Did the Introduction of the Code Frijns in 2009 Have An Impact on the Amount of Compensation of Directors of Dutch Listed Companies?
Max Lodewijk Kooij Student number: 10025154
Bachelor Thesis Finance and Organization Under guidance of: M. Koudstaal
Date: 02/02/2014
Index
1 Introduction 1
2 Theoretical Background 3
2.1 Corporate Governance 3
2.2 The Code Frijns 4
2.3 The Sarbanes-Oxley Act 5
2.4 Hypotheses 6 3 Research method 9 3.1 Data gathering 9 3.2 Operationalization of variables 11 3.3 Regression models 13 3.4 Descriptive statistics 15 4 Data analyses 22 4.1 Regression analyses 22
4.2 The ratio between Long and Short Term Variable Compensation 27
5 Conclusions 30
5.1 Limitations 31
References 32
1 Introduction
In the past decade there has been a public debate about the behavior of directors of Dutch companies. In particular, the amount of remuneration of these directors is a point of critique (e.g., NRC, 2004; EenVandaag, 2005 ). The discussion on corporate governance in the Netherlands was triggered by the Ahold accounting scandal in 2003. In contrast to Enron and Worldcom the accounting irregularities at Ahold did not lead to a bankruptcy for the
company in question. However, this fraud revealed that poor corporate governance was not only a problem in the USA. The Ahold scandal was an important trigger for regulatory changes in the Netherlands (Swagerman and Terpstra, 2007).
An important product of these regulatory changes is the Dutch Code Corporate Governance that was introduced in 2004 (hereinafter referred to as “Code Tabaksblat”). The Code Tabaksblat applies to all Dutch listed companies with registered offices in the
Netherlands (Commissie Corporate Governance, 2003). The Code Tabaksblat consists of principles and specific provisions for individuals and parties that are involved with a company. Companies that are subject to the Code Tabaksblat are obliged to state in their annual statements whether they comply with the Code Tabaksblat and if applicable, why and how they deviate from this Code. This principle is called ‘comply or explain’ (Commissie Corporate Governance, 2003).
In 2008 the Code Tabaksblat was evaluated and updated by the Monitoring Commissie Corporate Governance Code. The updated Code Corporate Governance (hereinafter referred to as “Code Frijns”) applies to financial years beginning on or after January 1, 2009. In comparison with the Code Tabaksblat the Code Frijns puts more emphasis on influencing the behavior of the executive board. Furthermore, the supervisory board is expected to fulfill a more active role in determining executive compensation (Staatsblad van het Koninkrijk der Nederlanden, 2009). In accordance to the Code
Tabaksblat, the Code Frijns must be complied with in conformity with the ‘comply or explain principle’ (Commissie Corporate Governance, 2008).
A reason for updating the Code Tabaksblat to the Code Frijns was the changing attitude towards good corporate governance. Moreover, the critique on the amount of the bonuses of directors of Dutch companies was incorporated in the code Frijns as well (Russel Advocaten, 2010). Although the guidelines for the payment of bonuses are tightened, the Code Frijns does not contain mandatory rules on the amount of bonuses and other variable compensation. Hence it is still questionable what the exact influence of the Code Frijns is on
the amount and structure of the remuneration of directors of Dutch listed companies, which is the central theme of this paper.
In their research on the effects of the Code Tabaksblat on CEO remuneration Otten, Schenk & Verwer (2008) cautiously conclude that the introduction of this code of conduct did not have a significant effect on the remuneration of CEO’s in the Netherlands. On the contrary, Swagerman & Terpstra (2009) conclude that the Code Tabaksblat, in combination with the IFRS 2 regulations, has led to a trend towards equity-based pay for executives in the Netherlands. However, they argue that the principles and provisions in the code Frijns might not be economically efficient.
The Monitoring Commissie Corporate Governance Code (2013) concludes in their final document that in general, the accountability for remuneration structures and the remuneration policy in the annual reports is not simple and transparent. However, the commission did not monitor the actual amount and structure of compensation of directors of Dutch listed companies, they only examined the developments that have occurred since the introduction of the code Frijns. Amongst other things, the Monitoring Commissie Corporate Governance Code monitored the compliance with the code Frijns, the accountability towards risk management and the functioning of supervisory boards.
The research of Otten, Schenk & Verwer (2008) and Swagerman & Terpstra (2009) on the Code Tabaksblat examined the amount and structure of compensation of directors of Dutch listed companies. However, these papers did not explore the exact impact of the Code Frijns. Since the guidelines on executive compensation became stricter in the Code Frijns, the Code Frijns potentially should have more influence on the amount and structure of
compensation of directors of Dutch listed companies than the Code Tabaksblat had. Furthermore, whilst monitoring the code Frijns the Monitoring Commissie Corporate
Governance Code (2013) did not monitor the actual amount and structure of compensation of directors of Dutch listed companies. Therefore, my research question is: Did the introduction of the Code Frijns in 2009 have an impact on the amount of compensation of directors of Dutch listed companies?
The structure of this thesis is as follows. The second chapter elaborates on the
theoretical background. First of all, the principle-agent theory will be set forth. Subsequently, the Code Frijns and Sarbanes-Oxley Act are set out. Lastly, the hypotheses will be
substantiated. Chapter three sets forth the research method and explains the variables in more detail. In chapter four the data will be analyzed by means of regression analyzes and charts. Chapter five consists of a summary and concludes.
2 Theoretical background
2.1 Corporate Governance
Corporate governance is the system of rules, processes and practices by which corporations are directed and controlled. In essence, corporate governance is about the relation between the board and the various stakeholders of the company.
Solving principal-agent problems is an important challenge in corporate governance (Jensen & Meckling, 1976). The essence in the principal-agent theory is the separation of ownership and control (Schleifer & Vishney, 1997). Principal-agent theory assumes a conflict of interests between the principal and the agent. Given that an asymmetry of information exists, the principal will try to persuade the agent to act in the principal’s best interest rather than in his own (Conyon, 1997).
Principal-agent problems occur in various relationships based on subordination, such as employer-employee, parliament-government, and doctor-patient. In the context of this thesis the company directors (agents) do have goals that conflict with the interest of the shareholders (principals) of the company (Hall & Liebman, 1998). To solve the conflict of interests, the principal and the agent ideally sign a complete contingent contract that specifies exactly what the agent does in all possible situations. This contract specifies how the profits of the company are allocated between the principal and the agent (Schleifer & Vishney, 1997).
Because most future contingencies cannot be foreseen, it is not possible to compose a complete contingent contract (Schleifer & Vishney, 1997). Therefore, the principle and the agent have to allocate the residual control rights (Grossman & Hart, 1986). Residual control rights are the rights to make decisions in situations that or not described in the contract. When put into the context of this thesis, Schleifer and Vishney (1997) state that directors typically do possess more expertise in managing the company than the shareholders do, together with the existence of information asymmetry, this causes directors typically ending up with the residual control rights. This gives the directors the opportunity to act out of self-interest which can cause an inefficient outcome of the principle-agent problem (Schleifer & Vishney, 1997).
According to Schleifer and Vishney (1997) a superior solution to the principle-agent problem between company directors and shareholders is to compose a highly contingent, long term incentive contract that aligns the interest of the directors with those of the shareholders. Such a contract will consist of a performance measure that is correlated with the quality of
the directors’ decisions and that is verifiable in court. Typically incentive contracts include stock options, share ownership or a treat of dismissal if the income of the firm is low (Jensen and Meckling, 1976; Fama, 1980).
Another solution for principle-agent problems between shareholders and directors is the separation of decision management and decision control (Fama & Jensen, 1983). Decision management is the determination and implementation of company policies. Directors that are concerned with decision control are expected to approve and monitor the determined
company policies. In a one-tier board, decision control and decision management is incorporated in one board. In a two-tier board, decision management is performed by the board of directors and decision control is performed by the supervisory board. Both the one-tier and two-one-tier board structure are used by Dutch companies. The Code Frijns contains specific provisions for companies with a one-tier structure board (Commissie Corporate Governance, 2008).
2.2 The Code Frijns
As stated above, the Code Tabaksblat was replaced by the Code Frijns in 2009. The Code Frijns has a greater focus on the behavior of directors, the supervisory board and
shareholders. Therefore the Code Frijns has a bigger emphasis on the actual effectuation of the tasks of various stakeholders instead of the accountability of those tasks afterwards (Commissie Corporate Governance, 2008).
The Code Frijns contains several significant adjustments on the determination of remuneration of directors. First of all, before determining directors’ remuneration, the supervisory board must analyze the possible outcomes of the variable compensation components and its implications for the remuneration of the directors. On the basis of this scenario analyzes, the supervisory board then determines the level and structure of the directors remuneration (Commissie Corporate Governance, 2008). On top of that the compensation ratios within the company must be taken into account. Besides, the variable part of the directors’ remuneration must be in line with the fixed part of the remuneration. Moreover, the Code Frijns prescribes that the variable part of the remuneration is based on long term incentives. The ‘claw back clause’ gives the supervisory board the possibility to recover variable compensation that is assigned according to incorrect financial information (Commissie Corporate Governance, 2008).
The Code Frijns, in comparison with the Code Tabaksblat, puts a greater emphasis on 4
the report of the remuneration policy. The remuneration report must contain an
understandable and transparent justification of the remuneration policy, wherein the total directors’ compensation is differentiated by the various short and long term components. The remuneration report must explain how the remuneration policy contributes to the realization of the long term goals of the company (Commissie Corporate Governance, 2008).
Furthermore, the remuneration report must contain accountability for the methods used for the evaluation of the performance measures for variable compensation. Last but not least, both the relation between the performance measures and strategy objectives and the relation between the remuneration and director performance must be accounted for.
2.3 The Sarbanes-Oxley Act
The Sarbanes-Oxley Act of 2002, commonly referred to as “SOX”, is a U.S. law on corporate governance. SOX, contains regulations about the corporate responsibility for financial
reports, internal control and prosecution in case of fraud, amongst other things. However, SOX does not include rules on executive compensation and its disclosure (SEC, 2002). Anticipating on the emerging discussion on excessive executive compensation, the Securities and Exchange Commission (SEC) adopted new compensation disclosure
guidelines for U.S. corporations in 2006. These guidelines, the Compensation Disclosure and Analysis (CD&A), are intended to provide investors with a clear and complete overview of the amount and underlying philosophy of director remuneration at U.S. companies (Dalton and Dalton, 2008). The CD&A must describe each element of compensation, the objectives of the company’s compensation programs and what these programs are designed to reward, the components of the compensation benchmark and the role of executive officers in determining their own compensation. Finally, The CD&A must describe the relationship between compensation and the long-term firm performance (SEC, 2006).
Comparatively, the CD&A and the Code Frijns both dictate comprehensive reporting on director remuneration. However, in contrast with the CD&A, the Code Frijns consists of principles and specific provisions on the amount, structure and determination of director remuneration (Commissie Corporate Governance, 2008). More explicitly, this means that unlike the determination of director remuneration at the U.S. benchmark companies, the determination of director remuneration at Dutch companies in the sample must be based on scenario-analyses. Furthermore, compensation ratios within the company must be taken into account, the variable part of the remuneration is ought to be based on long term incentives
and the variable part of the directors’ remuneration must be in line with the fixed part of the remuneration. Finally, the supervisory board of Dutch listed companies have the possibility to recover variable compensation that is assigned according to incorrect financial information (Commissie Corporate Governance, 2008).
2.4 Hypotheses
The Code Frijns states that the compensation policy of Dutch listed firms must foster the interest of these firms in the medium and long term. Whilst determining the structure and amount of director remuneration, the indicators that are relevant for the long term value of the company are taken into account. Furthermore, in contrast with the Code Tabaksblat, the Code Frijns states that firms must justify their performance criteria for variable income and report the relationship between remuneration and firm performance as well (Commissie Corporate Governance, 2008).
In short, the Code Frijns has a strong emphasis on pairing director compensation to both financial and non-financial performance measures. This indicates that there should be a relation between director compensation and firm performance at Dutch listed companies. In contrast, Cohen, Dey and Lys (2007) found that the pay-performance sensitivity declined after the introduction of the SOX-act, using a sample of 1279 U.S firms. However this survey did not take into account the CD&A guidelines that were introduced by the SEC in 2007. Otten, Schenk and Verwer (2008) studied the influence of the introduction of the Code Tabaksblat on the pay-performance sensitivity of Dutch listed firms. They conclude that there was no significant relation between directors’ compensation and firm performance between 1996 and 2005. However, this study only includes the first two years after the introduction of the Code Tabaksblat. Possibly, the effects of the introduction of the code Tabaksblat will become only gradually visible. If so, research on the introduction of the Code Tabaksblat that includes more sample years after the introduction of this code, might have a different conclusion than the paper of Otten, Schenk and Verwer (2008). Furthermore, Duffheus and Kabir (2003) did not find a relationship between total director compensation and various performance measures at Dutch listed companies. However, their research was done before the introduction of the Code Tabaksblat and their dataset only includes data of a one year period.
Various researches on non-Dutch companies found a positive relation between compensation and performance. Jensen & Murphy (1990), Conyon (1997), Hall & Liebman
(1998) and Murphy (1998) all found this positive relationship using the performance measure shareholder return. Moreover, Miller (1995) found a non-linear relationship between CEO compensation and the performance measures shareholder return, return on equity, revenue and profit. However these researches did not take into account the influence of corporate governance regulations on the pay-performance sensitivity.
Although these results are contradictory, I expect a relation between director
compensation and firm performance in my sample. Moreover, albeit that Cohen, Dey and Lys (2007) and Otten, Schenk and Verwer (2008) could not demonstrate that particular corporate governance regulations have a positive impact on pay-performance sensitivity, I expect that the Code Frijns has a positive impact on the relation between director compensation and firm performance at Dutch listed companies. Decisive in this prediction was the strong emphasis on pay-performance sensitivity by the Code Frijns.
H1: As a consequence of the introduction of the Code Frijns there will be a stronger relationship between executive pay and firm performance at Dutch listed firms.
An important motivation for introducing the Code Tabaksblat and the Code Frijns was the emerging discussion on the amount of director compensation in the Netherlands. The Code Frijns does not contain mandatory rules on the amount of bonuses and other variable
compensation. However, the guidelines on the bonus and variable compensation payments as well as the guidelines on the disclosure of director compensation might influence the level of total, bonus and variable compensation at Dutch listed companies.
Conyon (1997) illustrated that the introduction of a remuneration committee in the time span of 1988 until 1993 has a negative relation with the growth in director
compensation. As the Code Frijns prescribes a remuneration committee, this is a relevant finding. Furthermore, Joskow et al. (1993) found that CEO’s of regulated companies earn substantially less than CEO’s of unregulated firms. They set out that besides a potential indirect effect of regulation on compensation, political pressures do constrain CEO compensation.
Cohen, Dey and Lys (2007), using a sample of over 1200 U.S firms, found that overall compensation did not change after the introduction of SOX. However, they found an increase in salary and bonus compensation and a decrease in option compensation. Moreover, Otten, Schenk and Verwer (2008) conclude that the Code Tabaksblat did not have an explicit effect on the compensation of CEO’s in the Netherlands.
Because of the contradictory results of existing research on the relation between 7
regulation and director remuneration it is debatable that the Code Frijns did actually result in lower total, bonus and variable compensation at Dutch listed companies.
H2: The introduction of the Code Frijns resulted in lower overall bonuses for executives at Dutch listed firms.
The Code Frijns in contrast to the Code Tabaksblat prescribes that the variable part of the remuneration is based on long term incentives. Consequently, after the introduction of the Code Frijns a change in the ratio between Short Term and Long Term Variable
Compensation can be expected in a way that there will be relatively more Long Term Variable Compensation compared to Short Term Variable Compensation.
Research from the U.S. by Cohen, Dey and Lys (2007) shows that the proportion of incentive-based compensation of U.S. companies declined after the introduction of SOX. As SOX, in contrast to the Code Frijns, does not prescribes that the variable part of the
remuneration of directors should be based on long term incentives, I still expect to see a growth in the ratio between Short Term and Long Term Variable Compensation after the introduction of the Code Frijns.
H3: The introduction of the Code Frijns resulted in a growth in the ratio between Short Term and Long Term Variable Compensation.
3 Research method
3.1 Data gathering
This research makes use of data on the remuneration of CEO’s and CFO’s of Dutch and U.S. listed companies. The data on remuneration is split into a base salary, short term bonus, long term bonus, options, stocks, post employment benefits and other compensation. The Code Tabaksblat and the Code Frijns apply to all Dutch listed companies with registered offices in the Netherlands (Commissie Corporate Governance, 2009). Therefore, the dataset does only contain Dutch companies that were listed on the AEX-index or the AMX-index during the whole period in scope. Besides, these companies must have registered offices in the Netherlands.
Initially, the dataset composed for this research contained the 35 Dutch companies that were listed on the AEX-index or the AMX-index during the whole period 2005 until 2012. These companies are shown in appendix A.1. As the Code Frijns only applies for companies with registered offices in the Netherlands, Arcelor Mittal has been omitted from the dataset.
As this research examines the additional effects of the Code Frijns compared to the Code Tabaksblat, ideally the dataset contains all years after the Code Tabaksblat, including all years from 2004 until 2012. Besides, it is important to have a sample that is as large as possible. As considerably more companies were listed on AEX and AMX-index in the period from 2005 until 2012 compared to the period 2004 until 2012, the dataset contains Dutch companies that were listed on the AEX-index or AMX-index in the time span of 2005 until 2012.
A difficulty in this research is the fact that the introduction of the Code Frijns coincides with the start of the financial crisis in 2008. If the crisis causes lower firm
performance, this will affect the relation between performance and executive compensation. As explained in detail in section 3.3, I aim to control for the influences of the financial crisis in the performed regression analyses by adding U.S. listed companies to the dataset.
In order to measure the effects of the introduction of the code Frijns on the
relationship between executive pay and firm performance at Dutch listed firms, it is important that the U.S benchmark companies are comparable to the Dutch companies in the dataset. Only if the relation between executive pay and firm performance at both the Dutch and U.S. companies in the sample are affected by the same influences, it is possible to measure the additional effects that the introduction of the code Frijns had on the relation between executive pay and firm performance at Dutch listed companies. The same argumentation
applies for examining the effects of the introduction of the code Frijns on the amount of overall bonuses for executives at Dutch listed firms.
The dataset consists of U.S. firms that are listed on the S&P 500-index, because the required information on the remuneration of the directors of these companies is available in the ExecuComp database. To make sure that the Dutch companies are affected by the same market and industry influences as the U.S companies, these companies are matched on the basis of the Global Industry Classification Standard sub industry. If possible, the companies that were matched on this basis were matched based on firm size as well. Unfortunately, this was not always the case. Finally, like the Dutch companies, the U.S. companies were listed on the S&P 500-index in the time span of 2005 until 2012.
Appendix A.2 shows the U.S. companies that are matched with the Dutch listed companies. The company that has been matched with Arcelor Mittal, that is Applied
Materials, has not been omitted from the dataset because that would be on the expense of the number of observations in the dataset. A low number of observations might affect the
statistical power of the statistical tests performed by the models that are explained in detail in section 3.3. In other words, it is possible that if the number of observations would be larger, the models will observe effects of the introduction of the Code Frijns that are not visible in a smaller dataset. Furthermore, leaving Applied Materials in the dataset whilst omitting Arcelor Mittal, will not have a significant impact on the quality of the benchmark.
The data on the remuneration of CEO’s and CFO’s of Dutch companies, as well as their age and nationality are obtained from the annual reports of these companies. In case of the comparable U.S. listed companies these data are obtained from the ExecuComp database in WRDS. The sum of the variables base salary, short term bonus, long term bonus, options, stocks and other compensation obtained from ExecuComp did not always add up to the ExecuComp variable Total Compensation. This is problematic as the variable total
compensation is formulated as the sum of these variables. To account for this, I introduced a new variable called ‘Delta’. This variable is the difference between the sum of the variables base salary, short term bonus, long term bonus, options, stocks and other compensation less the variable total compensation of each U.S. director.
Furthermore, the ExecuComp database did not consist of a variable “Post Employment Benefits”. As the U.S companies are not subject to the same obligations concerning post employment benefits as the Dutch companies (United States Department of Labor, 2014), it is possible that these kind of payments are included in the variable “Other Compensation”.
Finally, as the remuneration variables in the ExecuComp database were reported in dollars, these variables were converted to euro’s using the exchange rate noted on 31th of December of the relevant year.
The performance measures used for the performance of the selected companies are Shareholder Return (TSR), Return on Equity (ROE), Return on Assets (ROA) and Operating Profit Margin (OPM). In accordance with the researches of Boyd (1994), Miller (1995) and Baker & Hall (2002), Net Sales or Revenue (hereinafter referred to as “Revenue”) is used as a measure of firm size. It represents gross sales and other operating revenue less discounts, returns and allowances. All the performance measures as well as the variable Revenue are obtained from DataStream.
3.2 Operationalization of variables
Both the first and second hypothesis will be tested using a model of Murphy (1998), which will be explained in detail in the next section. The relation between executive pay and firm performance at Dutch listed firms will be tested on the basis of the first hypothesis. The second hypothesis states that the introduction of the Code Frijns resulted in lower bonuses for executives at Dutch listed firms. As stated before, to measure firm performance the
performance measures Shareholder Return (TSR), Return on Equity (ROE), Return on Assets (ROA) and Operating Profit Margin (OPM) will be used.
According to the principle-agent theory, the principal should align the reward of the agent with a variable that is in the principal’s best interest such as shareholder return (o.a. Jensen and Murphy, 1990; Conyon, 1997). Shareholder Return is the sum of the capital gains and dividends on a stock and therefore a good measure of the financial gain of the principal (shareholder). The main drawback in using Shareholder Return as a signal for executive effort is that there are many factors which can influence stock prices that are outside the control of a executive (Conyon, 1997). Furthermore, using Shareholder Return as a performance measures gives executives the incentive to manipulate the short term share price. Nevertheless, Shareholder Return is used in earlier research on pay-performance sensitivity by inter alia, Jensen & Murphy (1990), Miller (1995), Conyon (1997), Hall & Liebman (1998) and Murphy (1998).
Return on Equity (ROE) is calculated by dividing Net income by Shareholder’s Equity. ROE measures how much profit a company generates with the money shareholders have invested. Return on Equity is used in comparable researches by Miller (1995) and
Duffheus & Kabir (2003). Return on Assets (ROA) is a comparable relative performance measure and is calculated by dividing Net Income by Total Assets and shows how effectively a company converts its investable money into net income. ROA in contrast to ROE takes into account how well a company uses its financing by loans. Together ROE and ROA give a clear view on management's effectiveness.
Finally the performance measure Operating Profit Margin (OPM) will be included. OPM is calculated as operating income divided by revenue. It is an indicator the efficiency of a company in its operations. In contrast with ROA and ROE, OPM does not account for the capital that is used to realize the company’s profit. OPM is used as a performance measure by Duffheus and Kabir (2003) and Otten, Schenk & Verwer (2008).
Executive compensation will be measured using the variables Total Compensation (TC), long term variable compensation (LTI) and short term variable compensation (STI). Total compensation is the sum of base salary, short term bonus, long term bonus, options, stocks, post employment benefits, other compensation and the variable ‘Delta’ for U.S. companies. Long term variable compensation consists of long term bonus, options and stocks. STI is equal to the year (short term) bonuses paid to executives. By splitting up executive compensation in the TC, LTI and STI it is possible to display the possible changes in the structure of the remuneration of directors of Dutch listed companies caused by the Code Frijns.
Next to the Code Frijns, the financial crisis and the various performance measures I control for other factors that might influence the amount of director compensation at Dutch listed companies. Like Aggerwal & Samwick (1999) did in their research on the impact of risk on executive compensation, year effects and executive fixed effects on executive compensation are included. The variable “ year effect” controls for the change in average levels of compensation during the sample years (Aggerwal & Samwick, 1999). The executive fixed effect allows for different average levels of compensation across the CEO’s and CFO’s in the sample (Aggerwal & Samwick, 1999).
The reply by Core & Guay (2002) to the paper of Aggerwal & Samwick (1999) stated that the variable firm size was omitted in the model that Aggerwal & Samwick (1999) used for their research on the impact of risk on executive compensation. Also, Boyd (1994) and Baker & Hall (2002) use firm size as a control variable in their research on CEO
compensation. In the underlying research, firm size is measured on the basis of total revenue. Finally the underlying research takes into account the effects of the age and nationality of executives on executive compensation.
3.3 Regression models
To provide an answer to the first and second hypothesis, an adjusted model of Murphy (1998) will be used. The first hypothesis states that the relation between executive pay and firm performance at Dutch listed firms is changed after the introduction of the Code Frijns. The effect of the Code Frijns on the amount of the overall bonuses for executives at Dutch listed firms will be tested based on the second hypothesis. The first model is as follows:
Log (director compensation) = β0 + β1 * (Performance Measure)ij+ β2 * (Code Frijns) + β3 * (Code Frijns) * (Performance Measure)ij + β4 * (crisis) +β5 * Log (Revenue)ij + εij (1)
Coefficient β1 estimates the relation between the compensation and performance of
CEO’s and CFO’s of Dutch and U.S. listed companies. Based on research of Jensen & Murphy (1990), Conyon (1997), Hall & Liebman (1998) and Murphy (1998) I expect to find a positive relation between director pay and firm performance. β3 is an estimate of the
influence of the introduction of the Code Frijns on the pay-performance sensitivity of Dutch listed companies. Therefore, β3 will be used to provide an answer to the first hypothesis. As
stated before, it is expected that the introduction of the Code Frijns has led to a stronger relationship between director pay and firm performance at Dutch listed firms. To avoid inter-correlation between the performance measures the regression will be run for each
performance measure separately.
The coefficient β2 is an estimate of the effects of the introduction of the Code Frijns
on the amount of director compensation. The regression will be run with the logarithm of the variables Total Compensation, Long Term Variable Compensation and Short Term Variable Compensation as a measure for director compensation. According to Murphy (1998) the logarithm of these variables must be taken because they are not normally distributed. As the overall bonus is split up into Long Term Variable Compensation and Short Term Variable Compensation, β2 can be used to provide an answer to the second hypothesis. As stated in
section 2.4, it is expected that the introduction of the Code Frijns resulted in lower bonuses for CEO’s and CFO’s of Dutch listed companies. Furthermore, log revenue is added to the model to account for the effects of firm size on director compensation.
A difficulty in this research is the fact that the introduction of the Code Frijns coincides with the start of the financial crisis in 2008. If the crisis causes lower firm
performance, this will affect the relation between performance and director compensation in 13
the model. Hence, I expect that the influence of the financial crisis on executive
compensation will be captured by the performance measure variables. To check the validity of this assumption the model contains a variable named Crisis that checks for the influence of the financial crisis.
As there was no introduction of an updated code of conduct in the United States in the period from 2009 until 2012, there should not be a significant change in executive
compensation of directors of U.S. companies in this period compared to the years before 2009. Besides, both Dutch and U.S. listed firms were subject to the financial crisis.
Therefore, it is possible to capture the effects of the financial crisis on both Dutch and U.S. listed firms apart from the effects of the Code Frijns on Dutch listed companies. This is done by generating the dummy variable Code Frijns that is one for Dutch listed companies in the years 2009 until 2012, and the dummy variable Crisis that is one after 2008 for both Dutch and U.S. listed firms. Hence, the ‘Crisis’ dummy estimates the effects of the financial crisis on director remuneration.
Subsequently, to prevent omitted variable bias, a robustness check will be performed. This will be done by including additional factors that might influence director compensation at Dutch and U.S. listed companies in the first model. The control variables that are added to this model are year effect, executive fixed effect, age and nationality. Consequently, the second model is as follows:
Log (director compensation) = β0 + β1 * (Performance Measure)ij+ β2 * (Code Frijns) + β3 * (Code Frijns) * (Performance Measure)ij+ β4 * (crisis) +β5 * Log (Revenue)ij+ β6
(Nationality) + β7 (age) +ηi + φi + εij (2)
In this specification, ηi is the executive fixed effect, φi is the year effect and εij is the error
term.
Finally, to analyze the hypothesis that the introduction of the Code Frijns resulted in a change in the ratio between Short Term and Long Term Compensation for directors of Dutch listed companies the following probit regression will be used:
Ratio = β0 + β1 * (Code Frijns) + β2 * (crisis) + φi + εij (3)
In this specification, Ratio is the ratio between Short Term and Long Term Compensation for directors of Dutch listed companies, coefficient β1 is an estimate of the effects of the
introduction of the Code Frijns on this ratio, β2 estimates the effects of the financial crisis on
the same ratio, φi is the year effect and εij is the error term.
3.4 Descriptive statistics
Table 1 provides a summary of the various compensation variables of CEO’s and CFO’s of Dutch listed companies. The number of observations consists of the number of directors that actually got the relevant compensation variable. For example, the variable mean stock shows the mean value of the stocks earned in case the relevant company issued an amount of stock to their directors. Also, the value minimum is not included if a director was not given the relevant compensation. The same applies to the tables 2, 3 and 4.
Table 1 shows that from the total amount of compensation paid out to directors of Dutch listed companies 31 percent consists of base salary, 26 percent consists of stocks, 24 percent consists of short term bonuses, and 7 percent consists of options. The average yearly compensation of directors of the Dutch listed companies in the sample is €1,883,096.
Noteworthy are the big differences between the values indicated as minimum and maximum. This indicates that the amount of compensation received by the various directors varies a lot.
Table 2 provides a summary of the various compensation variables of CEO’s and CFO’s of U.S. listed companies. Worrisome are the large amounts of mean, median, minimum and maximum Delta. This means that for a large amount of compensation of directors of the U.S. companies it is not known in what form this compensation is paid out. Table 1: Summary of compensation variables Dutch listed companies (in €)
Base salary Short term bonus Long term bonus Options Stocks Post employment benefits Other compensation Mean 588,511 490,501 176,076 456,864 724,057 172,288 127,753 Median 513,750 315,000 84,000 313,749 419,261 113,000 42,000 Minimum 103,000 2,187 22,000 - 340,274 - 546,000 - 323,687 2,000 Maximum 1,925,000 4,292,000 2,500,000 3,600,000 7,007,784 1,795,000 2,164,000 Number of Observations 538 489 33 152 364 497 252 15
When comparing the compensation variables of Dutch directors in table 1 with the
compensation variables of U.S. directors in table 2, it is remarkable that it is more common in the Netherlands to pay out a short term bonus. Also, U.S. directors are compensated with options more often. Finally, the amount of various compensation variables is much larger in the U.S. than in the Netherlands. A possible explanation for this is that according to
unreported tables the U.S. firms in the sample are bigger than the Dutch firms.
Table 3 shows that for 82 percent of the directors of Dutch listed companies in the sample a LTI was reported in the annual report. Even more worth mentioning, according to the annual reports, 91 percent of the directors in the sample got a STI. It is remarkable that the minimum LTI paid out was a negative amount.
The payment of a negative amount of compensation is reported ten times across the various compensation variables of all directors in the sample. A possible explanation for reporting Table 2: Summary of compensation variables U.S. listed companies (in €)
Base salary Short term bonus Long term
bonus Options Stocks
Other compensation Delta Mean 650,309 1,227,825 2,287,039 1,680,112 2,799,012 170,415 1,523,067 Median 591,919 752,400 1,109,937 1,098,553 1,351,206 88,618 954,688 Minimum 27,304 5,276 144,178 29,645 30,104 331 -5,484,903 Maximum 3,906,560 18,176,100 7,429,279 12,799,652 24,523,496 4,985,494 27,101,643 Number of Observations 560 189 17 348 370 545 482
Table 3: Summary of variable and total compensation Dutch listed companies (in €)
STI LTI TC Mean 490,501 771,779 1,883,096 Median 315,000 466,052 1,439,685 Minimum 2,187 - 546,000 215,000 Maximum 4,292,000 7,007,784 9,960,000 Number of observations 489 439 538 16
negative remuneration numbers by companies might be that there has been an adjustment on the valuation (in accordance with IFRS) of an earlier awarded compensation.
Table 4 provides a summary of the STI, LTI and TC awarded to directors of U.S. listed companies. Again it is shown that the amount of the various compensation variables of a U.S. directors is much larger than the amount of the various compensation variables of their Dutch colleagues. Remarkably, the highest amount of TC paid out to a director of a U.S. company is more than four times higher than the highest amount of TC paid out to a director of a Dutch company in the sample.
Table 5 summarizes the values of the performances measures Shareholder return, ROE, ROA and OPM of the Dutch listed companies in the sample. The number of observations does not sum up to the number of observations of director compensation of Dutch listed companies in the sample. This is because the relevant information on the performance measures was not always available. The same applies to table 6. The impact of the financial crisis on shareholder return is highly visible. In 2008, at the beginning of the crisis, slightly more than half of the Dutch companies in the sample had a negative
shareholder return. A year later all Dutch companies in the sample had a negative shareholder return. In 2010, the AEX- and AMX-index seemed to recover since only 4 out of 34
companies had a negative shareholder return.
Overall, the Dutch companies in the sample scored positive average performance measures during the sample period. The highest share holder return (187.78%) was made by Binckbank in 2006. Remarkably, the best performances on all performance measures were made before the start of the financial crisis in 2008. In contrast, all minimum performances were made between 2008 and 2010.
Table 4: Summary of variable and total compensation at U.S. listed companies (in €)
STI LTI TC Mean 1,227,825 3,037,641 5,324,172 Median 752,400 1,709,918 3,854,088 Minimum 5,276 32,942 359,677 Maximum 18,176,100 24,523,496 41,321,671 Number of observations 189 513 560 17
Table 5: Summary of performance measures of Dutch listed companies
TSR ROE ROA OPM
Mean 12.56 17.91 6.87 13.89 Median 10.30 15.96 6.34 9.30 Minimum -86.93 -87.66 -34.41 -9.34 Maximum 187.78 128.57 62.35 72.24 Observations 524 526 536 522
Table 6 summarizes the values of the performance measures Shareholder return, ROE, ROA and OPM of the U.S. listed companies in the sample. In accordance with the Dutch companies in the sample, the impact of the financial crisis on the shareholder return of the U.S. companies is highly visible as well. In 2008, 63% of the U.S. companies had a negative shareholder return. A year later, even 91% of the U.S. companies in the sample had a
negative shareholder return. Just like the AEX- and AMX-index in the Netherlands, the S&P 500-index seemed to recover in 2010, because less than 10% of the companies in the sample had a negative shareholder return.
Table 6: Summary of performance measures of U.S. listed companies
TSR ROE ROA OPM Mean 9,47 16,87 6,56 12,15 Median 8,04 15,24 6,62 11,79 Minimum -73,68 -91,93 -56,18 -23,78 Maximum 186,84 82,64 33,80 56,99 Observations 558 496 558 560
Notwithstanding the fact that Dutch companies in the sample scored slightly better average performance measures, the U.S. companies scored positive average performance measures during the sample period. Remarkably, the best and the worst performance on ROA were both made at the beginning of the crisis in 2008. Furthermore, the minimum
performances on TSR, ROE and ROA were made after the start of the financial crisis. Finally, the minimum performances on OPM was made in 2007, just before the start of the crisis.
By making use of charts, an overall view on the development of the various director compensation variables will be provided. The numbers in charts 1 to 4 are indexed, starting in 2005 with 100. Chart 1 shows that the average STI, average LTI and average TC were
fluctuating over the sample period. Whilst average LTI and average TC increased during the sample period, average STI decreased during the same period. After the introduction of the Code Frijns in 2009, al the measures of director compensation experienced a growth. The growth of the LTI from 2009 until 2012 is striking. However, this does not mean that these changes are caused by the Code Frijns.
The enormous spike in the LTI-line in 2010 is most striking. A part of this spike can be explained by the exceptional high amount of stock paid out to the CEO's and CFO's of both Reed Elsevier and Royal Dutch Shell in 2010, which was in total 293% higher than in 2009 and 113% less than in 2011. If the stock paid out to these directors in 2010 was in line with the years 2009 and 2010, the spike would be at around 165 index points. As the firm performance of both Reed Elsevier and Royal Dutch Shell was better in 2011 than it was in 2010, the explanation for the high amount of stock paid CEO's and CFO's of both Reed Elsevier and Royal Dutch Shell in 2010 cannot be found in firm performance. A possible explanation for this payment might be that these directors are somehow compensated for the relatively low stock payments in the years before 2009.
Chart 2 compares the development of the average STI, average LTI and average TC with the indexed course of the AEX-index. It is remarkable that the AEX-index progresses along the development of the average STI, average LTI and average TC variables until 2011. This indicates that the remuneration of directors of Dutch listed companies is dependent on the course of the AEX-index. Furthermore, in comparison with the various compensation variables, the steep fall of the AEX- index in 2009 indicates that the development of the various compensation variables are more stable then the course of the AEX stock exchange.
60 80 100 120 140 160 180 200 2005 2006 2007 2008 2009 2010 2011 2012 Chart 1: development of STI, LTI and TC at Dutch listed companies
Average STI Average LTI Average TC Crisis Frijns
Finally, in chart 3 the average STI, average LTI, and average TC are compared with the average yearly Costumer Price Index (CPI) in the Netherlands. The trend of the CPI line does not seem to be in line with the trend of the average STI and LTI lines. However, the TC line fluctuates around the CPI line and ends almost at the same point in 2012. This chart indicates that the variable remuneration of directors of Dutch listed companies is not dependent on the average CPI in the Netherlands. However, Total Compensation might depend on the CPI.
Although both the Netherlands and the U.S. were struck by the financial crisis, only Dutch listed companies were subject to the Code Frijns. Therefore, when comparing average Total Variable Compensation of the directors of the Dutch and U.S. listed companies in the sample, which is the sum of STI and LTI, I expect to see that the financial crisis has a negative impact on the average Total Variable Compensation of both the Dutch and U.S. companies. Moreover, in accordance with hypothesis 1, I expect that the introduction of the Code Frijns in 2009 in the Netherlands has an additional influence on the average Total
60 80 100 120 140 160 180 200 2005 2006 2007 2008 2009 2010 2011 2012
Chart 2: development of STI, LTI, TC at Dutch listed companies compared with AEX index
Average STI Average LTI Average TC AEX-index Crisis Frijns 60 80 100 120 140 160 180 200 2005 2006 2007 2008 2009 2010 2011 2012 Chart 3: development of STI, LTI, TC at Dutch listed companies compared
with CPI Average STI Average LTI Average TC CPI Frijns Crisis 20
Variable Compensation of directors of the Dutch listed companies compared to their U.S. colleagues.
Chart 4 shows the indexed total Variable Compensation of directors of the Dutch and U.S. listed companies in the sample. It shows that the Total Variable compensation of both the Dutch and U.S. companies are very fluctuating. However, they do not seem to move together. Furthermore, the chart contains the indexed course of the AEX- and S&P 500-index. The level of the S&P 500-index is growing during the whole period 2008 – 2012. In contrast, the level of the AEX-index still has ups and downs during this period. Taken together, when comparing the effects of the financial crisis and the Code Frijns on the Total Variable Compensation of directors of the Dutch and U.S. listed companies, there seems to be no additional effect of the Code Frijns on the Total Variable Compensation of directors of the Dutch listed companies visible.
40 60 80 100 120 140 160 180 2008 2009 2010 2011 2012
Chart 4: Total variable compensation and the financial crisis
Indexed total variable compensation NL Indexed total variable compensation U.S. AEX index
S&P 500 index Crisis Frijns
4 Data analyses
4.1 Regression analyses
Although the charts in chapter 3 provide an impression of the development of the displayed variables, they do not show the underlying causalities that caused these developments. Using Total Compensation as the measure for director compensation in model 1, I will test the first hypothesis that states that the relation between executive pay and firm performance at Dutch listed firms is changed after the introduction of the Code Frijns. Thereafter, I will test the second hypothesis, that states that the introduction of the Code Frijns resulted in lower bonuses for executives at Dutch listed firms, using Short Term Variable Compensation and Long Term variable Compensation as measures for director compensation in the first model. Table 7 summarizes the results of model 1 performed with the compensation variable Total Variable Compensation (TC). In this analysis, all the performance measures have a significant positive relation with the TC of directors of Dutch and U.S. listed companies. Furthermore, in all of the four regressions displayed in table 7, the Log(revenue) has a
significant positive impact on the Total Compensation of these directors. It is remarkable that the financial crisis had a significant positive impact on the TC of the directors in the sample. β3, the coefficient that is used to test the first hypothesis, indicates that introduction of the
Code Frijns did not lead to a significant stronger relation between firm performance and the total compensation of directors of Dutch listed companies.
Table 8 shows the results of model 1 performed with the compensation variable Short Term Variable Compensation (STI). The performance measures TSR, ROA and OPM have a significant positive relation with the STI of directors of the Dutch and U.S. listed companies
Table 7: TC Performance measure
TSR ROE ROA OPM
Performance 0.00214 *** 0.00490 *** 0.01510 *** 0.00819 *** Frijns (β2) -.71600 *** -.78500 *** -.72200 *** -.73800 *** Crisis 0.43000 *** 0.47900 *** 0.44200 *** 0.42100 *** Frijns*Performance (β3) -.00025 0.00220 0.00401 0.00045 Logrevenue 0.33500 *** 0.32500 *** 0.32300 *** 0.35200 *** Constant 9.31800 *** 9.40200 *** 9.43400 *** 8.98700 *** R-squared 0.482 0.503 0.492 0.484
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
in the sample. Furthermore, firm size (Logrevenue) and the variable Crisis have a significant positive relation with the STI of the directors in the sample. β2, the coefficient that is used to
test the second hypothesis, is significant and negative for all the performance measures. This means that the introduction of the Code Frijns has led to lower STI for directors of Dutch listed companies.
Table 9 shows the results of model 1 performed with the compensation variable Long Term Variable Compensation (LTI).
Here the performance measures ROE, ROA and OPM have a significant positive relation with the LTI of directors of Dutch and U.S. listed companies. Both firm size (Logrevenue) and the financial crisis have a significant positive impact on the LTI of directors in the sample. Because of the introduction of the Code Frijns, the relation between ROE and the LTI of directors of Dutch listed companies, as well as the relation between ROA and the LTI of these directors, became significantly stronger. Furthermore, β2, the coefficient that is used
to test the second hypothesis, indicates the introduction of the Code Frijns has led to lower
Table 8: STI Performance measure
TSR ROE ROA OPM
Performance 0.00442 *** 0.00426 0.01390 *** 0.00839 ** Frijns (β2) -.40100 *** -.40800 *** -0.50300 *** -.41500 *** Crisis 0.28000 ** 0.25600 * 0.25100 ** 0.24200 * Frijns*Performance (β3) -.000588 -.00124 0.0142 -.00459 Log(revenue) 0.42600 *** 0.41800 *** 0.40200 *** 0.43300 *** Constant 6.16600 *** 6.28700 *** 6.53500 *** 6.04600 *** R-squared 0.443 0.44200 0.42400 0.414
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
Table 9: LTI Performance measure
TSR ROE ROA OPM
Performance 0.00145 0.00626 ** 0.01510 *** 0.01810 *** Frijns (β2) -.26300 *** -.45100 *** -.39300 *** -.18100 *** Crisis 0.60900 *** 0.67000 *** 0.62400 *** 0.59800 *** Frijns*Performance (β3) 0.00190 0.010900 ** 0.03590 *** -0.00800 Logrevenue 0.37800 *** 0.36700 *** 0.35300 *** 0.38000 *** Constant 7.75900 *** 7.81900 *** 8.07500 *** 7.54500 *** R-squared 0.404 0.435 0.424 0.413
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
STI for directors of Dutch listed companies.
In conclusion, the outcome of model 1 suggests that the introduction of the Code Frijns did not lead to a stronger relationship between executive pay and firm performance at Dutch listed companies. Besides, the results of this regression analyses show that the
introduction of the Code Frijns has led to lower short and long term bonuses for executives of Dutch listed companies. As stated before, a robustness check will be done using the second model. Table 10, 11 and 12 summarize the results of the second model performed with the various performance measures and compensation variables. As the year effect consists of 7 dummy variables, these variables are omitted from the tables to keep them readable.
Table 10 summarizes the results of model 2 performed with the compensation variable TC. Most of the control variables, including Crisis and Logrevenue, have a significant effect on total compensation. All the performance measures have a significant positive relation with the TC of directors of Dutch and U.S. listed companies.
In accordance with model 1, in model 2 β3, the coefficient that is used to test the first
hypothesis, indicates that introduction of the Code Frijns did not lead to a significant stronger relation between firm performance and the Total Compensation of directors of Dutch listed companies.
Table 11 shows the results of model 2 performed with the compensation variable STI. It is remarkable that almost all added variables have a significant influence on STI. As a consequence, the variable Crisis does not have a significant influence on STI anymore. This is consistent with the assumption that the effects of the financial crisis are included in the Table 10: TC using model 2 Performance measure
TSR ROE ROA OPM
Performance 0.00189 ** 0.00511 *** 0.01480 *** 0.00721 *** Frijns (β2) -.69700 *** -.77400 *** -.69900 *** -.72900 *** Crisis 0.50800 *** 0.57400 *** 0.49500*** 0.45200 *** Frijns*Performance (β3) -.00019 0.00220 0.00327 0.00130 logrevenue 0.31900 *** 0.30400 *** 0.30500 *** 0.33300 *** Executive Fixed Effect -.24700 *** -.28400 *** -.23700 *** -.25200 *** Nationality 0.02840 0.05370 0.03980 0.04900 Age 0.01750 *** 0.01660 *** 0.01760 *** 0.01670 *** Constant 8.62600 *** 8.78500 *** 8.73400 *** 8.37900 ***
R-squared 0.503 0.529 0.515 0.506
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
various performance measures. Still, STI is significantly positive related to firm size. In contrast with the model displayed in table 8, now only the performance measures TSR and ROA have a significant positive relation with the STI of directors of the Dutch and U.S. listed companies. In accordance with table 8, β2, the coefficient that is used to test the second
hypothesis, indicates that the introduction of the Code Frijns has led to lower STI for directors of Dutch listed companies.
The results of model 2 performed with the compensation variable LTI are summarized in table 12.
Table 11: STI using model 2 Performance measure
TSR ROE ROA OPM
Performance 0.00458 *** 0.00397 0.01040 ** 0.00540 Frijns (β2) -.45900 *** -.50200 *** -.54900 *** -.49500 ***
Crisis 0.08190 -.01670 0.01620 -.02580
Frijns*Performance (β3) -.00034 0.00036 0.01300 -.00195 Logrevenue 0.39400 *** 0.38200 *** 0.37700 *** 0.40200 *** Executive Fixed Effect -0.17700 -.23000 ** -.25300 ** -.28200 ** Nationality 0.34600 *** 0.34700 *** 0.27300 ** 0.27400 ** Age 0.01550 *** 0.01560 *** 0.01460 ** 0.01370 ** Constant 6.04400 *** 6.23200 *** 6.37900 *** 6.06500 ***
R-squared 0.466 0.469 0.447 0.439
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
Table 12: LTI using model 2 Performance measure
TSR ROE ROA OPM
Performance 0.00172 0.00730*** 0.01670 *** 0.01750 *** Frijns (β2) -.17700 *** -.38100 *** -.30600 *** -.14100 *** Crisis 0.77000 *** 0.80700*** 0.75600 *** 0.70000 *** Frijns*Performance (β3) 0.00170 0.01080 ** 0.03420 *** -.00479 Logrevenue 0.36100 *** 0.34400 *** 0.33100 *** 0.36100 *** Executive Fixed Effect -.21200 * -.26400 ** -.19600 * -.22200 ** Nationality -.14300 -.10800 -.11400 -.09400 Age 0.02300 *** 0.02040 *** 0.02360 *** 0.02220 *** Constant 6.69200 *** 6.97400 *** 7.05700 *** 6.58900 ***
R-squared 0.423 0.453 0.444 0.431
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
Notwithstanding the fact that most of the added control variables are significantly related to LTI, the performance measures ROE, ROA and OPM do still have a significant positive relation with the LTI of directors of Dutch and U.S. listed companies. Also, firm size and the financial crisis have a significant positive effect on the long term variable compensation of these directors. More importantly, β2 has a significant negative relation with all the
performance measures. This indicates that the Code Frijns led to lower LTI for directors of Dutch listed companies
The R2 is the coefficient of determination andmeasures the proportion of the variance in the dependent variable that is explained by the independent variables. The regression analyses in this research have a R2 between 0.404 and 0.529. Moreover, the R2 of the
extensive models were greater than those of the initial models. Although the R2’s are medium to weak, the significant coefficients are still a good indicator of the mean change in the compensation variable caused by a unit of change in the relevant independent variable while holding the other independent variables constant.
To check for multi-collinearity, the Variance Inflation Factor (VIF) will be used as a measure of the degree of multi-collinearity between the various independent variables. This is important because as a result of collinearity, individual independent variables may not give valid estimations. However, within the sample data, multi-collinearity does not reduce the predictive power of the model as a whole. O'Brien (2007) states that both 5 and 10 are commonly used by researchers as a sign of severe multi-collinearity.
The fact that according to unreported tables the highest VIF of each independent variable predicted by model 1 was 2,8 indicates that there is no severe multi-collinearity between the various independent variables predicted by this model. However, the variable Crisis predicted by model 2 using STI as measure of director compensation has a VIF between 5 and 6 using all the performance measures separately. Therefore, it is possible that in these regressions the variable Crisis is not a reliable estimator of STI on itself. Finally, as the other VIF measures in this model are below 5, it seems that there is no severe multi-collinearity between the other independent variables predicted by model 2.
Based on the regression analyzes the first hypothesis is rejected. In both model 1 and 2, I did not find a significant stronger relationship between total executive compensation and firm performance at Dutch listed companies caused by the Code Frijns. However, 2
regressions found a significant stronger relationship between Long Term Variable
Compensation and firm performance at Dutch listed companies caused by the Code Frijns. Both model 1 and 2 found this stronger relation using the performance measures ROE and
ROA. To conclude, the introduction of the Code Frijns did not lead to a stronger relationship between executive pay and firm performance at Dutch listed firms.
Since both the first and second model found a significant negative relation between the introduction of the Code Frijns and the various compensation variables, it can be concluded that the introduction of the Code Frijns led to lower Short Term Variable Compensation, and Long Term Variable Compensation. Therefore the second hypothesis, that is: The introduction of the Code Frijns resulted in lower bonuses for executives at Dutch listed firms, is accepted.
As a consequence of adding U.S. listed companies to the dataset to check for the effects of the financial crisis on director remuneration, this research did examine the relation between director compensation and firm performance on a sample consisting of both Dutch and U.S. listed companies. To examine the relation between pay and performance for directors of Dutch listed companies, model 2 is run on the Total Compensation of the Dutch directors in the sample only. Table 13 summarizes the results of this regression. The
performance measures TSR, ROE and ROA have a significant positive relation with the Total Compensation of directors of Dutch listed companies. This indicates that there is a positive relation between pay and performance of CEO's and CFO's of Dutch listed companies. Finally, in accordance with table 10, table 13 shows that β3, the coefficient that is used to test
the first hypothesis, indicates that introduction of the Code Frijns did not lead to a significant stronger relation between firm performance and the Total Compensation of directors of Dutch listed companies.
4.2 The ratio between Long and Short Term Variable Compensation
The third hypothesis states that the introduction of the Code Frijns resulted in a change in the ratio between Long Term and Short Term Variable Compensation of directors of Dutch listed companies. This change is expected because the Code Frijns in contrast to the Code
Tabaksblat prescribes that the variable part of the remuneration is based on long term incentives.
Table 13: TC with only Dutch companies Performance measure
TSR ROE ROA OPM
Performance 0.00201** 0.00706*** 0.01440*** 0.00325 Frijns*Performance (β3) 0.00146 0.00116 0.00587 0.00339
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
Chart 5 shows the development of the ratio between Long and Short Term Variable Compensation of the directors of the Dutch listed companies in the sample, which is calculated by dividing Long Term Variable Compensation by the sum of Short and Long Term Variable Compensation. As the ratio grew from 0.54 to 0.64 during the sample period, the chart shows a trend towards relatively more Long Term Variable Compensation. In the years after the introduction of the Code Frijns, the ratio grew from 0.59 to 0.64. The growth of 8.5% indicates that the introduction of the Code Frijns resulted in a change in the ratio between Short Term and Long Term Compensation of directors of Dutch listed companies.
Chart 6 compares the development of the ratio between Long and Short Term Variable Compensation of the directors of the Dutch and U.S. companies in the sample. Surprisingly, between 2005 and 2007 there has been an enormous growth in the ratio between the Long and Short term Variable Compensation of directors of U.S. companies. A possible explanation for this is that the variable Stocks is considerably less times reported in 2005 and 2006 than in the subsequent years. Presumably, a large the amount of stock paid out to directors of the U.S firms in the dataset is reported in the Delta variable and therefore not
0,4 0,5 0,6 0,7
2005 2006 2007 2008 2009 2010 2011 2012
Chart 5: The ratio between Long and Short Term Variable Compensation Netherlands Frijns 0,4 0,5 0,6 0,7 0,8 0,9 1 2005 2006 2007 2008 2009 2010 2011 2012
Chart 6: A comparison of the ratio between Long and Short Term Variable Compensation at Dutch and U.S companies
Netherlands United states Frijns
visible in the Long Term Variable Compensation. As a consequence, the ratio between Long and Short Term Variable Compensation is significantly smaller in 2005 and 2006 than in the subsequent years. After 2007, the ratio between Long and Short Term Variable
Compensation of directors of U.S. listed companies remains stable.
A possible explanation for the difference in the ratio between Long and Short Term Variable Compensation paid out to U.S. and Dutch directors can be found in the difference in corporate governance systems between these countries. In contrast with the Netherlands, in the U.S. corporate governance system the markets are to the utmost extent responsible for corporate control (Conyon, 1999). This might explain why CEO's from the U.S. urn a higher fraction of their remuneration in the form of stocks and options than Dutch CEO's do
(Fernandes, Ferreira, Matos and Murphy, 2013). Because after all, the valuation of firm performance by the market is reflected in stock prices.
Although the charts 5 and 6 show that the ratio between the Long and Short Term Compensation of directors of Dutch listed companies is growing, they do not show whether this growth is a result of the introduction of the Code Frijns. Therefore, to test the third hypothesis, model 3 has been run. The results are summarized in table 14
.
Table 14 shows that β1, the coefficient that is used to provide an answer to the third
hypothesis is positive. However the positive relation between the introduction of the Code Frijns and the ratio between Long and Short Term Variable Compensation of directors of Dutch companies is not significant. Furthermore, the variable Crisis is positive and
insignificant is as well. To conclude, the hypothesis that the introduction of the Code Frijns resulted in a change in the ratio between Short Term and Long Term Compensation is rejected.
Table 14: Ratio between Long and Short Term Variable Compensation of directors of Dutch companies
Ratio
Frijns ( β1) 0.176
Crisis 0.199
Constant 0.674***
*** Coefficient is significant with p<0.01, ** Coefficient is significant with p<0.05, * Coefficient is significant with p<0.1
5 Conclusions
In the past decade the behavior and remuneration of directors of Dutch companies has been a point of critique. Responding to this public debate, the Code Tabaksblat was introduced in 2004. The Code Tabaksblat consists of principles and specific provisions for individuals and parties that are involved with a Dutch listed company (Commissie Corporate Governance, 2003). In 2009 the Code Tabaksblat was evaluated and updated to the Code Frijns.
Because of its strong emphasis on pay-performance sensitivity it was expected that the Code Frijns had a positive impact on the relation between director compensation and firm performance at Dutch listed companies. Notwithstanding the conflicting results of the researches of Joskow et al. (1993), Conyon (1997) , Cohen, Dey and Lys (2007) and Otten, Schenk and Verwer (2008), it was expected that the introduction of the Code Frijns resulted in lower bonuses for executives at Dutch listed firms. Finally, as the Code Frijns in contrast to the Code Tabaksblat prescribes that the variable part of the remuneration must be based on long term incentives (Commissie Corporate Governance, 2008), the Code Frijns was
expected to result in a change in the ratio between Short and Long Term Variable Compensation.
First, the development of the various compensation variables was examined using charts. I found a growth of all the measures of director compensation after the introduction of the Code Frijns in 2009. Moreover, I found that the AEX-index progresses along the
development of these variables until 2011 whilst the trend of the Costumer Pride Index seems to be in opposite of these variables. Finally, whilst comparing the effects of the financial crisis and the Code Frijns on the Total Variable Compensation of directors of the Dutch and U.S. listed companies, no additional effect of the Code Frijns on the total variable
compensation of directors of the Dutch listed companies was visible.
Furthermore, making use of models that are based on a model of Murphy (1998), I found that the introduction of the Code Frijns did not lead to a significant stronger
relationship between executive pay and firm performance at Dutch listed firms. Furthermore, the model showed that the introduction of the Code Frijns led to lower Short Term Variable Compensation, Long Term Variable Compensation and Total Compensation.
Finally, in the years after the introduction of the Code Frijns, the ratio between Short Term and Long Term Variable Compensation grew with 8.5%. Making use of model 3, I found that there is no significant relation between the introduction of the Code Frijns and the ratio between Long and Short Term Variable Compensation of directors of Dutch companies.
