The effect of executive remuneration
structure on firm performance: An
empirical research for Dutch firms
Abstract
This paper researches the effect of long-term variable remuneration for executives on shareholders’ value. Given that the common performance measure for conditional LTI grants is a proxy for shareholders’ value, it is expected that there is a positive relationship between conditional LTI grants and shareholders’ value. Furthermore it is expected that this
effect is stronger for CEO’s than for CFO’s. Results confirm both expectations. For CEO’s there is a positive significant relationship and for CFO’s there is a positive relationship but
not significant. Therefore the conclusion is that long-term variable remuneration is justifiable for CEO’s, but variable remuneration for CFO’s needs modification before it can
be justified. This conclusion contributes to the current debate in the Netherlands where variable remuneration is highly controversial.
Name:
Maxim van Acker
Student number:
10003633
Programme:
MSc BE Finance
Document:
Master Thesis
Date:
6
thJuly 2015
2 Statement of Originality
This document is written by Student Maxim van Acker who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of
3 Table of Contents 1 INTRODUCTION ... 5 2 LITERATURE REVIEW ... 6 2.1 Theories ... 7 2.2 Relevant Definitions ...11 2.3 Predictions ...12 3 METHODOLOGY... 13 3.1 Data collection ...13 3.2 Model ...15
4 DATA AND DESCRIPTIVE STATISTICS ... 16
4.1 Summary statistics per year ...16
4.2 Regression Samples ...20 5 RESULTS ... 22 5.1 Regression results ...22 5.1.1 CEO Sample ... 22 5.1.2 CFO Sample ... 24 5.1.3 Implications ... 28 5.2 Mean comparison ...29
6 ROBUSTNESS CHECKS AND ADDITIONAL RESULTS... 31
6.1.1 CEO sample ... 32 6.1.2 CFO sample ... 34 7 CONCLUSION ... 36 8 REFERENCES ... 38 9 APPENDIX ... 40 9.1 Relative TSR ...40
4
9.2 Sectors ...41
9.3 Datastream codes ...42
9.4 Robust standard error regressions ...43
5
1 Introduction
Since the Dutch corporate governance code was enacted in 2003, firms listed in the Netherlands with a balance sheet total larger than €500 million were forced to give detailed information on the remuneration of their top executives in the annual report (Corporate Governance Code, 2008). Executive remuneration in the Netherlands is a sensitive topic. The recent salary increase announcement for the ABN-AMRO top executives resulted in a lot of criticism and the increase was finally cancelled. Variable remuneration is even more a cause of incomprehension among the public. Since the corporate governance code was enacted remuneration policies are transparent and this causes a political and social debate. Boot states that this is a cultural difference with the United States where executive remuneration is far less a point of discussion and wealth is respected and admired (FD, 13-04-2015).
These recent events in the Netherlands make the optimal remuneration puzzle interesting once again and the detailed information that is now available for the Netherlands makes a new approach for solving this puzzle possible. This new approach focuses on the Long-Term incentive and the fact that the mean level of conditional long-term incentive grants was 37% shows the relevance of this focus. For 20% of the observations in this research this percentage was higher than 50%. Can the high levels of variable remuneration be justified by proving a positive effect on shareholders’ value?
The new approach in this paper tries to eliminate simultaneous causality problems related to the effect of variable remuneration on firm performance. Previous papers, like the paper by Jensen and Murphy (2010), suggest that variable equity-based compensation gives better incentives to executives compared to fixed cash-based compensation regarding firm-value. Their research was based on a US dataset and the results in this dataset are significant. Their results however lack the possibility for causal interpretation. In addition to this Murphy (1999) states that the most important and undiscovered area of research is on CEO incentives and subsequent company performance. Is there evidence that executives understand the effects of their actions on the share price? The results of this paper, using the new approach on a hand collected Dutch dataset, might be a big contribution in this recent debate. Since there are no databases available with the data necessary for this approach , the results will be a contribution to the scientific
6 literature that can be used in the debate. The current debate in the Netherlands lacks empirical results to support opinions. The research question in this paper therefore is: What is the effect of conditional LTI grants on shareholders’ value for Dutch firms?
There are two recent developments that make this focus even more interesting. First of all the executive compensation structure in the Netherlands is converging towards US
compensation structures. Is this convergence a positive development given the differences between the US and the Netherlands? Regulation, public opinion and characteristics of the compensation packages differ among the two, so prior research results based on US data might not be applicable for the Netherlands.
Secondly a new law regarding compensation for financial firm executives was enacted on the 7th of February 2015. The law was proposed on the 14th of June 2014. This law puts a 20% limit of the fixed base salary on variable compensation. This bonus legislation is the strictest legislation in Europe and in the USA the $500.000 cap is in practice less severe (Staatsblad, Wet van 28 januari 2015 beloningsbeleid financiële instellingen). The question is what the effect on the Dutch financial sector will be. And how will this affect the compensation structure?
Unfortunately there is no data available to actually observe these differences since the
enactments was in the year this research was done. Some companies however already adjusted their policies and therefore this research will form a basis for further research on this specific area.
In section 2 of this paper the relevant literature is discussed. In addition to this relevant definitions are explained and predictions are made. In section 3 the data collection and methodology is explained and in section 4 summary statistics for the subsamples are given. Section 5 discusses the regression results and interpretation and section 6 gives additional results and robustness checks. Finally the conclusion is presented in section 7.
2 Literature Review
In the first part of this section relevant literature is discussed. After the discussion of literature definitions of relevant terms are explained and predictions are made.
7 2.1 Theories
The relationship between CEO compensation and firm performance has been researched quite extensively. Finkelstein and Hambrick (1988) suggest three main channels through which executive compensation affects firm performance. They suggest that performance is affected indirectly. The CEO’s decision to join or stay is dependent of the compensation package and thus the firm’s ability to attract qualified executives is affected by compensation. This first channel is particularly important given the value of outside directors as stated by Hermalin (2005).
Hermalin states that appointing a new outside CEO has more uncertainty than appointing a new CEO from inside the organization. This is reasonable given that the insider is already monitored for some time and his capabilities are known. Comparing this situation to a financial option on some underlying value helps to understand why firms are hiring outside (foreign) CEO’s more frequently in the Netherlands. Hermalin’s explanation is that, as with a financial option, value increases when uncertainty increases. Therefore external CEO hires have more value for firms. Subsequently the external hire knows that he will be monitored extensively to determine his actual value. These events put upward pressure on the level of compensation and might result in better performance. In addition to this Conyon and Peck (1998) find that top management pay and company performance are more aligned if there is a large proportion of outside directors on a main board. Their results are based on empirical data for firms in the U.K. and confirm earlier results from USA. These results follow from the application of agency theory by Tosi and Gomez-Majia (1989). They argue that principles (owners) monitor agents (managers) and design
incentives to align interests of owners and managers. Since it is in the interest of owners to maximize performance, better alignment has a positive impact on performance. Higher levels of monitoring forces the agent to align to the interest of the principal and this subsequently leads to alignment of interests, which is maximizing performance (1989). If there is a lack of
monitoring, the agent might try to increase firm size since this is in his interest. This phenomenon is known as empire building.
Secondly the type of strategic decisions made by executives is affected by their
compensation (Finkelstein & Hambrick, 1988). Rappaport finds that research and development (R&D) expenditure is positively correlated with long-term conditional pay (1978). Since the focus
8 in this research is on conditional long-term incentives it is particularly interesting to extend this result to a correlation between long-term incentives and performance. Dechow and Sloan (1991) enforce the results suggested by Rappaport. They empirically show that R&D expenditure significantly decreases during the final years of a CEO tenure. They conclude that CEO’s try to boost short-term performance in their final years.
The third effect is the reaction of stakeholders, other top managers and regulatory
institutions. This last effect recently became clear in the Netherlands where salary increases and variable remuneration resulted in negative publicity and angry stakeholders. Gabaix and Landier (2006) find that CEO pay in 2003 is a sixfold of the pay level in 1980. They argue that this sharp increase is justified because market capitalization increased with the same pace in this period. Despite this argument, the public opinion in the Netherlands is against these high pay levels and firms can be affected by this.
The relationship between CEO compensation and firm performance is empirically tested by Jensen and Murphy (1990). They find a positive relationship between CEO compensation and firm performance. However, the relationship they find is weak. They suggest that there might be a stronger relationship between equity-based compensation and performance. This suggestion implies that not only the overall level of compensation matters. Indeed it seems that the structure of the compensation package has an effect on firm performance. Hall and Liebman (1997) argue that CEO compensation is in fact highly sensitive to stock performance. They state that the elasticity of CEO compensation to firm market value more than tripled from 1,2 to 3,9 in the 1980 – 1994 period. Hall and Liebman argue that the sharp increase in the value of stock options and shares awarded to CEO’s caused this increase in elasticity. This implies that firms can structure executive compensation in a way that optimizes the effect on performance. The
channels as suggested by Finkelstein and Hambrick (1988) give an idea of how firms can do this. The higher sensitivity implies that compensation designs increasingly aligns the interests of executives and shareholders.
Given that a higher level of compensation does not always mean better performance, the
structure of compensation might be a more effective instrument. Core, Holthausen and Larcker
9
regressed the excess on performance. They found a negative relationship, implicating that increasing the level of pay eventually leads to worse firm performance (1999).
Existing literature on the relationship between executive compensation structure shows that there is a positive relation between the percentage of equity-based compensation and firm performance. Mehran (1995) uses Tobin’s Q and Return on Assets (ROA) as proxies for firm performance in two separate OLS regressions where these proxies where the dependent variable and the independent variable of interest was the percentage of equity-based
compensation. Mehran’s results show for both regressions a positive and significant coefficient for this variable of interest. Mehran states that an important area for future research is the form of equity based compensation that has the greatest effect on firm performance. A problem in investigating this is the lack of data on these specific compensation details in existing data.
Murphy (1985) researched the reverse relationship, the effect of firm performance on executive compensation. He finds that firm performance positively affects total compensation. Due to SEC regulation it was not possible to fully distinguish the several components of the compensation. Despite this shortcoming it is very plausible, if not certain, that there is a
simultaneous causality problem when testing the relationship between compensation structure and firm performance. Proxies for firm performance are often also measures that determine whether an executive gets (equity-based) compensation which causes this simultaneous causality.
Mehran (1995) also finds determinants other than firm performance that might explain the percentage of equity-based compensation. He follows Jensen and Meckling (1976) in the argument that ownership structure, executive compensation and board compensation are all determined by each other and additional determinants like business risk. Taking a closer look at this relationship gives the following insights. First of all top executives are risk averse, as stated by Harris and Raviv (1979). Executives want their compensation structured in a way that they face less personal risk. Therefore executives will prefer cash compensation. Equity-based compensation, which is dependent of the stock market, will be more risky since executives cannot (fully) control the stock market. Shareholders on the other hand are considered as risk neutral. They can diversify firm specific risk away. Shareholders are not responsible for
10 structuring the executive compensation, but they are responsible for electing the board. The board on their turn is responsible for structuring the compensation for executives. Based on this reasoning Mehran hypothesizes that outside directors tend to maximize shareholders’ value by making compensation more equity-based. Inside directors are more sensitive to executive preferences and will thus make compensations more cash-based.
Another determinant of compensation structure is the incentives executives have to maximize firm value. Mehran states that the amount of stock an executive currently holds will influence future compensation plans. When an executive owns a small amount, equity-based compensation will be higher.
Van der Laan, van Ees and van Witteloostuijn (2010) state that their paper is among the first papers that document the pay-performance relationship for the Netherlands. One of the contributions of their research is the trend that share and options grants are getting more popular in the Netherlands. They state that share and option grants only became popular in the Netherlands since the turn of the millennium. This trend suggests convergence towards
American standards. This paper will extent the research as done by van der Laan, van Ees and van Witteloostuijn. Their timeframe was 2002-2006. If the trend continues after 2006 the results of this paper might help to confirm the convergence towards American compensation
structures. The important question that then remains is whether this convergence is desirable in terms of effectiveness. The empirical analysis of the Dutch pay-performance relationship will help to answer this question.
Remuneration for executives consists of several components, and to solve the
simultaneous causality problem presented by Mehran (1995) and prove the causal relationship as suggested by Jensen and Murphy (1990), this study focuses on three of these components. The first component is the fixed remuneration. This is basically the fixed annual salary an
executive receives. The second component is the Short-Term Incentive, henceforth STI. The third and last component is the Long-Term Incentive, Henceforth LTI. These components are disclosed and explained in the remuneration policies of each company.
11 2.2 Relevant Definitions
STI policies focus on a single year as performance period. The grant of this incentive is unconditional of future performance. A STI has one or more performance indicators. These indicators are disclosed in the remuneration policies and are set by the remuneration committee at the beginning of the performance year. At the end of the year actual performance is
measured and a corresponding STI is granted unconditionally. From these characteristics it is clear that causality between STI and firm goes both ways. If in a performance year the STI is high, the firm performance has to be high and vice versa. This is graphically shown in figure 2.1.
Figure 2.1
As shown by the timeline the STI grant is post performance period. The value of STI grant is by definition also dependent of the performance period. Therefore giving a causal interpretation for the effect of variable STI on performance is impossible.
An LTI is granted conditionally at the beginning of a three year performance period. An LTI also has one or more performance indicators. The value of the conditional grant is set by the remuneration committee and is often expressed as a percentage of base salary. These
characteristics implicate that firm performance does not influence the value of the conditional grant. Only the part that becomes unconditional after the three year performance period is influenced by firm performance. This is graphically shown in figure 2.2.
1 Year Performence Period
12 Figure 2.2
As the timeline shows, the grant is an event that takes place prior to the performance period. Therefore the value of this grant is independent of the actual performance. In the annual report of a company the fair value of a LTI grant is disclosed. The unconditional vesting is
dependent of the performance period and the amount that vests unconditionally is expressed as a percentage of the conditional grant. This leads to the key assumption that the conditional grant has an effect on performance, but performance does not affect the conditional LTI grant. Therefore the simultaneous causality problem for this part of the remuneration package is eliminated. In addition to this the value that becomes unconditional is dependent of the stock price, because LTI grants are equity based. The approach in this research focuses on this part of the remuneration package.
2.3 Predictions
The theories above predict that there is a relationship between remuneration structure and firm performance for firms based in the Netherlands. Using the characteristics of the remuneration packages the prediction is that there is a simultaneous causality problem for STI and firm performance, but there is no simultaneous causality problem for conditional LTI grants and firm performance in the same year. The predicted effect is expected to be stronger for CEO’s than for CFO’s. Another prediction is that the remuneration structure in the Netherlands is changing. The part of equity-based compensation is increasing for firms based in the Netherlands. These three predictions are captured in three hypotheses that are tested in section 5 of this paper. The hypotheses that are tested in this paper are introduced below.
Hypothesis 1: Conditional LTI grant as a percentage of total remuneration is positively related with shareholders’ value. It is expected that executives want to maximize the vesting value of
Conditional LTI Grant
3 Year Performence Period
13 their conditional grant by maximizing shareholders’ value as much as possible, given that proxies for shareholders’ value are the performance measures for their LTI. Hypothesis 2: The effect of conditional LTI grants is stronger for CEO’s than for CFO’s. The reasoning behind this is that CEO’s are better able to influence the most common performance measure for long-term incentives. Hypothesis 3: LTI as a percentages of total compensation is increasing over time. It is expected that the increase is stronger for CEO’s than for CFO’s and the reasoning follows from the
expectation that conditional LTI grants are more effective for CEO’s. When it is more effective for CEO’s, boards will increase this remuneration component more for CEO’s.
3 Methodology
In this section the methodology used in this paper is explained. The first part focuses on the data collection, which was a time-consuming and intensive process. For each year and company the annual report and remuneration policy were used to hand-collect the necessary info. In the second part the regression formula is given and the relevant variables are explained. 3.1 Data collection
The data used in this paper is hand collected from annual reports published by companies listed on the AEX and AMX. CEO’s and CFO’s for which data is available are included in the sample. The time period is 2012-2014. There is no database with executive remuneration available. Another difficulty is the way the several remuneration components are disclosed in the annual reports. Each company has a different way of reporting the board remuneration. As mentioned before the remuneration components of interest are the fixed annual salary, STI and conditional LTI grant. Collecting the fixed annual salary from annual reports does not cause problems. Collecting STI and LTI data requires additional explanation.
STI can be paid in several ways. Most common is payment in cash, but some companies pay part of the STI in shares. If the shares are unconditional of future performance it is
considered as a true STI payment. Some companies make part of their STI conditional on future performance. If this is the case it is considered as an LTI in this research. Because the value of
14 this grant is subject to performance in the previous year the characteristics are the same as with an LTI grant. The conditional grant is subject to the previous year, only the part that becomes unconditional is dependent of performance in the actual year. It is assumed that executives are aware of their remuneration policy and thus know that part of their STI is actually a conditional grant of shares or cash. This conditional part becomes unconditional if performance in the subsequent years is sufficient.
LTI grants can be done in three ways. Most common is a conditional grant of performance shares. After the performance period a percentage of these shares is awarded unconditionally to the executive. Another way is a grant of options that is conditional on future performance. Common to the performance shares a percentage of the conditional option grant is awarded to the executive after the three-year performance period. Less common is a grant in shares of which the unconditional part after three years is paid out in cash. In all LTI plans the percentage of the grant that becomes unconditional is dependent of the actual performance in the
performance period. The fair value at grant date is used as measure for LTI grant value. For options the Black and Scholes option valuation theory is used. Most companies disclose the fair value of a LTI grant in their report and otherwise the data necessary to do this is disclosed.
To test the effectiveness of remuneration structure an appropriate performance measure has to be selected. Since the conditional LTI plans eliminate simultaneous causality problems, the performance measure used is relative TSR. Firstly relative TSR is the most common LTI performance indicator used by the firms in the sample. Moreover it follows from Harris and Raviv (1979) that equity based compensation tends to maximize shareholders’ value more than other types of compensation. Relative TSR measures the increase in shareholders’ value relative to a chosen benchmark. The benchmark for the firms in this research is the index it is in. Daily closing prices are collected for the period between 01-01-2012 and 31-12-2014. For each year the median closing price in January and December is calculated. These median values are used to calculate the relative TSR by dividing the median share price in December minus January by January. Also the relative TSR for the three years prior to 2012 is calculated. These lagged values are used as controls.
15 One of the control variables used in this research is the Beta of a firm. The beta is used as risk measure. The closing prices are used to calculate the relative TSR and the returns for each company over the three year period. The covariance between the returns of the firm and index is divided by the variance of the index to get the beta. The beta is calculated separately for each year. To control for the growth opportunities of the firm the price-to-book ratio is included in the regression. The natural logarithm of assets is included to control for firm size and the long term debt over total assets controls for the leverage of the firm.
Firm dummies are included to control for firm specific events. To control for time effects dummies are created for 2012 and 2013. Datastream is used to collect data on the financial variables used in the regression. The element codes can be found in appendix 9.3.
3.2 Model
The first and main hypotheses that conditional LTI grant as a percentage of total remuneration positively affects firm performance is tested using an OLS regression. The regression formula is given in equation 3.1
𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑇𝑆𝑅𝑖 = 𝛽0+ 𝛽1%𝐿𝑇𝐼𝑖+ 𝛽2%𝑆𝑇𝐼1𝑖+ 𝛽3𝑃𝑟𝑖𝑐𝑒𝐵𝑜𝑜𝑘 2𝑖+ 𝛽4𝐿𝑜𝑛𝑔.𝑇𝑒𝑟𝑚.𝐷𝑒𝑏𝑡𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠 3𝑖+
𝛽5𝐿𝑛(𝐴𝑠𝑠𝑒𝑡𝑠)4𝑖+ 𝛽6𝑅𝑖𝑠𝑘5𝑖+ 𝐿𝐴𝐺𝐺𝐸𝐷. 𝑅𝑇𝑆𝑅 + 𝑇𝐼𝑀𝐸. 𝐸𝐹𝐹𝐸𝐶𝑇𝑆 + 𝐹𝐼𝑅𝑀. 𝐸𝐹𝐹𝐸𝐶𝑇𝑆 + 𝜀𝑖
(3.1)
Given the LTI characteristics a causal relationship between LTI and relative TSR is tested. It is hypothesized that an increase in LTI percentage causes an increase in relative TSR. Relative TSR and %STI are expected to be highly correlated. Because STI plans are an important incentive for firm performance, including this in the regression is necessary to control for omitted variable bias. Price-to-book value controls for growth opportunities. Firms with more growth
opportunities will in general experience a higher relative TSR. Other controls are the Long-Term debt relative to total assets and the natural logarithm of Total Assets. In the regression dummies are included to control for firm specific and time effects.
16 The third hypotheses that equity based compensation as a part of total remuneration is increasing in the Netherlands is tested by comparing means for the years in the sample. The differences are tested on significance.
4 Data and descriptive statistics
In this part several summary statistics will be presented. First the actual means for the relevant variables are given. In the second paragraph the descriptive statistics of both regression
subsamples are given.
4.1 Summary statistics per year
For the period 2012-2014 summary statistics are presented in the tables below. For the relevant variables the mean is given in each year. The statistics are presented separately for the AEX and AMX. For each index the summary statistics are given for the entire sample and the sample excluding firms where a CEO or CFO turnover occurred. If there was an appointment of a new executive in a specific year this might lead to an increase or decrease of compensation that would not have happened if the same executive had stayed in his position. If two executives worked more than three months in the same year and the compensation details are disclosed in the report, both executives are included in the sample. The Total LTI grant is for each executive the sum of shares, options and LTC.
Table 4.2 and table 4.4 show that the average total compensation increased 38,24% and 12,31% respectively for CEO’s and CFO’s in the AEX from 2012 to 2014. For the CEO’s and CFO’s in the AMX the average total compensation increased 44,88% and 27,75% respectively in the same period. This follows from the data in table 4.6 and 4.8. The increase for CEO’s is quite high and for the biggest part caused by an increase in LTI grants. The increase in LTI grants was 79,10% and 8,84% for respectively CEO’s and CFO’s in the AEX. For the CEO’s and CFO’s in the AMX the increase was 89,73% and 44,13% respectively. This increase is mainly due to big increases for just a few observations. In section 5 of this paper the increases will be tested on significance. The differences are calculated based on the sample of executives that stayed in
17 place. A CEO or CFO turnover often causes a change in the different compensation elements that makes comparison impossible.
In appendix 9.1 the relative TSR is showed in a table for the 2012-2014 period. The percentages are multiplied by 100. The relative TSR for PostNL was extremely high in 2013. The firm dummy will control for the firm specific events that caused this rapid increase in
shareholders’ value. Imtech and Pharming also experienced large volatility in shareholders’ value. Just like the case with PostNL these cases are not considered as outliers. In appendix 9.2 a table is included with information on the sector in which the companies are divided.
18
Table 4.1, AEX CEO complete sample
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI
2012 € 899.936,30 € 806.035,96 € 1.396.277,67 € 1.385.674,71 € 262.800,00 € 821.256,67 € 2.885.081,71 2013 € 910.175,68 € 780.026,73 € 1.824.329,84 € 1.877.012,19 € 271.200,00 € 1.400.275,00 € 3.297.502,54 2014 € 953.440,05 € 954.568,80 € 1.858.553,17 € 1.930.333,31 € 271.200,00 € 1.337.472,33 € 3.721.106,35
N(2012) = 23, N(2013) = 24, N(2014) = 22
Table 4.2, AEX CEO excluding CEO turnover
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI
2012 € 922.458,45 € 876.239,93 € 1.443.324,23 € 1.297.899,75 € 262.800,00 € 1.390.096,73 € 3.130.997,67 2013 € 933.748,23 € 880.320,23 € 2.060.863,09 € 1.929.832,46 € 271.200,00 € 3.231.000,00 € 3.716.403,62 2014 € 975.325,92 € 966.975,38 € 2.584.964,14 € 2.471.633,60 € 271.200,00 € 3.560.400,00 € 4.328.422,05
N(2012) = N(2013) = N(2014) = 13
Table 4.3, AEX CFO complete sample
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI
2012 € 589.368,16 € 458.758,33 € 831.823,40 € 862.069,34 € 92.700,00 € 263.700,92 € 1.724.464,18 2013 € 610.973,38 € 501.339,48 € 860.542,60 € 901.367,86 € 62.400,00 € 176.136,20 € 1.715.941,52 2014 € 600.441,38 € 492.661,60 € 1.135.605,30 € 1.086.354,80 € - € 337.019,05 € 1.910.013,57
N(2012) = 19, N(2013) = 24, N(2014) = 24
Table 4.4, AEX CFO excluding CFO turnover
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI
2012 € 642.888,89 € 580.680,00 € 987.751,15 € 1.053.089,06 € 92.700,00 € 136.774,55 € 1.927.299,78 2013 € 659.694,44 € 579.983,75 € 1.026.040,95 € 1.126.171,66 € 93.600,00 € 68.963,00 € 2.087.271,96 2014 € 683.055,56 € 591.690,63 € 1.075.033,10 € 1.218.764,40 € - € 34.457,00 € 2.164.587,75
19
Table 4.5, AMX CEO complete sample
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI 2012 € 518.202,92 € 224.767,43 € 317.570,48 € 269.548,59 € 261.925,58 € 345.941,63 € 986.677,40 2013 € 563.321,26 € 286.107,00 € 360.046,53 € 324.390,96 € 332.176,40 € 295.931,25 € 1.132.156,94 2014 € 574.656,90 € 324.455,48 € 555.117,80 € 413.731,34 € 329.483,67 € 773.185,71 € 1.308.536,58
N(2012) = 25, N(2013) = 23, N(2014) = 25
Table 4.6, AMX CEO excluding CEO turnover
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI 2012 € 534.534,22 € 224.302,71 € 367.468,86 € 271.631,96 € 344.916,67 € 345.941,63 € 1.035.880,67 2013 € 568.812,50 € 310.017,00 € 403.601,86 € 338.978,16 € 368.000,00 € 295.931,25 € 1.177.376,08 2014 € 587.873,17 € 401.322,12 € 694.462,84 € 453.308,54 € 388.673,33 € 773.185,71 € 1.500.820,97
N(2012) = N(2013) = N(2014) = 18
Table 4.7, AMX CFO complete sample
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI 2012 € 410.146,51 € 161.402,09 € 189.832,43 € 154.662,80 € 153.750,00 € 164.382,78 € 679.902,92 2013 € 432.726,18 € 196.257,27 € 191.143,97 € 184.606,28 € 157.398,00 € 162.201,43 € 719.500,03 2014 € 412.539,98 € 228.894,49 € 249.403,57 € 255.290,77 € 315.223,07 € 119.622,50 € 796.672,45
N(2012) = 22, N(2013) = 23, N(2014) = 23
Table 4.8, AMX CFO excluding CFO turnover
Average Total Fixed STI Total LTI Grant Shares LTC Options Total Fixed+STI+LTI 2012 € 394.629,50 € 176.140,15 € 200.292,59 € 162.281,70 € 153.750,00 € 211.075,00 € 651.152,41 2013 € 410.252,71 € 218.526,46 € 204.888,91 € 166.543,98 € 157.398,00 € 214.752,50 € 702.996,11 2014 € 427.022,50 € 275.278,08 € 288.684,55 € 307.303,74 € 315.223,07 € 174.400,00 € 831.864,32
20 4.2 Regression Samples
For the regressions all years are included in the sample. Separate regressions are performed for the CEO and CFO sample.
Table 4.9, CEO Sample
LTITotaal STITotaal N Valid 134 135 Missing 2 1 Mean ,35366 ,20580 Median ,33566 ,19831 Std. Deviation ,17806 ,12239 Minimum ,00000 ,00000 Maximum ,85465 ,56043
The mean percentage of conditional compensation for CEO’s is 35,4%. The unconditional short-term bonus is on average 20,6% of the total compensation package. Total compensation is as calculated in table 4.1 – 4.8. In this total amount pension and other compensation elements are not included. In the sample some executives do not receive a LTI grant. Due to regulations financial firms often do not have LTI grants.
Table 4.10, CFO Sample
LTITotaal STITotaal N Valid 122 122 Missing 1 1 Mean ,30061 ,19241 Median ,29450 ,19442 Std. Deviation ,18933 ,13210 Minimum ,00000 ,00000 Maximum ,80585 ,56521
For the CFO sample the mean percentage of conditional compensation is 30,1%. The
unconditional short-term bonus is on average 19,2%. The variables are computed in the same way as for the CEO sample. The dependent variable in the regressions is the relative TSR for the
21 company. Summary statistics for this variable can be found in the appendix. Summary statistics for the control variables can also be found in the appendix.
As hypothesized the unconditional STI bonus is correlated with the dependent variable, relative TSR. Causality in this relation is impossible to identify given the nature of the STI bonus. In the table below the actual correlation between STI and relative TSR is presented for the CEO and CFO sample.
Table 4.11, CEO correlations Pearson Correlation RTSR
STI/Total 0,227*
N 123
*. Correlation is significant at the 0.05 level (2-tailed).
For the CEO sample the Pearson Correlation coefficient is 0,227 and this coefficient is significant at the 0,05 level. This confirms the expectation that STI is correlated with relative TSR and justifies the assumption that relative TSR is an appropriate outcome measure. Since STI is determined by the relative TSR in the same year and the STI affects relative TSR it is impossible to determine causality in this relationship
Table 4.12, CFO correlations Pearson Correlation RTSR
STI/Total 0,223*
Observations 111
*. Correlation is significant at the 0.05 level (2-tailed).
For the CFO sample the correlation coefficient is 0,223 and this coefficient is also significant at the 0,05 level.
22
5 Results
In this section the results of both samples are reported separately. The regression model given in the methodology section is used to identify a causal relationship between the conditional LTI grant as a percentage of total compensation in that year.
5.1 Regression results
In total six regressions are performed. In the following paragraphs the results are presented and discussed for the CEO and CFO sample where normal standard errors are used. In appendix 9.4 regression output is presented when robust standard errors are used.
Before running the regressions the correlations between predictors that might be
correlated is checked. Since lagged values of the performance indicator (relative TSR) might be correlated with the conditional grant (%LTI).
5.1.1 CEO Sample
Since the conditional LTI grant might be dependent on performance in previous years (the board might grant more LTI when performance was high in previous years), correlations are reported in table 5.1 to check this. If the conditional LTI grant relative to total compensation is highly
correlated with a lagged value of stock performance, which is the performance indicator, these lags cannot be included as controls. When lagged values of stock performance are correlated with the conditional LTI grant and not correlated with performance in the future this enables the possibility for an instrumental variable regression. The lags that are positively correlated with conditional LTI grant but only the second lag is significantly correlated. Therefore an
instrumental variable regression using the lags of the performance measure (RTSR) as an
instrument is not possible since correlation is a necessary condition for an instrumental variable regression.
Table 5.1 shows that the second lag of relative TSR is significantly correlated with the conditional LTI grant relative to total compensation. The pearson correlation coefficient is 0.154 and the level of significance is 0,043. Before concluding that this results in problematic
23 Table 5.1 %LTI Pearson Correlation RTSR -1 Observations RTSR -2 Observations RTSR-3 Observations 0.107 123 0.154** 125 0.130 125 *** p<0.01, ** p<0.05, * p<0.1
The results of the regression for the CEO sample are presented in table 5.2. Relative TSR (RTSR) is the dependent variable in each regression. The relative TSR is multiplied by 100 to get the output as presented. Regression (1) did not include firm specific effects. For the CEO sample the coefficient for %LTI is 33,99 and is significant at the 0,05 level. This implicates that an
increase of 1 percentage point in the conditional LTI relative to total compensation causes a 0,3399 percentage point increase in relative TSR. The %STI has a coefficient of 61,71 and is significant at the 0,01 level. Causal interpretation however is not possible due to the
simultaneous causality problem. The R-squared is 0,195 and given the number of observations in this hand-collected dataset it is fair to assume R-squared will increase significantly when increasing the number of observations. Including firm specific effects increases the LTI
coefficient quite dramatically. Regression (2) shows that the coefficient for %LTI is now 87,02 and the level of significance is 0,05. R-squared increases to 0,615, thus the overall fit of the model increases significantly. A 1 percentage point increase in LTI grant now results in a 0,8702 percentage point increase in relative TSR. In regression (3) lagged values of relative TSR are included as additional control variables. This is done because performance indicators tend to be auto correlated. Indeed the lagged values have significant coefficient, lowering the %LTI
24 percentage point increase in LTI grants results in a 0,8283 percentage point increase in Relative TSR. The R-squared increases to 0,776 implicating that the overall fit of the model benefits from including the additional controls. There is no implication for problematic multicollinearity and because of the autocorrelation an instrumental variable regression is not possible.
5.1.2 CFO Sample
The same regressions are performed in the CFO sample. The results are presented in table 5.4. Contrary to the results for the CEO sample, there is no significant relationship between %LTI and relative TSR. In each of the regression the coefficient is positive but not significant. The
magnitude is also lower for each regression. As expected, %STI does have a significant
coefficient. Again causal interpretation is not possible. The coefficients for %LTI are 21,51, 22,66 and 13,32 in regression (1), (2) and (3) respectively. Adding the lagged values of RTSR decreased the magnitude and the coefficient is still not significant.
The first lag of relative TSR with a coefficient of -0.444 and the 2013 year dummy with a coefficient of 10,77 are the only control variables that have a p-value lower than 0.05 in model (3), where in the CEO sample Price-to-Book and Beta where the only variables with a p-value higher than 0.05. The only differences in the model are %STI and %LTI. Where for CEO’s almost all variables are significant predictors, for CFO’s the model has only three and one of them being the percentage of STI. The simultaneous causality for this variable is discussed before and the correlation in table 4.11 already confirmed the existence of this positive relationship in the CFO sample. Including the lagged values has no positive effect which might be caused by
multicollinearity. The correlation table for the CFO sample is reported in table 5.3. Both the second and third lag are significantly correlated with the p-value being lower than 0.05 and 0.01 respectively. The pearson correlation coefficients are 0.184 and 0.229 respectively. Given that the relationship between %LTI and relative TSR was not significant without the lags, the possibility of using the lags as an instrument for %LTI not exploited. No matter how good the instrument, the relationship is nonexistent. Using an instrument to eliminate potential simultaneous causality therefore makes no sense.
25 Table 5.3 %LTI Pearson Correlation RTSR -1 Observations RTSR -2 Observations RTSR-3 Observations 0.092 112 0.184** 113 0.229*** 113 *** p<0.01, ** p<0.05, * p<0.1
26 Table 5.2 (1) RTSR (2) RTSR (3) RTSR
VARIABLES CEO CEO CEO
%LTI 33.99** 87.02** 82.83*** (16.20) (37.56) (30.94) %STI 61.71*** 130.1*** 130.0*** (23.29) (40.66) (33.32) BETA -15.37* -15.10 -9.301 (8.043) (11.52) (9.399) Price-to-Book -3.294** -0.0628 -1.663 (1.382) (2.250) (1.832) Debt/Assets -32.75* -94.16 -137.4*** (17.89) (58.29) (48.78) LN(Assets) 1.303 36.90** 38.14*** (1.404) (14.16) (11.37) 2012 3.422 13.15** 11.04** (5.861) (5.584) (4.506) 2013 10.06* 15.35*** 12.12*** (5.721) (5.183) (4.143) RTSR, t-1 -0.479*** (0.0970) RTSR, t-2 -0.360*** (0.127) RTSR, t-3 0.165** (0.0813)
Firm Specific Effects NO YES YES
Constant -25.55 -461.4*** -504.1***
(25.08) (145.7) (119.0)
Observations 119 119 118
R-squared 0.195 0.615 0.776
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
27 Table 5.4 (1) RTSR (2) RTSR (3) RTSR
VARIABLES CFO CFO CFO
%LTI 21.51 22.66 13.32 (14.66) (19.53) (17.24) %STI 58.96*** 57.77** 49.61** (21.63) (28.18) (24.19) BETA -15.42* -24.84** -19.29* (8.243) (11.55) (10.58) Price-to-Book -3.474** -1.561 -2.970 (1.355) (2.396) (2.113) Debt/Assets -24.51 12.49 -50.53 (19.24) (67.53) (64.21) LN(Assets) -0.260 8.595 20.06 (1.563) (19.12) (16.89) 2012 11.14* 14.04** 10.06* (5.927) (5.698) (5.157) 2013 13.38** 13.39** 10.77** (5.502) (5.331) (4.621) RTSR, t-1 -0.444*** (0.102) RTSR, t-2 -0.254* (0.136) RTSR, t-3 0.112 (0.0885)
Firm Specific Effects NO YES YES
Constant 2.899 -150.4 -242.7
(25.37) (377.1) (223.1)
Observations 108 108 107
R-squared 0.196 0.616 0.740
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
28 5.1.3 Implications
The estimated coefficients for the %LTI are positive in both samples and significant in the CEO sample. The magnitude however might be overestimated due to omitted variables. The R-squared is quite high, implicating a good overall fit. Despite this good overall fit their might still be other factors that affect firm performance. As stated before, there are factors beyond control of executives that affect firm performance. Unobserved factors or omitted variables might result in an overestimation. The aim of this research is to prove the existence of a causal relationship between conditional LTI grant and firm performance. The regression results confirm this relationship since the overall fit of the model is good and the level of significance is high, but there might be a small overestimation of the coefficient. Regression results for the CEO and CFO sample with robust standard errors are presented in table 9.1 and 9.2 respectively in appendix 9.4. The coefficient for %LTI remains significant at the 0,05 level in model (3).
Given this result an interesting question is till what extent this incentive can be used and whether the use of it should differ among executives. A compensation package consisting of only conditional components is not realistic and very few, if any, executives would be willing to accept this. Therefore this result should not lead to an infinite increase of conditional LTI grants for CEO’s. In combination with previous models however, it can be very helpful in optimizing
compensation packages. The model developed by Core, Holthausen and Larcker (1999) can be
used to determine the level of pay that is justified by economic determinants to prevent excess pay that has a negative effect on firm performance. The second step is now to structure the compensation package, dividing the total level of compensation in fixed annual salary, short-term incentives and conditional long-short-term incentives. This research shows that conditional LTI grants affect shareholders’ value in a positive way and thus should be part of the overall compensation package for CEO’s. The coefficient for the CFO is smaller and not significant, implicating that conditional LTI grants are less effective for CFO’s.
Jensen and Murphy (1990) find that bonuses account for 50% of executive compensation packages, but the bonuses are awarded in ways that are not sensitive to performance. The characteristics of these bonuses partly explain their empirical finding. A common way of awarding bonuses in the United States used to be an unconditional stock option award at the
29
beginning of a performance period. The rationale behind this was that the options become worthless if performance is bad, so no costs for the company, and if performance is good both the firm and the executive benefit. The incentive thus is to perform because otherwise the options become worthless. This differs substantially from the characteristics of the conditional LTI award as researched in this paper. It seems that awarding long-term incentives can be done in a way that is sensitive to performance.
Yermack (1995) states that his findings are in the same spirit as the results found by Jensen and Murphy. In addition he claims that 6 out of 9 compensation theories are not supported by empirical data. Hall and Liebman (1997) show that compensation structure changed dramatically, increasing the sensitivity of compensation to stock performance. The results of this paper together with the results of Hall and Liebman imply that empirical data is now more in line with existing compensation theories.
The results presented in this section also have consequences for the Dutch financial sector. As stated in the introduction, a 20% cap on variable remuneration has recently been enacted. This means that financials in the Netherlands have limitations in structuring executive
compensation. Some of the consequences are less R&D expenditure (Rappaport, 1978) and more focus on short-term performance (Dechow & Sloan, 1991). The consequence that follows from the findings in this research is a limitation in maximizing shareholders’ value through optimizing the compensation structure. Although this paper did not determined an optimal level, it is reasonable to assume that a 20% cap on variable remuneration results in a non-optimal division in short-term, long-term and fixed incentives. Since this cap is introduced
recently, future research can determine the actual effect. Unfortunately the data to do this is not available yet.
5.2 Mean comparison
The third hypothesis of this research is that the compensation structure changed for executives, and more specifically that the conditional LTI component increased relative to STI and fixed salary. Given the results of the regression it is particularly interesting to see whether there is a difference between CEO’s and CFO’s in the change of compensation structure. Van der Laan,
30 van Ees and van Witteloostuijn (2010) find a change in structure in their research. In the period 2002-2006 they find an increase in equity based compensation.
In this paper the means are compared for 2012 and 2014. The variable that is tested is %LTI. Table 5.5 shows that the increase in mean %LTI is approximately 2,5 percentage point in the CEO sample. This change however is not significant given the standard error of 0,03789. The difference is tested assuming equal variances given that the significance for Levene’s test for equality of variances is 0,873. The hypotheses that the variances are equal is thus not rejected.
The same comparison of means is performed for the CFO sample. Statistics are presented in (2) of table 5.3. There is a decrease in the mean percentage conditional LTI of 0,6 percentage point. This change however is not significant given that the standard error of the difference is 0,04440. The difference is tested assuming equal variances given the significance of 0,524 for Levene’s test for equal variances.
Given this result the hypothesis that remuneration structure changed for executives is rejected. For both CEO’s and CFO’s the remuneration structure did not change on average in this sample. Looking at an individual level however there does seems to be a trend towards a larger part of conditional LTI in the compensation package. Over 50% of the companies
increased the conditional LTI component for their executives. In addition to this the difference on average might not be statistically significant, it is in line with expectations based on the regression results presented in section 5.1. Conditional LTI is an effective incentive for CEO’s and it is less effective for CFO’s. More than half of the CEO’s experienced an increase in conditional LTI grants relative to other components, but on average the increase of 2,5% is not significant. Since conditional LTI grants are less effective for CFO’s, the expectation is that for CFO’s the conditional LTI grant relative to other components increased at least less than for CFO’s. In fact CFO’s experienced a decrease of 0,6 percentage point, but this difference is not significant.
31 Table 5.5
%LTI %LTI
Statistics CEO CFO
Mean 2012 Mean 2014 Standard Deviation 2012 Standard Deviation 2014 0.34157 0.36682 0.16844 0.18608 0.31197 0.30684 0.18213 0.20621 Significance
Levene’s Test for Equality of Variances
0.873 0.524
Mean difference -0.025248 0.006133
(equal variances assumed) (0.03789) (0.04440)
Degrees of Freedom 86 76
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
6 Robustness checks and additional results
In this section the regression formula as presented in 3.1 is adjusted by changing the dependent variable. Return on equity (ROE) is now used as a proxy for firm performance. The right hand side of the equation does not change except for the lagged values. The ROE for the companies is retrieved from Datastream and regressed on the same variables as the model in 3.1. ROE and relative TSR are both proxies for the shareholders’ value and therefore this additional regression is performed. The difference between these proxies is that relative TSR is the most common performance indicator and ROE is not used as a performance indicator by the firms in the sample. For shareholders however return on equity is an important measure for performance.
Table 6.1 shows the results for the regression of ROE on its lagged values. Only t-1 has a significant coefficient. Regressions with t-2 and t-3 are performed but not reported in table 6.2 and 6.3. Since ROE is an accounting measures increases the possibility of multicollinearity.
32 Including higher lags had no impact on the other coefficients and its standard errors as already implicated by the results in table 6.1.
Table 6.1 (1) VARIABLES ROE ROE, t-1 0.728*** (0.0975) ROE, t-2 0.0945 (0.145) ROE, t-3 0.0680 (0.0599) Constant -2.814 (2.526) Observations 121 R-squared 0.412
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 6.1.1 CEO sample
Table 6.2 shows the regression output for the CEO sample. Model (1) is the regression without firm specific effects. In this regression %LTI coefficient of 39,04 is significant at the 0,05 level %STI is only significant at the 0,10 level with a coefficient of 44,44. Including firm dummies to control for firm specific effects causes an significant increase in the coefficient for %LTI and %STI. The output for this regression is in model (2). A 1 percentage point increase in these
compensation components now causes an increase of 0,639 and 0,849 in ROE respectively. Both coefficients are significant at the 0,05 level. Adding the first lag of ROE results as the output in model 3. The coefficient for %LTI decreased in magnitude to 57,78 but is now only significant at the 0,10 level. %STI remained significant at the 0,05 level with a coefficient of 71,55.
The interpretation is the same as the interpretation for the results in the previous section. Causal interpretations for the %STI is not possible given that performance and STI incentives influence each other. For %LTI there seems to be a causal relationship where a 1 percentage point increase in conditional LTI relative to total compensation causes a 0,639 percentage point increase in ROE in the model where no lagged values are included. Including lagged values of
33 ROE decreases the significance for %LTI and the coefficients for the lags are small and not
significant. The correlation between the first lag of ROE and %LTI might be the reason for this. Therefore the output as in model (2) is used.
The overall relationship between the variable of interest, %LTI, and firm performance is thus weaker with ROE as a proxy for performance. This means that the relative TSR, which is a proxy for performance but also the most common performance indicator for the conditional LTI grants, is affected more by the compensation structure. CEO’s thus seem to be able to influence measures of performance that are more beneficial for them.
Table 6.2
(1) (2) (3)
VARIABLES ROE ROE ROE
%LTI 39.04** 63.92** 57.78* (16.48) (31.82) (33.05) %STI 44.44* 84.93** 71.55** (23.70) (32.26) (34.21) BETA -19.93** -9.568 -12.30 (8.772) (8.910) (9.273) Price-to-Book -6.118*** -3.266 -2.832 (1.630) (2.134) (2.197) Debt/Assets -37.33* -226.7*** -218.3*** (20.52) (54.70) (56.08) LN(Assets) 1.873 37.54** 36.50** (1.456) (15.09) (15.46) 2012 2.039 6.178 7.043 (6.581) (4.435) (4.477) 2013 3.971 6.932 7.843* (6.435) (4.271) (4.340) ROE, t-1 0.0947 (0.107)
Firm Specific Effects NO YES YES
Constant -10.00 -465.4** -448.2**
(27.76) (176.9) (179.8)
Observations 118 118 115
R-squared 0.212 0.814 0.819
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
34 6.1.2 CFO sample
The results for the regression in the CFO sample are presented in table 6.3 and again model (1) is without firms specific effects, model (2) includes firms specific effects and (3) adds a lagged value of ROE. As with the CEO sample, including ROE seems to have a negative effects on the significance of the other variables. With ROE as a proxy for performance, the significance level for the %LTI coefficient is 0,10 for both model (1) and (2). The model with dummies to control for firm specific effects has a coefficient for %LTI of 26,94. Given that the significance is higher with ROE as a proxy than with Relative TSR implicates that CFO’s can influence this measure to a larger extent than the firms’ share price. The result is still not significant, but this implication seems fair. The %STI coefficient, as expected, is significant at the 0,05 level and the coefficient is 46,68. As with all the regression results presented so far, the simultaneous causality problem associated with this variable makes causal interpretation impossible.
35 Table 6.3
(1) (2) (3)
VARIABLES ROE ROE ROE
%LTI 33.50* 26.94* 22.47 (16.93) (16.05) (16.44) %STI 43.97* 46.68** 36.94 (24.54) (23.16) (23.95) BETA -20.92** -1.495 -1.251 (9.943) (9.277) (9.853) Price-to-Book -6.200*** -2.682 -2.519 (1.746) (1.947) (2.027) Debt/Assets -54.33** -283.1*** -296.1*** (24.82) (56.34) (59.46) LN(Assets) 0.278 41.72*** 45.75*** (1.658) (15.60) (16.01) 2012 1.629 4.393 5.646 (7.373) (4.663) (4.729) 2013 3.305 7.427* 8.104* (6.865) (4.343) (4.409) ROE, t-1 -0.0166 (0.113)
Firm Effects NO YES YES
Constant 23.53 -835.3** -608.0**
(28.57) (323.6) (230.7)
Observations 108 108 105
R-squared 0.199 0.832 0.838
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
36
7 Conclusion
This paper investigates the existing of a causal relationship between conditional LTI grants and the increase in shareholders’ value. A hand-collected dataset is created were the total executive remuneration is broken down into different components. The components of interest are the fixed salary, STI and conditional LTI grant. Using these variables the part of STI and LTI grant relative to the sum of all three is computed. These percentages are regressed on relative TSR and in the robustness section on other proxies for firm performance. The results of the regression on relative TSR, which is the performance indicator for most LTI grants, resulted in significant coefficients for the percentage STI and LTI. STI suffers from simultaneous causality problems by definition. The conditional LTI part does not have this problem given that the grant is independent of performance in that year. In the robustness section the same regression is performed but the dependent variable is now ROE.
The significant result for the LTI variable implicates that the LTI affects the performance indicator in a positive way for Dutch firms, as it is supposed to. In the CEO sample the coefficient for LTI is positive and significant . The dependent variable is relative TSR, both a proxy for
shareholders’ and the performance indicator on which the conditional LTI grant is dependent. For the CFO sample the regression does not result in a significant coefficient for LTI, implicating that CFO’s have less influence on the share price.
In addition a regression is performed on return on equity. Results show similar results as the regression on relative TSR for the CEO sample. For the CFO sample the LTI coefficient is now significant at the 0,10 level. This significance level is not high enough to draw conclusions, but it does implicate that CFO’s seem to influence ROE more than the share price. Given that share prices for a large part are uncontrollable by executives, it seems fair that the CEO has more control than the CFO. This follows from the regression results on the different proxies for performance in this paper. It is therefore concluded that variable remuneration can be justified for CEO’s given the effectiveness for this type of executive. For CFO’s the empirical results do not justify the long-term variable remuneration. Modifications are necessary to design long-term incentive plans in a way that is justifiable for CFO’s. Currently the variable remuneration they receive is dependent on measures that are beyond their control. Comparing the findings of this
37 paper to the study by Jensen and Murphy (1990) and the results from Hall and Liebman (1997) implies that the effectiveness of executive compensation has increased in the recent decades. This increased effectiveness is stronger for CEO’s than for CFO’s.
Optimizing compensation packages is a very challenging task. This paper divided the optimization process into two different parts. The first part is to determine the optimal level of
compensation. Core, Holthausen and Larcker showed that too much pay results in less
performance and too little compensation might result in an inability to attract the best
executives (1999). The second part in the optimization process is structuring the compensation package into fixed, STI and LTI components. This research focuses on the second part in the optimization process. The results do not provide a model to determine an optimal structure. The results show that a larger part of conditional compensation results in a higher relative TSR. Further research should focus on the optimal level and design of conditional compensation, this paper just shows the existence of the hypothesized relationship.
38
8 References
Black, F., & Scholes, M. (1973). The pricing of options and corporate liabilities. The journal of
political economy, 637-654.
Conyon, M. J., & Peck, S. I. (1998). Board control, remuneration committees, and top management compensation. Academy of Management Journal, 41(2), 146-157. Core, J. E., Holthausen, R. W., & Larcker, D. F. (1999). Corporate governance, chief executive
officer compensation, and firm performance. Journal of financial economics, 51(3), 371-406.
Dechow, P. M., & Sloan, R. G. (1991). Executive incentives and the horizon problem: An empirical investigation. Journal of accounting and Economics,14(1), 51-89.
Dutch Corporate Governance Code, 2008. Principles of good corporate governance and best practice provisions
Fama, E. F., & French, K. R. (2004). The capital asset pricing model: Theory and evidence. Journal
of Economic Perspectives, 18, 25-46.
Finkelstein, S., & Hambrick, D. C. (1988). Chief executive compensation: A synthesis and reconciliation. Strategic Management Journal, 9(6), 543-558.
Gabaix, X., & Landier, A. (2006). Why has CEO pay increased so much? (No. w12365). National Bureau of Economic Research.
Hall, B. J., & Liebman, J. B. (1997). Are CEOs really paid like bureaucrats?(No. w6213). National bureau of economic research.
Harris, M., & Raviv, A. (1979). Optimal incentive contracts with imperfect information.
Journal of economic theory, 20(2), 231-259.
Hermalin, B. E. (2005). Trends in corporate governance. The Journal of Finance, 60(5), 2351-2384.
Jensen, M. C., & Murphy, K. J. (1990). Performance pay and top-management incentives.
Journal of political economy, 225-264.
Jensen, M. C., & Murphy, K. J. (2010). CEO Incentives—It's Not How Much You Pay, But How*. Journal of Applied Corporate Finance, 22(1), 64-76.
39
Mehran, H. (1995). Executive compensation structure, ownership, and firm performance.
Journal of financial economics, 38(2), 163-184.
Murphy, K. J. (1985). Corporate performance and managerial remuneration: An empirical analysis. Journal of accounting and economics, 7(1), 11-42.
Murphy, K. J. (1999). Executive compensation. Handbook of labor economics,3, 2485-2563. Rappaport, A. (1978). Executive incentives vs corporate growth. Harvard Business Review,
56(4), 81-88.
Shleifer, A., & Vishny, R. W. (1986). Large shareholders and corporate control. The Journal of
Political Economy, 461-488.
Tosi Jr, H. L., & Gomez-Mejia, L. R. (1989). The decoupling of CEO pay and performance: An agency theory perspective. Administrative Science Quarterly, 169-189.
van der Laan, G., Van Ees, H., & Van Witteloostuijn, A. (2010). Is pay related to performance in The Netherlands? An analysis of Dutch executive compensation, 2002–2006. De
Economist, 158(2), 123-149.
Wet van 28 januari 2015 tot wijziging van de Wet op het financieel toezicht houdende regels
met betrekking tot het beloningsbeleid van financiële ondernemingen (Wet
beloningsbeleid financiële ondernemingen). Staatsblad, Nr. 45
Yermack, D. (1995). Do corporations award CEO stock options effectively?.Journal of financial
40
9 Appendix
In this section several appendices are presented. Throughout the paper references to these appendices are made. They are presented and explained in order of referencing.
9.1 Relative TSR
Table 9.1 Relative TSR AEX Table 9.2, Relative TSR AMX
Relative TSR 2012 2013 2014 Aalberts 1,550292 23,14095 2,358909 AMG -32,4396 -4,33302 -9,67156 Arcadis 17,69439 15,50761 -6,55746 ASMI -2,15049 -12,4389 34,15748 BAM -17,1724 -7,18315 -35,227 Binck -33,9623 -6,12928 -12,6595 Boskalis 3,87144 -2,12695 19,5628 Brunel 34,51295 0,426725 -40,1118 Ten Cate -23,6386 2,698144 -20,8969 Corbion 18,79689 -18,6647 -13,2987 Delta Lloyd -17,0103 12,64825 -0,8449 Eurocommercial 12,52208 -13,1 15,73642 Heijmans -26,7173 14,77405 -27,9281 NSI -42,6103 -40,7156 -19,8359 Pharming -75,1479 -56,1229 47,73033 TKH -1,78666 21,18468 -1,0366 TomTom 2,428899 19,84929 0,195268 USG -18,1726 34,74635 -17,2183 Vastned -10,3807 -11,6107 6,321401 Vopak 23,26594 -27,3402 2,50454 Wereldhave -18,8401 -2,63129 10,16889 Relative TSR 2012 2013 2014 Aegon 23,44254 19,62879 -13,0573 Ahold -9,25113 10,40915 -1,66608 Akzo 17,06122 -5,3006 -3,6651 ASML 34,50453 21,12267 29,28432 DSM 10,28733 11,7401 -13,5398 Fugro -16,7167 -13,6289 -64,9524 Gemalto 10,28733 11,7401 -13,5398 Heineken 30,94969 -16,4999 20,54199 ING 2,802964 14,03217 0,986043 KPN -58,1861 -20,6783 -7,13882 Philips 22,62763 11,41608 -17,1847 Postnl -14,535 105,6065 -29,668 Randstad -0,8469 41,7444 -21,8625 Reed 14,62957 22,32271 21,26957 Shell -13,4124 -16,2665 -0,69039 Imtech -31,6685 -84,5684 -95,8899 SBM -48,2707 22,12846 -39,9361 Unibail 20,67246 -7,12897 9,18519 Unilever 6,433051 -12,8885 8,67463 Wolters 3,323569 22,38507 10,2531
41
9.2 Sectors
AEX AMX
Company Sector Company Sector
AEGON Financial AALBERTS INDUSTR Industrial
AHOLD KON. Consumer Goods AMG Industrial
AKZO NOBEL Chemicals ARCADIS Services
ASML HOLDING Technology ASM INTERNATIONAL Technology
CORIO Real Estate BAM GROEP KON Construction
DSM Chemicals BINCKBANK Financial
FUGRO Oil & Gas BOSKALS WESTMNSTR
KON Offshore
GEMALTO Technology BRUNEL INTERN Services
HEINEKEN Consumer Goods CATE TEN KON. Industrial
ING GROEP Financial CORBION Food
KPN KON. Media & Telecom DELTA LLOYD Financial
PHILIPS ELECTR KON. Consumer Goods EUROCOMMERCIAL Real Estate
POSTNL Postal HEIJMANS Construction
RANDSTAD Services NIEUW STEEN INVSTM Real Estate
REED ELSEVIER Media NUTRECO Food
ROYAL DUTCH SHELL A Oil & Gas PHARMING GRP Biotech
ROYAL IMTECH Technology TKH GROEP Media & Telecom
SBM OFFSHORE Oil & Gas TOMTOM Consumer Goods
UNIBAIL-RODAMCO Real Estate UNIT 4 Services
UNILEVER Consumer Goods USG PEOPLE Services
WOLTERS KLUWER Media VASTNED Real Estate
ZIGGO Media & Telecom VOPAK Oil & Gas
42 9.3 Datastream codes AEX H:AGN,H:AH,H:AKZO,H:GTO,H:ASML,H:IM,H:DSM,H:FUG,H:HB,H:ING,H:KPN,H:PHIL,H:PNL,H:RAND,H:ELS,H:RDSA,H: SBMO,H:TNTE,H:UBL,H:UNIL,H:WSG AMX H:ALBI,H:AMG,H:HDJ,H:ASIN,H:BAM,H:BINC,H:BOSK,H:BRU,H:NTC,H:CRBN,H:DL,H:SIPF,H:HEI,H:NSI,H:NUO,H:PHAR, H:TKF,H:TOM,H:UNI4,H:USG,H:VAST,H:VPK,H:WH
Worldscope element codes
WC01201,WC03251,WC03255,WC08301,WC08376,EPS,WC07011,WC02999,DWSL,WC04890,WC04870,WC04860,P TBV
