Remuneration Shareholder dissent and company performance

Remuneration Shareholder dissent and company

performance

Supervisor: Prof.Dr. C.L.M. Hermes Author: B.D.S. Roebbers

2 Abstract

3 1. Introduction

Many financial analysts take indicators such as accounting scandals, falling sales or mediocre earnings into account when assessing stocks. In July 2020, the Financial Times reported that firms like Morgan Stanley and USB management had found a new possible indicator to sell a company’s stocks1_{; In addition to monitoring the usual indicators, they now use a newly}

established indicator, namely pay revolts. As far as the investment companies are concerned, these voting outcomes of more than 50 per cent on remuneration resolutions have detrimental effects on company performance. An analyst of Morgan Stanley stated that they consider ‘Say-on-Pay’ votes to be an important factor of shareholder dissatisfaction, as it reflects the broader governance and strategic execution of a company. The analyst’s statement is based on the fact that Morgan Stanley has discovered a relative share price underperformance to the S&P 500 benchmark, for companies who have lost on their Say-on-Pay vote.

When the majority of shareholders are against the remuneration resolution, this is considered to be a measure of shareholder dissatisfaction. Say-on-Pay rights can be described as the right to vote about the plan of executive remuneration concerning company performance. Shareholders are thus not only voting on the pay plan but also on the level of confidence the CEO generates and the subsequent value-added to the company. Therefore Say-on-Pay votes can be seen as an explicit vote of confidence in the management (Cotter, Palmiter and Thomas, 2013; Cunat, Gine and Guadalupe, 2016). These negative shareholder votes indicate a misalignment of incentives between management and shareholders and could result in a share price reduction.

The detrimental effects of voting outcomes amounting to more than 50 per cent on remuneration are already acknowledged. However, this raises the question as to whether remuneration resolutions voting results achieving less than 50 per cent could have adverse effects on stock prices. This could mean that a significant amount of disapproval Say-on-Pay votes against management could indicate a share price decrease. Hence, this research will focus on the effect of significant remuneration shareholder dissent on company performance. Shareholder dissent is considered to be significant when 20 per cent or more of the shareholders vote against the resolution. In other words, when more than 20 per cent of the

4 shareholders vote against the remuneration policy or reports, this implies that a significant number of shareholders are not satisfied with the governance of the company. Therefore the main research question of this study is as follows:

Does significant shareholder dissent on remuneration harm stock prices?

When considering the question as to whether significant shareholder dissent influences company performance, it can be argued that vast differences in governance exist among companies. These differences in management can impact the magnitude of the effect that significant shareholder dissent has on company performance. Consequently, the anticipated effect of shareholder dissent is expected to be more apparent in cases where more dissatisfaction with the governance of shareholders is predicted. In the Say-on-Pay legislation literature, some of these ex-ante characteristics are indicated. Companies with weak boardroom governance and inefficient remuneration are expected to profit from the legislation of Say-on-Pay. These researched companies are more likely to benefit from changes induced by shareholder pressure. Therefore, follow-up research questions have been constructed:

1) Are the stock prices more sensitive to shareholder dissent on remuneration when the executive compensation is inefficient?

2) Are the stock prices more sensitive to shareholder dissent on remuneration when the board room governance is weaker?

5 shareholder dissent on company performance are stronger or weaker for different characteristics of governance.

The main research question will be answered by performing an event study, using the underlying methodology of Mackinlay (1997). To answer the follow-up research questions a regression analysis will be used with the cumulative abnormal returns (CARs) as the dependent variable. In this manner, the influence of the variables for weak boardroom governance and inefficient remuneration can be determined on the shareholder reaction of companies.

The Investment Association, a public company in the United Kingdom, set up an ideal setting for this research. As a reaction to the so-called “corporate governance revolution” plans of Theresa May, the investment association assembled all the cases of shareholder dissent in one public register. Consequently, to overcome possible selection bias, the United Kingdom is the ideal setting for this research. Therefore, the data used in this research will only focus on companies listed in the London Stock Exchange (LSE).

6 2. Literature Review and Hypotheses

The first subsection will discuss agency theory and the link with Say-on-Pay. The second subsection will present earlier conducted research about the effect of Say-on-Pay regulation on company performance and will offer additional information on Say-on-Pay voting outcomes on company performance. The third subsection will provide a possible shareholder reaction to remuneration dissent for different characteristics of governance. Finally, the review will conclude with the hypotheses based on the literature discussed.

Agency Theory Say-on-Pay

Jensen and Meckling (1976) argued that the separation of control and ownership creates agency problems. Therefore, shareholders need mechanisms to monitor the behaviour of management to keep their interests aligned. This mechanism could be expressed in the ‘exit-voice’ model proposed by Hirschman (1970). In this model shareholders can intervene in the management using two methods; “exit” and “voice”. Shareholders can intervene in the management via exit by threatening to sell their shares. This will cause a drop in the share price, will negatively impact management concerning equity-based compensation, and will indirectly impact the company’s reputation (Admati and Pfleider, 2009). The threat of exiting gives shareholders the power to align their interest with management (Edmans, 2009). According to the ‘exit-voice’ model, shareholders can also intervene in management policies with ‘voice’. This ‘voice’ is characterized as shareholder activism. The introduction of Say-on-Pay can be seen as an extension to ‘voice’ in this model (Stahopoulous and Voulgaris, 2016). Say-on-Pay rights can be described as the right to vote on the plan of executive remuneration concerning company performance. These rights are one of the latest initiatives giving shareholders more ‘voice’ in the boardroom. In contrast to ‘exit’, shareholders with ‘voice’ can take direct actions and become more involved in the decision making of the company. As a consequence, when shareholders vote against the remuneration policy, they are trying to intervene in management policy by expressing their dissatisfaction. This dissatisfaction is based on the fact that the pay plan is not in line with the value-added to the company by management. When management refuses to listen to these dissatisfied and dissenting shareholders, the shareholders could feel they can not intervene with “voice”. In this case, the shareholders could feel ignored and will therefore be tempted to intervene with “exit”.

Effect Say-on-Pay on company performance

7 on company value in the United States. They discovered, with an event study approach, that the passing of the Say-on-Pay bill in 2010 has had a positive effect on company performance. Cai and Walkling (2011) argue that this positive effect is because Say-on-Pay makes it feasible to involve widely dispersed shareholder in the negotiation on remuneration. According to Jensen and Meckling (1976), the principal should design the compensation contracts for the agent. Furthermore, Cai and Walkling (2011) pinpointed some ex-ante characteristics which can affect the company more by the legislation of Say-on-Pay. Companies with inefficient executive compensation and inadequate boardroom guidance were found to be better off with the legislation. When the board deviates from good governance or an efficient compensation plan, shareholders will express their dissatisfaction which consequently may help improve governance and the remuneration design.

Larcker, Ormazabal, and Taylor (2011) extended the research of Cai and Walkling (2011) by investigating multiple legislative actions of governance with various event dates on the market, including Say-on-Pay. However, for the latter, they could not find any evidence for a positive or negative influence. In the United Kingdom setting Ferri and Maber (2013) examined the same effect to the legislation of Say-on-Pay in 2002. They found a positive market reaction for companies with controversial pay practices. However, they also found evidence proving that the introduction of Say-on-Pay can destroy value. Larger companies in their research were more prone to receive shareholder revolts, which they claimed indicated that the introduction of mandatory Say-on-Pay was not beneficial for all companies.

8 effect of shareholder dissent, since they only research Say-on-Pay proposals between 10 percentage points of the majority threshold between 45 and 55 per cent. They found an abnormal stock market returns of 1.8-2.7 per cent for Say-on-Pay proposal passed on the day of the vote, compared to one that failed.

Effect shareholder dissent on company performance

The effect of shareholder dissent on company performance is lacking in the current literature. Cunat, Gine and Guadalupe (2016) briefly touched upon the effect of losing a Say-on-Pay vote. They argued that the significant effect of Say-on-Pay, after losing a Say-on-Pay vote, can be interpreted as a very negative signal. This negative signal increases the probability of future shareholder actions. These actions could lead to the dismissal of the CEO or to filing lawsuits demanding further change. Ferri and Maber (2013) indicated that larger companies may not benefit from the introduction of Say-on-Pay, since they are more prone to experience shareholder revolts. Both these findings indicate that negative Say-on-Pay votes could have a negative influence on company performance. These Say-on-Pay votes can be seen as an explicit vote of confidence in the management (Cotter, Palmiter and Thomas, 2013; Cunat, Gine and Guadalupe, 2016). Say-on-Pay can be described as the right to vote for the plan of executive remuneration concerning company performance. Shareholders will vote against the resolution when they feel that the pay plan does not resemble the value-added to the company. These votes of no-confidence in management can be harmful to the company for two reasons; Firstly, this can be harmful since shareholders will sell their shares due to the disagreement. Secondly, this can be harmful to the company since these no-confidence votes can be interpreted as a poor signal for outside investors. In the research, a significant shareholder dissent is defined as when more than 20 per cent of the votes are against the resolution. Therefore this research will focus on the effects of a significant shareholder dissent on company performance.

Effect of different ex-ante characteristics of governance.

9 dissatisfaction of shareholders with the company management. Cai and Walkling (2011) indicated some of these ex-ante characteristics of governance. They found that companies with weak boardroom governance and inefficient remuneration were better off with the legislation of Say-on-Pay. Therefore, this research will focus on these two ex-ante characteristics of governance.

For the ex-ante characteristic of weak governance, two proxies have been chosen. These proxies include gender diversity and the size of the board. Previous studies indicated that women on boards improve monitoring for several reasons; women on boards are more independent and thus are better monitors (Adams, 2016). Kirsch (2017) argued that women are more diligent than men, which helps to improve monitoring. Therefore it could be argued that a more diverse board should lead to improved boardroom governance. For this reason, insufficient gender diversity can be considered as sufficient proxy for weak boardroom governance. Alkabani, Cuomo and Mallin (2019) found that firms with more women in the remuneration committee is negatively related to shareholder dissent on remuneration. They stated that more women in the remuneration committee minimise apprehension for the executive remuneration plans. Earlier research also indicated that more women on remuneration committees decrease the likelihood of extremely high executive compensation.

The other chosen proxy for weak governance is the number of board members. Shareholders associate large boards with poor communication, an inability to initiate strategic change, poor group cohesiveness and lack of control over the CEO (Dalton et al. 1990). Ryan and Wiggens (2004) found that shareholders have a negative opinion of large boards. Furthermore, Yermack (1996) found that companies with a larger board size have lower market valuation and Guest (2009) proved that the size of the board has a strong negative impact on profitability. This supports the argument that large boards have poor communication and subsequently less efficient decision-making. Therefore, the size of the board is considered to be a suitable proxy for weak boardroom governance.

10 significant amount of negative Say-on-Pay votes. Therefore the effects on company performance may be larger when the CEO-Worker gap ratio is higher. Since more negative Say-on-Pay votes are predicted, this may be an indicator of poor quality governance.

CEO equity-based pay is recommended in previous research to align interests, between management and shareholders (Fama and Jensen, 1983; Jensen and Murphy, 1990). Therefore we would expect that more equity-based pay would result in a reduction of shareholder dissent. Colins, Marquardt and Niu (2019) proved that shareholders on average vote more in favour of remuneration resolutions that are more equity-linked. Therefore it can be argued that more equity-based pay, results in more confidence in the governance. For this reason, the effect of more equity-based pay is expected to provide less remuneration dissent on company performance.

Hypotheses

The existing research determines the effect of Say-on-Pay regulation on company performance. Less study is focused on the effect of Say-on-Pay voting. However, the effect Say-on-Pay shareholder dissent is missing in the current literature. Based on the presented research, I posit that shareholder dissent on remuneration harms company performance. The reasoning being that these negative Say-on-Pay votes can be considered as a measure of the extent to which shareholder are not satisfied with the governance of the company. Therefore the first hypothesis is formulated as follows:

Hypothesis 1: Shareholder dissent on remuneration has a negative impact on the stock performance of listed companies.

When answering this hypothesis, it can be expected that vast differences will exist among companies. The characteristics of corporate governance differ among companies and are strongly related to the stock market. Two main ex-ante characteristics of governance are presented in the study, which is expected to have stronger effects. The first ex-ante characteristic, which is expected to have a stronger effect is weak boardroom governance. The proxy’s gender diversity and board size are introduced to measure the effect of weak boardroom governance. Therefore the following hypotheses are developed:

11 Hypothesis 2b: Boards with a larger size are more heavily impacted by the consequences of shareholder dissent on remuneration.

The second ex-ante characteristics which are expected to have a stronger effect is inefficient remuneration. Therefore the following hypothesis is constructed:

Hypothesis 3a: Companies with a larger CEO-Worker gap are more heavily impacted by the consequences of shareholder dissent on remuneration.

A possible solution recommended in the study could be to pay CEOs more in common stocks to align interests between management and shareholders. Therefore the last hypothesis is constructed as follows:

12 3. Methodology

For the first hypothesis, this research will make use of the event study methodology, described by Mackinlay (1997). The event study method is applied to determine the impact of a specific event on the value of a company (Mackinlay, 1997). Thus, after a significant shareholder dissent has taken place. In the research, this significant shareholder dissent is defined when 20 per cent or more of the shareholders vote against the resolution. An assumption in this research is that shareholder dissent could be seen as an unexpected event since the outcome of the votes is not known ex-ante to the company shareholders. Therefore is it not possible to completely predict the outcome of the vote. However, this is a key assumption since large proxy advisers, such as ISS and Glass Lewis, provide advice well in advance of the voting day. Whenever the proxy adviser recommends voting against the remuneration resolution, this could influence the effect on company performance on the voting day. These negative recommendations could already be incorporated into the stock price.

13 A key element of an event study is the calculation of the abnormal returns. Firstly, to calculate the abnormal returns, the continuously compounded returns (Rit) for stock returns and

market returns need to be calculated. This is done according to the following formula:

𝑅_𝑖𝑡 =𝑃𝑖𝑡− 𝑃𝑖𝑡−1 𝑃_𝑖𝑡−1 (1)

Where Pit represents the share price of the company (i) on the day (t) and Pit-1 stands for the

share price of the previous day of Pit. Hereupon the expected returns need to be determined.

This can be done in various manners. This research will use the market model described in Mackinlay (1997), which is estimated as:

𝑅_𝑖𝑡 = 𝛼_𝑖 + 𝛽_𝑖𝑅_𝑚𝑡+ 𝜀_𝑖𝑡 (2) E(ε_𝑖𝑡 = 0) 𝑣𝑎𝑟(ε_𝑖𝑡) = 𝜎ε𝑖𝑡

2

Where alpha, beta and sigma are the market model parameters and eta the error term. Rit

represents the estimated expected return and Rmt stands for the actual market portfolio

return. Alpha can be interpreted as the risk associated with the stock, Beta can be interpreted as the volatility of the systematic risk compared to the overall market.

This research will make use of the market model since this is a straightforward model and the most common model to use for an event study. Furthermore, according to Brown and Warner (1980, 1985), the market model is well specified and relatively more powerful than more complex methods.

After calculating the expected returns, the abnormal returns can be calculated following Mackinlay (1997) by the formula:

𝐴𝑅_𝑖𝜏 = 𝑅_𝑖𝜏− α̂ − β_𝑖 ̂ 𝑅_{𝑖 𝑚𝜏} (3)

With these abnormal returns, the average abnormal returns can be calculated to determine the effect for the whole sample following the formula:

𝐴𝐴𝑅𝑡 = 1 𝑁∑ 𝐴𝑅𝑖𝑡 𝑁 𝑖=1 (4)

14 for three days, in line with earlier conducted research to the effects of Say-on-Pay regulation. This event window is as short as possible, since choosing a long event window reduces the statistical power of the test (McWilliams and Siegel, 2017). Subsequently, the event window will be from day -1 to day 1. However, event windows of 11 trading days and 21 trading days are implemented as well. These event windows are implemented to adjust for the possibility that some information could already be incorporated in the prices.

Since the event window will consist of multiple days, it is important to not only investigate the average abnormal returns (AAR) but also the cumulative average abnormal returns (CAARs). The CAARs are beneficial in addition to the AARs for the reason that they provide more understanding of the effect of the aggregate abnormal returns. Before calculating the CAARs, first, the cumulative abnormal returns need to be calculated. These are calculated as follows:

𝐶𝐴𝑅𝑖(𝜏1, 𝜏2) = ∑ 𝐴𝑅𝑖𝑡 (5) 𝜏₂

𝑘=𝜏1

Where τ1 to τ2 represents the time interval, in this case, the event window.

Following the CARs, the CAARs are estimated as follows:

𝐶𝐴𝐴𝑅𝑖(𝜏1, 𝜏2) = 1 𝑁∑ 𝐶𝐴𝑅𝑖(𝜏1, 𝜏2) 𝑁 𝑖=1 (6)

A Jarque-Bera test is conducted to test for normality in the sample. This is done by using the formula described below:

𝐽𝐵 =𝑁 6(𝑆

2₊1

4(𝐾 − 3)

2_{) (7)}

15 The parametric test used for testing the AARs and CAARs is the cross-sectional T-test. This test is used due to the reliability and the simplicity for normally distributed samples. The test statistics for the AARs and the CAARs are calculated as follows:

𝑡_{𝐴𝐴𝑅𝑡} = √𝑁 𝐴𝐴𝑅𝑡

𝑆_{𝐴𝐴𝑅𝑡} (8) 𝑡𝐶𝐴𝐴𝑅𝑡 = √𝑁 𝐶𝐴𝐴𝑅𝑡

𝑆_{𝐶𝐴𝐴𝑅𝑡} (9)

Furthermore, a non-parametric test is implemented for both the abnormal returns and the cumulative abnormal returns, under the assumption of non-normality. The non-parametric test used in this research is the non-parametric rank test, introduced by Corrado (1989). The Corrado rank test first ranks the abnormal returns. After this is established, the average excess rankings can be calculated for each day. In this manner the Corrado rank test measures whether these average excess rankings are statistically significant. However, for the CAARs, the Corrado rank test is not applicable. Therefore the generalized rank test, introduced by Kolari and Pynnonen, 2011) is implemented to test the significance for the CAARs. The generalized rank test is a continuation of the Corrado rank test, however, it uses generalized abnormal returns instead of normal abnormal returns. In this manner, the cumulative abnormal returns can be tested for significance efficiently.

Regression analysis of CAR’s

For the other hypotheses, an ordinary least squares (OLS) regression will be used, with as dependent variable the Cumulative Abnormal Returns (CARs). The independent variables include: gender diversity, the board size, CEO-worker gap ratio and CEO equity-based pay. Subsequently, the OLS regression will measure whether the effect of remuneration shareholder dissent on the stock returns should differ depending on the characteristics of governance. To correct for possible heterogeneity robust standard errors are adopted. The following regression will be used:

𝐶𝐴𝑅_{𝑖(𝜏1,𝜏2}) = 𝛼 + 𝛽1𝐵𝐺𝑅𝑖 + 𝛽2𝐵𝑆𝑖+ 𝛽3𝐶𝑊𝑅𝑖+ 𝛽4𝐶𝐸𝑅𝑖 + 𝜀𝑖 (10)

Where CARi(τ1,τ2) stands for the cumulative abnormal returns of stock i in the event window

from day τ1 to day τ2. Alpha (α) represents the intercept and beta the standardized coefficient.

BGRi is the board gender ratio, BSi is the board size, CWRi is the CEO-Worker gap ratio, CERi is

16 Robustness checks

For the reason that the unexpectedness of a Say-on-Pay vote can be debatable, several robustness checks are added. The first robustness check implemented is to check whether more votes withheld has a stronger effect. Commonly an abstained vote is seen as a vote against. For example in 2014, Warren Buffet abstained to vote since he disagreed with the pay plan of Coca-Cola. According to the Financial Times2 _{he did not want to go to war, he}

merely wanted to express his disapproval. Furthermore, White (2008) argued that more and more companies are moving to majority standards, meaning that a resolution is only passed with a majority of the votes. Majority standards make abstaining votes much more relevant, as majority standards are less likely to be met. Therefore, White (2008) argued that withholding shareholders votes is an indication of shareholder discontent. The first robustness check will be performed by summing the votes against and votes withhold, in line with the research of Conyon and Sadler (2010). Consequently, this newly created variable will be regressed on the abnormal returns.

Cai and Walkling (2011) find that it is eventually up to the board to make Say-on-Pay value increase. Based on this finding it can be argued that the board’s reaction to shareholder dissent can be of utmost importance for the effect it has on company performance. Therefore the second robustness check will split the sample up in two different subsamples to illustrate the reaction of the board. The sample will be split up according to whether companies reappear for consecutive years. It can be argued that when a significant shareholder dissent reappears for consecutive years, the board’s reaction was not valuable. Therefore, whenever a company copes with a significant shareholder dissent for consecutive years, this is treated as a poor reaction of the board. When a shareholder dissent reoccurs, the management can be considered to be disappointing. Therefore, the abnormal returns should be more negative for companies impacted by significant shareholder dissent for consecutive years. The first year is used within the consecutive year period since the management for the first year is expected to be worse. In my sample of 239 events, this is the case for 35 events.

2 _{Bond, S. and Foley, S. (April 23, 2014) Buffett says Coke employee equity pay plan is ‘excessive’. Retrieved}

17 4. Data and Descriptive Statistics

The data required for the event study are the events of the shareholder dissent and the stock returns of the companies. The stock returns for the companies will be gathered from Thomson Reuters Eikon. For the shareholder dissent events, a public register3 _{will be used, set up by}

The investment association. The investment association is the trade body that represents investment managers in the United Kingdom. The members of the association vary from small, independent United Kingdom investment firms to global players. In 2017, as a reaction to the so-called “corporate governance revolution” plans of Theresa May, the investment association assembled all the cases of shareholder dissent in one public register. The investment association created this register for companies when a significant amount of shareholders voted against a resolution. This significant amount is defined as more than one-fifth of the total votes cast. Using this United Kingdom public register is the ideal setting for this research to avoid possible selection bias since all the possible events are assembled in this public register. Therefore the data used in this research will focus on companies listed in the London Stock Exchange (LSE) only.

The information given in the public register includes: the name of the company, meeting date, meeting type AGM/GM, resolution number & title, percentage voted for, percentage voted against, percentage voted withheld and percentage issued share capital voted. Further, it states whether the company issued a declaration stating the actions it was planning to take in order to understand the reasons behind the vote result. The period of the data is from 2017 until August 2020. The distribution of the events is displayed in Table 1. The public register exists of more than 1,044 events of shareholder dissent on different types of resolutions as stated above. For the approval of the director’s remuneration report, it provides 168 events. For the approval of the remuneration policy4 _{it provides 69 events. Further, I have collected}

2 events of other remuneration resolutions. Eventually, my final sample consists of all these remuneration events together, which equates to 239 events.

3 _{Link to this public register: https://www.theia.org/public-register}

18

Table 1: Significant shareholder dissent distributed per year

Year Events

2017 61

2018 60

2019 65

2020 53

The database BoardEx is used to obtain the data for the proxies of inefficient remuneration and weak governance. The selection criteria used to navigate within the database BoardEx include: Company Name, Company ISIN, Director type, Board ID, Total Compensation (in 000s), Equity Linked Remuneration Ratio, Gender Ratio and Board size. Total Compensation (in 000s) is measured using the total Salary plus Bonus. Equity linked remuneration is measured using the equity linked compensation as a proportion of total compensation of the individual. Gender diversity is given as the proportion of male directors divided by the board size. From these selection criteria, the variables gender diversity, board size, CEO-Worker gap and Equity linked remuneration ratio can be derived. To obtain the CEO-Worker gap ratio, the total compensation is divided by the average U.K. salary of £30,420 (according to research office for national statistics). Subsequently these variables are linked to the obtained list of events, by using the ISIN of the companies and event year.

The descriptive statistics are shown in Table 2, with the number of observations, mean, median, standard deviation, minimum and maximum. The correlation matrix of the variables is displayed in Table 3.

Table 2: Descriptive Statistics: independent variables regression analysis of the abnormal returns.

Independent Variables N Mean Median Sd Min Max

CEO-Worker Gap 187 48.54 39.38 35.01 2.663 174.2

Equity ratio 147 0.524 0.53 0.190 0 0.980

Gender diversity 187 0.770 0.75 0.114 0.455 1

19 Multicollinearity arises when two variables are highly correlated with each other. High correlation is commonly seen as above 0.7. From the correlation matrix it can be seen that there is no threat of multicollinearity. The highest correlation is between CEO-Worker gap and Board size with a coefficient of 0.4506.

Table 3: Correlation Matrix of the used variables.

CEO-Worker gap Equity ratio Gender diversity Board size

CEO-Worker gap 1

Equity ratio -0.2689 1

Gender diversity -0.0882 -0.0479 1

20 5. Empirical Results

In this section the event study results are outlined. Firstly, the significance of the AARs and the CAARs results are highlighted. Secondly, the results of the regression on the abnormal returns, thirdly, the results of the performed robustness checks, and lastly, the hypotheses and the literature results are analysed.

Main results event study

The Jarque-Bera test, shown in Table A1, reveals that the sample is not normally distributed. Therefore the non-parametric test is used in order to determine the statistical significance. In Table 4 below, the main results of the event study are stated and Table 5. The graphical results of the average abnormal returns are displayed in Figure A1.

Table 4: Average abnormal returns in largest event window

Day AAR

P-value (Parametric)

P-value

(non-parametric) Day AAR

P-value (parametric) P-value (non-parametric) -10 -0.02% 0.8727 0.7530 0 0.33% 0.1447 0.3787 -9 -0.30% 0.0397 0.1562 1 -0.30%*** 0.0722 0.0011 -8 0.08% 0.6256 0.6701 2 0.11% 0.4083 0.7989 -7 -0.29%** 0.0576 0.0139 3 0.20% 0.2225 0.1806 -6 0.09% 0.5542 0.3965 4 -0.17% 0.1958 0.5718 -5 -0.25%* 0.0853 0.0772 5 -0.16% 0.3365 0.2631 -4 0.02% 0.9044 0.9885 6 0.02% 0.9184 0.7773 -3 -0.24% 0.2540 0.6137 7 0.02% 0.9043 0.8837 -2 -0.09% 0.6507 0.5994 8 -0.29% 0.0607 0.2125 -1 -0.02% 0.8990 0.8230 9 0.10% 0.4970 0.3141 10 0.43% 0.0187 0.2883

Note: The significance is denoted as *** p<0.01, ** p<0.05, * p<0.1.

Table 4 shows that the returns for day -7, -5 and 1 are statistically different from zero at the 10 per cent significance level. On day -7, -5 and 1 the average abnormal returns are -0.29 per cent, -0.25 per cent and -0.30 per cent respectively. The abnormal returns are all negative for these specific days. The significant negative abnormal returns depict little effect of a significant shareholder dissent on the stock value in economic terms.

21 as statistically insignificant. On day 1 there is a statistically negative effect on the 1 per cent level of -0.30 per cent. This is in line with the hypothesis that shareholder dissent has negative effect on the stock performance. In economic terms a significant shareholder dissent decreases the stock value on day 1 by -0.30 per cent. The effect of -0.30 per cent seems negligible, however it can be argued that it has sufficient economic impact compared to earlier conducted research such as Cai and Walkling (2011) and Ferri and Maber (2013). Cai and Walkling (2011) found a significant positive abnormal returns of 0.50 percent with the introduction of Say-on-Pay. The reason that day -7 and -5 tested statistically significant on the 5 per cent and 10 per cent level respectively, could be due to the earlier released recommendation by proxy advisers. Therefore, the effect of a significant shareholder dissent could be distributed over the event window.

Table 5: Cumulative average abnormal returns for all event windows

Note: The significance is denoted as *** p<0.01, ** p<0.05, * p<0.1.

To further examine whether the effect is distributed over the event window, the CARs were investigated. Table 5 shows the cumulative abnormal returns for three different event windows. Two cumulative abnormal returns are negative, the three day event window (-1,1) is positive but close to zero. However, both the parametric and the non-parametric test yield that the results are not statistically different from zero. For the other event windows the CAARs are negative, however both the parametric and the non-parametric test reveal that the results are not statistically significant. The CAAR for the event window (-10,10) is the most negative. The results of the CARs in the period (-10,10) are graphically displayed in Figure A2. For the event window (-10,10) we observe negative CARs in the whole of the window, with a CAAR of -0.73 per cent. Furthermore, these negative CARs demonstrate that the effect of shareholder dissent on remuneration is distributed over multiple days.

Given that investors may have discussed the play plan informally prior to the annual meeting with the board or have received advice form large proxy advisors such as ISS and Glass Lewis. The 10 days before the event day is not large enough to capture the possible effect. Therefore, the estimation window is moved back by 10 days. In this manner it can be surmised that the significant shareholder has been incorporated into the market prices. The new researched

22 event windows are (-20,1), (-20,20) and (-10,1). The event windows (-20,-1) and (-10,1) have been stopped earlier for the possible information leak. The results of these extended event windows are shown in the appendix of Table A5. All the cumulative average abnormal returns in Table A5 are negative, which is in line with the hypothesis. The event window (-20,20) displays negative CAARs from day -9 onwards. The highest CAAR of -1.19 per cent is found on day 8. The result of the event window (-20,20) is graphically displayed in Figure A3. The parametric tests turn out to be insignificant for all the event windows. On the other hand, after performing the non-parametric tests, the event window (-10,1) is found to be significant at the 10 per cent level. The CAAR in the event window (-10,1) is -0.98 per cent and negative in the whole of the event window. This significant result confirms the suspicion that the results are influenced by a possible information leak. The significant result found maybe due to the shortening of the CARs to day 1. Therefore, it is thought that investors may have discussed the pay plan prior to the annual meeting or have received advise from large proxy advisors.

After examining the AARs and the CARs, a regression analyses was performed on the CARs and the AAR of day 1. The performed regressions measure whether the effect of remuneration shareholder dissent on the stock returns differs depending on the characteristics of governance. The results of this regression can be seen in Table 6 below:

Table 6: abnormal returns regression analysis for the three different event windows and day 1 Variables CAR (-1,1) CAR (-5,5) CAR (-10,10) AR (1) CEO-Worker Gap 3.52e-05 -0.000172 -0.000166 0.000127

(0.000103) (0.000183) (0.000216) (7.72e-05) Equity ratio 0.00806 -0.0218 -0.0272 0.00477 (0.0191) (0.0321) (0.0421) (0.0138) Gender diversity -0.0718* -0.0888 -0.185*** 0.000730 (0.0398) (0.0568) (0.0694) (0.0294) Board size -0.00397 -0.000154 -0.00323 -0.00344* (0.00278) (0.00370) (0.00421) (0.00201) Constant 0.0841* 0.0899 0.194** 0.0200 (0.0455) (0.0633) (0.0773) (0.0360) Observations 147 147 147 147 R-squared 0.039 0.023 0.051 0.039

23 For the three day event window (-1,1) on Table 6, it can be ascertained that the CEO-Worker gap ratio and equity ratio have a positive influence on the CARs. The coefficient of the CEO-Worker gap ratio is not in line with the expectations; however it is rather small and not statistically significant. The coefficient of the equity ratio is positive and thus in line with the expectations. Nonetheless, this coefficient is not significantly different from zero. On the other hand, the variable gender diversity has negative influence on the CARs. This is in line with the expectations since a more gender diverse should lead to lower CARs. Only the coefficient for gender diversity is significantly different from zero at the 10 per cent level. Hence, if gender diversity goes up by 1 unit the CARs decreases by -0.0718. For the eleven day event window (-5,5) none of the coefficients are proven to be significant. In the twenty-one day event window (-10,10), again only gender diversity is found to be statistically significant at the 1 per cent level. Therefore, if the proportion of men divided by board size goes up by 1 unit, the CARs will decrease by -0.185. This coefficient of -0.185 seems economically significant. On day 1, only the variable board size is found to be significant at the 10 per cent level. This illustrates that when board size goes up by 1 person, the abnormal returns will decrease by -0.0034. Similarly, as in the other event windows, this coefficient of board size depicts little effect and thus seems economically insignificant.

Robustness checks

As a robustness check, the percentage of votes withheld on the resolutions in the dataset is regressed on the CARs and ARs of day 1. As described earlier, withholding of shareholders votes is an indication of shareholder discontent. The anticipated effect is that an increased amount of withheld votes has a stronger negative impact on the abnormal returns. The results of the extended regression can be seen in Table A4. Contradictory to expectation, the results reveal that abstained votes have a positive impact on the magnitude of the abnormal returns in the event window (-10,10) and on day 1. This indicates that higher shareholder dissent is positively correlated with the abnormal returns, however the results are tested insignificant. The results reveal that abstained votes have a stronger negative impact on the magnitude of the CARs in the event windows (-1,1) and (-5,5). Nonetheless, these results are tested insignificant as well. Therefore, I cannot conclude that more votes withheld has a stronger negative effect on the magnitude of the CARs and ARs.

24 be more negative for these companies impacted by significant shareholder dissent in consecutive years. The results of this sample split are shown in Table A3. The result is contradictory to expectation. However both the parametric and the non-parametric test are not statistically significant. Therefore, I have to remain inconclusive about the effect of these disappointing boards. This contradictory and inconclusive result is probably due to the small sample size of the recurring significant shareholder dissent in consecutive years.5

Discussion of the results

The first hypothesis predicted a negative impact of significant shareholder dissent on remuneration on the stock performance of listed companies. This should have resulted in negative AARs and CAARs. The AARs on days 7, -5 and 1 tested statistically significant showing negative abnormal returns. In terms of economic significance, it can be argued that the results are negligible. As the days -7 and -5 tested statistically significant and both yield negative abnormal returns, the CARs were also analysed. The most negative abnormal returns are shown in the event window (-10,10), with a CAAR of -0.73 per cent. This result illustrates that the negative effect of a significant shareholder dissent on remuneration is more distributed over the event windows. Nonetheless, all the event windows tested statistically insignificant for the CARs. Therefore, from the main event windows it cannot be concluded that a significant shareholder dissent on remuneration has a negative effect on company performance. Furthermore, it seems that the effect is large enough to have economic significance.

However, due to the suspicion that 10 days prior to the event window is probably not long enough to capture the possible effect, the estimation window was moved back by 10 days. Moreover, some of the event windows were stopped earlier than planned to avoid the possible information leak. The non-parametric test is found to be significant at the 10 per cent level of the event window (-10,1). The event window (-10,1) has a CAAR of -0.98 per cent. This shows that in this event window the average stock returns decreased by almost 1 per cent. Therefore, it can be concluded that a significant shareholder dissent on remuneration has negative effect on company performance. However the results should be interpreted with caution, since the event window (-10,1) is the only event window that tests significantly and only at the 10 per cent level. The negative abnormal returns of almost 1 per cent give every indication of being economically significant.

25 Based on the existing literature, these results are not surprising. Cunat, Gine and Guadalupe (2016) indicated that losing a Say-on-Pay vote can be interpreted as a negative signal for the company, as Say-on-Pay votes can be seen as an explicit vote of management confidence (Cotter, Palmiter and Thomas, 2013; Cunat, Gine and Guadalupe, 2016). As expected, the shareholder will vote against the resolution when they feel that the pay plan does not reflect the value-added to the company. These votes of no-confidence are subsequently harmful to the company. The inconclusive effect found in the main event windows can be due to the fact that investors may have discussed the pay plan prior to the annual board meeting (during informal discussions), or they may have received advice from large proxy advisors such as ISS and Glass Lewis. Consequently, a prematurely cut short estimation window and event windows were considered. In the event window (-10,1) a significant result is found at the 10 per cent level. It is not surprising that this result is only significant at the 10 per cent level. This can be due to the assumption made that shareholder dissent can be seen as an unexpected event. Cunat, Gine and Guadalupe (2016) dealt with this assumption in their research to find a clean causal estimate for the market reaction on Say-on-Pay. Unfortunately, for this research their regression discontinuity design is not compatible with my data set. For further research, a sample around the significant shareholder dissent threshold of 20 per cent could provide a cleaner causal estimate. It can be argued that a significant shareholder dissent is predictable. By comparing the observations on both sides of the threshold, treatment and control groups are generated that are akin in observable and unobservable characteristics. In this manner there is dealt with the infeasible unexpectedness. However, by only comparing the observations close to the threshold of 20 per cent, larger shareholder dissent under the majority standards of 50 per cent are neglected.

The second hypothesis predicted that the stock prices are more sensitive to shareholder dissent when the board room governance is weaker. Gender diversity yield a significant negative coefficient at the 10 per cent level in the three day event window (-1,1). Furthermore, gender diversity is found to be statistically significant at the 1 per cent level in the twenty-one day event window (-10,10). The coefficient of board size is negligible, and, therefore, the effect of board size on the CAR may be seen as economically insignificant. On the other hand, the coefficient of gender diversity appears to be economically significant.

26 according to Gompers, Ishii and Metrick (2003), corporate governance is strongly connected with the stock returns. However, this is not in line with Cai and Walkling (2011), who found that companies with weak boardroom governance and inefficient remuneration were better off with the legislation of Say-on-Pay. They reasoned that these companies are more likely to benefit from changes induced by shareholder pressure. On the other hand, the results are in line with Alkabini, Cuomo and Malin (2019) who found that companies with more women in the remuneration committee are negatively related to shareholder dissent on remuneration. Furthermore, this is in line with Ryan and Wiggens (2004) who found that shareholders have a negative opinion about large boards. To summarise, it has been established that companies with weaker governance, particularly with gender diversity, are more heavily impacted by shareholder dissent on remuneration.

The third hypothesis predicted that the stock prices are more sensitive to shareholder dissent on remuneration when the executive compensation is inefficient. This should have resulted in significant coefficients for the proxies of inefficient compensation. However, both the CEO-Worker gap and the equity ratio turn out to be insignificant. Therefore, my results are inconclusive about the stock price reaction to shareholder dissent on remuneration when the executive compensation is inefficient.

27 6. Conclusion

The current study on Say-on-pay mainly focuses on the value added from Say-on-Pay regulation and neglects the possible effects of shareholder dissent on company performance. I therefore conducted an event study concerning the effect a significant remuneration shareholder dissent has on company performance.

Say-on-Pay rights can be described as the right to vote on the plan of executive remuneration relating to company performance. Shareholders are thus not only voting on the pay plan, but also on the level of confidence the CEO generates and the subsequent value added to the company. Therefore Say-on-Pay votes are considered to be a measure of shareholder dissatisfaction. These negative shareholder votes indicate a misalignment of incentives between management and shareholders and could result in a decrease in share price.

This research found significant negative abnormal returns for day -7, -5 and 1. In terms of earlier conducted research the size of the results seem economically significant (Cai and Walkling, 2011; Ferri and Maber, 2013). As days -7 and -5 test statistically significant and both yield abnormal returns, the CARs were also examined. The most negative abnormal returns are displayed by the event window (-10,10), with a CAAR of -0.73 per cent. This result shows that the negative effect of a significant shareholder dissent on remuneration is more distributed over the event window.

Hereby, triggers the surmise that the effect of a significant shareholder dissent on remuneration has been earlier incorporated in the market price. Given that the investors may have informally discussed the pay plan prior to the annual board meeting or have received advice from large proxy advisors. Therefore additional event windows were researched, with statistically significant results for the event window (-10,1) at the 10 per cent level. The CAAR of -0.98 per cent proves that a significant shareholder dissent on remuneration has a negative effect on the stock price. However this result should be interpreted with caution, since this is the only event window which tests significant. Furthermore, this event window is only tested significant at the 10 per cent level.

28 board room governance is weaker. Gender diversity yield a significant negative coefficient at the 10 per cent level in the three day event window (-1,1). Furthermore, gender diversity is found to be statistically significant at the 5 per cent level in the twenty-one day event window (-10,10). The coefficient of board size is negligible, which means that the effect of board size on the CAR may be seen as economically insignificant. On the other hand, the coefficient of gender diversity seems economically significant.

My research is in contrast with the existing literature on the impact of Say-on-Pay regulation. (Cai and Walkling 2011; Ferri and Maber, 2013). These papers proved that the introduction of Say-on-Pay can be seen as value-creating mechanism. Cai and Walkling (2011) reasoned that increased shareholder pressure should induce better performance for companies. However, my research is in line with more recent literature (Cunat, Gine and Guadalupe, 2016), that briefly touched upon the effect of losing a Say-on-Pay vote. They argued that the significant effect of Say-on-Pay, after losing a Say-on-Pay vote, can be interpreted as a very negative signal.

29 References

Adams, R. B. (2016). Women on boards: The superheroes of tomorrow? The Leadership Quarterly, 27(3), 371–386.

Admati, A. R. & Pfleiderer, P. 2009. The “Wall Street Walk” and shareholder activism: Exit as a form of voice, The Review of Financial Studies, 22, 2645–2685.

Alkalbani N., Cuoma F., Mallin C. (2019), Gender diversity and Say-on-Pay: Evidence from UK remuneration committees, Corporate Governance: An International Review, 27 (3), 378-400 Alissa, W. (2015). Boards' response to shareholders' dissatisfaction: The case of shareholders' say on

pay in the UK. The European Accounting Review, 24(4), 727–752.

Bond, S. and Foley, S. (April 23, 2014) Buffett says Coke employee equity pay plan is ‘excessive’. Retrieved from: https://www.ft.com/content/8d6e7a3e-cb1d-11e3-ba9d-00144feabdc0

Brown, S., Warner, J., 1980. Measuring security price performance. Journal of Financial Economics, 8, 205-258.

Brown, S., Warner, J., 1985. Using daily stock returns. Journal of Financial Economics, 14, 3-31. Cai, J. and Walkling, R. (2011) Shareholders’ Say-on-Pay: Does it create value? Journal of Financial

and Quantitative Analysis, 46, 229–339.

Collins, D., Marquardt, B. and Niu, X., 2019. Equity‐based incentives and shareholder say‐on‐ pay. Journal of Business Finance & Accounting, 46(5-6), pp.739-761.

Conyon, M. and G. Sadler (2010). Shareholder voting and directors’ remuneration report legislation: Say on pay in the UK. Corporate Governance: An International Review, 18 (4), 296–312 Crawford S., Nelson K.K, Rountree B. (2020) Mind the Gap: CEO-Employee Pay Ratios and

Shareholder Say-on-Pay Votes, available at:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3088052

Cunat V., Giné, M. and Guadalupe, M. (2012) The Vote Is Cast: The Effect of Corporate Governance on Shareholder Value. The Journal of Finance, 67 (5), p1943 - 1977

Cunat, V., Giné, M. and Guadalupe, M. (2016) Say pays! Shareholder voice and firm performance. Review of finance, 20 (5), p1799-1834

Dalton, D. R., C. M. Daily, J. Johnson, A. Ellstrand. (1999) Number of directors and financial performance: A meta-analysis. Academic Management Journal, 42(6) 674-686.

Edmans, A. 2009. Blockholder trading, market efficiency, and managerial myopia, The Journal of

Finance, 64, 2481–2513

Ertimur, Y., Ferri, F. and Oesch, D., (2013). Shareholder Votes and Proxy Advisors: Evidence from Say on Pay. Journal of Accounting Research, 51(5), pp.951-996.

Ferri, F. and Maber, D. (2013) Say on pay votes and CEO compensation: Evidence from the UK,

30 Fama, E. and Jensen, M., 1983. Separation of Ownership and Control. The Journal of Law and

Economics, 26(2), pp.301-325.

Jensen, M. C. & Meckling, W. H. (1976) Theory of the firm: Managerial behavior, agency costs and ownership structure, Journal of Financial Economics, 3, 305–360.

Jensen, M. C. and Murphy, K. J. (1990). ‘Performance pay and top-management incentives’. Journal

of Political Economy, 98, 225–64.

Kirsch, A. (2017). The gender composition of corporate boards: A review and research agenda. The

Leadership Quarterly, 29, 346–364

Kolari, J.W., Pynnonen, S., 2011. Nonparametric rank tests for event studies. Journal of Empirical

Finance, 18 (5), p953–971.

Larcker, D. F., Ormazabal, G., and Taylor, D. J. (2011) The market reaction to corporate governance regulation, Journal of Financial Economics, 101, 431–448.

MacKinlay, A. Craig, 1997. ‘Event studies in economics and finance’, Journal of Economic

Literature, 35 (1), p13-39.

Mooney, A. (July 19 2020) A new sell signal? Stocks underperform after pay revolt, says UBS. Retrieved from: https://www.ft.com/content/6c2df0d9-6dd0-47e3-a2fd-ec8dfd89ddba

Ryan, H. E., Jr., R. A. Wiggins III. 2004. Who is in whose pocket? Director compensation, board independence, and barriers to effective monitoring. Journal of Financial Economics, 73 (3) 497-524

Stathopoulos K., Voulgaris G., (2016) The importance of Shareholder Activism: The Case of Say-on-Pay, Corporate Governance: An International Review, 24 (3), 359-370

Thomas, D. (January 6 2020) Top UK chiefs paid 117 times more than the average worker in 2018. Retrieved from: https://www.ft.com/content/9d971c48-2fa5-11ea-a329-0bcf87a328f2

Yermack, D. (1996) Higher Market Valuation of Companies with a Small Board of Directors. Journal of

31 Appendix

Table A1: Jarque-Bera normality test for the cumulative abnormal returns in three different event windows.

Figure A1: average abnormal returns in the event window (-10,10).

Figure A2: cumulative abnormal returns in the event window (-10,10).

-0,40% -0,30% -0,20% -0,10% 0,00% 0,10% 0,20% 0,30% 0,40% 0,50% -15 -10 -5 0 5 10 15 Ave rag e Ab n o rm al Ret u rn s Days

AAR (-10,10)

CAAR (-10,10)

32

Table A2: Cumulative abnormal returns for additional event windows.

Period CAAR P-value (parametric test) P-value (non-parametric test)

(-20,20) -0.98% 0.2525 0.4837

(-20,1) -0.92% 0.2126 0.1295

(-10,1) -0.98% 0.1186 0.0793*

Note: Significance is denoted as *** p<0.01, ** p<0.05, * p<0.1.

Table A3: Robustness check for subsample of reappearance in consecutive years.

(-10,10) (-5,5) (-1,1)

Occasional:

CAAR -0.89% -0.51% -0.04%

P-value (parametric) 0.2740 0.4479 0.9139

P-value (non-parametric) 0.6967 0.5702 0.3595 Recurring in consecutive years:

CAAR 0.16% -0.87% 0.33%

P-value (parametric) 0.9076 0.5818 0.6057

P-value (non-parametric) 0.9575 0.8191 0.8100 Note: Occasional exists of 204 events and recurring in consecutive years exists of 35 events. Significance is denoted as *** p<0.01, ** p<0.05, * p<0.1.

Table A4: Cumulative abnormal returns regression analysis for three different event windows, including the percentage of votes against plus the percentage of votes withheld.

Variables CAR (-1,1) CAR (-5,5) CAR (-10,10) AR (1)

CEO-Worker Gap 3.69e-05 -0.000171 -0.000175 0.000122

(0.000105) (0.000185) (0.000215) (7.68e-05) Equity ratio 0.00800 -0.0218 -0.0268 0.00499 (0.0192) (0.0322) (0.0421) (0.0137) Gender diversity -0.0723* -0.0890 -0.182*** 0.00249 (0.0402) (0.0567) (0.0695) (0.0285) Board size -0.00399 -0.000160 -0.00315 -0.00339* (0.00282) (0.00370) (0.00418) (0.00196) % Votes against + % Votes Withheld -0.00622 -0.00226 0.0317 0.0204

(0.0375) (0.0536) (0.0689) (0.0340)

Constant 0.0868* 0.0908 0.181** 0.0114

(0.0495) (0.0641) (0.0796) (0.0317)

Observations 147 147 147 147

R-squared 0.040 0.023 0.052 0.044

33

Figure A3: Cumulative abnormal returns in the event window (-20,20)

Table A6: Robustness check for subsample of first time events and multiple time events.

CAR (-10,10) (-5,5) (-1,1)

First time events

CAAR 0.27% 0.23% 0.17%

P-value (parametric) 0.7617 0.7555 0.6992

P-value (non-parametric 0.9568 0.5266 0.1779 Multiple time events

CAAR -2.43% -1.89% -0.24%

P-value (parametric) 0.0417** 0.0907* 0.5631 P-value (non-parametric) 0.0858* 0.1278 0.2240

Note: First time events exist of 150 events, companies that thereafter again cope with significant shareholder dissent exists of 89 events. Significance is denoted as *** p<0.01, ** p<0.05, * p<0.1.

Table A5: Robustness check for subsample of the remuneration policy and remuneration report. CAR (-10,10) (-5,5) (-1,1) Remuneration Policy: CAAR -2.26% -1.06% -0.88% P-value (parametric) 0.0719* 0.2065 0.1218 P-value (non-parametric) 0.1233 0.5393 0.1620 Remuneration Report: CAAR -0.11% -0.36% 0.38% P-value (parametric) 0.8966 0.6531 0.3215 P-value (non-parametric) 0.8886 0.1757 0.2224

Note: Remuneration Policy exists of 69 events and Remuneration Report exists of 168 events. Significance is denoted as *** p<0.01, ** p<0.05, * p<0.1.