Clawback provisions and executive compensation

Amsterdam Business School

Clawback provisions and executive compensation

Name: Michiel Sijp

Student number: 11160055

Thesis supervisor: dhr. dr. P. (Pouyan) Ghazizadeh Date: 24-06-2018

Word count: 12,481 (15,479 including references and appendices) MSc Accountancy & Control, specialization Control

Statement of Originality

This document is written by student Michiel Sijp who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

Using a sample of executives working for the S&P 1500 firms from 2016, this study examines whether the presence of clawbacks in a firm leads to the payment of a higher amount of total compensation and a higher proportion less-risky types of compensation. While prior research on executive compensation is limited and has focused on the amount of compensation, few address the relation between the presence of clawbacks and different types of executive compensation. Therefore I examine the relation between the presence of clawbacks and different types of compensation, specifically the proportion fixed pay compared to the proportion performance-based pay and the proportion debt-like compensation compared to the proportion equity-like compensation.

Recent years have witnessed an increased interest in ‘clawbacks’, or compensation recovery provisions, especially since the adoption of the Sarbanes-Oxley Act in 2002 (SOX) and the Dodd-Frank Act (DFA) in 2007. The clawback provision clause in these acts enables companies to recover already payed bonuses to executives, when they restate their financial statements as a result of misconduct. These clawback provisions should minimalize the agency problem. However, because clawbacks tie the executives’ pay to the long-term firm performance, the executive sustains the risk of higher pay volatility. This risk must be compensated for upfront by a risk premium. I expect this will be in the shape of a higher total compensation as well as a higher proportion fixed pay and debt-like compensation of total compensation.

I find that firms that employ clawbacks do compensate their executives by granting a higher amount of total compensation. They do not, however, compensate for the increased risk by granting higher proportions fixed pay and debt-like compensation.

Keywords: clawback provision, clawbacks, NEO compensation, executive compensation, risk aversion,

efficient market hypotheses, expectancy theory, equity-like compensation, debt-like compensation, fixed pay, salary, performance based pay, pay-for-performance, risk premium

Index

2.4 Risk aversion theory and efficient market hypotheses ... 7

2.6 Hypotheses development ... 8 3 Research methodology ... 9 3.1 Sample selection ... 9 3.2 Empirical method... 11 3.3 Control variables ... 12 3.4 Regression models ... 13 4 Results ... 16 4.1 Descriptive statistics ... 16 4.2 Regression analyses ... 19 4.2.1 Regression analysis H1 ... 19 4.2.2 Regression analysis H2 ... 21 4.2.3 Regression analysis H3 ... 23

5 Summery and Discussion ... 26

5.1 Summary ... 26

5.2 Conclusion ... 29

5.3 Limitations ... 30

5.4 Contribution and future studies ... 31

6 References ... 33

Appendices ... 39

Appendix 1: Examples of clawback provisions ... 39

Appendix 2: Sample categorized by industry ... 42

Appendix 3: S&P 1500 categorized by industry ... 43

1

1

Introduction

This study examines the influence of clawbacks on the total amount and structure of executive compensation on the basis of the following research question: Is the presence of clawback provisions

accompanied by a higher amount of total compensation and a less risky compensation structure with more emphasis on fixed pay and debt-like compensation?

‘Clawbacks’, or compensation recovery provisions, have been around at least since the 1800’s as a response to regulatory changes and to perceived failures in corporate governance (Babenko, Bennett, Bizjak, & Coles, 2017; Bloom, 2017). Recent years have witnessed an increased interest in clawbacks, especially since the adoption of the Sarbanes-Oxley Act in 2002 (SOX) and the Dodd-Frank Act (DFA) in 2007. In case of the latter as a response to earnings manipulation and fraud during the financial crisis. (Babenko, Bennett, Bizjak, & Coles, 2017; Dehaan, Hodge, & Shevlin, 2013; Fried & Shilon, 2012).

Both acts pose stricter rules on stock market listed companies with respect to corporate governance and accountability. Their aim is to enhance the company’s reliability and to discourage the type of risk-taking that prefaced the financial crisis. The clawback provision clause in these acts enables companies to recover already payed bonuses to Named Executive Officers (NEO’s), when they restate their financial statements as a result of misconduct (Lin, 2017; Chan, Chen, Chen, & Yu, 2012); acting as a corporate governance mechanism to retain executives from publishing misstated accounting information (Dehaan, et al., 2013).

Because of this increased interest, research on clawbacks has also flourished in recent years, not in the least because there are doubts about the effectiveness of clawbacks (Dehaan, et al., 2013). While the motivation for the implementation of these provisions lies in the critique on (the establishment of the amount of) NEO compensation, the question remains if clawbacks ensure fair game. While research on clawbacks is very diverse, ranging from the influence of governance structure to the influence on financial reporting, stock value, or firm performance (Addy, Chu, & Yoder, 2014; Babenko, Bennett, Bizjak, & Coles, 2012; Babenko, et al., 2017; Chan, et al. 2012; Davis-Friday, Fried, & Jenkins, 2011), less is known about the influence on different types of executive compensation.

Therefore motivation for this paper is threefold. First, since clawback policies are compensation contract modifications that can significantly modify the risk for NEO's in expected compensation, understanding the impact of clawbacks on NEO compensation is essential to understanding the overall impact of clawback provisions (Dehaan et al., 2013). Secondly, while prior research on NEO compensation focuses on the amount of compensation, few address the different types of compensation NEO’s receive while clawback provisions are in effect (Bhagat & Romano, 2009; Brown, Davis-Friday,

2

& Guler, 2011; Chen, Greene, & Owers, 2014; Dehaan et al., 2013). There has been some research on the specific types of incentives that Section 9541 _{of the DFA outlines as being covered by the clawback} regulation, but according to Pizoha (2013) there is still a wide range of compensation types which have not been included in research. Consequently, what has been lacking in literature is more detailed and comprehensive research on this issue. Thirdly, there’s a direct call for more comprehensive research on the matter (Pizoha, 2013).

Thus, building on prior research the aim of the current study is therefore not only to examine the influence of clawbacks on the total amount of NEO compensation, but also on the ratio fixed pay and performance-based compensation and on the ratio debt-like and equity-like compensation (Babenko et al., 2017; Chan et al., 2012). This research might bring new information to light that may influence management decisions on implementing clawback provisions and it may also inform regulators how to effectively implement clawback policies in firms, keeping the balance between shareholders and executives.

This paper seeks to achieve the above aims by examining data from several stock market listed American companies. Before this data collection process and the research design is discussed in more detail, an overview of the key concepts and relevant theoretical perspectives will be provided. Following the research design, the results will be presented and linked to the theoretical perspectives provided earlier in the paper. The final section of this paper will consist of concluding arguments, the limitations, contribution and recommendations for further research.

3

2

Literature review

In the literature review I will present the most important theoretical concepts related to the hypotheses: NEO compensation, the agency theory, equity- and debt-like compensation, clawback provisions, the risk aversion theory and efficient market hypotheses.

2.1 NEO compensation

According to Globler (2005) most payments to executives consist of a base salary and a performance-based part. This performance-performance-based part can take up many forms and has existed at least since the 1980’s (Bryson, Freeman, Lucifora, Pellizzari, & Perotin, 2012). As a reward for individual performance, it is usually subdivided into merit and bonus pay, where merit pay refers to an annually cumulative increase in base salary used to reward past performance and bonus pay is a single payment used to recognize past performance which has to be earned each year (Gerhart et al., 1992). Examples of bonus pay, a central issue in this paper, are an annual bonus in cash, stocks and options (Frydman & Jelter, 2010; Geiler & Renneboog, 2010; Globler, 2005; Pyzoha, 2013). Whereas stocks are a claim to a share in the assets and earnings of a company (Harrison & Kreps, 1978); and stock option consists of the possibility, not the obligation, to purchase a stock at a previously agreed upon fixed price, for example the current stock price (Gerhart, Milkovich, & Murray, 1992; Gobler, 2005). In case of the latter the executive receives a part of the generated profit he contributed to. When the stock price increases, the executive still has the right to buy the stock for the previously agreed upon price.

Hence, NEO’s are not just compensated by means of a regular salary, they’re also rewarded based on their own and the company’s performance. Research indicates approximately ninety percent of American companies choose to pay-for-performance in some degree (Bryson et al., 2012). Mainly because performance-based pay positively influences the company’s performance. As employees tend to be rational and self-interested when it comes to money, pay-for-performance appears to be a motivator: since monetary rewards are most effective to improve employee performance and employee performance influences the firm’s success, pay is related to employee performance (Gerhart, Milkovich, & Murray, 1992; Jensen & Meckling, 1976; Kuvaas, 2006). This is endorsed by the expectancy theory, which states intrinsic motivation is based on the person’s ability and the motivational forces (Gerhart & Rynes, 2003). The latter consists of expectancy (the expected performance related to the made effort), instrumentality (the sufficiency of the reward in relation to the demonstrated performance) and valence (the value of the pay in relation to other, mostly negative, outcomes). The employee will therefore exert effort based on the expected reward.

4

Even though it’s a widespread practice, performance-based pay has proven to be a sensitive and complicated issue (Frydman & Jenter, 2010; Rodgers & Gago, 2003). It has not only been subject to criticism in mainstream media and politics, research has far from exhausted the subject (DiDonato, 2014; Fisher, 2016). Performance-based compensation is believed by its critics to be negatively related to performance, demotivating, has a negative influence on cooperation and is difficult to implement and monitor (Ariely, Gneezy, Loewenstein, & Mazar, 2009; Deci & Ryan, 1980; Lawler, 2000; Pfeffer, 1998). The most common criticism is the fact that NEO’s are rewarded for performance achievements they not only decide upon, but also report on themselves (Kroos Schabus, & Verbeeten, 2017). This dual role of the executive enables him to engage in accounting manipulations fairly easily to further his own interest instead of the company’s or the shareholders’ interest. Some critics therefore belief the high amount of current incentive compensation is the result of financial manipulation by powerful managers motivated to enhance their own income, increasing the imbalance between the shareholder’s and executive’s interest (rent-seeking; Pyzoha, 2013). These critiques have exacerbated during the past few decades, especially since the amount of performance-based compensation paid to executives has grown dramatically since the mid 1970’s (Frydman & Jelter, 2010).

2.2 Agency theory

The mentioned imbalance or conflict of interest between the shareholder and executive, also known as ‘agency problem’, has been a central subject of discussion since key research by Ross (1973) and Jensen and Meckling (1976). Jensen and Meckling (1976) define an agency relationship as “a contract in which one or more persons (the principal(s)) engage another person (the agent) to take actions on behalf of the principal(s) which involves the delegation of some decision-making authority to the agent.” (p. 308). So we’re dealing with an agency relationship when one party (the agent) acts on behalf of another party (the principle) in a particular domain of decision making (Ross, 1973). An agency problem occurs when a conflict of interest arises within this relationship: one assumes both parties will act solely out of self-interest to further their own utility, and these self-interests are not congruent (Ross, 1973). So, the problem with an agency relationship is the fact that both parties seek to maximise their own personal gain (Jensen &Meckling, 1976).

Perrow (1986) adds to this by stating an agency relationship will not only be disturbed because of self-interest; it can be because of ineffectiveness of surveillance. This is due to the fact that the agent has expert knowledge at his disposal the principle lacks. This concerns knowledge necessary to evaluate if the agent is doing what he’s supposed to do, as well as knowledge related to the content of the task at hand (Sharma, 1997). Since this knowledge asymmetry is characteristic for an agency relationship, it’s difficult to implement effective monitoring mechanisms (Shapiro, 2005). Jensen and Meckling (1976)

5

state therefore that there is good reason to assume the agent will not always act in the best interest of the principal. The executive does not bear the full consequences of his actions, has effective control and consequently the incentive and ability to arrange for benefits at the expense of the shareholders.

As a counter measurement the principal is inclined to provide certain incentives to encourage the agent to act in the principle’s interest (Jensen & Meckling, 1976). Performance-based compensation is an example of such counter measurements, although research has shown not all types of performance-based compensation are effective (Edmans & Liu, 2011; Sundaram & Yermack, 2007). For example, equity-like compensation increases the executive’s tendency towards risk-taking behaviour (Edmans, 2010; Sundaram & Yermack, 2007). His objective is to increase the solvency of the company. Consequently, equity-holding executives tend to perform best when the company thrives, seeing their effort translated in the increase of the company’s value while it’s already solvent. In case of a bankruptcy, the equity-holding executive as well as shareholders are left with no payoffs, regardless of the liquidation value. The executive will therefore in all probability choose solvency over liquidation value, going against the interest of the shareholders. Nevertheless, according to years of research, NEO’s are most frequently compensated with equity-like instruments, such as restricted stock, stock options, and other instruments whose value is tied to future equity returns.

‘Inside debt’ moderates these risk-shifting incentives and therefore reduces the riskiness of the company’s outside debt. ‘Inside debt’, or an intracompany IOU, means debt held by a manager or executive in their own company: they get paid in fixed sums (of cash) in the future (Edmans, 2010; Jensen & Meckling, 1976; Sundaram & Yermack, 2007). Examples are defined benefit pensions and deferred compensation. It encourages the executive to be concerned with the value of the company in bad times, because it yields a positive payment in case of a bankruptcy related to the liquidation value (Edmans & Liu, 2011). Debt-like compensation therefore reduces the agency problem (Sundaram & Yermack, 2007).

Another example of a counter measurement is the employment of certain governance mechanisms (Bakke, Mahmudi, & Virani, 2017; Chen, Greene, & Owers, 2014; Dehaan, Hodge, & Shevlin, 2013; Li, 2014). Governance mechanisms are even considered to be one of the most common and effective measures to mitigate the agency problem, because it forces executives to act in the interest of the shareholders by means of monitoring and verifying their performance. (Daily, Dalton, & Cannella, 2003; Geiler & Renneboog, 2010; Li, 2014; Prescott & Vann, 2018). The extent to which it restores the balance between agent and principle might differ based on the type of governance mechanism in use and the characteristics of the concerning firm (Bakke, Mahmudi, & Virani, 2017; Chen, Greene, & Owers, 2014; Dehaan, Hodge, & Shevlin, 2013; Li, 2014). Clawbacks are a fairly recent governance mechanism deployed to solve the agency problem.

6 2.3 Clawback provisions

Even though clawbacks have been around for a long time, they’ve only become more customary since the introduction of the SOX in 2002 which allows the Securities and Exchange Commission (SEC) to force American stock market listed firms to recover paid NEO compensations when, in case of a restatement, the amount of compensation appears to have been based on false information due to misconduct (Li, 2014; Prescott & Vann, 2018). It’s usually a contractual agreement about the hypothetical repayment of performance-based compensation in case of fraud (Brown et al., 2011). The SOX regulation was put into place because of concerns about excessive NEO compensations and the apparent lack of correlation between the compensation and performance in practice. These clawback provisions should minimalize an executive’s tendency to manipulate the financial reports and counter the fact that performance-based compensation tends to motivate executives to make riskier investment choices to increase the firm’s value (Kroos, Schabus, & Verbeeten, 2017; Pyzoha, 2013). This is because the performance-based pay increases with the riskiness of firm’s assets (Jensen & Meckling, 1976).

Because the SOX doesn’t explicitly mandate all companies to implement a clawback policy and because the enforcement could only be enacted by the SEC when the restatement is based on fraud, the SEC only incidentally put the provision to use. Some companies chose to adopt a policy voluntarily (Li, 2014; Prescott & Vann, 2018). Employing voluntary clawback provisions strengthens the credibility of the firm’s governance and reputation towards stockholders (Brown et al., 2011).

But as a result of the worldwide economic crisis in 2007, the DFA came in to practice in 2010. In the years leading up to the crisis, executives had received extremely large bonuses. Because of the crisis these compensations were no longer reasonable, making it necessary to adjust the governance mechanisms regulating these incentives to avoid excessive compensations like the ones prefacing the financial crisis. This second Act therefore forced firms to adopt a clawback policy, which made the firms themselves, instead of the SEC, responsible for recovering excess pay (Fried & Shilon, 2012; Chan et al., 2012). Contrary to the SOX, the DFA is mandatory2_{, although a company can give substance to the} policy at their own discretion. Besides that, the excess pay does not have to be the result of fraud or explicit misconduct of the executive to have to be recovered after a restatement. The fact that firms can decide how to give substance to the clawback policy themselves, causes big differences between companies on this matter (Addy et al., 2014).

2 _{The DFA consists of several sections. Not all sections are implemented yet, including the section 953a on}

pay-for-performance disclosure and 954 on the recoupment of executive compensation (SEC, 2016). Therefore, it’s considered good practise to implement clawbacks according to the proposed Dodd-Frank provisions, but not yet mandatory.

7 2.4 Risk aversion theory and efficient market hypotheses

In all cases however, even though clawbacks are a solution to the agency problem, they cause a new problem: because clawbacks tie the executives’ pay to the long-term firm performance, the executive sustains the risk of higher pay volatility (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010). While firms without a clawback policy can’t alter the NEO compensation afterwards, for firms with a clawback policy the executives’ pay is susceptible to being reclaimed by the firm. As a result executives in firms using clawback provisions have an increased risk, because they ‘are not in perfect control of the clawback trigger’ (Kroos et al., 2017).

According to the risk aversion theory, founded on the prominent work of Arrow (1965) and Pratt (1964) (Menezes & Hanson, 1970) people “prefer to receive a sure amount equal to the expected value of the risk, to the risk itself”, that is people prefer to receive a guaranteed fixed salary over an uncertain performance-based monetary reward (Menezes & Hanson, 1970). Thus, people will prefer a low-risk job with a lower amount of fixed pay over a high risk job with a higher amount of volatility. Due to clawbacks the executive has higher pay volatility.

In an efficient market all known information is reflected in the price. ‘Efficient market’ refers to the fact that “a capital market is said to be efficient if it fully and correctly reflects all relevant information in determining security prices.” (Malkiel, 1989). In this case the executive knows working for a different firm reduces the incurred risk. This risk of higher pay volatility must be compensated for upfront in order to stay competitive in the market for managerial talent (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013). This means the implementation of clawback provisions could be considered to involve a risk premium. A risk premium is a compensation for risk (Amihud, 2002).

8 2.6 Hypotheses development

Because an executive might have to repay (part of) his compensation in case of a clawback, NEO’s working for firms with clawbacks have an increased risk (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013; Malkiel, 1989). In an efficient market this risk must be compensated for with a risk premium. Therefore, I expect that this risk will be compensated by a larger total compensation.

H1: Total compensation will be higher for NEO’s working for firms that apply clawbacks, than for NEO’s working for firms without clawbacks.

Besides that, performance-based incentives, such as stock options have an increased risk opposed to fixed pay (Geiler & Renneboog, 2010; Shen, Gentry, & Tosi, 2010). While, according to the risk aversion theory, NEO’s do not like variability in their compensation and clawbacks increase this risk involved in performance-based pay; I expect NEO’s could use the less-risky nature of fixed pay to reduce their exposure to recoupment. In this case the proportion of fixed pay to be higher opposed to the proportion performance-based compensation in a firm that has implemented clawbacks (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Eisenhardt, 1989).

H2: NEO’s working for a firm that has implemented clawbacks will receive a higher proportion fixed pay opposed to the proportion performance-based pay in comparison with NEO’s working for firms without clawbacks.

Thirdly, not every form of performance-based pay is perceived to be subject to the same amount of risk (Geiler & Renneboog, 2010). Firms might choose to compensate the executives’ risk premium, due to the presence of clawbacks, by means that involve the least risk. That is to say, since equity-like compensation encompasses more risk than debt-like compensation, I expect NEO’s could use the less-risky nature of debt-like compensation to reduce their exposure to recoupment. In this case the proportion debt-like compensation opposed to the proportion equity-like compensation would be higher for NEO’s that work at firms with clawbacks (Jensen & Meckling, 1976; Krishnaswami & Yaman, 2008).

H3: NEO’s working for a firm that has implemented clawbacks will receive a higher proportion debt-like compensation opposed to the proportion equity-like compensation in comparison with NEO’s working for firms without clawbacks.

9

3 Research methodology

In this chapter the used methods will be described. First the sample selection, data collection and manipulation will be discussed. Subsequently, a paragraph about the empirical method will follow as well as the control variables.

3.1 Sample selection

In order to investigate the previous hypotheses this research examined the rich data of a sample of executives from the S&P 1500 companies of fiscal year 2016, using information provided by WRDS in the databases Compustat and Execucomp. Fiscal year 2016 was chosen because it was the most recent data available. Using the most recent data would minimize any possible lag between implementation of clawbacks and its effects. To collect the data from Compustat the following path was used: Compustat / North America – Annual updates / Index Constituents [date range 2016; TIC = I0020]; for Execucomp: Compustat / Daily updates / Execucomp – Monthly updates / Annual compensation [date range 2016; TIC’s of S&P1500 uploaded in .txt-format].

Information on all but one variable, namely clawbacks, was extracted from these databases. Because I had no access to the database containing information on clawbacks (IncentiveLab), the presence of clawbacks was identified by manually searching the Edgar-database by using keywords to search through the 10-K-, 8-K- and DEF14A-forms. Like Babenko et al. (2017) and Bakke et al. (2017) I’ve presumed that when no results were found on variants of clawbacks, recoupment and recovery of incentives in the mentioned forms in the Edgar-database, the firm does not employ clawbacks3_.

To extract information about the content of the concerning clawback provisions, I’ve read through the descriptions in Edgar; hereby establishing i.e. the type of employee it applies to, the duration of the clawback provision in place, the types of compensation it encompasses, the cause of the clawback and any mentioning of the relation to the SOX or the DFA. Examples of some representative descriptions extracted from Edgar are found in Appendix 1.

Because the data on the presence of clawbacks in firms of the S&P 1500 was manually collected, and because the amount of firms without clawbacks is limited, random selection of the sample of firms

3 _{Because firms use different types of descriptions for clawbacks, I’ve used a variety of keywords to search the 8K-, 10K- and}

DEF14A-forms of the S&P1500-companies in EDGAR. Following instructions on Boolean strings in EDGAR, I’ve searched on clawback OR claw back OR recoupment OR recover* OR recoup*. Since clawbacks can also be used in private equity funds, I’ve verified the result is used in the terms indicated in this study.

10

within the S&P 1500 data was not possible4_{. Instead the sample of 183 firms was selected by first} manually checking for firms without clawbacks. Table 1 shows a summary of the data collection and manipulation that followed. After checking for clawbacks, 82 firms without information on total assets for 2016, not in Compustat nor in manually searched annual reports, were removed from the sample. Thirdly, I excluded 323 financial firms with Standard Industrial Classification code (SIC-code) 60-67 (Finance Insurance and Real estate), because compensation within these firms is more regulated.

Using this data a match was formed between the companies with and without clawbacks to ensure both groups would approximately contain the same industries and firms of similar size. To match firms concerning industries, SIC-codes were used. I tried to include as a broad a spectrum of industries as possible. To measure firm size the firm’s book assets extracted from Compustat were used. This match enabled the comparison of NEO compensation between companies with and without clawbacks.

Table 1

Data collection and manipulation

Manipulation of data firms observations

S&P 1500 1601 1601

- firms without information on AT 82 82

- financial firms (SIC-code 60-67) 323 323

- firms without match in industry code and firm size 1008 1008

Subtotal 188 188

Merge databases 188 1514

- executives without (valid) data (TDC = blank or ‘0’) 5 605

Total 183 909

Most of the remaining 1196 firms could be matched with a firm within the same industry (SIC-code) with a similar total assets amount. If this was not possible, a match was formed with a company with similar total assets within a similar industry. An assessment about the similarity of industries was made by manually checking the SIC-codes by industry on siccode.com. Subcategories from the same industry were matched. In 1008 cases it was not possible to find a realistic match. These firms were

11

excluded from the sample, leaving a sample of 188 firms. Firms without clawbacks were not excluded from the sample in order to function as a control group. Table 8 in Appendix 2 shows the firm sample categorised by industry. Table 9 in Appendix 3 shows the S&P 1500 companies categorized by industry. The proportions of the industries included in the firm sample corresponds with the proportion in the S&P 1500 as much as possible, considering the fact that I had to include all available and valid firms without clawbacks due to their limited number.

After matching and manipulating the sample, 188 firms remained (see Table 1). Data on the firms and executives in these firms was collected from Compustat and Execucomp using the paths mentioned at the beginning of this paragraph. The data definitions of all collected variables from Compustat and Execucomp are found in Table 10 in Appendix 4.

Both datasets were merged on a one to one bases on key variable TICKER5_{. I removed double} observations, observations with missing data on total compensation (TDC1 = blank or ‘0’) and on the variable ID-number for each executive/company combination (CO_PER_ROL = blank or ‘0’). In some cases the number of stocks was negative. These values have been replaced by absolute values. In order to test the hypotheses I finally generated variables based on the variables found in the databases. The hypotheses were tested with data on 909 observations from 183 firms.

3.2 Empirical method

As a dependant variable for H1, total NEO compensation was measured using the natural logarithm of data-item TDC1 in Execucomp (LNTDC1). I’ve used the natural logarithm because TDC1 was skewed (skewness = 8.04, p < .001). In Execucomp total compensation for the individual year is comprised of the following: Salary, Bonus, Long term incentive plans, Other Annual, Total Value of Restricted Stock Granted, Total Value of Stock Options Granted (using Black-Scholes6_{), and All Other Total.}

For fixed pay, the dependent variable in H2, the variable PERCTFIXED was used, expressed as a percentage of TDC1 (SALARY / TDC17_{). I’ve presumed salary is the fixed part of total compensation,}

5 _{“A ticker symbol is an arrangement of characters (usually letters) representing a particular security listed on an exchange or}

otherwise traded publicly.” (https://www.investopedia.com/terms/t/tickersymbol.asp#ixzz5HrOhSAb5)

6 _{“The Black Scholes model, also known as the Black-Scholes-Merton model, is a model of price variation over time of}

financial instruments such as stocks that can, among other things, be used to determine the price of a European call option. The model assumes the price of heavily traded assets follows a geometric Brownian motion with constant drift and volatility. When applied to a stock option, the model incorporates the constant price variation of the stock, the time value of money, the option's strike price and the time to the option's expiry.”

(https://www.investopedia.com/terms/b/blackscholes.asp#ixzz5Hud2RJLJ)

12

because it’s the risk-free part of the compensation and not subject to recoupment. The amount of the remaining parts of compensation depends on the executive’s and firm’s performance. They are more volatile and are therefore considered to be riskier than the fixed part of compensation. It should be noted however, not all of the remaining parts of total compensation are unequivocally performance-based. For example, it’s possible part of deferred earnings is not performance-based, but this information is not specified within the used variable. Nonetheless I had no reason to assume this hypothetical fixed part of deferred earnings substitutes for the fixed part in my conceptual framework. Therefore,I’ve presumed the remainder of TDC1 to be the performance-based part.

To test H3 the dependent variable debt-like compensation had to be expressed as a percentage of the performance-based part of total compensation (PERCTDEBT = (DEFER_EARNINGS_TOT + PENSIONNQDC) /( DEFER_EARNINGS_TOT + PENSIONNQDC + OPTIONAWARDS + STOCKAWARDS)). Debt-like compensation consists of deferred earnings (DEFER_EARNINGS_TOT) and Change in Pension Value & Nonqualified Deferred Compensation Earnings (PENSIONNQDC). Equity-like compensation consists of Option Awards (OPTIONAWARDS) and Stock Awards (STOCKAWARDS)8_.

3.3 Control variables

All dependent variables are subject to a lot of different factors of influence, like firms size and complexity. These factors will be added to the regression formula as control variables. They have been included to measure whether part of the variance of the amount of total compensation, percentage debt-like compensation and percentage fixed compensation are (partly) caused by subsequently the presence of clawbacks, stock valuation, firm leverage, and firm size. These control variables were extracted from Compustat.

Stock valuation was measured using Tobin’s Q, by dividing market value of assets by book value of assets (TQ = (AT + (CSHO * PRCC_F) - CEQ) / AT) (Babenko et al., 2017). According to Bai, Liu, Lu, Song and Zhang (2004) there’s a positive relationship between stock valuation and executive compensation. It’s even strongly positively associated with performance-based incentives (Sanders & Boivie, 2004). This means firms with a higher stock market valuation tend to grant their executives a higher amount of total compensation and reward executives with a higher proportion performance-based incentives. I presume therefore, directly related to all three hypotheses, that a firm’s stock

13

valuation explains part of the variance of total compensation, of the proportion performance-based compensation and of the proportion equity-like compensation.

Firm leverage was measured through dividing current plus long-term liabilities by total liabilities (LEV = (DLTT + DLC) / LSE). Prior research shows the firm’s leverage is related to the amount and structure of executive compensation (Frydman, & Molloy, 2011; Graham, Li, & Qui, 2011). Research has also found a positive relation between the firm’s leverage and the NEO’s debt/equity ratio (Bryan, Nash, & Patel, 2006; Sundaram & Yermack, 2007). I presume therefore, related to all three hypotheses, that firm leverage explains part of the variance of total compensation, of the proportion fixed pay and of the debt-like part of compensation.

Firm size was measured using the natural logarithm of a firm´s total book assets (LNAT). I use the natural logarithm because AT is skewed (skewness = 6.34, p < .001). Prior research shows that firm size is positively related to compensation height as well as compensation structure (Conyon & Murphy, 2000; Eaton & Rosen, 1983; Mehran, 1995). More specific, NEO’s working for larger firms with more growth opportunities, tend to receive a higher amount of compensation (Ozkan, 2007). Besides that, the proportion equity-like compensation is positively related to firm size (Eaton & Rosen, 1983; Mehran, 1995). I presume therefore, directly related to H1 and H3 and indirectly to H2, that firm size explains part of the variance of total compensation and of the equity-like part of compensation.

Initially I wanted to include firm complexity, NEO-tenure, institutional ownership, firm performance and firm’s risk as control variables. These variables are frequently used as control variables for research relating executive compensation (Gerhart & Rynes, 2003). For example for firm complexity I intended to use the variable research and development expenditures (R&D). Prior research suggests a link between R&D expenditures and compensation height and structure (Clinch, 1991; Milkovich, Gerhart, & Hannon, 1991; Smith, & Watts, 1992). More specific, ‘high tech’ firms with high R&D expenditures grant higher executive compensation and tend to use more performance-based compensation, like stock option plans. Nonetheless, they were not included, because too few observations were made available in Compustat and Execucomp to perform a reasonable regression.

Running the test with or without R&D expenditures as control variable did not change the result much. So, R&D expenditures has little explanatory power. It does however reduce the number of observations to 511. Since I value having more observations more than the little explanatory power R&D expenditures would add, I chose not to include this control variable.

3.4 Regression models

Using the described data, I’ve tested H1 with an Ordinary Least Square analysis (OLS) with the natural logarithm of total compensation as dependent variable, clawback as an independent variable and stock

14

valuation (TQ), firm leverage (LEV) and the natural logarithm of firm size (LNAT) as the control variables:

LNTDC1i1*CB i + 2*TQ i + 3*LEV i + 4*LNAT i + i .

Where, for fiscal year 2016 and person-year i:

CB = dummy: if clawback provision is present: 1; if clawback provision is not present: 0 TQ = total market value of firm divided by total asset value of firm

LEV = current plus long-term liabilities divided by total liabilities LNAT = natural logarithm of total assets

I predict the variable CB to have a positive relation with the dependent variable total compensation, due to a higher pay volatility that must be compensated for with a risk premium (Amihud, 2002; Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013).

Since proportion fixed is a fractional variable, H2 has been regressed using a fractional probit regression with the dependent variable percentage salary of total compensation (PERCTFIXED):

PERCTFIXED i1*CB i + 2*TQ i + 3*LEV i + 4*LNAT i + i

I predict the variable CB to have a positive relation with the dependent variable proportion fixed pay, due to the fact that performance-based incentives have an increased risk opposed to fixed pay (Geiler & Renneboog, 2010; Shen, Gentry, & Tosi, 2010). To compensate for the higher risk involved in clawbacks, I expect executives to get paid in a higher proportion fixed pay (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Eisenhardt, 1989).

For H3 I’ve also used a fractional probit regression with the dependent variable percentage debt-like of total compensation (PERCTDEBT):

PERCTDEBT i1*CB i + 2*TQ i + 3*LEV i + 4*LNAT i + i

I predict the variable CB to have a positive relation with the dependent variable percentage debt-like of total compensation, due to the fact that debt-debt-like compensation is inherently less risky than equity-like compensation (Geiler & Renneboog, 2010; Jensen & Meckling, 1976; Krishnaswami & Yaman, 2008).

15

Because the dataset contains different unique observations from the same firm, repeated deviation from the regression line should be considered. Therefore cluster robust standard errors have been used to correct for this possible bias in standard error.

16

4 Results

In this chapter the results will be presented and clarified. To test the three described hypotheses I’ve performed several regression analyses. Before describing these analyses and findings, the descriptive statistics will be presented.

4.1 Descriptive statistics

Since extreme data, or outliers, can influence the descriptive statistics and regression coefficients, continuous variables with outliers have been winsorized where needed, before regression analyses were performed. With winsorization the extreme observation values in the smallest percentile are replaced by the smallest value in the second percentile. Extreme observation in the largest percentile are replaced by the largest value of the ninety-ninth percentile. By doing so the influence of extreme observations is reduced. LNTDC1 and PERCTDEBT have been winsorized at one and ninety-nine percent. The variable LEV has relatively more outliers and has therefore been winsorized at five and ninety-five percent.

Table 2 shows the amount of person-year observations (N), the mean, standard deviation, minimum and maximum value of the dependent variables total compensation, proportion fixed and debt-like compensation and the control variables stock valuation (TQ), firm leverage (LEV) and firm size (LNAT) based on the presence of clawback provisions (CB). A t-test was performed to assess whether the mean of the different variables differs significantly between the two groups. The t-values are presented in the last column, including the level of significance indicated by asterisk(s).

Table 2

Descriptive statistics: summary statistics of full sample

With Clawbacks Without Clawbacks

Variable N Mean Std. Min. Max. N Mean Std. Min. Max. T-test

LNTDC1 470 7.42 .81 5.38 9.86 439 7.25 .93 5.38 9.86 3.01** PERCTFIXED 470 .30 .17 0 1 439 .36 .22 0 1 4.14** PERCTDEBT 451 .08 .19 -.01 1 408 .12 .25 -.01 1 2.23* CB 470 439 TQ 470 2.04 1.13 .27 6.99 433 2.13 1.07 .60 5.58 1.24 LEV 470 .25 .21 0 .64 439 .25 .19 0 .64 .61 LNAT 470 7.27 1.1 4.48 10.78 439 7.28 1.11 5.04 11.08 .27

17

470 person-year observations are from companies with clawback provisions and 439 are from companies without clawbacks. This difference is due to the fact that not all companies employ the same amount of executives. Besides that only singular person-year observations (data on one executive) were removed while manipulating the data, not all observations on the firm and the firm’s match. This might have caused disproportionate distribution of person-year observations among both groups.

The mean of LNTDC1 for firms with clawbacks is $ 1,669,000 (e7.42_{), and $ 1,408,000 (e}7.25_{) for} firms without clawbacks. This means that on average executives working for firms with clawbacks receive $261,000 more total compensation. This is significantly higher than their counterparts in the control group on a 1% significance level. This suggests that firms with clawbacks grant their executives a higher amount of total compensation than firms without clawbacks. This evidence is consistent with prior research which states that because of the increased risk due to clawbacks, in an efficient market executives must be compensated upfront with a risk premium (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013; Malkiel, 1989). It’s also consistent with H1: Total

compensation will be higher for NEO’s working for firms that apply clawbacks, than for NEO’s working for firms without clawbacks. However, the effect of 19% found in this data is much higher than I would expect were

it compensation for risk alone. This indicates that other unseen factors that are correlated with both an increase in total compensation and the presence of clawbacks might influence these results.

The mean of proportion fixed pay (PERCTFIXED) of total compensation is 30% for firms with clawbacks, while the mean of fixed pay is 36% for firms without clawbacks. This means that on average executives working for firms with clawbacks receive 6 percentage point less proportion fixed pay of total compensation on a 1% significance level. Interestingly, the mean of SALARY does not significantly differ between firms with clawbacks ($ 481,000) and firms without clawbacks ($ 472,000) (t = .56, p = .29). This suggests the increase in total compensation related to the presence of clawbacks consists of performance-based pay. This evidence is in contrast with my expectations based on prior research, which suggests that executives do not like the volatility and higher risk involved in performance-based compensation increased by the presence of clawbacks (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Eisenhardt, 1989; Geiler & Renneboog, 2010; Shen, Gentry, & Tosi, 2010). The results also contradict H2: NEO’s working for a firm that has implemented clawbacks will receive a higher proportion fixed pay

opposed to the proportion performance-based pay in comparison with NEO’s working for firms without clawbacks.

The mean of proportion debt-like compensation (PERCTDEBT) is 8% for firms with clawbacks, and 12% for firms without clawbacks. This means that on average executives working for firms with clawbacks receive 4 percentage point less proportion debt-like compensation of total compensation on a 5% significance level. This indicates executives working for firms with clawbacks receive a significantly lower proportion debt-like compensation than executives working for firms

18

without clawbacks. This evidence contradicts my expectation that (inherently less risky) debt-like compensation is used as a method to counterbalance the increased risk due to clawbacks (Geiler & Renneboog, 2010; Jensen & Meckling, 1976; Krishnaswami & Yaman, 2008). The results contradict H3:

NEO’s working for a firm that has implemented clawbacks will receive a higher proportion debt-like compensation opposed to the proportion equity-like compensation in comparison with NEO’s working for firms without clawbacks.

The included control variables do not differ significantly between firms with and without clawbacks. This was to be expected since firms in both groups were matched according to firm size and industry.

Table 3 shows the Spearman’s rank-order correlation coefficients between the dependent and independent variables used in the regression models. I’ve used a Spearman’s instead of the more commonly used Pearson’s correlation, because the assumptions for the Pearson’s were markedly violated. Hence, some variables were not normally distributed. First of all the table shows the presence of clawbacks is positively significantly related to total compensation. This means firms with clawbacks reward their executives significantly more than firms without clawbacks. This is consistent with H1. LNTDC1 shows a significant negative relation with PERCTFIXED, indicating an increase in total compensation consists of performance-based compensation, as described before.

Table 3

Spearman’s correlation between variables

Variable LNTDC1 PERCTFIXED PERCTDEBT CB TQ LEV LNAT

LNTDC1 1 PERCTFIXED -.86** 1 PERCTDEBT .03 .04 1 CB .09** -.11** -.02 1 TQ .03 -.08* -.09** -.04 1 LEV .16** -.17** .10** .01 -.24** 1 LNAT .43** -.34** .30** -.02 -.32* .39* 1

**,* Significant on a .01, .05 level, respectively.

Secondly, according to Table 3 the presence of clawbacks is negatively significantly related to proportions fixed pay. This means executives working for firms with clawbacks receive a significantly lower proportion fixed pay than executives working for firms without clawbacks. This is in contrast to H2. Furthermore, the Spearman’s correlation between presence of clawbacks and PERCTDEBT is not significant, unlike shown in Table 2. This suggests there is no significant difference in proportions debt-like compensation between groups. Based on this ambiguous evidence no conclusions can be drawn about H3. Finally, the control variables do not show sign of high correlation, meaning multicollinearity is

19

not a problem. A Variance Inflation Factor (VIF) test has also been done to confirm that there is no high correlation between two or more independent variables. Table 4 shows the mean VIF value is 1.09 and the highest VIF value is 1. This confirms my assumption that there is no sign of multicollinearity.

Table 4

Variance inflation factors

Variable VIF 1/VIF

CB 1 .99

TQ 1.11 .90

LEV 1.13 .89

LNAT 1.12 .89

Mean VIF 1.09

However, these findings are preliminary and do not clarify the variance in the dependent variables is necessarily caused by the presence of clawbacks. Conclusions can only be drawn after controlling for other factors that may influence total compensation, proportion fixed pay and proportion debt-like compensation in a regression model.

4.2 Regression analyses

4.2.1 Regression analysis H1

To test whether part of the variance in total compensation is explained by other factors than the presence of clawback provisions (H1) I’ve used an Ordinary Least Square analysis (OLS) with the natural logarithm of total compensation as dependent variable, clawback as an independent variable and stock valuation (TQ), firm leverage (LEV) and the natural logarithm of firm size (LNAT) as the control variables. The equation below was used to estimate the coefficients of the presence of clawbacks and the control variables. The coefficient ₁ was used to test whether the first hypothesis is correct:

LNTDC1i1*CB i + 2*TQ i + 3*LEV i + 4*LNAT i + i

Preliminary analyses were conducted to ensure no violation of the assumptions of normality, linearity, multicollinearity and homoscedasticity were present. The Shapiro-Wilk test indicated the residuals were not normally distributed (W = .99, p < .001). However, visual inspection of the Kernel density estimate suggested residuals were not problematically non-normally distributed. None of the other assumptions for multiple regression were violated. The model as a whole was significant and

20

explained 27% of the variance in total compensation (F = 82.92, p < .001). This indicated my model has a better fit to the dataset than the intercept-only model.

Table 5 shows Tobin´s Q was found to significantly predict total compensation (b = .16, p < .001). This suggests that firms that are valued higher on the stock market grant their NEO’s a higher amount of total compensation. An increase in one unit of TQ was associated with an increase of 16% in total compensation. This means, for example, an increase in stock valuation ratio from 1 to 29 _is associated with an average increase of 16% total compensation. This is consistent with prior research (Bai et al., 2004; Sanders & Boivie, 2004).

Table 5

Regression results of the natural logarithm of total compensation

With Clawbacks

Variable Exp. sign Coefficient Robust p value

INTERCEPT 3.9294 .000** CB + .1935 .009** TQ + .1588 .000** LEV + .1431 .426 LNAT + .4045 .000** Number of observations 903 Adjusted R2 _26.56%

**,* Significant on a .01, .05 level, respectively.

Leverage was not a significant predictor (b = .14, p = .43), which indicates that the firm’s proportion outside-debt was not associated with total compensation received by NEO’s employed by that firm. Additional statistical information about the regression can be found in Table 5. Due to ambiguous evidence on the relation between leverage and total compensation, no conclusions could be drawn about this relation.

LNAT was found to be a significant predictor (b = .40, p < .001), which suggests larger firms gave higher total compensation to their NEO’s. An increase of AT by 1% was associated with an increase of .4% in total compensation. This is consistent with prior research (Conyon & Murphy, 2000; Eaton & Rosen, 1983; Mehran, 1995; Ozkan, 2007).

21

When total compensation was predicted, it was found to differ significantly depending on whether firms implemented clawbacks. Firms with clawback provisions reported higher total compensation (b = .19, p = .009) than firms without. This means an increase of one unit in clawback provisions, switching from firms without clawbacks to firms with clawbacks, was associated with an average increase of 19% total compensation when all other variables are held constant. Given the mean of total compensation is $ 1,408,000 (e7.25_{) for firms without clawbacks, this would mean an increase of $} 267,000. The test was also performed on TDC1 giving a coefficient of 226,000. This supports the previous findings related to LNTDC1. The null-hypothesis can therefore be rejected. The findings support the alternative hypothesis (H1): Total compensation will be higher for NEO’s working for firms that apply

clawbacks, than for NEO’s working for firms without clawbacks. . This is consistent with prior research (Chen et

al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013; Malkiel, 1989).

4.2.2 Regression analysis H2

To test whether part of the variance in percentage salary of total compensation could be explained by other factors than the presence of clawback provisions (H2) I’ve used a fractional probit regression with the percentage salary of total compensation (PERCTFIXED) as dependent variable, clawback as an independent variable and stock valuation (TQ), firm leverage (LEV) and the natural logarithm of firm size (LNAT) as the control variables. For H2 a fractional probit regression was used instead of an OLS, because the dependent variable was fractional and skewed (skewness = 1.26, p < .001). This skewness was not a problem, given the fact that the fractional probit regression is a non-parametric test, and therefore robust against violations of assumptions. The equation below was used to estimate the coefficients of the presence of clawbacks and the control variables. The coefficient 1 was used to test whether the second hypothesis is correct:

PERCTFIXED i1*CB i + 2*TQ i + 3*LEV i + 4*LNAT i + I

None of the assumptions for the test were violated. The full model containing all predictors was significant (χ2 _{(4, N = 903) = 213.84, p < .001). Using a probit regression, the marginal effect}10 _of changing a variable on the dependent variable was not constant. It depends on the coefficient, the value of the independent variable and all other variables. To calculate the impact of the independent variable on the dependent variable the values for all other variables were held at their means. The coefficient

22

results given were the marginal effect on the z-score, when all other variables are held constant at their mean; which makes them a meaningless predictor of proportion fixed pay of total compensation. Therefore, in Table 6 in the last column the marginal effect in percentage points of proportion fixed pay of total compensation was given.

Table 6

Regression results of the proportion fixed pay of total compensation

With Clawbacks Variable

Exp. sign Coefficient Robust p value

Perc. point increase/ Elasticity

INTERCEPT 1.5166 .000** CB + -.1638 .008** -.0291 TQ - -.1170 .000** -.0843 LEV + -.3324 .018* -.0284 LNAT - -.2141 .000** -.2141 per 1% AT Number of observations 903

**,* Significant on a .01, .05 level, respectively.

NEO’s received significantly less proportion fixed pay in firms with a higher market valuation (b = -.12, p < .001). An increase of one unit TQ was associated with a decrease of 12% in z-score of proportion fixed pay of total compensation. This translates to .084 percentage point decrease in proportion fixed pay per 1% increase of Tobin’s Q. This means firms with higher stock valuation grant their executives a smaller proportion fixed pay of total compensation. This is consistent with prior research which has shown a positive relation between stock valuation and performance based-compensation (Sanders & Boivie, 2004).

Leverage was found to have a significantly negative relation with proportion fixed pay (b = -.33,

p = .02), which indicates that highly leveraged firms tend to give their NEO’s a smaller proportion fixed

pay of total compensation. This means, when leverage increases with one percent, proportion fixed pay is decreased by .03 percentage point. This is in contrast with prior literature (Bryan, Nash, & Patel, 2006; Sundaram & Yermack, 2007).

The natural logarithm of firm size was significantly negatively associated with proportion fixed say (b = -.21, p < .001), which indicates larger firms tend to give their NEO’s a smaller proportion fixed pay of total compensation. An increase of AT by 1%, proportions fixed pay decreases by .21 percentage point. This is consistent with prior research which has shown a positive relation between firm size and performance based-compensation (Eaton & Rosen, 1983; Mehran, 1995).

Firms that implemented clawback provisions gave significantly less proportion fixed pay to their NEO’s (b = -.16, p = .01). This means an increase of one unit in clawback provisions, switching from

23

firms without clawbacks to firms with clawbacks, was associated with an average decrease of 16% in z-score, or 2.9 percentage point proportion fixed pay of total compensation when all other variables are held constant. Given the mean of total compensation is 36% for firms without clawbacks, this would mean a decrease to 33.1%. This means the results supports neither the null-hypothesis nor the alternative hypothesis (H2): NEO’s working for a firm that has implemented clawbacks will receive a higher

proportion fixed pay opposed to the proportion performance-based pay in comparison with NEO’s working for firms without clawbacks. This is contrast with prior research (Chen et al., 2014; Cheng, Hong, & Scheinkman,

2010; Eisenhardt, 1989).

4.2.3 Regression analysis H3

To test whether part of the variance in percentage debt-like of total compensation is explained by other factors than the presence of clawback provisions (H3) I’ve used a fractional probit regression with the percentage debt-like of total compensation (PERCTDEBT) as dependent variable, clawback as an independent variable and stock valuation (TQ), firm leverage (LEV) and the natural logarithm of firm size (LNAT) as the control variables. The fractional probit regression was used instead of an OLS, because the dependent variable was fractional and skewed (skewness = 2.35, p < .001). This skewness was not a problem, given the fact that the fractional probit regression is a non-parametric test, and therefore robust against violations of assumptions. To test H3 only 840 observations could be used, because some observations had a negative value and the dependent variable of this regression can only take values on a zero to one scale. This could introduce some selection bias. The equation below was used to estimate the coefficients of the presence of clawbacks and the control variables. The coefficient

₁ was used to test whether the third hypothesis is correct:

PERCTDEBT i1*CB i + 2*TQ i + 3*LEV i + 4*LNAT i + i

During preliminary analysis, the assumption of homoscedasticity for H3 was found to be violated (χ2 _{(4, N = 840) = 67.1, p < .001), meaning the error variance in proportion debt-like was not} constant across the range of observations. This means the parameter estimates could be biased. Therefore, a fractional heteroscedastic probit regression would be considered more suited to predict proportion debt-like of total compensation. This full model containing all predictors was unfortunately not significant (χ2 _{(4, N = 840) = 6.1, p < .19). The model as a whole could not significantly predict the} outcome of proportion debt-like of total compensation, therefore I present the results of the fractional probit regression below. These results, considering the heteroscedasticity of the model, should be interpreted with caution.

24

Table 7 shows NEO’s received significantly less proportion debt-like compensation of total compensation in firms with a higher market valuation (b = -.16, p < .001). An increase of one unit TQ is associated with a decrease of 16% in z-score of proportion debt-like compensation of total compensation. This translates to .05 percentage point decrease in proportion debt-like compensation per 1% increase of Tobin’s Q. This means firms with higher stock valuation grant their executives a smaller proportion debt-like compensation of total compensation. This is consistent with prior research which has shown a positive relation between stock valuation and equity-like compensation (Sanders & Boivie, 2004).

Table 7

Regression results of the proportion debt-like compensation of total compensation

With Clawbacks

Variable Exp. sign Coefficient Robust p value Perc. point increase/ Elasticity

INTERCEPT -1.6234 .000** CB + -.1997 .019* -.0158 TQ - -1.1600 .000** -.0498 LEV + -.2639 .314 -.0116 LNAT - .1101 .004** .1101 per 1% AT Number of observations 840

**,* Significant on a .01, .05 level, respectively.

Leverage was found to have a non-significant negative relation with proportion debt-like compensation (b = -.264, p < .31), so no prediction could be made about this relation.

The natural logarithm of firm size was significantly negatively associated with proportion debt-like compensation (b = -.11, p = .004), which indicates larger firms tend to give their NEO’s a smaller proportion debt-like compensation of total compensation. An increase of AT by 1% is associated with a decrease in proportion debt-like compensation by .11 percentage point. This is consistent with prior research which has shown a positive relation between firm size and equity-like compensation (Eaton & Rosen, 1983; Mehran, 1995).

Firms that have implemented clawback provisions gave significantly less debt-like compensation to their NEO’s (b = -.1997, p = .02). This means an increase of one unit in clawback provisions, switching from firms without clawbacks to firms with clawbacks, was associated with an average decrease of 20% in z-score, or 1.6 percentage point proportion debt-like compensation of total compensation when all other variables are held constant. Given the mean of total compensation is 12% for firms without clawbacks, this would mean a decrease to 10.4%. This means the results supports neither the null-hypothesis nor the alternative hypothesis (H3): NEO’s working for a firm that has

25

implemented clawbacks will receive a higher proportion debt-like compensation opposed to the proportion equity-like compensation in comparison with NEO’s working for firms without clawbacks. This is in contrast with prior

26

5 Summery and Discussion

5.1 Summary

This study examines the influence of clawbacks on the total amount and structure of NEO compensation. Prior research presents the employment of governance mechanisms, such as the implementation of clawbacks, as one of the most common and effective solution to the agency problem (Bakke, Mahmudi, & Virani, 2017; Chen, Greene, & Owers, 2014; Dehaan, Hodge, & Shevlin, 2013; Li, 2014). The reason for this is the fact that it forces executives to act in the interest of the shareholders by means of monitoring and verifying their performance, aiming to restore the balance between agent and principle (Daily, Dalton, & Cannella, 2003; Geiler & Renneboog, 2010; Li, 2014; Prescott & Vann, 2018).

Even though clawbacks are a solution to the agency problem, they cause a new problem: because clawbacks tie the executives’ pay to the long-term firm performance, the executive sustains the risk of higher pay volatility (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010). While firms without a clawback policy can’t alter the NEO compensation afterwards, for firms with a clawback policy the executives’ pay is susceptible to being reclaimed by the firm. As a result executives in firms using clawback provisions have an increased risk, because they ‘are not in perfect control of the clawback trigger’ (Kroos et al., 2017). According to the risk aversion theory this risk, in an efficient market, must be compensated for upfront (Amihud, 2002; Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013). This means the implementation of clawback provisions might involve a risk premium.

Because of this higher pay volatility due to the presence of clawbacks, I expect executives to be compensated in the shape of a higher amount of total compensation and a different, less risky compensation structure. In order to examine whether executives working for firms that have clawback provisions in effect receive more total compensation and are compensated in a different way, I developed the following research question: Is the presence of clawback provisions accompanied by a higher amount

of total compensation and a less risky compensation structure with more emphasis on fixed pay and debt-like compensation?

In order to answer this research question, I developed three hypotheses. The first hypothesis is concerned with the influence of clawbacks on the amount of compensation, wherein I expect that executives will be compensated for the increased risk due to the presence of clawbacks by a larger total compensation (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Iskandar-Datta & Jia, 2013; Malkiel, 1989):

27

H1: Total compensation will be higher for NEO’s working for firms that apply clawbacks, than for NEO’s working for firms without clawbacks.

The second hypothesis is concerned with the influence of clawbacks on the proportion of fixed pay of total compensation. Performance-based incentives, such as stock options have an increased risk opposed to fixed pay (Geiler & Renneboog, 2010; Shen, Gentry, & Tosi, 2010). While, according to the risk aversion theory, NEO’s do not like variability in their compensation and clawbacks increase this risk involved in performance-based pay; I expect the proportion of fixed pay to be higher opposed to the proportion performance-based compensation when a firm has implemented clawbacks (Chen et al., 2014; Cheng, Hong, & Scheinkman, 2010; Eisenhardt, 1989):

H2: NEO’s working for a firm that has implemented clawbacks will receive a higher proportion fixed pay opposed to the proportion performance-based pay in comparison with NEO’s working for firms without clawbacks.

The third hypothesis is concerned with the influence of clawbacks on the proportion debt-like compensation of total compensation. Not every form of performance-based pay is perceived to be subject to the same amount of risk (Geiler & Renneboog, 2010). Therefore firms might choose to compensate the executives’ risk premium, due to the presence of clawbacks, by means that involve the least risk. That is to say, since equity-like compensation encompasses more risk than debt-like compensation, I expect a higher proportion of debt-like compensation opposed to the proportion equity-like compensation (Jensen & Meckling, 1976; Krishnaswami & Yaman, 2008).

H3: NEO’s working for a firm that has implemented clawbacks will receive a higher proportion debt-like compensation opposed to the proportion equity-like compensation in comparison with NEO’s working for firms without clawbacks.

Information on all but one variable, namely clawbacks, was extracted from Compustat and Execucomp. The data on the presence of clawbacks in firms of the S&P 1500 was manually collected. After checking for clawbacks and various necessary data manipulation (Table 1), a sample of 909 person-year observations in 183 S&P 1500 firms remained. This sample contained 470 person-person-year observations from companies with clawback provisions and 439 from companies without clawbacks. With these observations a t-test was performed to examine if the mean of total compensation, proportion fixed pay