1 8. BIG4t 0.03** 0.07*** 0.01 0.03* 0.02 -0.03** 0.23*** 1 9. LOSSt+1;t+2;t+3 -0.02 -0.03** 0.00 0.01 -0.00 -0.00 0.03** 0.00 1 10. ROAt 0.01 0.07 -0.00 -0.01 -0.06*** -0.04*** -0.06*** -0.02 -0.15*** 1 11. MTBt 0.02 0.04*** 0.01 -0.02 -0.04** -0.01 -0.04** 0.04*** -0.05*** 0.40*** 1 12. BOARDSIZEt 0.15*** 0.19*** 0.00 0.01 -0.02* -0.02 0.60*** 0.19*** -0.00 -0.06*** -0.02* 1 13. CEO-DUALITYt -0.02 0.06*** 0.00 -0.00 -0.02 -0.04*** 0.13*** -0.00 -0.01 0.04*** -0.01 -0.01 1 14. CEO-TENUREt -0.09*** -0.12*** -0.02 -0.03** -0.00 -0.01 -0.14*** -0.17*** 0.02 0.03* -0.03** -0.16*** 0.32*** 1

(32)

32 The different categories of clawback provisions are all negatively related to each other. This relationship is in line with the theory, implying that they are substitutes. There aren’t any firms that have in place more than one clawback provision.

It seems that none of the control variables really explain the variable error restatement, as the largest variable has only a coefficient of (-)0.06. Furthermore the tested relationship

between ERRORRESt and CB(ROBUST)t is negative, suggesting that the adoption of a robust

clawback lead to a reduction in error restatements. However according to the matrix error restatements only reduce with a coefficient of 0.1.

Instead of finding a relationship between error restatements and its dependent variables, there seems to be a relation between the adoption of a robust clawback and the size and boardsize of a firm. Respectively represented by a coefficient of 0.17 and 0.15.

The positive correlation between robust clawbacks and firm size suggests that a smaller firm is less likely to adopt a clawback provision. This relationship could be explained by the theory that smaller firms are expected to have less restatements (Woodland and Reynolds, 2003), therefore the need for clawback adoption could be lower. Further elaboration on the possible determinants of clawback adoption is discussed in the section that covers the results of hypothesis three and four.

4.2 Descriptive Statistics – Sample 2

Table 7 gives an oversight of the distribution, mean and standard deviation from the different types of clawback provisions for sample 2.

TABLE 7: CLAWBACK DESCRIPTIVES – SAMPLE 2

CB(MISCONDUCT)t 2964 0.243 0.431 0 1

CB(ROBUST)t 2964 0.199 0.399 0 1

CB(NONCOMPETE)t 2964 0.012 0.108 0 1

CB(OTHER)t 2964 0.013 0.113 0 1

The distribution of the different sort of clawbacks provisions between firm observations is generally the same as under sample one, although there are slightly more observations that are having a clawback provision (even though the number of observations is lower in sample two). This could be partly due to different firm year observations, but also because different control variables lead to a different process of dropping missing values. As provided in table 8 below,