THE POWER OF THE FAT CAT

THE POWER OF THE FAT CAT

The effects of excess pay and CEO power on impression management

Eveline Vaessen Oudezijds Kolk 73 1012 AL Amsterdam e.g.vaessen@st.rug.nl 0636159967 S2503441 Rijksuniversiteit Groningen

Faculteit Economie & Bedrijfskunde Master Accountancy & Controlling Track Accountancy

1

THE POWER OF THE FAT CAT

The effects of excess pay and CEO power on impression management

ABSTRACT

This paper draws on the agency theory to explore the impact of excess executive compensation and CEO power on impression management in the remuneration report of listed financial institutions in the UK. The results indicate that firms with excess CEO compensation employ strategic noise as well as less readable remuneration reports to obfuscate shareholders. Some evidence was found that remuneration report were also less readable when a powerful CEO was present.

KEYWORDS: executive remuneration, impression management, CEO power, managerial power theory, readability, strategic noise

2

TABLE OF CONTENTS

II. THEORY & HYPOTHESIS 7

2.1 AGENCY THEORY 7 2.2 SAY ON PAY 8 2.3 IMPRESSION MANAGEMENT 9 2.4 CEO POWER 10 III. METHOD 13 3.1 SAMPLE 13 3.2 DEPENDENT VARIABLE 13 3.3 INDEPENDENT VARIABLES 14 3.4 CONTROL VARIABLES 15 3.5 REGRESSION MODEL 16 IV. RESULTS 17 4.1 DESCRIPTIVE STATISTICS 17 4.2 MAIN RESULTS 18 4.3 ADDITIONAL ANALYSES 20 V. DISCUSSION 24

5.1 CONCLUSION AND DISCUSSION 24

5.2 THEORETICAL AND PRACTICAL IMPLICATIONS 25

5.3 LIMITATIONS AND AVENUES FOR FURTHER RESEARCH 25

3

I.

INTRODUCTION

“The only way to reduce fat cats is by having more watchdogs”1

The financial crisis has led to lots of criticism about the huge sums of remuneration paid to the CEO in times of low performance. The media has described executive remuneration as "fat cat pay"2 and "inflammatory"3, illustrating that emotions run high when CEO pay is the subject. The financial industry in particular has been a prime target, which may be a result of their executives receiving excessive pay while their banks had to be bailed out by the government. As policy makers may have perceived that the financial crisis was a result of bad corporate governance4, their response was to reform governance.

Because excessive CEO pay is not in line with the best interest of shareholders, corporate governance mechanisms are put in place to limit the agency costs (Osma & Guillamón-Saorín, 2011; Bednar, 2012). Outside directors are one example of such governance mechanisms, as they are expected to represent shareholders' interest. They however have incentives to support pay packages favorable to the executive. When they gain a reputation of a director who defies pay packages preferred by CEO s, it could impair their prospect to join other boards (Bebchuk & Fried, 2005). Collegiality, friendship and loyalty could explain why these directors may be unwilling to fight the CEO about the remuneration. It may be clear that directors favoring the CEO with excess pay do not act in the best interest of the shareholders nor act as a good corporate governance mechanism.

Recent efforts to reinforce corporate governance have focused on enhancing shareholders’ ability to influence corporate policy (Armstrong, Gow, & Larcker, 2013; Mason, 2012). In this light, the United Kingdom introduced say on pay in 2002 as a governance mechanism to limit the agency costs borne by shareholders. Say on pay gives shareholders the right to vote on the remuneration report at the general meeting, voting for or against the remuneration of the CEO. Initially put forward as a non-binding measure, its goals were to encourage shareholder participation in corporate governance, to protect their rights to the residual earnings of the firm, to confine excessive CEO pay and to help reduce CEO ’s incentives to chase short-term profits (Mason, Palmon, & Sudit, 2012). As the crisis progressed, public astonishment of the enormous amounts of CEO compensation increased. As a result, say on pay regulations became binding in the UK in 2013 implying that majority voting dissent on the Directors Remuneration Report (DRR) eliminates the possibility to grant the remuneration to the directors.

Research suggests that say on pay is an important governance mechanism as it helps reducing the CEO ’s influence over the board by empowering the Board of Directors (BoD) in their negotiations with the CEO (Correa & Lel, 2013; Mason et al., 2012; Burns & Minnick, 2013). According to Bebchuk (2007) voting dissent can act as a constraint and lead to more efficient bargaining between the board and the CEO. Correa & Lel (2013) provide evidence that say on pay is associated with a lower level of CEO compensation. They also find that executive pay 1 http://www.theguardian.com/commentisfree/2009/may/25/editorial-corporate-governance-pay-bonus retrieved 23-02-2015 2 http://www.theguardian.com/business/2012/jan/07/david-cameron-fat-cat-pay retrieved 16-02-2015 3 http://www.theguardian.com/business/2014/nov/25/bg-pay-deal-excessive-institute-of-directors retrieved 21-02-2015 4 http://www.theguardian.com/books/2015/feb/22/how-good-we-can-be-hutton-review-back-future-socialism-peter-hain-blue-labour retrieved 23-02-2015

4

is lower for firms with more CEO power following the say on pay regulation. Whether or not boards of directors will actually use their increased bargaining position remains unsure since they also have incentives to side with the CEO.

While the implementation of say on pay generally led to an increased market value (Cuñat, Gine, & Guadalupe, 2013), voting dissent can have a negative impact on a firm. This was even the case in the beginning when the vote was not binding. The costs derived from voting dissent can be significant since it requires engagement and action from the board (Krause, Whitler, & Semadeni, 2014) and has consequences for the reputation of the company and its directors. Even though say on pay voting outcomes are not-binding, high voting dissent has to be taken seriously (Kardis II, Earle, & Smith, 2012) and can have far-reaching effects (Krause et al., 2014).

It is in the firms’ interest to obtain an approving vote from its shareholders in order to avoid renegotiation costs and loss of reputation. Previous research shows that shareholders disapprove of excessive pay levels as well as CEO pay with a weak pay for performance sensitivity (Cai & Walkling, 2011; Armstrong et al., 2013; Carter & Zamora, 2007; Conyon & Sadler, 2010; Cotter, Palmiter, & Thomas, 2012; Kimmey, 2013; Kimbro & Xu, 2013; Ferri & Maber, 2013). This offers the directors who approve excess CEO pay with a motive to engage in impression management in order to influence the opinion of the shareholders and swing the votes on the say on pay resolution (Laksmana, Tietz, & Yang, 2012; Krause et al., 2014). Impression management can be described as the actions of an organization to shape stakeholders impressions and ensure they have a positive image of the firm (Hooghiemstra, 2000; Neu, Warsame, & Pedwell, 1998). Various impression management techniques can be used in an attempt to influence shareholders impressions (Graffin, Carpenter, & Boivie, 2011; Elsbach, Sutton, & Principe, 1998; Neu et al., 1998), including strategic noise, readability, deceptive language and positive/negative tone. I will look at strategic noise and readability.

1.1 RESEARCH QUESTION

Boyd (1994) argues that the CEO will try to maximise his compensation and his ability to do so depends on if he manages to dominate the BoD. In accordance with this reasoning he finds that CEO compensation is higher in firms with lower level of controls. Hill & Phan (1991) suggest that over time the CEO can circumvent governance mechanisms and is able to adjust his pay to his own preferences. This is consistent with the managerial power theory, which predicts that powerful CEOs are able to obtain a higher amount of compensation (Bebchuk & Fried, 2005). Morse et al. (2011) report a direct relation between CEO power and higher compensation. They reason that the CEO will use his influence over the board to adapt better-performing measures in his incentive contract and subsequently leaks firm value over time. Research thus suggests that weak governance allows the executive to gain power (Abernethy, Kuang, & Qin, 2013) and can lead to value-destroying pay practices (Mason, 2012; Core, Holthausen, & Larcker, 1999; Boyd, 1994; Morse, Nanda, & Seru, 2011). When the CEO has substantial power over the BoD, he is in a postition to obtain more remuneration. Since his reputation is also dependent on that of the firm, a powerful CEO could stimulate the directors to employ impression management. In addition, the fact that there is a powerful CEO present implies that the governance mechanisms in place are not effective (van Essen, Otten, & Carberry, 2015). Prior research shows that effective governance mechanisms are associated with less impression management (Osma & Guillamón-Saorín, 2011). To the extent effective governance is lacking, enables directors to employ impression management since ineffective governance mechanisms signifies less monitoring. As a result an opportunity for

self-5

interested behavior of board members is created. I will therefore examine the following research questions:

1. What is the influence of CEO excess pay on impression management in the remuneration report?

2. What is the influence of CEO power on impression management in the remuneration report?

1.2 RELEVANCE

This research contributes to the literature on the effectiveness of corporate governance and the use of impression management in several ways. First, I will determine if directors will engage in impression management when they present information about an excessive pay package. I provide policymakers and shareholders with an insight in how directors may try to manage impressions and influence the say on pay votes. This can create awareness for shareholders when reading the remuneration report in preparation of their decision to vote for or against the say on pay resolution, consistent with previous research that found that more easily readable disclosures increases investors’ reliance on the disclosure (Rennekamp, 2012) and their credibility assessment of the firm (Tan, Wang, & Zhou, 2015).

Second, by examining whether directors who award excess pay to the CEO will engage in impression management, a possible undesirable side-effect of say on pay can be explored. When the board of directors is able to predict shareholder voting behavior Krause et al. (2014) suggest that they can either manage the communication of the CEO compensation or adjust the CEO’s compensation package in order to prevent voting dissent. Say on pay thus provides an additional incentive for the BoD to employ impression management, since it is in their best interest to obtain an approving vote. This is obviously not in line with the desired effect of enhanced corporate governance. By addressing this, regulators will gain a better comprehension of how say on pay effects the governance positively as well as negatively. Since recommendations in corporate governance codes involve implementation costs for firms, it is of interest to determine how effective the say on pay resolution is as a governance mechanism (Mather & Ramsay, 2007).

Third, this study addresses the financial industry within U.K., which is a setting of increased public scrutiny of CEO pay. Especially the banking industry received lots of criticism by the media and politicians after the financial crisis (Chen, Zhang, & Xiao, 2011), as CEO s of banks bailed out by the government obtained extravagant bonuses in the past years. Laksmana et al. (2012) looked at the readability of the CD&A of companies listed on the American stock market and found that CD&As with CEO pay above the benchmark are more difficult to read. By focusing on financial firms in the U.K. I address the increased incentives of the directors to engage in impression management since a no-vote on the say on pay resolution has serious consequences for the reputation of the firm. Previous research often left the financial industry out of the sample due to the specific regulation or did not present it as a separate group, thereby overlooking an important group of firms in which empowering shareholders might be particularly important to curb excessive CEO pay (Conyon & Sadler, 2010; Ferri & Maber, 2013; Cai & Walkling, 2011).

Fourth, previous studies have addressed the CEO s incentives to use impression management (Laksmana et al., 2012) and the ability of corporate governance to limit self-serving behavior of the CEO. I combine these elements to provide an insight in how directors are managing shareholder impressions and how CEO power enhances the use of impression management.

6

To the best of my knowledge no paper has looked at the effect of CEO power on the use of impression management by directors. Two papers have looked at the constraining power of corporate governance mechanisms on impression management, where corporate governance can be seen as the opposite of CEO power. Osma & Guillamón-Saorín (2011) studied the association between corporate governance and impression management in annual results press releases of Spanish firms. They found that strong governance limits impression management and interpret the quality of the board of directors as a possible indicator of the quality of information disclosed by a company. Mather & Ramsay (2007) find that boards with more independent directors appear to be more effective in limiting opportunistic impression management by new CEO s in the period of their appointment. This research is the first to examine the influence of CEO power on impression management in the executive remuneration report. Therefore it offers a new insight on the use of impression management in the remuneration report and the reinforcing effect of CEO power on the use of impression management.

1.3 STRUCTURE

The remainder of this paper proceeds as follows. Section II will outline the theory resulting in the hypotheses. Section III describes how the key variables are measured and presents the research design of the tests. Section IV will present the findings of the statistics and test the hypotheses. Finally, in Section V the conclusions, limitations and avenues for further research will be discussed.

7

II. THEORY & HYPOTHESIS

2.1 AGENCY THEORY

The basic premise of Agency theory is that the agent (CEO) will not always act in the best interest of the principal (shareholders) without the appropriate incentives and monitoring costs (Jensen & Meckling, 1976; Fama, 1980). The CEO will act in self-interest and therefore tries to extract as much compensation possible. The board of directors represents the interest of the shareholders and acts as a key governance mechanism by monitoring the CEO and by setting pay levels accordingly (Fama & Jensen, 1983). Indeed, in support of this conjecture, research shows that poor governance is associated with greater agency problems and excessive compensation (Core et al., 1999). Although according to the agency logic directors are expected to serve in shareholders’ interest by negotiating an executive pay package that will encourage the CEO to act in shareholders’ interests, more recent scholars like Bebchuk & Fried (2005) argue that this might not always be the case. Specifically, directors have incentives to accept a compensation package that benefits the CEO but is not in the best interest of the shareholders (Bebchuk & Fried, 2005; van Essen et al., 2015). This reasoning is based on the managerial power theory which predicts that when CEO s have more power over the board of directors, they are able to extract higher pay (van Essen et al., 2015). These incentives of directors include self-interest5 _{as well as social and psychological factors. First,}

friendship & loyalty might prevent directors to block excessive pay arrangements. Also, due to prior social connections and because the CEO will often have been involved in recruiting the director. Collegiality too might play a role in this, since directors will probably treat the CEO as their respectful leader. Second, favoring the CEO only results in a small cost for the directors since they usually only own a small part of the firm’s shares. The direct costs of the devaluation of their shares probably will not outweigh the benefit of being on the good side of the CEO. In addition to the direct costs there is a threat of reputational costs for the director, since favoring the CEO does not fit in his description as an effective monitor (Hunton & Rose, 2008; Srinivasan, 2005; Dyck, Volchkova, & Zingales, 2006). This point will be addressed at a later stage. The third incentive of the directors is the desire to pay their CEO more than their equivalent in other firms, known as pay ratcheting (Leech, 2001). Arguments given to justify remuneration ratcheting include the incorporation of compensation consultants’ compensation advise, acting on the wish to become the number one in the industry by setting pay accordingly, attracting the best executives and setting pay as an international comparator in order to maintain the CEO (Tricker, 2003). This was perfectly illustrated by a non-executive director of the bailed-out Royal Bank of Scotland who also chairs the remuneration committee: “…we must do what it takes to attract and keep the people who will help us achieve our goals”6_{. The proposed bonus plan was subsequently blocked by}

the government.

Another reason why CEOs may be awarded a pay package that is not "at arm's length" involves reputational issues. Theory shows that reputation may affect directors decision making regarding CEO pay in two ways. The difference is their reputation in the eyes of the

5 _{Directors who side with the CEO might expect that they have a better chance at being re-elected, and therefore secure}

financial benefits. Moreover, when they support a pay package favorable to the CEO they might as well reasonably expect the CEO to support higher director compensation in return (Bebchuk & Fried, 2005)

6 _{http://www.theguardian.com/business/2014/apr/25/royal-bank-of-scotland-bonus-plan-blocked-by-government retrieved}

8

executives vis-à-vis their reputation in society. First, directors will want to please the CEO in order be re-elected and secure his options to join other boards. The directorship offers a certain prestige and substantial financial benefit, which will contribute to directors wanting to be re-elected. Although shareholders officially select individual directors, the board presents the potential directors to the shareholders. Since the CEO has considerable influence over the board, Bebchuk & Fried (2005) argue that directors will avoid to displease the CEO. Furthermore, being known as a director who obstructs the CEO from obtaining substantial pay packages could hurt one’s chances to join other boards. Tosi et al. (2000) point out that in firms where the CEO receives high compensation, the directors often do too. But second, directors have a reputation as effective monitors to protect. This is because their public reputation could just as well reduce their future career opportunities (Hunton & Rose, 2008; Srinivasan, 2005; Dyck et al., 2006). Bebchuk & Fried (2005) describe these economic and social costs as outrage costs, reflecting society’s judgment. Dyck et al. (2006) outline this outrage as negative media which questions if directors are capable at their job. Outside directors will want to protect their reputation as independent decision makers (Fama & Jensen, 1983; van Essen et al., 2015). Since supporting the CEO on his pay package increases the risk of reputational costs, directors have an incentive to ‘camouflage’ the compensation arrangements (Bebchuk & Fried, 2005).

2.2 SAY ON PAY

Increased public outcry of excessive CEO pay (Abernethy et al., 2013) led to the aspiration to enhance shareholders’ ability to influence corporate policy (Armstrong et al., 2013; Mason, 2012). This resulted in the introduction of say on pay as a governance mechanism in the United Kingdom. In 2002 it was at first promoted as a non-binding advisory vote. Despite the non-binding say on pay vote remuneration packages remained tremendous, especially in the eyes of the public at a time of financial crisis. Therefore, in 2013 the votes on the say on pay regulation became binding. This implies that when the Directors Remuneration Report (DRR) receives majority voting dissent, the pay package cannot be awarded to the CEO. The effectiveness of say on pay is widely discussed. Some argue that shareholder voting does not affect the CEO pay but rather the composition of the pay package (Mason et al., 2012) and that binding votes will lead to lower CEO pay (Balsam, Gordon, & Kwack, 2013). While others are convinced that it will lead to a decrease in pay and that only advisory votes will increase the pay performance sensitivity (Correa & Lel, 2013). While the evidence remains inconclusive about these issues, researchers largely agree that say on pay empowers the board’s bargaining position in their negotiations with the CEO about his compensation (Burns & Minnick, 2013; Bebchuk, 2007; van Essen et al., 2015). This could lead to the idea that as boards can negotiate more effectively, subsequently pay will decrease. This is consistent with the finding that the implementation of say on pay generates heightened market value, as lower pay is not in the best interest of shareholders (Cuñat et al., 2013). Ferri & Maber (2013) advocate that the threat of no-votes votes on the advisory say on pay resolution may affect CEO pay ex ante, as well as ex post. Which will arguably improve a firms’ corporate governance practices and shareholder value.

Say on pay improves the shareholder position and can result in more effective corporate governance but also leads to costs, as voting dissent requires engagement and response from the board (Krause et al., 2014). Advisory as well as binding votes have a substantial impact on a firm’s governance and the reputation of its directors (Ertimur, Ferri, & Stubben, 2010; Del Guercio, Seery, & Woidtke, 2008). Subsequently directors’ reputation in the labor market will be affected by the votes on the say on pay resolution and it can decrease their future career options when shareholders vote against the pay package the directors presented to them. The

9

power of the media and public judgment hits directors at their weakest spot (Del Guercio et al., 2008) as there is no insurance policy that saves a directors’ reputation (Ferri & Maber, 2013). Directors will thus take actions to avoid public humiliation and the associated damage to their reputation (Del Guercio et al., 2008). Researchers stress the importance of voting dissent (Kardis II et al., 2012) as it can have significant effects (Krause et al., 2014). Companies are urged to open a dialogue with shareholders in the extent of maintaining good shareholder relations (Kardis II et al., 2012). However, when directors present a compensation package favorable to the CEO , the expected voting dissent will impose costs on them and the company.

2.3 IMPRESSION MANAGEMENT

Shareholders will use their vote on the DRR to express their disapproval of excessive remuneration (Alissa, 2009). Research shows that shareholders mainly disapprove of high and excessive CEO rewards (Carter & Zamora, 2007; Kimmey, 2013; Krause et al., 2014; Conyon & Sadler, 2010; Armstrong et al., 2013; Kimbro & Xu, 2013), weak performance sensitivity (Carter & Zamora, 2007) and poor firm performance (Krause et al., 2014; Cotter, Palmiter, & Thomas, 2012). As shareholders will vote against the DRR when CEO pay is excessive, directors may want to communicate compensation policies that are seen as more appropriate or influence shareholders’ opinion in a different manner to avert voting dissent (Conyon & Sadler, 2010; Bebchuk & Fried, 2005). For instance when the CEO uses his influence to shift the performance measures towards more stable and better performing measures (Morse et al., 2011). But there are other ways to influence the shareholders’ opinion, one of them being the use of impression management. Through the use of impression management, directors can attempt to prevent reputational damage. Previous research has associated impression management with executive remuneration. Impression management can be defined as intentionally seeking to influence stakeholders’ reaction (Elsbach et al., 1998; Goffman, 1959), or as an individual’s effort to maximize social approval (Schlenker & Weigold, 1992). Moreover, Gellerman (1986) cites several rationalities that can lead to misconduct, one rationality being that harmful conduct is actually safe when it will not be discovered. This reasoning could apply to the situation of the directors, supporting the pay package favorable to the CEO because they believe that with the use of impression management, their wrongdoing will not be discovered. Therefore I argue that remunerating an excessive pay package to the CEO will increase the incentives of the directors to cover it up, in order to maintain their reputation as expert monitors.

Impression management takes on several forms. One can influence another by simply formulating performance in a very positive tone or attribute detrimental events to external forces and take credit for favorable outcomes (Bettman & Weitz, 1983). Another way to manage investors' impressions involves the use of deceptive language, which Larcker & Zakolyukina (2012) characterize by more references to general knowledge, less modest positive emotions and fewer references to shareholder value. In this research I will focus on the following two forms of impression management. First, directors may try to overwhelm shareholders with information by releasing several announcements or reports simultaneously (Graffin et al., 2011). By doing so it requires more effort and it is more complicated to form conclusions about a single piece of information. Graffin et al. (2011) describe that strategic noise takes place when multiple important events are announced at the same time, when the firm controls the timing of the announcement and the announcements are not related to the initial event. Second, by reducing the readability of information, firms may try to obfuscate shareholders (Bloomfield, 2002). Obfuscation is described by Courtis (2004) as “a writing technique that obscures the intended message, or confuses, distracts or perplexes readers,

10

leaving them bewildered or muddled”. Long sentences, complex grammatical structures and confusing vocabulary can discourage or prevent the reader from fully comprehending the information presented. Tan et al. (2015) find that less readable disclosures obfuscate investors when benchmark performance is inconsistent. In similar light, Laksmana et al. (2012) find that the average CD&A is very difficult to read and especially for those firms who pay their CEO above the benchmark compensation level. Rennekamp (2012) uncovers that more readable disclosure prompt more extreme reactions. Following this reasoning, directors can distort shareholders opinion by providing less readable reports. Strategic noise and readability are tactics to alter shareholders perceptions and subsequently prevent voting dissent. It serves directors in an attempt to maintain their reputation and status.

As discussed, the board of directors has incentives to support an excessive pay package in favor of the CEO. This is not in the best interest of shareholders and they will reasonably vote against the DRR. Voting dissent will probably hurt the reputation of directors and their future career opportunities. The threat of no-votes hence offers directors an incentive to mislead the shareholders and engage in impression management. This leads to the following hypothesis:

H1 CEO excess pay is positively associated with impression management

2.4 CEO POWER

As mentioned earlier, corporate governance mechanisms can enhance the negotiating power of the board when bargaining with the CEO about executive remuneration (van Essen et al., 2015). However, this relation is twofold as power can be seen as a relationship and ineffective governance mechanisms thus offer an opportunity to the CEO to gain power (van Essen et al., 2015). The managerial power theory predicts that when a CEO has more power over the board, he will be in a better position to negotiate for more compensation and pay that is less sensitive to corporate performance (Bebchuk & Fried, 2005). A powerful CEO will thus use his influence to select and compensate directors, and subsequently exploit personal bonds with them (Boyd, 1994; Baysinger & Hoskisson, 1990). Bebchuk & Fried (2005) mention the existence of golden parachutes evidence of the managerial power theory, as it is common to assign golden parachutes even to directors being fired that have performed extremely poor (Chen et al., 2011). Other scholars found that powerful CEO s achieve higher compensation, supporting the managerial power theory (Morse et al., 2011; Adams, Almeida, & Ferreira, 2005). Morse et al. (2011) found a direct relation between CEO power and higher pay as well as that a powerful CEO can use his position to direct his performance assessment towards better performing measures. Tosi et al. (2000) explain the decoupling of CEO pay from performance by reasoning that executives are risk averse, and thus prefer to link their compensation towards a more stable factor such as firm size. Research to date thus provides evidence for the inference that a powerful CEO can secure more compensation that is less sensitive to performance.

Several factors contribute the CEOs’ ability to achieve power over the board. His ability to do so is highly related to the extent and effectiveness of corporate governance mechanisms, as CEO s are better able to circumvent board monitoring when their influence increases (Boyd, 1994). Various board and ownership characteristics contribute to CEO power:

CEO duality is often mentioned as indicator that strong executive power is present. After a period of poor performance dual CEO s are less likely to be replaced, showing a weakened monitoring role of the board of directors (Goyal & Park, 2001). CEO duality is therefore cited as cause for a decrease in board independence and effective monitoring (Mather & Ramsay,

11

2007). For that reason the Cadbury Committee Report recommends separation of the CEO and chairperson positions (Daily & Dalton, 1995). When a CEO is also the chairman of the board his power increases significantly (Boyd, 1994) due to several reasons. First, since the CEO -chairman organizes board meetings and is responsible for preparing the board meetings’ agenda, he controls the information supply to the BoD (van Essen et al., 2015; Pearce II & Zahra, 1991). Second, the dual role of the chairman and CEO can be regarded as the highest position in an organizations’ hierarchy and therefore the CEO gains more power (Boyd, 1994; Zajac & Westphal, 1995). Prior research shows that CEO duality is associated with more earnings management (Dechow, Sloan, & Sweeney, 1996), lower pay-performance sensitivity (Goyal & Park, 2001) and higher total pay (van Essen et al., 2015; Cyert, Kang, & Kumar, 2002). Implying that the increased mandate and power can lead to less effective corporate governance and more influence of the CEO in the pay setting process (Goyal & Park, 2001). Third, CEO ’s influence in the nomination process of new directors increases when CEO duality exists (Westphal & Zajac, 1995). As the UK Corporate Governance Code states that a chief executive should not simultaneously be chairman of the same company, it is likely that there will not be many companies who combined these roles (Financial Reporting Council, 2014). However, it does occur and will therefore be included as component of CEO power.

The tenure of the CEO is also seen as an important factor in the process of establishing a power position (Goyal & Park, 2001). Directors may feel more inclined to follow the CEO s judgment when the CEO has served the firm for a longer time because he has more experience and status within the organization (van Essen et al., 2015; Bebchuk & Fried, 2003). Moreover, directors who have been appointed after the CEO took seat may feel a certain degree of gratitude and loyalty towards the CEO , preventing them from fighting the CEO about executive remuneration (Bebchuk & Fried, 2003). Shivdasani & Yermack (1997) argue that directors who have been appointed by the current CEO are less likely to monitor him. Consistent with this argumentation Core et al. (1999) find that executive remuneration is higher when a larger amount of outside directors are appointed by the CEO. Moreover, Hill & Phan (1991) find that pay increases over the tenure of a CEO and is more closely tied to the preferences of the CEO.

The size of the board can be of influence in the CEOs’ ability to gain power. Researchers found that large boards are associated with higher executive compensation (van Essen et al., 2015; Core et al., 1999). They reason that it takes more time and effort to align all directors in large boards (Mather & Ramsay, 2007; Core et al.,1999) and that therefore they can be more easily dominated by the CEO as compared to smaller boards (van Essen et al., 2015; Zahra & Pearce, 1989; Jensen, 1993). Dowell, Shackell, & Stuart (2011) argue that small boards are more competent to monitor management because the individuals on the board have a greater sense of responsibility and they make faster decisions. In this consideration it is argued that small boards are more likely to remove poor performing executives (Certo, Daily, & Dalton, 2001) and bargain CEO pay with a greater pay-performance sensitivity (Boyd, 1994).

Another factor contributing to CEO power relates to a lack of board independence. Independent directors are assumed to be better able to monitor management than dependent directors. As directors who never worked at the company nor had family or business relations with the firm, are less bound to the CEO (Bednar, 2012). The existence of independent directors is a key corporate governance mechanism and at least half of the board is required to be independent in order to comply with the UK Corporate Governance Code (Financial Reporting Council, 2014). For these reasons the ratio of independent directors to the total

12

number of directors is taken as proxy for board independence (Boyd, 1994; Certo et al., 2001; Bednar, 2012; Osma & Guillamón-Saorín, 2011). The lack of independent directors can thus increase the power of the CEO.

The above mentioned board characteristics have an influence on the CEOs’ ability to gain power and influence in setting his compensation. Conform the managerial power perspective, previous research has established that CEO power is positively related to executive remuneration (van Essen et al., 2015; Boyd, 1994; Morse et al., 2011; Adams et al., 2005). I posit that CEO power will amplify the positive association between excessive pay and impression management due to several reasons. First, excessive remuneration will probably damage the CEO ’s reputation. Through voting dissent or simply publicity of the disproportionate remuneration, the firms’ and CEOs’ reputation can be damaged. This is illustrated by the Guardian as they headline with “furious investors attack WPP and RBS over excessive bonuses and pay”7_{. The CEO could therefore encourage the directors to use}

impression management in the DDR in order to avert reputation damage. Second, research implies that strong governance limits impression management employed by management (Osma & Guillamón-Saorín, 2011). As corporate governance and CEO power are interrelated (van Essen et al., 2015), CEO power will offer an opportunity to employ impression management as there is less monitoring on the actions of the board. The main role of the board of directors is to monitor management, but as they are the ones covering up their own misconduct, monitoring is in fact minimal. The above mentioned board characteristics thus both decrease monitoring and empower the CEO. Therefore, I expect CEO power to positively moderate the association between CEO excess pay and impression management.

H2 The positive effect of CEO excess pay on impression management is positively moderated by CEO power

13

III. METHOD

3.1 SAMPLE

This research will focus on financial institutions in the UK that are listed on the London Stock Exchange. The UK especially presents an interesting sample to investigate executive remuneration and impression management because the say on pay regulation has been into effect since 2002. The presence of the regulation could provide directors incentives to employ impression management in order to avoid voting dissent. The financial sector is of particular interest due to the public outcry over CEO pay. This tendency was already notable before the financial crisis, as politicians and the media raised the question to what extent executive remuneration is appropriate (Tosi & Gomez-Mejia, 1989). The financial crisis further increased the astonishment of the enormous remuneration paid to the CEO as the media headlines with “lives are being trashed by klepto-remuneration”8_{. This paper includes the}

years 2002 till 2014, starting with the year say on pay was introduced in the UK. The data is collected from several sources. Data on executive remuneration, board structure stems from BoardEx9_{. Capital market and performance data is obtained from Datastream. The sample}

exists of 602 firm-year observations from 77 firms. After eliminating firm-year observations that contained missing values the remaining observations accounted for 500 firm-year combinations. Continuous variables have been winsorized at three times the standard deviation to limit the influence of outliers (Field, 2009).

3.2 DEPENDENT VARIABLE

In this research two measures of impression management will be used to determine if the directors try to influence shareholders’ reactions. For all of the firm-year samples the remuneration report was extracted from the firm’s annual report. The tables, graphs and tabulated texts in the remuneration report were excluded from the text. First, to determine the readability or obfuscation I use the Fog index (FOG). The Fog index determines the number of years an average reader would need as formal education to understand the text (Laksmana, et al., 2012). A higher Fog score thus constitutes a less readable text. The index takes the number of complex words10 _{and the amount of words per sentence into account. It is}

calculated as follows:

Second, strategic noise (STRATEGIC_NOISE) will be used as proxy for impression management to determine if firms simultaneously release several announcements in the period when the DRR is released. I use the database LexisNexus to look at the released announcements of the London Stock Exchange. Data within LexisNexus will be obtained from three sources: London Stock Exchange Aggregated Regulatory News Service, London Stock Exchange Corporate Actions Service and London Stock Exchange – Board Memberships. These are all available sources in LexisNexus to obtain announcements of the London Stock Exchange. With a fixed search query all strategic press releases of

8 _{http://www.theguardian.com/commentisfree/2015/mar/31/wealth-creators-klepto-rewards-bosses retrieved April 21, 2015} 9 _{BoardEx is an independent research company that gathers detailed information on company boards, corporate governance,}

remuneration and directors. Information is collected from public sources. Reliability of the information has been confirmed by prior studies (Ferri & Maber, 2013; Mason, 2012)

14

confounding events of the firms in the sample are examined. The search query can be viewed in figure 1. A confounding event can be described as a significant event that can autonomously influence a firm’s stock price (Graffin et al., 2011). The objective of impression management in the DRR, as defined in this research, is to influence shareholders impressions to obtain an approving vote on the say on pay resolution. When strategic noise is used as mean to influence the shareholders, the release of other strategic announcements will thus be prior to the say on pay voting. Therefore the period from the start of the new fiscal year till the date of the annual general meeting will be taken as window for the use of strategic noise.

3.3 INDEPENDENT VARIABLES

The independent variables consist of CEO excess pay and the moderating variable CEO power. Their influence on impression management will be researched.

CEO excess pay

To determine the excessive component of the executive remuneration I will follow the method of Core & Guay (1999). They define excess pay as actual compensation minus the expected compensation (Core, Guay, & Larcker, 2008). The expected compensation will be determined by a regression of total executive pay and several determinants/factors contributing to executive pay. The following factors will be included in the regression: natural logarithm of total assets, earnings per share, return on assets, market-to-book ratio, natural logarithm of the age of the CEO , natural logarithm of the tenure of the CEO , percentage of dependent directors on the board and natural logarithm of the number of boardmembers. The regression will result in a error term. A positive term signifies an overpaid CEO and a negative term signifies that the CEO is underpaid. To account for both cases, two variables will be included. When the error term < 0, the CEO is underpaid (UNDER_PAY) and when the value is > 0 it will take a value of zero. In the opposite case, when the error term > 0, the CEO is overpaid (EXCESS_PAY) and when the value is < 0 it will take a value of zero. Cases where the error term is equal to zero means the economic determinants correctly predicted the pay level of the CEO. I will threat those cases as if there is no excess pay present.

CEO power

In this research CEO power is constructed of several elements, I will discuss them all. First, the dummy variable CEO duality is measured by the presence of a CEO who is also chairman. The CEO will then be able to set the agenda and decide what to discuss, which greatly strengthens his power. In those cases CEO duality (CEO_DUALITY) will take the value of 1, where a CEO is not simultaneously chairman the value will be 0 (Adams et al., 2005). Second, the tenure of the CEO is also expected to influence the ability of the CEO to gain

HEADLINE(enterprise name) AND HEADLINE(dividend OR repurchase OR sale OR restructure OR resignation OR resign OR acquisition OR(capital w/1 reorganisation) OR(compulsory w/4 shares) OR (conversion w/2 securities) OR(directorate w/1 change) OR disposal OR(dividend w/1 declaration) OR (exchange w/1 suspension) OR(Exchange w/1 restoration) OR(issue w/2 debt) OR(issue w/2 equity) OR(intention w/2 float) OR(merger w/1 update) OR(offer w/1 document) OR(notice w/2 merger) OR(product w/1 launch) OR(publication w/2 prospectus) OR(re w/1 agreement) OR(re w/1 alliance) OR(re w/1 contract) OR(joint w/1 venture) OR(research w/1 update) OR(restructure w/1 proposal) OR(tender w/1 offer) OR(possible w/1 offer) OR(statement w/2 suspension) OR(tender w/1 offer) OR(own w/1 shares) OR(treasury w/1 stock))

15

power. New directors could feel a certain degree of loyalty to the sitting CEO for being appointed to the board. The tenure of the CEO will be measured by the number of years an individual is the CEO of the company (CEO_TENURE). Third, prior research suggests that larger boards are less efficient as it takes more time to reach a decision due to dispersed opinions. This could enable the CEO to gain more power. Therefore board size will be measured by the number of directors serving on the board (BOARD_SIZE). Last, board independence (BOARD_INDEPENDENCE) will be measured by the ratio of independent directors to the total number of directors (van Essen et al., 2015). I will combine the several measures of CEO power into an aggregate CEO power index (CEO_POWER) (Adams et al., 2005). This will be accomplished by calculating the z-score of all the individual variables and subsequently aggregate the four variables into one measure for CEO power.

3.4 CONTROL VARIABLES

Following prior research on impression management (i.e. Laksmana et al., 2012; Merkl-Davies & Brennan, 2007) several firm characteristics will be taken into account as control variables. In addition, ownership concentration will also be used as control variable in order to account for outside monitoring. To control for industry- and year-fixed effects, industry and year dummies are included in the regression.

Firm characteristics

First, firm size (SIZE) will be included as natural logarithm of total assets. Larger firms are more sensitive to the media and are therefore more likely to encouter political costs (Watts & Zimmerman, 1986). Because of the threat of political costs they are more likely to submit a more readable DRR than smaller firms (Laksmana, Tietz, & Yang, 2012). This is consistent with Baker III & Kare (1992) who find that president’s letters of larger firms are more readable. However, because larger firms are more complex they could also have a more complex compensation plan. This is in line with research of Merkl-Davies (2007) who concludes that the chairman’s report of larger firms is more difficult to read. The possible effect of firm size can therefore not yet be predicted.

Second, the market-to-book ratio (MTB) is a proxy for the potential growth and investment opportunities of a firm (Laksmana et al., 2012). Firms that are expanding may have a more complex environment and therefore also more complex compensation plans. The market-to-book ratio is measured by the market value of the ordinary equity at year end divided by the equity value on the balance sheet at year end.

Third, the return on equity (ROE) measured as net income divided by the average of last and current year’s equity will be included as profitability measure. Firms with poor performance are more likely to conceal their performance than firms with good performance. Subramanian, Insley & Blackwell (1993) find that that annual reports of profitable firms are more readable (Li, 2008). Consistent with research of Courtis (2004), who finds that bad news is associated with a lower reading ease. However, other researchers find no significant difference in readability or conclude that there is no association (Courtis, 1995; Clathworthy & Jones, 2001).

Ownership concentration

Concentration of ownership is likely to provoke close monitoring as large owners have the resources and incentives to do so (Shleifer & Vishny, 1986; van Essen et al., 2015; Cyert et al., 2002). An increase in monitoring will affect the CEOs’ power and reduce his influence over CEO compensation (Bebchuk & Fried, 2003; Hill & Phan, 1991). Several researchers

16

have elaborated on this reasoning. Cyert et al., (2002) report that an increase in the equity ownership of the largest shareholder can lead to a lower amount of executive pay. Benz, Kucher, & Stutzer (2002) find that shareholder concentration is negatively associated with the number of options granted to top management. Concentrated ownership is also expected to increase the pay-performance sensitivity, as owners are believed to encourage value-enhancing decisions (Shleifer & Vishny, 1986; van Essen et al., 2015). Furthermore, when firms have external shareholders with a smaller than 5 percent holding, executives receive more luck-based pay due to profits caused by external factors (Bertrand & Mullainathan, 2000). Showing that when shareholders are too dispersed they have limited reason to monitor the CEO since the monitoring costs will likely exceed the benefits (Tosi, Werner, Katz, & Gomez-Mejia, 2000; van Essen et al., 2015), creating an opportunity for the CEO to gain influence. Although CEO power is generally conceived to be relative to internal corporate governance mechaniusms, ownership concentration (external corporate governance) could also have a negative effect on CEO power (Abernethy et al., 2013). Concentration of ownership (MAJOR_SHARE) will take the value of 0 when there is no shareholder present with a > 5% shareholding, the value of 1 when there are owners with a ≥ 5% shareholding but smaller as 10%, and the value of 2 when there are owners with a shareholding ≥ 10%.

3.5 REGRESSION MODEL

Both hypotheses will be tested on two forms of impression management; readability (FOG) and strategic noise (STRATEGIC_NOISE). To determine if CEO excess pay is positively associated with impression management I will perform a linear regression. Subsequently I will test if CEO power positively moderates the positive association between CEO excess pay and impression management. The formula for this regression is as follows:

17

IV. RESULTS

To determine whether CEO excess pay is positively associated with impression management a statistical analysis will be performed. Before executing the analysis all continuous variables have been winsorized at three times the standard deviation and the natural logarithm of the values have been calculated in order to reduce the influence of outliers. Due to the date of collecting the data and the availability of annual reports over 2014 and 2013, the sample collected from these years was quite small. There were 10 observations of 2014 and 30 of 2013. Moreover, the say on pay regulation changed as of 2013 when the say on pay vote became binding in the UK. For these reasons I combined the years 2013 and 2014 and labelled them all as if these observations were made in 2013, resulting in 40 observations of the year 2013.

4.1 DESCRIPTIVE STATISTICS

Table 1 presents the descriptive statistics for the total sample of 500 firm-year combinations. Consistent with prior evidence, the average remuneration reports is very difficult to read (Laksmana, Tietz, & Yang, 2012). The FOG index shows an average score of 17,44; implying that the average reader needs 17 years of formal education to understand a remuneration report. This is equivalent to the reading level of a post-graduate, which accounts for only 11% of the working population aged 26 to 60 in the UK (Lindley & Machin, 2013). Looking at the descriptive statistics of STRATGIC_NOISE you can see that the average amount of strategic announcements is almost 3 (2,91). There are firms that did not employ strategic noise (there is minimum of 0) and firms that did excessively, as the maximum value is 69 announcements. Dummy variables were made for EXCESS_PAY and CEO_POWER in order to correctly present the descriptive statistics of these variables. EXCESS_PAY was labelled with a 0 when the CEO was underpaid and labelled with a 1 in those cases that the CEO was overpaid. Whether the CEO was under or overpaid was determined by the residual of a regression of total annual compensation and several economic determinants contributing to executive compensation. Table 1 shows that 47% of the CEO’s in the sample are overpaid. The dummy variable CEO_POWER for the descriptive statistics was labelled with a value of 1 when a powerful CEO was present, and labelled with a 0 when there was no powerful CEO. Table 1 shows that in 49% of the observations a strong CEO was present. Looking at the control variables, the descriptive statistics show that in most firms of the sample there was a shareholder with a ≥ 5% holding (the mean has a value of 1,62).

Variable Mean Std. Dev. Min Max

FOG 17,44 1,49 12,16 21,34 STRATEGIC_NOISE 2,91 6,27 - 69,00 EXCESS_PAY 0,47 0,50 - 1,00 CEO_POWER 0,49 0,50 - 1,00 SIZE 14,63 2,33 7,72 20,54 MTB 2,44 4,32 -68,62 28,20 ROE 15,11 28,83 -273,16 97,18 MAJOR_SHARE 1,62 0,61 - 2,00 TABLE 1 Descriptive statistics

18 Correlations

Table 2 shows that some correlations among independent variables are significant. Excess pay is significantly correlated with the FOG index (r = 0.134; p < 0.01) as well as with strategic noise (r = 0.123; p < 0.01). This could be an indicator that the first hypotheses can be accepted, but a regression will be performed to make a definite conclusion. A check for multicollinearity shows that there is a correlation coefficient of 0,687 between the return on equity and the market-to-book ratio. This points out that multicollinearity between these variables potentially could be a problem as the value is close to the 0,7 limit. But looking at the more precise variance inflation factor there can be determined that there will be no problem with multicollinearity, as the VIF does not exceeds 2, and is thus well below the value of 10.

4.2 MAIN RESULTS

To determine if CEO excess pay is positively associated with impression management multiple regression analysis were performed. Both forms of impression management will be tested separately. All explaining variables have been standardized before the analysis with the interaction terms was carried out in order to make the interpretation of the main results easier. The results on the Fog index can be viewed in Table 3 and the results on strategic noise are presented in Table 4.

Test of control variables

As shown in model 1 of Table 3 and 4 firm size proved significant and has a positive effect on the readability of the remuneration report and on strategic noise. This is consistent with research of Merkl-Davies (2007) who found that the chairmans report of larger firms are more difficult to read. As predicted the return on equity rations proved to be negatively associated with the FOG index, as shown in Table 3. This is previously supported by Courtis (2004). Other control variables did not prove to be significant. Only 2% (R2 = 0.020, p < .01) of the variation in readability can be explained by the control variables and 5,7% (R2 = 0.057, p < .001) of the variation in strategic noise can be explained by the control variables.

Test of Hypothesis 1

Hypothesis 1 states that the CEO excess pay is positively associated with impression management. The impact of CEO excess pay on the FOG readability index will be first discussed and subsequently the effects on strategic noise. Table 3, model 2, presents the assessment of excess CEO pay on the FOG readability index. The control variables and

EXCESS_PAY in model 2 explain 3,7% (R2 = 0.037, p < .01) of the variation in the FOG

Variables 1. 2. 3. 4. 5. 6. 7. 8. 1. FOG 1 2. STRATEGIC_NOISE ,032 1 3. EXCESS_PAY ,134** ,123** 1 4. CEO_POWER -,145** -,073 -,117** 1 5. SIZE ,128** ,253** -,030 -,064 1 6. MTB -,023 ,002 ,035 -,017 -,023 1 7. ROE -,084 ,027 ,029 ,057 ,015 ,687** 1 8. MAJOR_SHARE ,013 -,070 -,006 ,016 -,213** -,091* -,145** 1

** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)

TABLE 2 Correlation matrix

19

readability index. The relationship between excess pay and the FOG index is significantly positive at the .01 level (β = .351). These results are consistent with the results of Laksmana et al. (2012) who found that the Compensation Discussion and Analysis in a US sample is less readable when executives are overpaid. Looking at Table 4, model 2, the effects of excess pay on the use of strategic noise can be viewed. The variables in model 2 explain 7,3% (R2 = 0.073, p < .001) of the total variation in strategic noise. Model 2 shows a significantly positive association between excess pay and the use of strategic noise (β = 1.374, p < .01). Therefore Hypothesis 1 is supported, as excess pay is positively associated with both forms of impression management.

Test of Hypothesis 2

Hypothesis 2 tests whether the positive association of excess pay on impression management is positively moderated by CEO power. The effects of the intercept CEO power will be first discussed for the FOG index and subsequently for strategic noise. Table 3, model 3, shows that all the independent variables explain 4,7% (R2 = 0.047, p < .001) of the variance in the FOG index. As predicted, the intercept CEO power positive influences the FOG index (β = .111, p > .1). However the association is not significant and therefore does not contribute to Hypothesis 2. Model 3 in Table 3 also includes CEO power as separate variable, showing that CEO power negatively influences the FOG index at the .05 level (β = -.361, p < .05). This is not in line with the expectations nor indirectly supports the managerial power theory. Table 4, model 3, shows the effects of the intercept on strategic noise. The control variables, excess pay, CEO power and intercept explain 7,1% (R2 = 0.071, p < .001) of the variation in strategic noise. The coefficient EXCESS_PAY x CEO_POWER negatively influences the use of strategic noise but is not significant (β = -.476, p > .1). The influence of CEO power on strategic noise is also not significant (β = -.415, p > .1). Hypothesis 2 will have to be rejected as the intercept CEO power has no significant influence on the association between excess pay and both proxy’s for impression management.

Variable INTERCEPT 16,071 *** 17,307 *** 17,322 *** SIZE ,088 ** ,213 ** ,196 ** MTB ,025 ,103 ,084 ROE -,007 * -,198 * -,173 MAJOR SHARE ,072 ,046 ,045 EXCESS_PAY (H1) ,351 ** ,323 ** CEO_POWER -,361 * EXCESS_PAY x CEO_POWER (H2) ,111 Adjusted R-squared ,020 ,037 ,047 F-statistics 3,481 ** 4,869 *** 4,485 *** Highest VIF 1,919 1,919 1,944

Year effects YES YES YES

Industry effects YES YES YES

Model 1 TABLE 3

Results of Regression Analysis for FOG

Model 2 Model 3

* Correlation is significant at the 0.05 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed) *** Correlation is significant at the 0.001 level (2-tailed)

20

4.3 ADDITIONAL ANALYSES

To determine the robustness of the results, several additional analysis will be performed. Other measures for readability, strategic noise and excess pay will be included to test of the previously made conclusions will hold.

Readability

Consistent with Laksmana et al. (2012) I will look at different readability measures to determine if the conclusions of paragraph 4.2 remains the same. A regression analysis has been performed using the SMOG readability index and the Flesch reading ease index. The SMOG index is used to describe how many years of education an individual needs in order to understand the text. The results of the regression on the SMOG index have been displayed in Table 5. From the results of the additional analysis can derived that the conclusions are re-affirmed, Hypothesis 1 is still supported and even more significant (β = .284, p < .001) and Hypothesis 2 is not supported (β = .131, p > .1). The effect of CEO power on the SMOG index is again negatively significant (β = -.274, p < .05). The Flesch Reading Ease index varies from 0 till 100, where a lower score equals lower readability. An index score between 0 and 30 equals the reading level of a university graduate. The results of the regression with the Flesch Reading Ease index are presented in Table 6. The coefficient with respect to excess pay is negative. This is due to the scaling of the Flesch Reading Ease index, as a lower value equals a lower readability whereas for the Fog index or strategic noise a higher index equals a lower readability. Hence, the conclusions remains the same as the regressions with Fog and Strategic noise and Hypothesis 1 is still supported (β = -1.334, p <.01) and Hypothesis 2 is again rejected (β = -.143, p > .1). The effect of CEO power on the Flesch reading index is again significant (β = 1.222, p < .05). Variable INTERCEPT -6,75 ** 2,384 *** 2,398 *** SIZE 0,671 *** 1,589 *** 1,59 *** MTB -0,022 -0,119 -0,144 ROE 0,007 0,193 0,222 MAJOR SHARE -0,135 -0,075 -0,059 EXCESS_PAY (H1) 1,374 ** 1,291 ** CEO_POWER -0,415 EXCESS_PAY x CEO_POWER (H2) -0,476 Adjusted R-squared 0,057 0,073 0,071 F-statistics 8,574 *** 8,803 *** 6,458 *** Highest VIF 1,919 1,919 1,944

Year effects YES YES YES

Industry effects YES YES YES

* Correlation is significant at the 0.05 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed) *** Correlation is significant at the 0.001 level (2-tailed)

Model 1 Model 2 TABLE 4

Results of Regression Analysis for Strategic Noise

21 Strategic noise

Previous research of Graffin et al. (2012) measures the use of impression management by the number of confounding announcements within the plus and minus one day period of the researched event. This three day window is considerably smaller compared to the researched period of 4 to 5 months in this research (from the publication date of the annual report till the date of the general meeting). It could be that firms deliberately release strategic

Variables INTERCEPT 14,979 *** 16,028 *** 16,039 *** SIZE ,075 *** ,180 *** ,166 ** MTB ,023 ,094 ,079 ROE -,006 * -,161 * -,142 * MAJOR SHARE ,054 ,035 ,032 EXCESS_PAY (H1) ,284 *** ,267 ** CEO_POWER -,274 * EXCESS_PAY x CEO_POWER (H2) ,131 Adjusted R-squared 0,027 ,050 ,059 F-statistics 4,509 ** 6,200 *** 5,468 *** Highest VIF 1,919 1,919 1,944

Year effects YES YES YES

Industry effects YES YES YES

Variables INTERCEPT 35,959*** 32,496*** 32,448*** SIZE -,257* -,624** -,575* MTB -,114 -,473 -,405 ROE ,025* ,730* ,644* MAJOR SHARE -,189 -,124 -,126 EXCESS_PAY (H1) -1,334** -1,221** CEO_POWER 1,222 * EXCESS_PAY x CEO_POWER (H2) -,143 Adjusted R-squared 0,014 ,035 ,044 F-statistics 2,769 * 4,619 *** 4,304 *** Highest VIF 1,919 1,919 1,944

Year effects YES YES YES

Industry effects YES YES YES

* Correlation is significant at the 0.05 level (2-tailed)

TABLE 5

TABLE 6

Results of Regression Analysis for FleschRR

Model 1 Model 2 Model 3 Model 2

Model 1 Model 3

*** Correlation is significant at the 0.001 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)

Results of Regression Analysis for SMOG

*** Correlation is significant at the 0.001 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed)

22

announcements in the days prior to the annual meeting to manage the impressions of the shareholders. To see if the results will persist, a window of 7 days will be used for an additional regression. The results of the regression analysis are shown in Table 7. Model 2 shows Hypothesis 1 is significant at the .01 level (β = .173). Model 3 reveals that Hypothesis 2 can be rejected (β = -.185, p < .1), consistent with the prior regression results.

Excess pay

Another way of measuring if a CEO is excessively paid is by calculating the average compensation by sector and to determine if the CEO is paid above or below the average. The sample consists of three sectors; banks, insurance and other finance. The variable

EXCESS_PAY will take the ln value of the excessive amount of compensation, compared to

the average sector wage and zero in those cases where the CEO is underpaid, compared to the average sector wage. The results of the effect of excess pay on readability remains positively significant at the .01 level for the FOG index as well as for strategic noise, as can be viewed in Table 8 (β = .206), respectively Table 9 (β = 1.018). The results for the second hypothesis remain the same, both tables show that the effect of CEO power on the positive association between excess pay and strategic noise is not significant (β = -.276, p > .1 and respectively β = .283, p > .1). The separate variable CEO power also has no effect on both readability measures, in contrast with previous regressions. This implies that no definite conclusion can be made of the effect of CEO power on readability. All additional test however do support the evidence from paragraph 4.2, Hypothesis 1 can therefore be accepted.

Variables INTERCEPT -,609 * ,222 *** ,220 *** SIZE ,054 ** ,128 ** ,135 *** MTB -,014 -,064 -,063 ROE ,003 ,090 ,088 MAJOR SHARE ,061 ,038 ,044 EXCESS_PAY (H1) ,173 ** ,161 * CEO_POWER ,032 EXCESS_PAY x CEO_POWER (H2) -0,185 Adjusted R-squared 0,021 ,035 ,035 F-statistics 3,663 ** 4,593 *** 3,548 ** Highest VIF 1,919 1,919 1,944

Year effects YES YES YES

Industry effects YES YES YES

*** Correlation is significant at the 0.001 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)

TABLE 7

Results of Regression Analysis for Short Window

23 Variables INTERCEPT 16,071 *** 17,441 *** 17,469 *** SIZE ,088 ** ,129 ,130 MTB ,025 ,115 ,122 ROE -,007 * -,212 * -,239 * MAJOR SHARE ,072 ,045 ,056 EXCESS_PAY (H1) ,206 ** ,224 ** CEO_POWER ,031 EXCESS_PAY x CEO_POWER (H2) -,276 Adjusted R-squared ,020 ,034 ,037 F-statistics 3,481 ** 4,509 *** 3,711 ** Highest VIF 1,919 1,927 1,996

Year effects YES YES YES

Industry effects YES YES YES

Variables INTERCEPT -6,750 ** 2,908 *** 2,880 *** SIZE 0,671 *** 1,184 *** 1,322 *** MTB -,022 -,067 -,100 ROE ,007 ,120 ,201 MAJOR SHARE -,135 -,080 -,087 EXCESS_PAY (H1) 1,018 ** 1,004 ** CEO_POWER -,625 EXCESS_PAY x CEO_POWER (H2) ,283 Adjusted R-squared ,057 ,078 ,077 F-statistics 8,574 *** 9,457 *** 6,945 *** Highest VIF 1,919 1,899 1,996

Year effects YES YES YES

Industry effects YES YES YES

TABLE 8

Results of Regression Analysis for industry excess pay on FOG Model 1 Model 2 Model 3

** Correlation is significant at the 0.01 level (2-tailed) *** Correlation is significant at the 0.001 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)

TABLE 9

Results of Regression Analysis for industry excess pay on strategic noise Model 1 Model 2 Model 3

*** Correlation is significant at the 0.001 level (2-tailed) * Correlation is significant at the 0.05 level (2-tailed)