Boards protecting themselves

(1)

Boards protecting themselves

The use of strategic noise to influence say-on-pay voting in the UK

Name: Jordy Tijhuis Student number: S3812340

University of Groningen Economics and business MSc A&C (Controlling)

Supervisor: prof. dr. R.B.H. Hooghiemstra Co-assessor: dr. S. Wang

Date: 19-06-2020 Word count: 12,090 words

(2)

Abstract

Starting from 2002, the UK introduced say-on-pay legislation that enabled shareholders to cast a non-binding vote on the firm’s compensation report, thereby indicating their (dis)agreement with the proposed CEO remuneration. The say-on-pay legislation concerns UK quoted companies and was changed to binding in 2013, indicating that firms have to act upon high levels of dissent opposed to ignoring it. As firms are aiming at the lowest shareholder revolts possible, they are actively seeking for ways to reduce the levels of dissent. One of these ways is through the introduction of strategic noise, which entails confounding publications published in the period surrounding the voting event. This thesis examines the relationship between the levels of shareholder dissent and excessive CEO compensation, with the possible moderating effect of strategic noise. The sample contains 711 observations of firms active in the UK financial industry in the years 2001-2019. The results show that there is a positive and significant relationship between shareholder dissent and excessive CEO compensation, this suggests that higher shareholder dissent is found at firms that compensate their CEOs more excessively. The results also show that there is no significant relationship between shareholder dissent and strategic noise. Additionally, the moderating effect of strategic noise on the relationship between excessive compensation and dissent could not be proven. The results are robust for log-transforming the dependent variable, a longer timeframe for strategic noise, introducing the resolution order as control variable, and including dummies for different firm-year combinations. It indicates that strategic noise does not influence shareholders within the UK financial industry.

Keywords: say-on-pay; shareholder dissent; excessive CEO compensation; strategic noise; impression management; financial industry; United Kingdom

(3)

3.1 Sample selection and data sources 12 3.2 Dependent variable 12 3.3 Independent variable 12 3.3 Moderating variable 13 3.4 Control variables 13 3.5 Empirical model 14 4. Results 15 4.1 Descriptive statistics 15 4.2 Testing hypotheses 18 4.3 Robustness analyses 20 4.3.1 Strategic noise timeframe 20

4.3.2 Resolution order 20

4.3.3 Transforming DISSENT 20

4.3.4 Timing effect of legislative changes 20

4.3.5 Change in data gathering method 21

4.4 Discussion 21 5. Conclusion 24

5.1 Conclusion 24 5.2 Limitations and future research 25 Literature 27

(4)

3

1. Introduction

Within the UK, there is a growing concern about rising CEO compensation, fuelling increased shareholder dissatisfaction. Data from remuneration reports of firms listed throughout Europe suggests that, from a compensation point of view, CEOs are best to accept executive positions in the United Kingdom (UK) (Thomas, 2019). This data shows that the average CEO of a FTSE 250 company earned £2.1 million in 2018, rising from £1.9 million in 2016. CEO pay packages of FTSE 100 companies even reached £4.7 million, but these pay packages remained relatively stable over the past years (Chartered Institute of Personnel and Development, 2019). As a result of the high compensations CEOs receive, the media regularly reports on excessive CEO compensation.

One of the most reported and criticized industries is the financial industry because it is seen as instigator of the financial crisis and because society has lost its trust in the industry ever since. Yet, the compensation CEOs of banks, insurance companies, and other financial firms receive, are among the highest in the UK corporate sector. Shareholders accompany the rising compensation packages with rising dissent, sometimes referred to as ‘shareholder spring’ or ‘shareholder revolution’. One of the severest shareholder revolts found, in the UK financial sector, was with insurance company Aviva in 2012. Shareholders responded with 59% of the votes casted against the proposed remuneration report (Dunkley, 2012). One of the largest British banks, Barclays, also faced a major backlash following a controversial pay package in 2012. Specifically, shareholders did not think a £17.7 million pay package was acceptable for the CEO of Barclays in the aftermath of the financial crisis and, therefore, responded with 32% of the votes against the remuneration report (Wilson, 2012).

Cases like Aviva, Barclays, and others illustrate the effects of providing shareholders with influence on corporate policy via so-called ‘say-on-pay’ legislation. The UK was one of the first countries to adopt say-on-pay legislation, which provided (starting from 2002) shareholders with the ability to cast a non-binding vote regarding the firm’s remuneration policy. The non-non-binding nature of the vote indicates that companies are not required to act upon voting outcomes, even if it involves high voting dissent (Alissa, 2015). In the first years after the adoption of say-on-pay, shareholders were quick to respond with high voting dissent (Ferri & Maber, 2013). After the introduction phase, shareholder dissent decreased to an average of 10% and continued to decline (Conyon & Sadler, 2010). As compensation packages continued to grow in the years following, it seemed that the non-binding vote did not have the anticipated effect of limiting excessive CEO compensation. Therefore, the UK government changed the vote to binding in 2013, to improve engagement and strengthen voting rights (Alissa, 2015; The Companies Regulation 2019 No. 970, 2019).

According to prior research, shareholders’ voting behaviour at the annual general meeting (AGM) is affected by factors such as ownership, board size, board independence, remuneration committees, price-book value, firm performance, current compensation, prior compensation, and compensation relative to peers (e.g., Alissa, 2015; Stathopoulos & Voulgaris, 2016; Cucari, 2019). Research suggests that shareholders critically assess company performance and compensation packages prior to voting, but it also reveals that shareholders often lack sufficient information to base their votes upon (Krause, Whitler, & Samadeni, 2014; Correa & Lel, 2016; Obermann & Velte, 2018; Cucari, 2019). Secondly, grasping the complexity of compensation packages is hard for unexperienced shareholders (Hooghiemstra, Kuang, & Qin, 2017). Following agency theory, shareholders experience information asymmetry and in order to reduce this, shareholders have to rely on both internal and external information. Companies themselves provide a lot of information through investor pages on their public websites to inform investors, media, and other stakeholders. As this is the most informed source, shareholders rely heavily on the information corporations publish. Agency theory describes that actors act of out self-interest and aim at achieving their personal goals instead of the overall company goals. Therefore, the information published can be biased (Jensen & Meckling, 1976; Eisenhardt, 1989).

(5)

4 Companies can choose to publish what, when, and how they see fit, thereby influencing the stakeholders’ knowledge base.

A high voting dissent can lead to negative publicity as shareholders are in overall disagreement with company policies and it is likely that the media, driven by ‘sensationalism’ (Core, Guay, & Larcker, 2008), will cover it. Obviously, this can be detrimental to directors’ and CEOs’ reputations, reputations they have built for years. As the position of individuals in the labour market for directors highly depends on having a sound reputation as experts in decision control, CEOs and directors have strong incentives to avoid bad publicity resulting from high voting dissent (Arthaud-Day, Certo, Dalton, & Dalton, 2006; Kang, 2016). Indeed, previous research indicates that boards will likely do anything to limit shareholder dissent (Ertimur, Ferri, & Maber, 2012; Kaplan, Samuels, & Cohen, 2014; Alissa, 2015). While shareholder dissent does not necessarily lead to lower total compensation, it can lead to a change in compensation structure as a way to influence shareholders’ voting behaviour in order to lower the overall dissent and prevent negative publicity (Conyon & Sadler, 2010; Ferri & Maber, 2013; Alissa, 2015; Correa & Lel, 2016).

As prior research shows, influencing shareholders to minimize dissent can be done through the information a firm publishes itself or, more in general, via impression management. For instance, Hooghiemstra et al. (2017), show that the readability of remuneration reports can influence the vote shareholders cast. In this study, I will build upon this research by considering another way as to how firms can try to influence shareholders’ perception, namely through strategic noise. As a concept, first introduced by Graffin, Carpenter, & Boivie (2011), strategic noise refers to a form of anticipatory impression management, where boards try to manage shareholders’ perceptions by influencing the information surrounding the event when resistance is expected (Elsbach, Sutton, & Principe, 1998; Graffin et al., 2011; Graffin, Haleblian, & Kiley, 2016). Graffin et al. (2011) have shown how boards turn to the use of strategic noise to obfuscate difficult CEO successions. In this thesis, I will focus on the role of strategic noise as means of boards to influence voting outcomes. Strategic noise can be used in the period surrounding the say-on-pay voting, by obfuscating the voting event with the publication of other important events and information. Examples of confounding events are events that can individually influence share prices and include: the announcement of mergers and acquisitions, earnings, or changes in dividend policy. By focusing on alternative events, shareholders are proposed to overlook the excessive CEO compensation (Kaplan et al., 2014).

The goal of this thesis is to find what is the effect of strategic noise on shareholders, to gain understanding to what extent shareholders can be influenced by the companies at stake. Therefore, I propose the following research question:

To what extent does strategic noise affect shareholders’ say-on-pay votes, in case of overpaid CEOs?

A sample of the UK financial industry, containing 711 observations from 83 firms in the period of 2001-2019, is investigated. Shareholder dissent is measured by dividing the votes against and abstained by the total casted votes at the AGM. Whether a CEO is overpaid is measured in conformance with Core, Holthausen, and Larcker (1999) by regressing an expected level of pay and comparing it to actual compensation. Strategic noise is measured by the number of confounding publications in the short period surrounding the AGM. Three analyses are performed to answer the research question. The first analysis examines the relationship between shareholder dissent and excessive compensation. The second analysis examines the relationship between shareholder dissent and strategic noise. The third analysis examines whether strategic noise can moderate the relationship between shareholder dissent and excessive compensation.

(6)

5 The results show that shareholder dissent is positively associated with excessive compensation. It indicates that excessive compensation leads to higher levels of dissent. Additionally, in contrary to the expectations, there is neither evidence that strategic noise influences dissent nor that it moderates the relationship between excessive compensation and dissent in the UK financial industry.

This thesis aims at making three contributions to the existing literature. First, it adds to the literature of say-on-pay as there is limited knowledge about how companies can influence shareholders’ voting behaviour. While previous research has focused on the effects of high compensation (Krause et al., 2014; Correa & Lel, 2016), company performance (Correa & Lel, 2016), proxy advisors (Ertimur, Ferri, & Oesch, 2013), and the media (Hooghiemstra, Kuang, & Qin, 2015), the company itself as determinant of voting behaviour is not assessed as extensively. As shareholders rely on company information and companies try to limit negative publicity, it is likely that these firms will try to influence voting behaviour. Hooghiemstra et al. (2017) did find readability of remuneration reports as a way companies could influence shareholders. While these reports are important, companies can already influence shareholders in earlier stages using strategic noise. It is important to gain more knowledge about the way shareholders are influenced, as the say-on-pay regulation was created to increase shareholder rights and power. Shareholders need to be aware of the steering behaviour towards the company’s goal, to critically assess what information sources are biased. As a result, researchers can assess whether there are other forms of influencing behaviour, to gain more understanding about how votes are determined. For practice it could lead to shareholders seeing strategic noise as “cheap talk” (Merkl-Davies & Brennan, 2011). By expanding on the influences that external factors have on shareholders and how this is reflected in the level of acceptance for remuneration reports, valuable knowledge will be added to the existing literature.

Second, it adds to the literature on strategic noise, as there is no application yet to say-on-pay. The overarching concept of impression management has been applied to topics as CEO compensation (Chng, Rodgers, Shih, & Song, 2015), discretionary disclosures (Merkl-Davies & Brennan, 2011), report readability (Hooghiemstra et al., 2017), and CEO succession (Graffin et al., 2011). Researchers found that impression management is mainly expected when boards and CEOs are concerned for their reputation and is used effectively in most cases, with the exception of sophisticated investors, whom seem to be able to see through attempts of managing behaviour (Merkl-Davies & Brennan, 2011; Chng et al., 2015; Hooghiemstra et al., 2017). In contrast, strategic noise has been used by Graffin et al. (2011) to study the effects on CEO succession. A lack of attention for the effects of strategic noise indicates a lack of knowledge as to how companies (can) use strategic noise to influence shareholders. This lack of knowledge can lead to researchers not considering its influences, while previous papers have indicated significant effects. Therefore, this thesis adds say-on-pay to the known relations of strategic noise and creates a path for future research to address new areas and broaden its known use. Third, it adds to the literature regarding say-on-pay by focusing on the financial industry. While the financial industry receives one of the highest compensations within the UK and scrutiny is high, on-pay literature on the UK financial industry is scarce. To the best of my knowledge, there is no say-on-pay research specifically focusing on the UK financial industry. Many studies did include the financial industry in their samples, but not in isolation (Ferri & Maber, 2013; Ertimur et al., 2013; Alissa, 2015). These firms were embedded in indexes and, therefore, the influence of the industry on the outcomes is not clear. Yet others, like Hooghiemstra et al. (2017) specifically exclude the financial industry because of regulations and different compensation structures. Say-on-pay literature in the United States did consider the financial industry (Crawford, Nelson, & Rountree, 2018), but regulations and culture differ to the UK, hence, the results do not generalize. Due to the high degree of criticism directed towards the financial industry, it is expected that this industry receives high voting dissents as well. Therefore, it would be interesting to see whether the financial industry responds with more strategic noise in the periods prior to the voting and adds to the literature by expanding the scope and application of research.

(7)

6

2. Theory & hypotheses development

2.1 Agency theory

Agency theory describes the contract between principals and agents. At its core, agency theory describes the problem of separation of ownership (principals) and control (agents). While both contracted parties have interest in the success of the company, their interests might not be aligned if they maximize utility (Jensen & Meckling, 1976). Whereas shareholders seek for value maximization, the CEO (as agent) might pursue personal goals as achieving bonuses or status. Shareholders do not know whether the goals of the CEO are the same as theirs (adverse selection) and whether the CEO practises what he preaches (moral hazard). The issues as described in agency theory are the result of information asymmetry. CEOs have superior information that is not shared with shareholders, while shareholders rely heavily on shared information. This introduces the risk that CEOs act out of self-interest instead of acting upon the shareholders’ best self-interests (Jensen & Meckling, 1976; Eisenhardt, 1989).

Solving problems resulting from the separation of ownership and control is considered difficult and involves agency costs. These are the costs of measures implemented to ensure goal alignment and information provision. Eisenhardt (1989) regards information as commodity that can be purchased and therefore, shareholders can possess all information if they can reach an agreement regarding the costs. Shareholders need to align interest of both parties by introducing monitoring and incentives (Jensen & Meckling, 1976; Eisenhardt, 1989). Monitoring can be through boards and committees, or by reporting on pre-defined measures to follow progress on goals. Literature on incentive alignment is focused on CEO compensation as alignment tool. CEO pay can be based upon performance measures that represent the shareholders’ goals and, therefore, incentivise CEOs to behave as induced by its shareholders (Jensen & Meckling, 1976; Eisenhardt, 1989).

2.2 CEO compensation

Within the corporate governance literature, there is much debate about CEO compensation. According to agency theory, CEO compensation is a method to reduce moral hazard and, therefore, compensation packages need to be carefully compiled based on set goals and actual performance (Eisenhardt, 1989). The packages themselves can have a large impact on the CEOs’ wealth and therefore CEOs will do whatever is needed to achieve and protect these bonuses (Buck, Bruce, Main, & Udeuni, 2003).

Prior research has explored what determines the amount of CEO compensation. This line of research shows that the level of CEO compensation depends on factors like: luck (Bertrand & Mullainathan, 2001), CEO power (Morse, Nanda, & Seru, 2011; Abernethy, Kuang, & Qin, 2015), peer companies included in the benchmark (Bizjak, Lemmon, & Naveen, 2008; Fong, Misangyi, & Tosi, 2010; Bizjak, Lemmon, & Nguyen, 2011), firm size (Tosi, Werner, Katz, & Gomez-Mejia, 2000; Gabaix & Landier, 2008), industry, and firm performance. For shareholders, one of the important factors is firm performance, as this can be used to align interests and incentivize CEOs based on their actions and decisions. While it seems that firm performance should be the most important determinant for CEO compensation, it is not always the case. According to different studies, bad firm performance does not necessarily lead to lower pay and changes in pay packages are just partially explained by (bad) firm performance (Tosi et al., 2000; Shaw & Zhang, 2010; Morse et al., 2011).

While the composition of pay packages seems quite different across countries, UK research reveals that compensation packages are mainly composed of a base salary and long-term incentive plan (LTIP). Based upon descriptive statistics of Gregg, Jewell, and Tonks (2012), concerning four major UK firms, pay packages are comprised of 28.2% salaries and pensions, 20.6% non-equity bonuses, and 51.2%

(8)

7 equity bonuses (e.g. options and LTIP). Additionally, pay packages of four major UK banks are comparable, with the difference that banks have higher short-term bonuses (Gregg et al., 2012). Despite the fact that scientific evidence shows that short-termism of the banking industry did not cause the financial crisis (Gregg et al., 2012), society still considers the current remunerations too high with the financial crisis as underlying reason (Buck et al., 2003; Fernande, Ferreira, Matos, & Murphy, 2009; Gregg et al., 2012).

2.3 Say-on-pay

Concerns as briefly discussed in the prior sub-section and the unexplainable rising compensation executives of firm received, led to rising public criticism of executive pay practices and a call for help to governments. In response, countries around the world have adopted say-on-pay-regulations. Simply put, ‘say-on-pay’ aims to give shareholders the right to vote about the compensation policies, in order to better align the objectives of CEOs and shareholders. Countries developed several different forms, varying from binding to non-binding and yearly votes to only voting when changes occurred.

One of the first countries to adopt say-on-pay, in 2002, was the UK. The goal for the UK was to strengthen relations between the boards and shareholders, and to make shareholders participate in the remuneration processes after years of complaining about excessive compensation structures (Alissa, 2015). In 2002 the UK started with a non-binding, advisory vote that shareholders of quoted companies were able to cast. The introduction of the say-on-pay legislation was thought to lead to: lower compensation, compensation based upon performance, less CEO influence in determining pay packages, higher shareholder satisfaction, and using CEO compensation more effectively as alignment tool (Alissa, 2015). In the years following the introduction, the UK government kept receiving signals of boards’ ignorance to shareholder dissent. CEO compensation continued to rise (Shue & Townsend, 2017; Chartered Institute of Personnel and Development, 2019), performance-based pay led to larger goal misalignment, and CEOs still have high power over their own pay (Abernethy et al., 2015). Shareholders were not taken seriously, therefore, the voting process was changed to binding in 2013 and has shown to be more successful. The UK implementation of the legislation requires the AGM to vote on the policies that have changed. If a remuneration policy did not change, shareholders at least have to vote once every three years about the existing policies (Alissa, 2015; The Companies Regulation 2019 No. 970, 2019).

For shareholders to achieve change, they can vote ‘against’ the proposed remuneration report and thereby voice their disagreement with the current policy. The threshold for failing the voting requires at least half of the casted votes against the proposition and entails continuation of the old policies, or compulsory changes to the new ones (The Companies Regulation 2019 No. 970, 2019). Experiencing a voting dissent of more than 50% is exceptional, as shown by, for instance, Gregory-Smith, Thompson, and Wright (2013) and Hooghiemstra et al. (2017) in studies regarding UK FTSE 350 companies. The highest average dissent was found in 2012, with maximum peaks of 15%, while the average dissent from 1998 to 2012 was around 5-10%. As experiencing a voting dissent of more than 50% is rare, the literature considers high shareholder dissent when more than 20% of the votes are against the proposed policy (Ferri & Maber, 2013). It was discovered that dissents of 20% and more, already lead to high policy change rates as boards seek to reduce its negative consequences (Del Guercio, Wallis, & Woidtke, 2008). It indicates that the current 6-8% average dissents are considered acceptable and it is expected that boards and CEOs will not address these issues.

Reasons for shareholders to vote against the propositions are generally limited to monetary and non-monetary factors. First, the non-monetary factors, these were endorsed by Alissa (2015) and entail compensation. It was found that shareholders’ dissent is higher when CEOs are excessively paid and it is assumed that investors are sophisticated enough to assess whether the proposed compensation is excess, based on: peers, performance, and the industry (Laksmana, Tietz, & Yang, 2012; Alissa, 2015;

(9)

8 Stathopoulos & Voulgaris, 2016). Balsam, Boone, Liu, and Yin (2016) confirm these findings and add that shareholders consider large increases, especially when not explained by economic factors, as negative. More specifically, high levels of ‘other compensation’ (private jets, club membership, etc.) lead to higher disapproval of the remuneration packages (Balsam et al., 2016). Second, there are non-monetary factors for shareholders to vote against the proposed remuneration report. Lower ownership dispersion (Judge, Gaur, & Muller-Kahle, 2010; Cucari, 2019), bigger boards (Del Guercio et al., 2008; Cucari, 2019), higher independence (Del Guercio et al., 2008; Cucari, 2019), more active committees (Cucari, 2019), and higher price-book values lead to lower levels of dissent (Cucari, 2019), as shareholders are more confident of boards and CEOs having the right intentions. Various authors have investigated the connection between poor performance and shareholder dissent, but the results were contradictory. While Alissa (2015) found no relationship, Ferri and Maber (2013) did find that there is a negative response and higher dissent to high CEO pay combined with poor performance. As Conyon and Sadler (2010) and Del Guercio et al. (2008) have found comparable results to Ferri and Maber (2013), firm performance is considered influential to the shareholders vote.

The fact that shareholders consider high CEO remuneration as determinant for their vote is important, especially as this was the initial cause for countries to adopt the legislation. According to the agency theory, compensation packages are carefully constructed to align interests between shareholders and CEOs. Boards within organizations create the proposed remuneration packages and, as a result of say-on-pay legislation, shareholders have to approve them. Criticism directed towards the basic ideas of agency theory indicates that compensation packages enhance the alignment issue as managers try to influence their own pay (Bebchuck & Fried, 2003; Morse et al., 2011; Abernethy et al., 2015). Research to the link between say-on-pay and CEO compensation has already shown that high total compensation is an indication that shareholders will vote against the proposed remuneration report (Conyon & Sadler, 2010; Alissa, 2015).

While total compensation gives an indication of overpayment (excessive compensation), shareholders do not necessarily derive overpayment from total compensation. When high compensation is combined with excellent firm performance, shareholders will not respond with high voting dissents (Ferri & Maber, 2013; Krause et al., 2014). Voting dissent will only be high when compensation is considered excessive, defined as the amount of CEO compensation that cannot be explained by performance, peers, board independence, and other economic or company variables (Core et al., 1999). Research conducted by Ferri and Maber (2013), and Krause et al. (2014) gained insight into the effects of excessive pay packages combined with performance. They describe that shareholders respond with higher levels of dissent when parts of the CEO compensation cannot be explained by economic factors, hence, is considered to be excessive. Similar results were described by Gregory-Smith et al. (2013), Krause et al. (2014), Alissa (2015), and Alkalbani, Cuomo, and Mallin (2019). In summary, shareholders have responded with higher dissent in the past in the case that CEOs received excessive compensation packages. While these results did not represent the financial industry, it would be expected that data from this industry confirms the prior findings. Therefore, I propose the following hypothesis:

Hypothesis 1. CEOs with excessive compensation packages are associated with higher levels of shareholder voting dissent.

While shareholders base their votes upon compensation, performance, and monitoring abilities, these variables are hard to determine. And while investors differ in their degrees of sophistication (e.g. Alissa, 2015; Hooghiemstra et al., 2017), a considerable number of investors lack the time, knowledge, and information to critically assess whether the proposed compensation is excessive. In the past it was found that total compensation did not always change when shareholders shared their dissent with new policies. Rather than changing the amount of pay, boards tried to trick shareholders by changing

(10)

9 the structure of the compensation (Ferri & Maber, 2013; Abernethy et al., 2015). The new structure would likely be approved by shareholders, as they were not sophisticated enough to assess whether it is fair (Perry & Zenner, 2001; Buck et al., 2003; Hooghiemstra et al., 2017). Both because of a lack of knowledge and the complexity of compensation packages, it is hard for shareholders to properly assess what vote to cast. It is therefore found that, among others, shareholders rely heavily on proxy advisors (Ertimur et al., 2013), the media (Bednar, 2012; Hooghiemstra et al., 2015), and the firm itself as main sources to base their votes upon. The reliance of shareholders on information provided by the firm, makes them vulnerable to distortions in the voting processes (Jensen & Meckling, 1976; Eisenhardt, 1989; Kaplan et al., 2014). In the next sub-section, I will focus on how firms and their directors make use of this opportunity by engaging in one specific type of impression management, namely: strategic noise.

2.4 Anticipatory impression management & strategic noise

A way of influencing shareholders is through impression management. As adopted from the social psychology and defined by Goffman (1959), impression management is the expression of the self for an audience. As agency theory describes, actors want to achieve their own goals and, thus, act out of self-interest. In cases of poor performance, there will be negative consequences for achieving those self-interests and actors will try to influence the counterparty by using impression management (Merkl-Davies & Brennan, 2011). Impression management has been applied as a way of explaining managers´ behaviour after the occurrence of a certain event (e.g. not reaching performance goals) and is largely used to maintain status-quo. When it became clear to managers that impression management worked in those cases where investors were less sophisticated, researchers found that it was used in advance of important events as well. This is when Elsbach et al. (1998) supplemented the main idea with anticipatory impression management (AIM), where targeted users are influenced for future events. AIM was used in cases that management, boards, or CEOs expected a negative response to a certain event and, therefore, tried to influence shareholders to reduce the negative effects.

The literature of AIM has a broad scope of applications, but has not been linked to say-on-pay much. Hooghiemstra et al. (2017) investigated the effects of readability of director remuneration reports, and found lower levels of dissent when the report was more difficult to read. In conformance with Merkl-Davies and Brennan (2011), whom stated that sophisticated investors see impression management as “cheap talk”, it was found that institutional investors were not affected by the lower readability of reports and that they would increase their voting dissents instead (Hooghiemstra et al., 2017). A form of AIM that has not been applied to say-on-pay, introduced by Graffin et al. (2011), is strategic noise. Strategic noise is the tactic used to confound the response of shareholders to a certain event by announcing other important events. By timing announcements and publications in the period surrounding the important or focal event, strategic noise can help reduce the potentially negative effects of the important event. For instance, to confound and, hence, soften shareholders’ reaction, concerning the appointment of a new COO, a firm may announce that it embarked on new investments in R&D in the same period. In their research, Graffin et al. (2011) found that confounding publications (e.g. positive publications concerning: dividend, earnings, or M&A) in the period surrounding CEO succession lead to higher acceptance of the new CEO. It was used by firms when boards expected high resistance to the event. In situations where outgoing CEOs have high tenures or when pay gaps between CEOs and top management are significant, firms generate more noise and as a result experienced higher acceptance to successions. The study by Graffin et al. (2011) reveals that shareholders are highly influenceable with the use of confounding publications. Both the research to the readability of reports and strategic noise are proven to be effective practices of firms in order to influence shareholders’ responses.

(11)

10 In the case of say-on-pay, firms and individuals in the upper-echelons face considerable scrutiny regarding remuneration packages. In the short period after the introduction of say on pay, shareholder dissent was high and the media extensively reported negatively about pay packages at firms and the role of CEOs and boards (Conyon & Sadler, 2010; Ferri & Maber, 2013). Agency theory describes that actors value their own interests above those of the firms and shareholders. Having a role as CEO or board member comes with status, and those members of society are chosen because of their prior accomplishments and built reputations (Nguyen & Nielsen, 2010). For those individuals, status and reputation are important and they are seeking to protect them at all cost. While the firms’ reputation is important, reputation of individual CEOs and board members is even more important. Prior research indicates that boards face reputational damage following negative attention, for instance as a result of fraud within their company (Fich & Shivdasani, 2007). Research regarding boards responding to media attention and shareholder activism shows that boards follow the media and shareholders’ criticism towards their company by addressing the discussed issues (Del Guercio et al., 2008; Ertimur, Ferri, & Muslu, 2011; Bednar, 2012; Bednar, Boivie, & Prince, 2013). The same is the case with CEO compensation, as excessive CEO compensation has led to negative media coverage in the past (Core & Guay, 2010). When high dissent is expected, boards might try to influence shareholders in responding differently. It all indicates that boards try to prevent negative media coverage, possibly through strategic noise, as it destroys their reputations and status.

While it is anticipated that boards will use strategic noise to reduce the negative effects of shareholder dissatisfaction, it is not known whether shareholders respond to strategic noise as expected. There are three general explanations for shareholders to respond with lower levels of dissent. First, shareholders rely on information that is available due to the issues of information asymmetry (Jensen & Meckling, 1976; Eisenhardt, 1989). The most important stream of information comes from the firms themselves, as they possess all required information. The downside of this information stream is that it introduces the risk that shareholders become a target for strategic noise. Firms decide what information is published and how pronounced the more negative publications are. Shareholders are possibly only assessing all information that the firms want them to see. Second, shareholders are not able to cope with large streams of information as it leads to information overload (Kaplan et al., 2014). Individuals can only focus on a limited number of issues simultaneously, and with higher use of strategic noise this becomes more difficult. The number of issues and events rise and shareholders have to choose what requires their focus. Shareholders might not be able to assess all information regarding CEO remuneration effectively anymore. Third, shareholders lack time to assess all aspects of CEO remuneration already. The lack of time and resources are an issue for shareholders and when strategic noise is introduced, even more of these issues arise. Strategic noise will lead to an enhanced number of publications that are focused on confounding shareholders. With a rising lack of time and resources, shareholders will have more difficulty recognizing strategic noise and will more likely be affected by it. Prior research discovered that shareholders do not respond with high levels of dissent when excessive compensation is combined with high firm performance (Ferri & Maber, 2013; Krause et al., 2014). This indicates that there are reasons for shareholders to lower their dissent, namely when it independently leads to value enhancement of stock prices (Graffin et al., 2011). The focus of ‘strategic-noise-like’-publications is to retain a positive view when a negative response is anticipated, the ‘strategic-noise-like’-publications therefore will always be positive to ensure low shareholder dissent. Examples of specific announcements that change stock prices, are: mergers and acquisitions (M&As) and earnings announcements. Indeed, M&A-announcements were found to result in a 3-6% increase of stock prices on the day of the announcement (Barraclough, Robinson, Smith & Whaley, 2013). For earnings announcements there also is a large span of literature indicating that it leads to higher stock prices on average (Patell & Wolfson, 1981), with the exception of late announcements as they most likely contain more bad news (Kross, 1982). Accordingly, these announcements can influence shareholders’ perceptions on what remuneration is appropriate, as these kinds of publications lead to enhanced stock values and, therefore, higher performance for shareholders. As a result, shareholders think that

(12)

11 the excessive compensation is more appropriate and they will respond with lower levels of dissent in comparison to cases with lower firm performance (Graffin et al., 2011; Ferri & Maber, 2013; Krause et al., 2014). Publications involving these topics might complicate the shareholders’ ability to objectively decide whether CEO remuneration is considered fair.

Because of the aforementioned reasons, it is expected that shareholders will have high difficulty assessing say-on-pay reports when strategic noise is introduced. In line with the results of prior research (Graffin et al., 2011), it is expected that strategic noise will lead to higher shareholder acceptance, hence, lower levels of voting dissent. Therefore, I propose the following hypothesis:

Hypothesis 2. The use of strategic noise is associated with lower levels of voting dissent.

When boards expect high resistance for a proposed remuneration package, strategic noise will be introduced to protect their reputations. If excessive compensation leads to higher voting dissent and voting dissent can become lower when strategic noise is introduced, it is expected that combining these effects helps to understand why shareholders do not always respond negatively to excessive compensation. If strategic noise is introduced, the relationship between excessive compensation and shareholder dissent is not as clear. Shareholders are not able to objectively assess excessive remuneration and, therefore, it does not lead to higher shareholder dissents immediately. Strategic noise can reduce the negative effects of excessive compensation on shareholder dissent. Based upon this argumentation, I propose the following hypothesis:

Hypothesis 3. The use of strategic noise weakens the relationship between excessive CEO compensation and higher levels of voting dissent.

(13)

12

3. Research design

3.1 Sample selection and data sources

The initial sample for this research is comprised of firms within the UK financial industry, in the period of 2001-2019. For companies to be required to have a voting process concerning director remuneration reports, they have to be quoted on the UK stock exchange (The Companies Regulation 2019 No. 970, 2019). The sample contains all quoted or formerly-quoted companies in the financial industry of the UK, with data starting from shortly before the implementation period. The sample consists of 711 observations from 83 different firms, after dropping firm-years due to incomplete data. This broad timeframe enables comparisons between the different stages of the legislation (i.e. advisory votes and binding votes).

Data was gathered from various sources and merged through their unique id’s (ISIN, company ID or director ID). For the amount of shareholder dissent, data was manually gathered from the firms’ AGM results. One of the voting resolutions relates to the remuneration report at fiscal year-end; these voting results (in terms of % in favor and % against) are used to construct my dependent variable. Data regarding CEO compensation comes from BoardEx. By linking multiple databases and distinguishing the UK and the rest of the world (note that some UK-listed financial firms are formally headquartered in ‘tax havens’ like Jersey and the Bermudas), almost all CEOs could be linked to their compensations. The last main variable was strategic noise, hence, the amount of confounding publications surrounding the period of the event. All publications were gathered through LexisNexis.

Next to these main variables, other variables were collected for control purposes. Data regarding the consultants, board meetings, and remuneration committee was gathered manually from the annual reports. All information about characteristics of the CEOs and boards is obtained through BoardEx. Lastly, data regarding performance and the financial situation of the companies is gathered through Eikon. To mitigate the (negative) effect of outliers, all continuous variables are winsorized for the upper and lower 1%-range. Appendix A reports the entire list of used variables.

3.2 Dependent variable

The dependent variable in this research is shareholder dissent. The results of the say-on-pay voting can produce four outcomes, votes can: agree, disagree, be withheld and be discretionary. Withheld votes are seen as against votes, as investors consciously choose to be present at the meeting and not vote for this resolution. Conyon and Sadler (2010) describe it as indicating that shareholders are against the resolution. In conformance with numerous authors that have collected the same variable (e.g. Conyon & Sadler, 2010; Ferri & Maber, 2013; Alissa, 2015; Hooghiemstra, 2017), shareholder dissent therefore is the sum of withheld votes and those against the remuneration report as a percentage of the total casted votes (DISSENT).

3.3 Independent variable

The independent variable in this research is excessive compensation. Measuring excessive compensation is done in accordance to researchers Core et al. (1999) whom stated that shareholders expect a certain level of compensation based upon CEO, board, and firm characteristics. Therefore, I will use the following variables to explain the expected level of compensation: current total compensation (COMPENSATION) as baseline, number of board members (BOARD_SIZE) to indicate governance measures, percentage of independent board members (IND_BOARD) to indicate effectiveness of the board, sales (SALES) to explain size, ROA (ROA) to explain performance, market-to-book value (MTB) to explain growth potential, annual stock returns (RET) to explain performance for shareholders, volatility of the annual stock returns (VOL) to explain riskiness, CEO tenure (TENURE) and CEO age (AGE). In my regression I use the natural logarithm of COMPENSATION, BOARD_SIZE, SALES, TENURE and AGE, to alleviate heteroskedasticity issues. The expected compensation is

(14)

13 measured separately per year and results in 17 levels of expected compensation that are compared to their years’ actual compensation. By measuring each year separately, expected and actual compensation are more accurate and closely related. The years 2001-2002 and 2018-2019 are combined, as both 2001 and 2019 have not enough observations for comparable results.

Using the ordinary least squares (OLS) linear regression, CEOs with compensation structures that are not in line with the regression are considered as overpay (OVERPAID) or underpay (UNDERPAID). Underpay or overpay is calculated by comparing the natural logarithm of actual compensation (salaries, bonuses and others) with the expected compensation (using the regression approach discussed earlier). An error term above zero indicates excessive compensation, an error term below zero indicates underpayment. Defining the variables separately is important to understand the effects of both situations, with the main focus on overpaid CEOs.

3.3 Moderating variable

The moderating variable is strategic noise and is measured by the sum of all confounding publications. Publications are considered confounding, based upon Graffin et al. (2011), when they are published in the period surrounding the event. Required is that they must be under control of the firm and they do not have the purpose to help understand the remuneration report. This narrowed the number of categories down to six, concerning: acquisitions, board/management changes, equity and debt transactions, dividend, joint ventures and other announcements. The period in which strategic noise is measured in the main model, is a short timeframe. The window concerns the 7 days prior to the event date and therefore generates the same period for each firm-year combination.

3.4 Control variables

To control for the possibility that variability in my dependent variable (DISSENT) is affected by other differences between firms, I include the following control variables: CEO years at the company (YEARS_IN_COMPANY), CEO tenure (TENURE), CEO gender (GENDER), board size (BOARD_SIZE), percentage of independent board members (IND_BOARD), number of remuneration committee meetings (RC_MEET), number of pay consultants (CONSULTANTS), market-to-book value (MTB), stock volatility (VOL) and the firm-years (YEAR_F.E.) to alleviate influences from year-to-year.

Taking into account CEO characteristics as gender, tenure, and years at the company is based on prior research describing its effects on CEO compensation and company performance. CEOs with higher tenures and years at the company were found to receive higher levels of compensation (Hogan & McPheters, 1980; McKnight, Tomkins, Weir, & Hobson, 2000; McKnight & Tomkins, 2004) and male CEOs tend to receive more compensation than their female peers, therefore shareholders might think that higher compensation is more appropriate for longer tenured male CEOs and thus leads to lower dissent (Kullich, Trojanowski, Ryan, Haslam, & Renneboog, 2011; Perryman, Fernando, & Tripathy, 2016; Fortin, Bell, & Böhm, 2017). Additionally, firms with longer CEO tenures have lower firm performance (Nguyen, Rahman, & Zhao, 2018). Krause et al. (2014) found that high CEO compensation was accepted by shareholders when firm performance was high. In this case, the CEO characteristics indicate higher compensation combined with poorer firm performance and therefore, shareholder dissent is expected to be higher. Hence, it is important to control for these variables.

Board size is used to explain the influence of the governance structures in place. Core et al. (1999) describe that firms with weak corporate governance have lower performance. Additionally, Gomez-Mejia and Wiseman (1997), and Fahlenbrach (2008) describe that bigger boards indicate less effective oversight. This could also indicate that boards with more meetings, are more thorough in discussing the issues at sight. Especially for the remuneration committee, more meetings could indicate that the committee takes CEO compensation seriously and tries to behave in the best interest of shareholders. Taken these effects together indicates that board size and committee meetings are important

(15)

14 determinants, with both positive and negative influences according to prior research. This indirectly results in effects on the levels of dissent, as shareholders base their votes upon performance (Krause et al., 2014) and therefore have to be included as control variables.

A third type of influential factor is the pay consultants, as they (in)directly influence the levels of dissent. If firms could develop pay packages that are acceptable for all involved parties, they would not have to worry about shareholder dissent. But as firms want to forego the negative effects of high dissent (e.g. negative media attention), this job is outsourced to pay consultants. These consultants are experienced in developing attractive pay packages that would lead to lower levels of dissent. Therefore, pay consultants are contracted by firms to reduce dissent, this is why the number of pay consultants involved is included as control variable (Murphy & Sandino, 2010; Voulgaris, Stathopoulos, & Walker, 2010; Armstrong, Ittner, & Larcker, 2012).

Last factors are concerning firm performance. In the past, size, and growth opportunity were seen as important determinants for firm performance (Murphy, 1984; Core et al., 1999; Cuñat, Giné, & Guadalupe, 2015), while size (SALES) did not have an influence on the regression model, growth opportunity (MTB) does. Additionally, higher firm performance was found to lead to lower dissent (Graffin et al., 2011; Ferri & Maber, 2013; Krause et al., 2014; Cuñat et al., 2015). Hence, market-to-book value (to indicate growth opportunity) and VOL (to stock volatility, and therefore risk) are included as control variables, as they are thought to explain lower levels of dissent.

3.5 Empirical model

In this thesis I argue that excessive compensation structures lead to higher levels of shareholder dissent, that the use of strategic noise will lead to lower levels of dissent, and that the use of strategic noise weakens the relationship between shareholder dissent and excessive compensation.

The first hypothesis explains the relation between higher shareholder dissent and excessive compensation. When compensation is considered excessive, based upon the earlier mentioned nine indicators, the variable OVERPAID has a positive value.

DISSENTi,t = β0 + β1OVERPAID i,t + β2UNDERPAID i,t +

∑

βkCONTROLS k,i,t + ɛi,t

The second hypothesis seeks to explain whether strategic noise is influencing the levels of dissent. Strategic noise is defined as the number of confounding publications in the week before the AGM voting event. A higher amount of publications will most likely lead to more shareholder confusion.

DISSENTi,t = β0 + β1SHORT_NOISE i,t +

∑

βkCONTROLS k,i,t + ɛi,t

The third hypothesis combines the prior two models to seek whether strategic noise moderates the relationship between shareholder dissent and excessive compensation. This is accomplished by including SHORT_NOISE and the interaction term with OVERPAID.

DISSENTi,t = β0 + β1OVERPAID i,t + β2SHORT _NOISE i,t + β3UNDERPAID i,t +

(16)

15

4. Results

4.1 Descriptive statistics

Following 711 observations, regarding 83 firms from the financial industry in the period of 2001-2019, table 2 presents an overview of all compensation and levels of dissent over the years. While shifting a bit from year to year, the base salary (plus bonus) and equity compensation remain rather stable. Both compensation parts vary from 45-55% each year, similar to the findings of Gregg et al. (2012). While Ferri and Maber (2013), and Abernethy et al. (2015) suspected a shift from base salary to more equity-based compensation due to large shareholder revolts and changes in the legislation, this is not observed for the financial industry.

As table 2 shows, shareholder dissent peaked in 2002, 2009-2011 and 2013. The possible explanations are that most companies adopted the legislation in 2002 or 2003, which led to shareholders immediately and furiously responding to the misuse of power by companies in the years prior (Ferri & Maber, 2013). In 2009-2011 the high revolt might be as a result of the financial crisis and high levels of total compensation in those years (Gregory-Smith, Thompson, & Wright, 2013). In 2013 the vote became mandatory for companies to follow, which might have led to shareholders trying to exert their influence on firms’ pay practices again.

Table 2. Descriptive statistics: compensation and dissent 2001-20191

YEAR

Base

compensation

Equity

compensation

Total

compensation

DISSENT

2001

734.67

326.67 1,061.67

4.6%

2002

1,255.69

2,026.31

3,544.77

10.9%

2003

1,719.39

1,425.62

3,144.96

9.3%

2004

1,944.85

1,589.44

3,534.22

6.8%

2005

1,767.35

2,023.63

3,793.33

5.1%

2006

2,391.98

2,434.25

4,850.48

5.7%

2007

3,129.12

2,664.74

6,011.68

7.7%

2008

2,162.57

1,925.43

4,191.55

8.1%

2009

1,644.28

1,987.00

3,631.34

9.1%

2010

1,717.16

3,192.89

5,034.96

11.2%

2011

2,099.49

2,963.23

5,177.54

10.5%

2012

1,871.15

3,386.05

5,389.12

8.9%

2013

2,261.38

2,931.28

5,320.48

10.8%

2014

2,210.09

2,421.16

4,679.63

7.9%

2015

2,012.71

1,676.19

3,761.02

8.5%

2016

1,879.14

1,643.00

3,522.14

6.7%

2017

1,865.83

2,215.88

4,206.81

6.2%

2018

1,556.59

2,097.32

3,772.73

9.4%

2019

1,358.29

2,209.14

3,923.00

9.6%

Total

2,000.50

2,288.43

4,375.13

8.2%

(17)

16 Table 3 contains descriptive statistics of all regression variables2_{. The table shows an average dissent}

of 8.2%, the exact number Hooghiemstra et al. (2017) found. Similarly, Conyon and Sadler (2010) reported levels from 7-10% between 2007, Alissa (2015) reported 5.8% dissent between 2002-2012 and Gregory-Smith, Thompson, and Wright (2013) show averages from 3-8% in 1998-2002-2012. The independent variable SHORT_NOISE shows a mean of 0.305 and indicates that the average firm has 0.305 confounding publications in the 7 days prior to the AGM voting event. While this number seems rather low, these are only publications that were discovered to confound shareholders according to prior research (Graffin et al., 2011). Shareholders considering these publications are more prone to become influenced and change their votes in the companies’ best interest.

The average CEO in the financial industry sample is male (98.9%) and has a tenure of 5.4 years, while active within the company for 9.4 years. It indicates that a large fraction of the CEOs is selected from the existing workforce. Oversight within the firms is through boards, which differ in size from 2 to 22 members. The average board consists of 10.2 members, of which 54.3% are independent. Due to the introduction of the say-on-pay legislation in 2002, firms are required to have remuneration

committees. In the sample, their meetings took place 1 to 15 times a year. On average they met 5.3 times per year. Additionally, on average a company contracts services 1.3 consultants, indicating that most companies do use consultants to advise them about the appropriate compensation structures. The market-to-book value is 2.6, indicating the book value of assets of the companies are valued higher than the entire company itself; it also indicates high growth opportunity for investors. The variable volatility reports a mean of 5.4 and shows that the standard deviation of share prices was 13.5%.

Table 3. Descriptive statistics: DISSENT, OVERPAID and control variables

Variables

Obs

Mean

Median Std. Dev.

Min.

Max.

DISSENT

711 0.082

0.057

0.077

0.001

0.348 OVERPAID

711 0.280

0.000

0.438

0.000

2.385 UNDERPAID

711 -0.389

-0.018

0.636 -4.565

0.000 SHORT_NOISE

711 0.305

0.000

0.779

0.000

10.000 YEARS_IN_COMPANY

711 2.041

2.067

0.805

0.000

3.875 TENURE

711 1.567

1.589

0.757

0.000

3.374 GENDER

711 0.989

1.000

0.106

0.000

1.000 BOARD_SIZE

711 2.282

2.197

0.296

0.693

3.091 IND_BOARD

711 0.543

0.556

0.148

0.000

1.000 RC_MEET

711 5.300

5.000

2.243

1.000

15.000 CONSULTANTS

711 1.290

1.000

1.013

0.000

6.000 MTB

711 2.598

1.780

3.055

0.370

22.066 VOL

711 5.378

6.000

3.120

1.000

13.000

Please refer to appendix A for a description of all variables

In table 4, the Pearson correlation matrix is reported. The table shows four positively correlated variables with DISSENT at the 1% significance level (p < .01), these are: OVERPAID, SHORT_NOISE, IND_BOARD and CONSULTANTS. Additionally, there are two positively correlating variables at the 5% significance level (p < .05), these are: YEARS_IN_COMPANY and RC_MEET. As all correlations reported in table 4 remain below the absolute value of 0.7, multicollinearity should not be a problem in this study. This is confirmed when looking at the maximum VIFs in table 5, that all remain below 3.

2_{The descriptive statistics of YEARS_IN_COMPANY, TENURE and BOARD_SIZE is shown in table 3 as a natural}

(18)

17 Table 4. Pearson correlation matrix

A B C D E F _G H I J K _L _M DISSENT _{A 1.000} OVERPAID _{B 0.098** 1.000} UNDERPAID _{C 0.033 0.393** 1.000} SHORT_NOISE _{D 0.097** 0.109** 0.118** 1.000} YEARS_IN_COMPANY _{E 0.076* - 0.014 - 0.001 0.059 1.000} TENURE _{F 0.001 - 0.003 - 0.003 - 0.029 0.645** 1.000} GENDER G - 0.040 _0.031 - 0.020 _0.025 _0.047 _0.019 _1.000 BOARD_SIZE _{H 0.027 0.006 0.035 0.079* 0.188** 0.003 - 0.025 1.000} IND_BOARD _{I 0.118** - 0.089* - 0.147** 0.066 - 0.062 - 0.102** - 0.006 0.131** 1.000} - RC_MEET _{J 0.095* - 0.099** - 0.009 0.090* 0.042 0.008 0.002 0.197** 0.331** 1.000} CONSULTANTS _{K 0.100** 0.121** 0.126** 0.127** - 0.046 - 0.029 0.017 0.227** 0.152** 0.216** 1.000} MTB L - 0.043 _0.012 - 0.043 - 0.030 _0.038 _0.042 _{0.008 - 0.195** - 0.035} _{- 0.074* - 0.107** 1.000} VOL _{M 0.059 - 0.108** - 0.108** 0.009 - 0.081* - 0.111** 0.004 - 0.176** 0.258** 0.071 - 0.066 - 0.039 1.000} *, ** respectively indicate significance at the 5% and 1% level

(19)

18

4.2 Testing hypotheses

In testing the hypotheses, multiple regression analysis is used to determine the effects of the dependent variable (DISSENT), independent variables (OVERPAID and SHORT_NOISE) and the control variables (YEARS_IN_COMPANY, TENURE, GENDER, BOARD_SIZE, IND_BOARD, RC_MEET, CONSULTANTS, MTB, VOL, and YEARS_F.E.).

Model 1 from table 5 presents the effects of the control variables on the levels of dissent. It shows that the control variables explain 7.9% (R2_{= 0.079) of all variance in the levels of dissent. The results indicate}

that YEARS_IN_COMPANY (p < .01), TENURE (p < .1) and CONSULTANTS (p < .05) are associated with dissent levels.

Model 2 analyzes the first hypothesis: “CEOs with excessive compensation packages are associated with higher levels of shareholder voting dissent”. In addition to the first model, the variables OVERPAID and UNDERPAID are introduced, where UNDERPAID is used to check whether lower than expected CEO compensation lowers dissent. OVERPAID shows a positive and significant (p < .05) coefficient. One standard deviation increase of OVERPAID leads to an increase of 0.740% (0.017 * 0.438 * 100%) in DISSENT (Hooghiemstra et al., 2017). It shows that excessive CEO compensation leads to higher levels of dissent, hence, hypothesis 1 is accepted. Similarly, and as expected, UNDERPAID shows no significant impact with DISSENT; this is consistent with Laksmana et al. (2012) and Hooghiemstra et al. (2017).

Model 3 analyzes the second hypothesis: “The use of strategic noise is associated with lower levels of voting dissent”. This model replaces the variables OVERPAID and UNDERPAID with SHORT_NOISE to indicate the relation between the levels of shareholder dissent and use of strategic noise. SHORT_NOISE shows no significant results and therefore deviates from the expected relationship as stated in the hypothesis, hence, the second hypothesis is rejected.

Model 4 analyzes the third hypothesis: “The use of strategic noise weakens the relationship between excessive CEO compensation and higher levels of voting dissent”. This model includes all variables that were used in the previous models to explain possible joint relationships. There is also an interaction term added for the variables OVERPAID and SHORT_NOISE to include the possible effects between the two independent variables and shareholder dissent. The independent variables OVERPAID and SHORT_NOISE, and the interaction term show no significant relationship. These results show no indications for the moderating relationship of strategic noise on the positive relationship between shareholder dissent and excessive compensation. Hence, the third hypothesis is rejected.

(20)

19 Table 5. Regression results, hypotheses 1-3.

Variables

Model 1

Model 2

Model 3

Model 4

OVERPAID

0.017**

0.009 (0.008)

(0.009)

SHORT_NOISE

0.010

0.003 (0.006)

(0.006)

OVERPAID x SHORT_NOISE

0.015 (0.012)

UNDERPAID

0.007

0.007 (0.010)

(0.010)

YEARS_IN_COMPANY

0.015***

0.014***

(0.005)

TENURE

-0.011*

-0.010

(0.006)

GENDER

-0.061

-0.062

(0.039)

(0.040)

(0.039)

(0.040)

BOARD_SIZE

-0.024

-0.023

-0.024

-0.022

(0.016)

(0.015)

IND_BOARD

0.037

0.039

0.035

0.039 (0.038)

(0.036)

(0.037)

(0.036)

RC_MEET

0.001

0.000

0.001 (0.002)

(0.002)

CONSULTANTS

0.008**

0.007*

0.006 (0.004)

(0.004)

MTB

-0.001

-0.002

-0.001

(0.001)

VOL

0.000

0.001

0.000

0.001 (0.001)

(0.001)

YEAR_F.E.

Yes

Constant

0.127**

0.126**

0.129**

0.126**

(0.058)

(0.057)

(0.058)

(0.057)

Observations

711

711 R-squared

0.079

0.088

0.085

0.095 Prob > F

0.000

0.000 Highest VIF

1.91

1.93

2.40

*, **, *** respectively indicate significance at the 10%, 5% and 1% level (two-tailed). Numbers in parentheses are the standard errors, clustered by firm.

(21)

20

4.3 Robustness analyses

4.3.1 Strategic noise timeframe

In the current models, strategic noise is defined as the number of confounding publications in the last 7 days prior to the voting event. Instead of the last week, a longer time period can be taken into account, to check whether this influences the prior results. The proposed long timeframe is from the fiscal year-end until the event date.

Similar results are yielded when changing the timeframe wherein noise is measured. Model 2 in table 6 still reports a positive and significant (p < .05) coefficient for independent variable OVERPAID. Hence, hypothesis 1 is still accepted in the adjusted regression. When LONG_NOISE is introduced, the model shows no significant relationships. Hence, hypothesis 2 is still not accepted. Model 4 from table 6, shows that OVERPAID has a positive and significant (p < .05) coefficient, LONG_NOISE and the interaction term are not significant. Hence, hypothesis 3 is still not accepted. The main results are robust for changing the noise timeframe, it does not change the levels of dissent.

4.3.2 Resolution order

During the AGM, shareholders have the ability to vote regarding various statements, one of them being the director remuneration report (DDR). As the DDR is just one of the statements and as it is not regulated in which order these statements have to be voted for, companies treat the order differently. Sometimes the order even changes between years with the same firm. While not indicated by previous research yet, it could be possible that the order has an influence on the way shareholders vote. Therefore, this robustness test includes the specific spot (RES_NUMBER) as control variable.

Equal to the main regression, hypotheses 1 can be accepted with a positive and significant (p < .1) coefficient of the independent variable OVERPAID. For hypothesis 2, SHORT_NOISE is not significant, hence, hypothesis 2 is rejected. Hypothesis 3 also does not result in significant values and therefore is rejected as well. The main results are robust for including the resolution order, indicating that the shareholder dissent is not influenced by the order of the resolutions.

4.3.3 Transforming DISSENT

Following research by Hooghiemstra et al. (2017), log-transforming the dependent variable DISSENT [DISSENT / (1 – DISSENT)] might yield different results as DISSENT is a value from 0 to 1. The transformed variable will take on different values and, as such, influence the results.

After log-transforming the variable and regressing the models from the main regression, the results differ slightly. A noticeable and relatively large difference is the amount of variance of LOG_DISSENT explained by the model, it rose with 4-5 percentage points. In the second model, OVERPAID is still positive and significant (p < .01), resulting that hypothesis 1 is accepted. Model 3 shows no significance for SHORT_NOISE and therefore hypothesis 2 is rejected. Model 4 shows a significant result for OVERPAID, but not for the other independent variables. Hence, hypothesis 3 is rejected. Overall, the result for all hypotheses is that all results are robust for changing the dependent variable.

4.3.4 Timing effect of legislative changes

In the past, casted shareholder votes against a resolution did not lead to an obligation to change compensation height and structures. Due to changes in 2013, all companies with high levels of dissent had the obligation to make changes when the dissent-threshold of 50% was reached. Although the sample of the financial industry only contains one instance where this threshold was reached (GoshawK Insurance in 2005 with 55.7% against votes), shareholders might have changed their responses after the legislative changes. To check for these changes, I included a dummy with the value 1 for all firm-year combinations after 2014, as the legislation was introduced in October 2013 and the sample does not contain companies starting their fiscal year in the months in between October and

Boards protecting themselves