Master thesis
CEO compensation at sustainable companies
Name: Joeri Buiten
Student number: 11399074
Thesis supervisor: Alexandros Sikalidis Date: 26 June 2017
Word count: 14.196
MSc Accountancy & Control, specialization Control
Statement of Originality
This document is written by student Joeri Buiten 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 the work, not for the contents.
Abstract
Social responsible investing (SRI) is becoming more important for asset managers each year. Sustainability is becoming common in investing decision. In this thesis, the focus is not on the investor’ side, but on the single most important person in a company for strategic decision making, the CEO. This study examines the effect of environmental, social and governance (ESG) criteria on the level and structure of CEO compensation.
Prior research found that each of those three factors have a significant effect on CEO compensation. However, in some other studies that effect wasn’t found. In this thesis sustainability is measured by using the overall ESG-score, which incorporates all different criteria in one variable.
The ESG-score is found to have significant impact on the structure of CEO compensation, not on the level. ESG-scores vary from 0 to 100%. For every 1pp increase in the ESG-score, the fixed salary as percentage of total compensation drops with 0,018%.
This means that the higher the ESG-score, the lower the percentage of fixed compensation and the higher the variable compensation. Prior research found that SRI has a positive effect on returns and performance, possibly the CEOs of sustainable companies receive extra compensation because of the social performance in alignment with shareholders’ interest.
Contents 1 Introduction ... 5 2 Literature overview ... 8 2.1 Expectations ... 10 3 Research Design ... 12 3.1 Variables ... 12 3.2 Model ... 15 4 Sample Selection ... 17 4.1 Sample selection ... 17 4.2 Definitions of variables ... 17 4.3 Descriptives ... 19 5 Findings ... 22 5.1 Model 1 ... 24 5.2 Model 2 ... 27 5.3 Model 3 ... 30 6 Conclusion ... 33
6.1 Limitations and recommendations ... 34
References ... 35
Appendices ... 40
Appendix 1: Normality test ... 40
Appendix 2: Total Pearson Correlation ... 42
1 Introduction
In recent years a lot of research has been done on social responsible investing. Especially on social responsible investing and the effect on profitability for shareholders. Where it was long supposed by the public that investors should choose between profitability and sustainability, recent research shows that sustainable investing could be even more profitable on the long run (Clark, Feiner, & Viehs, 2015). One of the reasons for these higher return is the focus on long-term targets and goals, where traditional companies often have more short-long-term targets and goals (Clark, Feiner, & Viehs, 2015). In prior studies that impact is not always found (Galema, Plantinga, & Scholtens, 2008).
In another research, the authors found a positive effect of reduction of emissions on market value (Konar & Cohen, 2001). The costs of implementing solutions for new environmental regulations are often overestimated (Porter & Van der Linde, 1995). That’s because prior studies did not consider the positive effect of the innovation this brings on the resource productivity (Porter & Van der Linde, 1995). However, their view is criticized by others (Palmer, Oates, & Portney, 1995). Bansal & Roth (2000) did find three motivations for ecological innovation; competitiveness, legitimation and ecological responsibility.
In a study of 2003 the authors found that stocks from firms with high shareholder rights outperformed firms that had lower rights for shareholders (Gompers, Ishii, & Metrick, 2003). The higher shareholder rights can be seen as better governance. Another study did not find this relationship (Core, Guay, & Rusticus, 2006). In their research, Core et al. (2006) did examine the unexpected earnings around earnings announcements. They did not find earnings surprises specifically to occur with companies with lower governance. They do mention that the companies with lower governance underperform, but this is no surprise (Core, Guay, & Rusticus, 2006).
An empirical study done on German listed companies did find that 12% of abnormal earnings were possible when buying high corporate governance companies and short selling low corporate governance companies (Drobetz & Zimmerman, 2004). Contrary, another study about European companies did find a negative relationship between governance and performance (Bauer, Guenster, & Otten, 2004).
Enthused by this topic and papers, I wanted to do a research about a specific topic within this domain that adds to the existing knowledge such as the paper of Berrone & Gomez-Mejia
topics in environmental, social and governance discussed are all assessed on an individual basis, not as a group of coherent factors, and do examine the effect on performance or stock returns. The effect of this group of factors on executive compensation is not as much researched as the performance and returns. Off course, a lot of research has been done on single factor effects, for example, governance (Conyon M. J., 1997), (Core, Holthausen, & Larcker, 1999) & (Talley & Johnsen, 2004) and even on environmental performance (Berrone & Gomez-Mejia, 2009) & (Mahoney & Thorn, 2006).
A recent study (Van Duuren, Plantinga, & Scholtens, 2016) found that using ESG-factors is becoming increasingly important for asset managers. They also found that asset managers see the selection process of companies using these ESG-factors as a form of fundamental investing, albeit that this effect is greater in Europe than in the US (Van Duuren, Plantinga, & Scholtens, 2016). As this importance should or could translate to management of these companies, it could be the case that ESG-factors become part of the performance measures for executives and ultimately in executive compensation. As found in earlier research, institutional investors have influence on the structure and level of executive compensation (David, Kochhar, & Levitas, 1998).
In this research, the focus will be on this structure and level of executive compensation, CEOs in particular. The CEO is the most important person in the company for policy making and strategic decisions. As ESG-related issues are becoming more important, the focus of (some) CEOs may be on these issues. As said, sustainable companies are often more long-term oriented than traditional companies. The question arises why CEOs choose this way and if it goes hand in hand with different compensation plans.
CEOs could choose to make the sustainable choice because it is the rational way of making more profit over a longer period. Or is it because they believe in a better world? It could also just be some form of legitimacy as found as one of the motivations for ecological innovation (Bansal & Roth, 2000).
In this research about the compensation programs I hope to find an answer about the determinants of CEO compensation in relation to sustainability. It could be the first step in a research about CEO characteristics at sustainable companies. Does it take a different kind of CEO to be the leader of a sustainable company and do they have different compensation programs?
If such a difference exists, this could, off course, have multiple causes. In recent years, bonusses and other incentive programs – such as stock option programs – have been in the news, mostly because people find them (too) high or inappropriate due to the economic circumstances. So, a cause could be that CEOs of sustainable companies don’t want such publicity. They choose for less performance bonusses to avoid bad publicity. Another reason could be that they have a long-term focus and are highly committed to s characteristics their sustainable companies and don’t want that kind of compensation programs.
With this research, I hope to find the evidence that ESG-factors have a significant effect on CEO compensation. The theoretical basis for the why and the direction of this effect will be made in the literature overview.
This leads to the research question of this thesis:
“How does the level of sustainability of a company influence the level and structure of CEO compensation.” Structure
The rest of this thesis is structured as follows; first a literature overview will provide a theoretical construct for the research question, that is followed by the research design where the model and variables will be introduced, sample selection, the findings of the model will be analyzed and the thesis ends with a conclusion.
2 Literature overview
A good starting point for the literature overview is the existing literature about executive compensation. In the past, a lot of research on this subject has been done and it is necessary to get a comprehensive view on this matter. Otherwise the scientific environment of executive compensation is not clear.
Murphy (1999) summarized and shows earlier empirical and theoretical research on CEO compensation. Executives receive more compensation as the size of the firm increases and base salaries are normally determined by benchmarking with industry averages (Murphy, 1999), which means a part of the compensation is based on the industry. Contrary, in another study the size of a company is not to be found of significant effect on CEO (Deckop J. R., 1988). In the study of Deckop (1988) the profit as percentage of sales is found to be of significant effect. Also, CEOs recruited from outside the company earn significantly more than CEOs that were promoted from inside the company.
In other studies the positive relation between size and CEO compensation is also found (Livne, Markarian, & Milne, 2011), (Kostiuk, 1990), (Tosi, Werner, Katz, & Gomez-Mejia, 2000), (Finkelstein & Hambrick, 1989) & (Larcker, Lambert, & Weigelt, 1991). To extent to which size is important differs from study to study, but the consensus is that size does matter.
Other literature also back the construct that industry has a significant effect on CEO compensation (Aggarwal & Samwick, 1999), (Kostiuk, 1990). Another study does puts that in perspective, as peers are selected in a way that best suits the board (Porac, Wade, & Pollock, 1999). There even exists something that is called the Lake Wobegon Effect, no board wants its CEO’ compensation to be in the bottom-half of the peer group, resulting in rising compensation for CEOs (Hayes & Schaefer, 2009).
As size and industry matter, one should especially expect from performance, as a lot of CEOs have incentive plans that are (partly) based on financial performance measures (Ittner, Larcker, & Rajan, 1997). In prior studies ROE (Finkelstein & Hambrick, 1989) and stock returns (Leone, Wu, & Zimmerman, 2006) are found to be of significant impact on CEO compensation. In another research, the effect is found to be significant, but small (Tosi, Werner, Katz, & Gomez-Mejia, 2000).
These previous paragraphs give a good view that multiple variables do affect CEO compensation. This should be considered while constructing the model. In the research design
all control variables will be discussed in more depth. The rest of this literature overview will be focused on sustainability and what the expected effect of sustainability on CEO compensation is. As mentioned in the introduction, Bansal & Roth (2000) did find three motivations for ecological innovation; competitiveness, legitimation and ecological responsibility. The first and latter one are easy to understand, the middle one is more interesting.
Bourdieu (1984) did define legitimacy as “[a]n institution, action or usage which is dominant, but not
recognized as such, that is to say, which is tacitly accepted”. In more recent research this is described as “something is considered ‘legitimate’ in a particular domain when it is generally and implicitly acknowledged as valid and thus deemed consonant with broadly accepted norms, values and beliefs of that domain. This definition draws attention to the collective construction of legitimacy, where presumptions about the social acceptability of prevailing beliefs and patterns of behavior take precedence over individual endorsement and censure” (Andon,
Free, & Sivabalan, 2014).
Legitimation would - so to say - mean that the board chose to do so to avoid interference of stakeholders or to satisfy stakeholders without believing it really is necessary or good for the company. As this is with environmental behavior, one would expect this to be also true for environmental reporting. In a study conducted in Australia on the influence of the WWF on mineral companies the authors found that the mineral companies changed their reporting behavior because of the attention of the WWF (Deegan & Blomquist, 2006). Legitimacy and stakeholder management were the main reasons this occurred (Deegan & Blomquist, 2006). In another study Guthrie & Parker (1989) did not find legitimacy as one of the primary explanations for corporate social reporting at a mining company in Australia.
Maybe the main reasons for providing CSRs aren’t clear yet, CSR has a positive effect on CEO compensation in Canadian firms (Mahoney & Thorn, 2006). This is not confirmed for US companies by the other research (Cai, Jo, & Pan, 2011), as they found a negative relation between CSR and executive total and cash compensation. In another research with US companies the authors found that the focus of CEO’ pay, short vs. long, did have a negative/positive correlation with CSR (Deckop, Merriman, & Gupta, 2006).
To align CEOs with shareholder interest, companies provide incentives for CEOs to increase social performance (Hong, Li, & Minor, 2016). At those companies, CEO’s would earn more when the social performance goes up.
China (Conyon & Lerong, 2011). The latter also found a positive relation between governance and pay-for performance. Something that is confirmed for US companies in another study (Hartzell & Starks, 2003). Concluding, as a standalone measure governance has a negative relation with total compensation and a negative relation with fixed salary as percentage of total compensation.
Another theory that is important to understand CEO compensation is the principal-agent problem. Shareholders have information asymmetry with CEOs and they can’t directly control the actions of the CEO’s, arising moral hazard (Dechow & Sloan, 1991) & (Page jr, 1991). In their study Dechow & Sloan (1991) found that CEOs cut expenditures on R&D in their final years in office. They do this to cut costs, which positively affects the short-term performance and their compensation (Dechow & Sloan, 1991). They also found that this behavior doesn’t occur when CEOs own stock themselves. This alignment with stockholders’ interest makes the CEO behave in a way that’s best for the shareholder (Dechow & Sloan, 1991). However, a side note is placed that even this alignment is a principal-agent problem itself (Bebchuk & Fried, 2003). Donaldson & David (1991) did find no support for this agency theory in their study on governance and returns.
As SRI has become more important over the past years, alignment in social performance and executive compensation is to be expected to be one of the pursuits of shareholders. This will result in CEO’ performance measures to be based on social factors to some extent. As those incentives are not fixed, this means it has a negative impact on fixed salary as a percentage of total compensation.
2.1 Expectations
To recapitulate; given this literature overview, I expect that the level of sustainability has a positive effect on the total compensation of CEOs. This is because of the goal congruence between shareholders and CEOs and the found positive relations for CSR and social performances.
On the other hand, as the prior literature suggests that governance has a negative relation with the level of compensation, it is also possible that the relation is offset or turned negative as the effect of governance might be more important or strong. Though, for measuring sustainability, an equal-weighted score will be used (see research design). Governance only accounts for a part of the score and I don’t expect it to be dominant in the overall score.
For the structure, I do expect a negative effect on the fixed salary as percentage of total compensation. When developing the research idea, I thought it to be positive, because of legitimacy and stakeholder management. However, the goal congruence between shareholders and CEOs and findings from prior literature changes that view. Especially because this study will be based on US companies, where SRI and CSR isn’t as common as in Europe (Van Duuren, Plantinga, & Scholtens, 2016).
3 Research Design
In the introduction, the research question is stated as:
“How does the level of sustainability of a company influence the level and structure of CEO compensation.”
To investigate this research question, a regression model will be set up. The research question incorporates two dependent variables, as the research question is about the level and structure of CEO compensation. To be able to measure sustainability a proxy must be chosen and relevant control variables need to be included in the model. The next paragraphs will describe all variables
3.1 Variables
Level of CEO compensation
The first dependent variable is the level of CEO compensation. This compromises all compensation CEOs receive. As the compensation for CEOs can differ a lot, the log of the value will be used, like done in other studies (Livne, Markarian, & Milne, 2011) & (Finkelstein & Hambrick, 1989). In Model 3 the level will also be used as an independent variable. Prior literature suggests that to a certain level CEOs wish salaries, but above that they don’t prefer receiving more fixed salary (Murphy, 1999).
Structure of CEO compensation
The second dependent variable is the structure of CEO compensation. CEO compensation packages contain five basic components: salary, annual bonus, payouts from long-term incentive plans, restricted option grants, and restricted stock grants (Frydman & Jenter, 2010). In addition, CEOs often receive contributions to defined-benefit pension plans, various perquisites, and, in case of their departure, severance payments (Frydman & Jenter, 2010). As the measure of the structure could be chosen in different ways, the fixed salary is most easy to grasp. The other four components are (to some extent) variable. To measure structure, the salary as percentage of total compensation will be used.
Sustainability
Sustainability could be measured in a thousand different ways. In this research, the average-weighted overall ESG-score will be used. This score does include environmentally, socially and governance factors. Because it uses over 280 KPI’s, it gives a comprehensive view on the sustainability of a company. As found in prior studies, the level and structure of executive compensation is influenced by (one of) these three factor (Core, Holthausen, & Larcker, 1999),
(Mahoney & Thorn, 2006), (Talley & Johnsen, 2004), (Conyon M. J., 1997) & (Berrone & Gomez-Mejia, 2009)
The level of structure of compensation are not influenced by sustainability solely. So, it is important to use control variables in the research to have an accurate and comprehensive model. The next nine control variables are included in the model.
Firm size
As firms come in different sizes, the size of a company can affect CEO compensation. One would expect that the larger the company, the more responsibilities a CEO has and thus raising its compensation package. Prior research found that firm size has a significant impact on the compensation of CEOs of US banks (Livne, Markarian, & Milne, 2011). Also in other researches the relation between size and executive compensation is found to be significant (Kostiuk, 1990), (Tosi, Werner, Katz, & Gomez-Mejia, 2000), (Finkelstein & Hambrick, 1989) & (Larcker, Lambert, & Weigelt, 1991). Larcker et al. (1991) do mention that the found correlation is smaller than found in prior research, but still significant. They state that other drivers are probably more important for executive’ compensation.
There are a lot of different ways to measure size; sales, market capitalization, and assets are a few examples. I will use total assets. Because assets from firms could vary from millions to billions I will use the log of total assets, as done in the research of Livne et al. (2011). As they mention, this will also control for visibility and information asymmetry in capital markets.
Firm industry
Just like size, industry could have effect on CEO compensation as well. This relationship is found in prior research (Aggarwal & Samwick, 1999) and acknowledged by Kostiuk (1990). To classify firms in different industries I will use the GIC sectors to categorize them. The use of GIC sectors to differentiate between industries is done in prior research (Benson & Davidson, 2010), (Chan, Lakonishok, & Swaminathan, 2007) & (Bhojrai, Lee, & Oler, 2003). The latter found that GICS is better than NAICS in capital market research. Where the results of Chan et al. (2007) not totally agree on GICS as being the best differentiator, as they found that grouping based on statistical cluster analysis was better in their research of return comovement.
In the model, the 11 GICS sectors are incorporated using dummy variables. One of the sectors should be chosen as reference sector, as only 10 dummies should be used in the model. Information Technology (sector 45) is chosen as the reference sector. As this sector has higher
than average usage of stock options (Anderson, Banker, & Ravindran, 2000), this is a good sector to have as a reference. The model should find significant differences from this sector.
Investment/growth opportunities
Investment/growth opportunities could have effect on the level and structure of compensation plans. As set out by Smith and Watts (1992) investment opportunities can have effect on CEO performance measures, resulting in higher variable compensation. Later research find evidence that is consistent with that view (Baber & Janakiraman, 1996). To measure the investment opportunity, I use the log of cash on balance.
Market to book ratio
Another way of measuring growth opportunities is with the market to book ratio (Livne, Markarian, & Milne, 2011) & (Houston & James, 1995). Firms with higher market to book ratios have lower financing costs and do borrow more (Chen & Zhao, 2006). That money can be used for investments and growth.
CEO tenure
CEO tenure is one of the other control variables I’ve included in the model. As CEOs work longer at a firm, this may change their compensation package or CEOs. Finkelstein and Hambrick (1989) found that the salary of CEOs will increase with tenure, but may decline after a longer period of time. After further investigating this relation they found that the total compensation did not decrease, but the mix of compensation did change. So, this variable is important for the level and structure of compensation, as also suggested by other research (Hill & Phan, 1991).
ROE
Prior research found that a higher ROE has a significant effect on CEO compensation (Finkelstein & Hambrick, 1989). Other research puts the effect of firm performance in perspective, as firm size accounts as eight times more in the variance of CEO compensation (Tosi, Werner, Katz, & Gomez-Mejia, 2000). Overall it is good to include the ROE in the model.
Leverage
Leverage is used and found significant to have a significant effect on CEO compensation in prior research (Livne, Markarian, & Milne, 2011). It does control for risk-taking by the company (Aggarwal & Samwick, 1999) & (Livne, Markarian, & Milne, 2011).
For shareholders, stock return must be the absolute number one economic pursuit. To capture the effect of stock returns on compensation this is also included in the model. Prior research found that the compensation is more sensitive to negative stock returns as to positive stock return (Leone, Wu, & Zimmerman, 2006). The total return – which includes capital gains and dividends – is used, like in the paper of Livne et al. (2011).
Year
As data from five consecutive years is used, it should be excluded that any effect found is the result of time effects. In certain years the circumstances could be different from others, this could lead to biased results, as time could be interfering. To control for time, dummy variables for each year will be included. As with the industry, one year should be the reference year and is not included as a dummy in the regression model. As no useful estimate could be made about differences between the years, simply the first year of the range, 2011, will be used.
3.2 Model
Because the research question incorporates the two independent variables, two variants of the model are necessary to analyze the effects on the level and structure of compensation. As the level of compensation could also affect the structure of compensation, a third variant on the model is introduced to take this into account. The variants of the model are named Model 1, 2 and 3. Below the formulas for each of these variants are stated. In Table 1 the definitions for each of the abbreviations are given.
Model 1
LTC = b0 + b1 ESG + b2 LTA + b3 LCASH + b4 ROE + b5 LEV + b6 MB + b7 RET+ b8 TEN
+ b9 Y2012 + b10 Y2013 + b11 Y2014 + b12 Y2015 + b13 S10 + b14 S15 + b15 S20 + b16 S25 +
b17S30 + b18 S35 + b19 S40 + b20 S50 + b21 S55 + b22 S60 + ε
Model 2
PFS = b0 + b1 ESG + b2 LTA + b3 LCASH + b4 ROE + b5 LEV + b6 MB + b7 RET+ b8 TEN
+ b9 Y2012 + b10 Y2013 + b11 Y2014 + b12 Y2015 + b13 S10 + b14 S15 + b15 S20 + b16 S25 +
Model 3
PFS = b0 + b1 LTC + b2 ESG + b3 LTA + b4 LCASH + b5 ROE + b6 LEV + b7 MB + b8 RET+
b9TEN + b10 Y2012 + b11 Y2013 + b12 Y2014 + b13 Y2015 + b14 S10 + b15 S15 + b16 S20 + b17
S25 + b18S30 + b19 S35 + b20 S40 + b21 S50 + b22 S55 + b23 S60 + ε
4 Sample Selection
4.1 Sample selectionData for the analysis is derived from three databases. Compustat and Execucomp via Wharton Research Data Services (WRDS) and Asset 4 from DataStream. After collection, the data is joined in Excel and used for analysis with SPSS.
The population for this research is defined by the availability of data from these different sources. Most of Compustat’ data is available for thousands of companies around the world. However, the Asset 4 database only has 3.500 companies worldwide in its universe and the data from the Execucomp database is limited to the 1.000 largest US companies. This limits the number of observations to a maximum of 1.000 per year.
For this research, I will use data from five years, as that’s the maximum amount of years that data is retrievable from the Asset 4 database. This means the maximum number of observations is 5.000. After joining the information from the different databases about 3,750 observations do exists in the all three databases. For a given population of 3.750 the proper n = 350 (Krejcie & Morgan, 1970).
However, because the sectors are divided in 11 different variables, the sample should be of sufficient size to give significant results in the analysis per industry. Therefore, I want to increase the n and include about 900 observations in the analysis, that’s on average 80 per industry, which should be enough observations to reach significant results. To select them out of 3.750, I used the randomize function (=RAND) of Excel. After this selection some companies had missing information for particular years, those specific years are deleted from the dataset. In the end, the sample contains 855 observations.
4.2 Definitions of variables
Table 1, on the next page, gives you an overview of the variables used in the analysis. Per variable the label for SPSS, definition and source is stated.
Table 1
Variable label Definition Source
LTC Log of all compensation like salary,
bonusses and stocks/option. Execucomp – Log TDC1
PFS Percentage of fixed salary in total
compensation Execucomp – SALARY / TDC1
ESG Equal-weighted overall score on
economics, environmental, social and governance aspects. From 0 to 100%.
DataStream Asset 4 - A4IR
LTA Log of total assets Compustat – Log AT
LCASH Log of cash on balance Compustat – Log CH
ROE The net income divided by
shareholder’s equity Compustat – NI / SEQ
LEV Leverage, total liabilities divided by
total assets Compustat – LT / AT
MB Market value divided by book value Compustat – MKVALT / SEQ
RET Total return for shareholders Execucomp – TRS1YR
TEN Tenure; total years a CEO is in charge Execucomp – FYEAR – Date Became CEO
Y20111 Fiscal year 2011 Compustat - FYEAR
Y2012 Fiscal year 2012 Compustat - FYEAR
Y2013 Fiscal year 2013 Compustat - FYEAR
Y2014 Fiscal year 2014 Compustat - FYEAR
Y2015 Fiscal year 2015 Compustat - FYEAR
S10 GICS-sector 10 Energy Compustat – GSECTOR
S15 GICS-sector 15 Materials Compustat – GSECTOR
S20 GICS-sector 20 Industrials Compustat – GSECTOR
S25 GICS-sector 25 Consumer
Discretionary Compustat – GSECTOR
S30 GICS-sector 30 Consumer staples Compustat – GSECTOR
S35 GICS-sector 35 Health Care Compustat – GSECTOR
S40 GICS-sector 40 Financials Compustat – GSECTOR
S452 GICS-sector 45 Information
Technology Compustat – GSECTOR
S50 GICS-sector 50 Telecommunication
Services Compustat – GSECTOR
S55 GICS-sector 55 Utilities Compustat – GSECTOR
Table 2
Descriptive Statistics
Mean Std. Deviation Median N
LTC 6,86 0,29 6,87 855 PFS 0,15 0,09 0,13 855 ESG 0,56 0,27 0,57 855 LTA 9,89 0,58 9,82 855 LCASH 8,55 0,71 8,60 855 ROE 0,16 0,46 0,12 855 LEV 0,51 0,27 0,54 855 MB 3,82 6,24 2,44 855 RET 0,14 0,31 0,10 855 TEN 7,01 6,53 5 855 Y20111 178 Y2012 167 Y2013 169 Y2014 172 Y2015 169 S10 78 S15 68 S20 129 S25 144 S30 38 S35 106 S40 80 S452 115 S50 7 S55 41 S60 49
1 2011 is the reference year in this analysis. ² S45 (Information Technology) is the reference sector. Therefore, both
variables are not included in the regression coefficients. They are included in this table to provide information about the number of observations.
4.3 Descriptives
In Table 2 the mean, median, standard deviation and number of observations are stated. The mean of the log of total compensation is 6,86, that is 7,2 million dollars. The 0,15 at percentage fixed salary means that the mean is 15% of total compensation consists of salary. The average ESG-score is 56% and that is close to the median of 57%. Total assets are on average 7,8 billion, cash 355 million and shareholders return on equity is 16%, with a standard deviation of 0,46.
We can see that the observations are equally spread amongst the years. With the sectors however, we do not have a lot of observations at S30, S55 and S60; Consumer Staples, Utilities and Real Estate. The telecommunication Services sector has only 7 observations. That is not enough to reach significance for that sector. I do keep that in mind when I interpret the findings of the analysis.
In Table 3 I did put in a part of the Pearson Correlation to analyze the correlations between the different variables. In this overview, the dummy variables for the years and sectors are not included. The years and sectors that that significant correlations with other variables, did all have relatively low correlations. All correlations are included in appendix 2.
From the information in Table 3 we can see that the amount of cash and assets are significantly correlated, and the level of 0,601 is relatively high. Above that, total compensation and percentage of fixed salary are negatively correlated. The level of -0,756 is high, but not greater than -0,8. Therefore we can still use it in our Model 3.
In Table 3 we see that the ESG-score is negatively related with PFS. Suggesting that the higher ESG, the lower the fixed salary is as percentage of total compensation. That is consistent with what we did expect from the theory. Next to that, we find that the ESG-score is positively related with total compensation. Which would mean the higher the ESG-score, the higher the total compensation. In the findings section, we will have a closer look at the total models and see if this relationship exists and is significant.
Table 3
Pearson Correlation
PFS LTC ESG LTA LCASH ROE LEV MB RET TEN
LTC -0,756** 1,000 ESG -0,200** 0,272** 1,000 LTA -0,258** 0,528** 0,325** 1,000 LCASH -0,226** 0,465** 0,313** 0,601** 1,000 ROE 0,009 0,009 0,069** -0,040 0,020 1,000 LEV -0,010 0,098** 0,118** 0,207** 0,075** 0,119** 1,000 MB -0,050* 0,077** 0,087** -0,075** 0,033 0,471** 0,216** 1,000 RET -0,043 0,079** 0,001 0,004 0,100** 0,096** 0,186** 0,168** 1,000 TEN 0,075** 0,047* -0,092** -0,031 -0,016 0,057** -0,036 0,055** 0,019 1,000
To distinguish between significant and non-significant values I used ** for significances at the 5% level and * for significances at the 10% level. For definitions on the variables, see Table 1.
5 Findings
In this chapter I will discuss the results. Before that, we’ll check if we can assume normality at the dependent variables, LTC and PFS.
In appendix 1 the histograms of LTC and PFS are included. It looks like the LTC has a good fit for normality. For PFS this is more difficult to examine, as the bars differ more from the normality line.
Therefore, the Q-Q plots are also included. For LTC we see confirmed that it looks like a normal distribution, as almost no difference with the straight line is observable. Also, the skewness of -0,222 and kurtosis of 1,202 don’t give a reason for that we can’t assume normality. For the Q-Q plot of PFS most observations seem to be around the straight line, but at the beginning and end they differ from it. From the Q-Q plot we can see that the data is somewhat right skewed. This is confirmed by the skewness of 3,343 in Table 4. The Kurtosis for PFS is 20,413. Meaning that observations are grouped around the mean, resulting in a high peak. Despite these not-so-perfect statistics, the observation of the histogram and Q-Q plot don’t alarm enough to stop assuming normality for this analysis.
Table 4 Descriptives Statistic Std. Error LTC Mean 6,856288 0,0098023 95% Confidence Interval for Mean Lower Bound 6,837049 Upper Bound 6,875528 5% Trimmed Mean 6,858926 Median 6,871187 Variance 0,082 Std. Deviation 0,2866241 Minimum 5,4255 Maximum 7,8168 Range 2,3913 Interquartile Range 0,3713 Skewness -0,222 0,084 Kurtosis 1,202 0,167 PFS Mean 0,151801 0,0031565 95% Confidence Interval for Mean Lower Bound 0,145606 Upper Bound 0,157996 5% Trimmed Mean 0,142220 Median 0,130314 Variance 0,009 Std. Deviation 0,0922978 Minimum 0,0000 Maximum 1,0000 Range 1,0000 Interquartile Range 0,0844 Skewness 3,343 0,084 Kurtosis 20,413 0,167
5.1 Model 1
The first model in the regression is the effect of the ESG-rating and the other variables on the total compensation of CEOs. As mentioned in the research design, the tested model is:
LTC = b0 + b1 ESG + b2 LTA + b3 LCASH + b4 ROE + b5 LEV + b6 MB + b7 RET+ b8 TEN
+ b9 Y2012 + b10 Y2013 + b11 Y2014 + b12 Y2015 + b13 S10 + b14 S15 + b15 S20 + b16 S25 +
b17S30 + b18 S35 + b19 S40 + b20 S50 + b21 S55 + b22 S60 + ε
Table 5
Model Summaryb
Model R R
Square Adjusted R Square Std. Error of the Estimate
1 ,652a 0,424 0,409 0,220
a. Predictors: (Constant), S60, RET, Y2014, TEN, LTA, S30, S50, S15, ROE, S35, S55, Y2012, S20, LEV, S40, ESG, S10, MB, Y2013, S25, LCASH, Y2015
b. Dependent Variable: LTC
In table 5 we can examine the explanatory power of this model. The r² of 0,424 indicates that 42,4% of the variation of LTC is explained by Model 1, if the model is deemed to be significant. That is a decent, not high, score. The adjusted r² is somewhat lower at 0,409.
Table 6 ANOVAa
Model Sum of
Squares df Square Mean F Sig.
1 Regression 29,781 22 1,354 27,894 ,000b
Residual 40,378 832 0,049 Total 70,159 854
a. Dependent Variable: LTC
b. Predictors: (Constant), S60, RET, Y2014, TEN, LTA, S30, S50, S15, ROE, S35, S55, Y2012, S20, LEV, S40, ESG, S10, MB, Y2013, S25, LCASH, Y2015
Table 6 shows us the model has an F-value of 27,894 and the significance level is well below 1%. So that means that we have a significant model with an r² of 42,4%. So, the next step is analyzing the individual variables in the model.
As we can see in Table 7, the multicollinearity of the variables in this model is not high. All variables score a VIF between 1 and 3. From 5 or 10 multicollinearity is regarded to be high. The constant of the model is 3,885. That is about 8k in dollars.
As this research is about the effect of sustainability on CEO compensation, the most interesting variable is ESG. In Model 1 this variable is not significant. Meaning that I did not find a relationship between the ESG-score and the total level of compensation.
The model did find some other variables that are significant. The amount of assets positively effects the total level of compensation. This is as expected by literature and found in earlier research of (Tosi, Werner, Katz, & Gomez-Mejia, 2000), (Livne, Markarian, & Milne, 2011) & (Finkelstein & Hambrick, 1989). The larger the company, the higher the compensation. For every 1 point the log of total assets increases, the log of total compensation grows with 0,255. The level of cash, which could indicate power to invest as discussed in the literature overview, also has a positive effect on LTC.
ROE, LEV and RET are not significant according to this regression. MB and TEN are. The effect of both MB and TEN is 0,003. Meaning that for every extra year of tenure and for every point higher in the M/B-ratio, the log of the total level of compensation grows with 0,003. None of the years is significantly different from the reference year of 2011, none of them are even close to significance. At the industries however, there are some significant differences from the Information Technology sector. S15 and S35 (Materials and Health Care) are just not significant at the 5% level, they are at the 10% level. Sectors 40, 50, 55 and 60 (Financials, Telecommunication Services, Utilities and Real Estate) are significant at the 5% level. The coefficients of those four sectors are all negative. Indicating that the CEOs in those industries are paid less than the CEOs working in Information Technology. Especially those working in the Financials, with an effect of -0,249. As we saw in the descriptives of Table 2, the average mean of LTC is 6,86. As -0,249 doesn’t seem much, it’s the difference between 7,2 million and 4,1 million dollars when calculating from the mean.
Table 7 Coefficientsa Model Unstandardized Coefficients Standardized Coefficients
t Sig. Correlations Collinearity Statistics
B Std.
Error
Beta Zero-order Partial Part Tolerance VIF
1 (Constant) 3,885 0,147 26,468 0,000 ESG 0,040 0,033 0,038 1,241 0,215 0,272 0,043 0,033 0,754 1,327 LTA 0,255 0,021 0,514 12,379 0,000 0,528 0,394 0,326 0,400 2,497 LCASH 0,052 0,016 0,128 3,272 0,001 0,465 0,113 0,086 0,449 2,226 ROE -0,015 0,019 -0,024 -0,774 0,439 0,009 -0,027 -0,020 0,733 1,364 LEV -0,009 0,044 -0,008 -0,209 0,835 0,098 -0,007 -0,005 0,419 2,386 MB 0,003 0,001 0,064 2,005 0,045 0,077 0,069 0,053 0,681 1,469 RET 0,027 0,029 0,029 0,944 0,345 0,079 0,033 0,025 0,736 1,358 TEN 0,003 0,001 0,059 2,170 0,030 0,047 0,075 0,057 0,932 1,072 Y2012 -0,028 0,030 -0,038 -0,911 0,362 -0,061 -0,032 -0,024 0,395 2,532 Y2013 0,010 0,031 0,014 0,324 0,746 0,034 0,011 0,009 0,367 2,722 Y2014 -0,017 0,030 -0,023 -0,548 0,584 -0,007 -0,019 -0,014 0,388 2,578 Y2015 0,008 0,032 0,011 0,242 0,809 0,067 0,008 0,006 0,357 2,804 S10 -0,006 0,036 -0,006 -0,153 0,878 0,061 -0,005 -0,004 0,529 1,889 S15 -0,068 0,035 -0,064 -1,957 0,051 -0,094 -0,068 -0,051 0,650 1,539 S20 -0,011 0,029 -0,014 -0,372 0,710 0,003 -0,013 -0,010 0,520 1,922 S25 0,046 0,029 0,060 1,590 0,112 0,070 0,055 0,042 0,494 2,024 S30 -0,001 0,043 -0,001 -0,028 0,978 0,082 -0,001 -0,001 0,719 1,392 S35 0,060 0,031 0,069 1,948 0,052 0,145 0,067 0,051 0,559 1,788 S40 -0,249 0,035 -0,253 -7,123 0,000 -0,137 -0,240 -0,187 0,546 1,830 S50 -0,059 0,088 -0,019 -0,674 0,500 0,073 -0,023 -0,018 0,904 1,106 S55 -0,184 0,045 -0,137 -4,080 0,000 -0,059 -0,140 -0,107 0,611 1,637 S60 -0,102 0,041 -0,083 -2,472 0,014 -0,113 -0,085 -0,065 0,617 1,621 a. Dependent Variable: LTC
5.2 Model 2
The second model in the regression is the effect of ESG and the other variables on the percentage of fixed salary in the total compensation of CEOs. As mentioned in the research design, the tested model is:
PFS = b0 + b1 ESG + b2 LTA + b3 LCASH + b4 ROE + b5 LEV + b6 MB + b7 RET+ b8 TEN
+ b9 Y2012 + b10 Y2013 + b11 Y2014 + b12 Y2015 + b13 S10 + b14 S15 + b15 S20 + b16 S25 +
b17S30 + b18 S35 + b19 S40 + b20 S50 + b21 S55 + b22 S60 + ε
Table 8
Model Summaryb
Model R R
Square Adjusted R Square the Estimate Std. Error of
2 ,389a 0,151 0,129 0,09
a. Predictors: (Constant), S60, RET, Y2014, TEN, LTA, S30, S50, S15, ROE, S35, S55, Y2012, S20, LEV, S40, ESG, S10, MB, Y2013, S25, LCASH, Y2015
b. Dependent Variable: PFS
In Table 8 we can examine the explanatory power of this model. The r² of 0,151 indicates that 15,1% of the variation of LTC is explained by Model 1, if the model is deemed to be significant. That is a low score. This model doesn’t explain the variation in PFS sufficiently. The adjusted r² is as low as 0,129.
Table 9 ANOVAa
Model Sum of
Squares df Square Mean F Sig.
2 Regression 1,098 22 0,050 6,725 ,000b
Residual 6,177 832 0,007 Total 7,275 854
a. Dependent Variable: PFS
b. Predictors: (Constant), S60, RET, Y2014, TEN, LTA, S30, S50, S15, ROE, S35, S55, Y2012, S20, LEV, S40, ESG, S10, MB, Y2013, S25, LCASH, Y2015
the individual variables in the model to see which ones are significant. But with the side note that the regression model only covers 15,1% of the variances.
As we can see in Table 10, the multicollinearity of the variables in this model is not high. All variables score a VIF between 1 and 3. From 5 or 10 multicollinearity is regarded to be high. The constant of the model is 0,597. 0,597 stands for 59,7% of total compensation is paid as a fixed salary.
As this research is also about the effect of sustainability on the structure of CEO compensation, the most interesting variable is ESG. In Model 2 this variable is significant at the 5% level. Meaning that I did find a relationship between the ESG-score and structure of compensation. The found relationship is consistent with what was expected from the literature. The higher the ESG-score, the lower the percentage of fixed salary. An increase of 1 in ESG (which is 100%) means the level of salary is 3% lower. The more realistic difference of 25% means 0,75% lower salary as a total of CEO compensation.
The model did find some other variables that are significant. The amount of assets negatively effects the total level of compensation, contrary to the positive effect of Model 1. Suggesting that the percentage of salary drops when firms are larger. The effect is -4,4% per 1 point in LTA. The level of cash in a company has no significant relationship in this model, so that is different from the first model.
ROE, LEV, RET, MB and TEN are not significantly correlated at the 5% level. TEN is at the 10% level with a significance of 0,054. The supposed effect is low, with a 0,1% increase per extra year of tenure.
None of the years is significantly different from the reference year of 2011, none of them are even close to significance. At the industries however, there are some significant differences from the Information Technology sector. 4 sectors are even significant at 1% level, only S30 (Consumer Staples) is significant at the 5% level, with an effect of 3,8%. S15, S25, S40 and S50 (Materials, Consumer Discretionary, Financial Services and Utilities) are the other significant sectors. With respectively effects of 4,6%, 3,2%, 3,8% and even higher effects of 6,8% and 6,5% for the last two industries. In Model 1 we found out that the sectors that differ significantly have negative effects, resulting in lower compensation. Here we see all positive effects, resulting in a higher fixed compensation.
Table 10 Coefficientsa
Model Unstandardized
Coefficients Standardized Coefficients t Sig. Correlations Collinearity Statistics
B Std.
Error
Beta Zero-order Partial Part Tolerance VIF
2 (Constant) 0,597 0,057 10,403 0,000 ESG -0,030 0,013 -0,087 -2,354 0,019 -0,200 -0,081 -0,075 0,754 1,327 LTA -0,044 0,008 -0,278 -5,515 0,000 -0,258 -0,188 -0,176 0,400 2,497 LCASH -0,003 0,006 -0,022 -0,452 0,651 -0,226 -0,016 -0,014 0,449 2,226 ROE 0,002 0,007 0,012 0,311 0,756 0,009 0,011 0,010 0,733 1,364 LEV 0,024 0,017 0,069 1,407 0,160 -0,010 0,049 0,045 0,419 2,386 MB -0,001 0,001 -0,064 -1,645 0,100 -0,050 -0,057 -0,053 0,681 1,469 RET -0,009 0,011 -0,031 -0,832 0,406 -0,043 -0,029 -0,027 0,736 1,358 TEN 0,001 0,000 0,064 1,926 0,054 0,075 0,067 0,062 0,932 1,072 Y2012 -0,003 0,012 -0,014 -0,281 0,779 0,045 -0,010 -0,009 0,395 2,532 Y2013 -0,015 0,012 -0,063 -1,198 0,231 -0,044 -0,041 -0,038 0,367 2,722 Y2014 -0,002 0,012 -0,009 -0,167 0,868 0,023 -0,006 -0,005 0,388 2,578 Y2015 -0,014 0,012 -0,061 -1,145 0,252 -0,054 -0,040 -0,037 0,357 2,804 S10 0,010 0,014 0,033 0,741 0,459 -0,082 0,026 0,024 0,529 1,889 S15 0,046 0,014 0,135 3,407 0,001 0,078 0,117 0,109 0,650 1,539 S20 0,018 0,011 0,072 1,618 0,106 -0,030 0,056 0,052 0,520 1,922 S25 0,032 0,011 0,128 2,813 0,005 0,061 0,097 0,090 0,494 2,024 S30 0,038 0,017 0,086 2,279 0,023 -0,014 0,079 0,073 0,719 1,392 S35 0,008 0,012 0,028 0,656 0,512 -0,084 0,023 0,021 0,559 1,788 S40 0,068 0,014 0,216 5,000 0,000 0,121 0,171 0,160 0,546 1,830 S50 0,035 0,034 0,034 1,023 0,307 -0,031 0,035 0,033 0,904 1,106 S55 0,065 0,018 0,150 3,672 0,000 0,042 0,126 0,117 0,611 1,637 S60 0,026 0,016 0,065 1,593 0,112 0,027 0,055 0,051 0,617 1,621
5.3 Model 3
The third model in the regression has one difference with Model 2. In this model, the log of total compensation is added in the model to test if this has a significant impact on the percentage of fixed salary in total compensation. As mentioned in the research design, the tested model is: PFS = b0 + b1 LTC + b2 ESG + b3 LTA + b4 LCASH + b5 ROE + b6 LEV + b7 MB + b8
RET+ b9TEN + b10 Y2012 + b11 Y2013 + b12 Y2014 + b13 Y2015 + b14 S10 + b15 S15 + b16 S20
+ b17 S25 + b18S30 + b19 S35 + b20 S40 + b21 S50 + b22 S55 + b23 S60 + ε
Table 11
Model Summaryb
Model R R
Square Adjusted R Square the Estimate Std. Error of
3 ,810a 0,655 0,646 0,0549212
a. Predictors: (Constant), S60, RET, Y2014, TEN, LTA, S30, S50, S15, ROE, S35, S55, Y2012, S20, LEV, S40, ESG, S10, MB, Y2013, LTC, S25, LCASH, Y2015 b. Dependent Variable: PFS
In Table 11 we can examine the explanatory power of this model. The r² of 0,655 indicates that it explains over 0,5 more than Model 2. 65,5% of the variances is explained with this model. Making it a better fit than Model 2. If the model is deemed to be significant, we have a model that has a moderately high score.
Table 12 ANOVAa
Model Sum of
Squares df Square Mean F Sig.
3 Regression 4,769 23 0,207 68,735 ,000b
Residual 2,507 831 0,003 Total 7,275 854
a. Dependent Variable: PFS
b. Predictors: (Constant), S60, RET, Y2014, TEN, LTA, S30, S50, S15, ROE, S35, S55, Y2012, S20, LEV, S40, ESG, S10, MB, Y2013, LTC, S25, LCASH, Y2015
Table 12 shows us the model has an F-value of 68,735 and the significance level is well below 1%. This is the highest score of the three models tested. So that means that we have a significant model with an r² of 65,5%. So, the next step is analyzing the individual variables in the model.
As we can see in Table 13, the multicollinearity of the variables in this model is not high. All variables score a VIF between 1 and 3. From 5 or 10 multicollinearity is regarded to be high. The constant of the model is 1,797, which means 179,7% of the compensation is fixed salary. As this is not possible, most coefficients should be negative.
As LTC is added in this model, it is indeed significant at the 1% level and that for every 1 point higher in LTC, the percentage drops with 30%, that’s quite a lot given the average of LTC of above 6.
In this model, the effect of the ESG-rating stays significant at the 5% level. The effect is somewhat lower. In Model 2 the coefficient was -0,03, now it is -0,018, it’s still negative. Meaning that the higher the rating, the lower the fixed salary as a percentage of total compensation. This is consistent with my expectations that I established after the literature overview.
The model did find some other variables that are also significant. The amount of assets no longer has a negative effect, but it is positive now with a coefficient of 0,032. The amount of cash on balance is also significant again, with a coefficient of 0,013. That’s opposite to the results of Model 2. As this model has a higher r², this model is the better fit.
ROE, LEV, RET, MB are still not significant at the 5% level, LEV is close with a significance of 5,1%. It’s effect not so much, with a coefficient of 0,021, as score are only around 0,5.
TEN is significant at the 1% level and the effect is an increase in fixed salary of 0,2% per year. That’s higher than in Model 2.
None of the years is significantly different from the reference year of 2011, the significance scores are lower than in Model 2. But still not close to the 5% level.
At the sectors however, there are some significant differences from the Information Technology sector. It is also interesting to see that it isn’t all the same sectors as in Model 2 and even the direction is different. In this model, all coefficients are positive. This is probably because this model has more explanatory power and the effect of LTC was partly incorporated in the sectors in Model 2. As the LTC is now included in the model, this should be the more realistic answer. The positive sectors are those from S15 through S35 (Materials, Industrials, Consumer Discretionary, Consumer Staples, Health Care). The effect ranges from 1,5% to 3,8%.
Table 13 Coefficientsa
Model Unstandardized
Coefficients Standardized Coefficients t Sig. Correlations Collinearity Statistics
B Std.
Error
Beta Zero-order Partial Part Tolerance VIF
3 (Constant) 1,769 0,050 35,608 0,000 LTC -0,301 0,009 -0,936 -34,882 0,000 -0,756 -0,771 -0,710 0,576 1,738 ESG -0,018 0,008 -0,051 -2,189 0,029 -0,200 -0,076 -0,045 0,752 1,329 LTA 0,032 0,006 0,203 5,805 0,000 -0,258 0,197 0,118 0,338 2,957 LCASH 0,013 0,004 0,099 3,227 0,001 -0,226 0,111 0,066 0,443 2,255 ROE -0,002 0,005 -0,011 -0,448 0,654 0,009 -0,016 -0,009 0,732 1,365 LEV 0,021 0,011 0,061 1,954 0,051 -0,010 0,068 0,040 0,419 2,386 MB -0,000 0,000 -0,004 -0,156 0,876 -0,050 -0,005 -0,003 0,678 1,476 RET -0,001 0,007 -0,004 -0,163 0,870 -0,043 -0,006 -0,003 0,735 1,360 TEN 0,002 0,000 0,119 5,629 0,000 0,075 0,192 0,115 0,927 1,078 Y2012 -0,012 0,008 -0,050 -1,542 0,124 0,045 -0,053 -0,031 0,394 2,535 Y2013 -0,012 0,008 -0,050 -1,487 0,137 -0,044 -0,052 -0,030 0,367 2,722 Y2014 -0,007 0,008 -0,030 -0,924 0,356 0,023 -0,032 -0,019 0,388 2,579 Y2015 -0,012 0,008 -0,051 -1,504 0,133 -0,054 -0,052 -0,031 0,357 2,804 S10 0,009 0,009 0,027 0,978 0,329 -0,082 0,034 0,020 0,529 1,889 S15 0,026 0,009 0,075 2,971 0,003 0,078 0,103 0,061 0,647 1,546 S20 0,015 0,007 0,059 2,089 0,037 -0,030 0,072 0,043 0,520 1,922 S25 0,045 0,007 0,184 6,326 0,000 0,061 0,214 0,129 0,492 2,030 S30 0,038 0,011 0,085 3,541 0,000 -0,014 0,122 0,072 0,719 1,392 S35 0,026 0,008 0,092 3,377 0,001 -0,084 0,116 0,069 0,557 1,796 S40 -0,007 0,009 -0,021 -0,747 0,455 0,121 -0,026 -0,015 0,515 1,942 S50 0,017 0,022 0,017 0,790 0,430 -0,031 0,027 0,016 0,904 1,107 S55 0,009 0,011 0,022 0,818 0,414 0,042 0,028 0,017 0,599 1,670 S60 -0,005 0,010 -0,013 -0,489 0,625 0,027 -0,017 -0,010 0,612 1,633 a. Dependent Variable: PFS
6 Conclusion
The goal of this thesis is the assess the effect of sustainability on the level and structure of CEO compensation. Based on prior literature the prediction is that sustainability has a positive effect on the level and a negative effect on the structure. As the r² of Model 2 is too low with 15,1%, this conclusion is based on Models 1 and 3.
For the level of CEO compensation Model 1 did not find a significant relation with sustainability. This can have multiple causes. One of them that the effect does not exist. Another reason that’s discussed in the literate overview could be that the positive effect of social performance is offset by the negative effect that is found in multiple studies. Further research with separate variables could possible give a solution.
Model 1 did find a positive effect of assets, cash, market-to-book ratio and tenure on the level of compensation. That is consistent with prior research as discussed in this thesis. No effect is found from a specific year, this may be because no big impact event or crisis occurred during this period. Consistent with literature the industry does have an impact on the level. CEOs of Financials and Utilities seem to earn significantly less than other sectors. The Financials are maybe somewhat remarkable, possibly that can be the heritage of the financial crisis.
For the structure of CEO compensation Model 3 did find a negative effect of sustainability. That is consistent with the prediction. However, the effect is relatively small. For every 1pp increase in the ESG-score, the fixed salary as percentage of total compensation drops with 0,018%. With an r² of 65,5% this models has moderately good explanatory power. As no in-depth research is done on the why this negative relation exists, we can only assume that the prior literature is right on those reasons of, for example, principle-agency problems and goal congruence, the pay-for-performance idea.
In Model 3 the single most important variable that explains the structure is the level of compensation. The higher the compensation, the lower the percentage of fixed salary. This effect could be explained by the idea that above a certain level CEOs don’t ‘need’ to receive more fixed salary.
Other significant results occurred at the assets, cash and tenure. Years were still not significant and at the industries some were significant. The Consumer Discretionary sector seems to pay the highest fixed percentages.
6.1 Limitations and recommendations
This research has multiple limitations, which directly lead to recommendations for further research. I mention some of the major ones.
1. How to measure sustainability?
In this research, an equal-weighted overall ESG-score is used to proxy for sustainability. There are more ways to measure sustainability, different metrics even that measure the same. Also, one of the factors can offset the effect of the other. Meaning that one or more of the factors incorporated in the ESG-score can be of significant impact, but that we don’t find that because the use of the overall score. Future research can be done on focus on this differences by testing different measures for sustainability and including each aspect as a standalone variable.
2. Industries; best-in-class
ESG-providers usually use best-in-class approaches, as opposed to absolute sustainability figures. This means that in relatively polluting sectors, one could still reach a very high score, just by being better than competitors. This means that industries influence the results. This should be further investigated as it’s note sure that this best-in-class approach corresponds with the actual effects.
3. Limited to listed largest listed US companies
Because of data availability this study is conducted on the largest US listed companies. For smaller or private companies or companies outside of the US the results can differ. Especially a research in Europe could be interesting, as SRI investing is more important there and compensation programs are somewhat soberer.
4. Limited timeframe
Data availability restricted this research on the last five years. Using a longer timeframe of, for example, 20 years can grasp the effects of sustainability even better. As the timeframe will start before sustainability became as important as nowadays.
5. Structure limited to fixed
The only aspect of compensation assessed for the structure is salary. To give a more comprehensive view, all aspects of compensation - such as bonusses and stock options – should be investigated. It can be the case that companies with higher ESG-score have more option incentives and less cash bonusses.
References
Aggarwal, R. K., & Samwick, A. A. (1999). Executive Compensation, Strategic Competition, and Relative Performance Evaluation: Theory and Evidence. The Journal of Finance, 54(6), 1999-2043.
Anderson, M. C., Banker, R. D., & Ravindran, S. (2000). Executive Compensation in the Information Technology Industry. Management Science, 46(4), 530-547.
Andon, P., Free, C., & Sivabalan, P. (2014). The legitimacy of new assurance providers: Making the cap fit. Accounting, Organizations and Society, 39, 75-96.
Baber, W. R., & Janakiraman, S. N. (1996). Investment opportunities and the structure of executive compensation. Journal of Accounting and Economics, 21(3), 297-318.
Bansal, P., & Roth, K. (2000). Why Companies Go Green: A Model of Ecological Responsiveness. The Academy of Management Journal, 43(4), 717-736.
Bauer, R., Guenster, N., & Otten, R. (2004). Empirical evidence on corporate governance in Europe: The effect on stock returns, firm value and performance. Journal of Asset
Management, 5(2), 91-104.
Bebchuk, L., & Fried, J. M. (2003). Executive Compensation as an Agency Problem. Journal of
Economic Perspectives, 17(3), 71-92.
Benson, B. W., & Davidson, W. N. (2010). The Relation between Stakeholder Management Firm Value, and CEO Compensation: A Test of Enlightened Value Maximization. Financial
Management,, 39(3), 929-963.
Berrone, P., & Gomez-Mejia, L. R. (2009). Environmental Performance and Executive Compensation: An Integrated Agency-Institutional Perspective. The Academy of
Management Journal, 52(1), 103-126.
Bhojrai, S., Lee, C. M., & Oler, D. K. (2003). What's My Line? A Comparison of Industry Classification Schemes for Capital Market Research. Journal of Accounting Research, 41(5), 745-774.
Bourdieu, P. (1984). Distinction: A social critique of the judgment of taste. Harvard University
C.ore, J. E., Guay, W. R., & Rusticus, T. O. (2006). Does Weak Governance Cause Weak Stock Returns? An Examination of Firm Operating Performance and Investors Expectations.
The Journal of Finance, 61(2), 655-687.
Cai, Y., Jo, H., & Pan, C. (2011). Vice or Virtue? The Impact of Corporate Social Responsibility on Executive Compensation. Journal of Business Ethics, 104(2), 15-173.
Chan, L. K., Lakonishok, J., & Swaminathan, B. (2007). Industry Classifications and Return Comovement. Financial Analysts Journal, 63(6), 56-70.
Chen, L., & Zhao, X. (2006). n the relation between the market-to-book ratio, growth opportunity, and leverage ratio. Finance Research Letters, 3(4), 253-266.
Clark, G., Feiner, A., & Viehs, M. (2015, March 5). From the Stockholder to the Stakeholder: How
Sustainability Can Drive Financial Outperformance. Retrieved from SSRN:
https://ssrn.com/abstract=2508281
Conyon, M. J. (1997). Corporate governance and executive compensation. International Journal of
Industrial Organization, 15(4), 493-509.
Conyon, M. J., & Lerong, H. (2011). Executive compensation and corporate governance in China. Journal of Corporate Finance, 17(4), 1158-1175.
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. David, P., Kochhar, R., & Levitas, E. (1998). The effect of institutional investors on the level and
mix of CEO compensation. Academy of Management Journal, 41(2), 200-208.
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.
Deckop, J. R. (1988). Determinants of Chief Executive Officer Compensation. Industrial and
Labor Relation Review, 41(2), 215-226.
Deckop, J., Merriman, K., & Gupta, S. (2006). The Effects of CEO Pay Structure on Corporate Social Performance. Journal of Management, 32(3), 329-342.
Deegan, C., & Blomquist, C. (2006). Stakeholder influence on corporate reporting: An exploration of the interaction between WWF-Australia and the Australian minerals industry. Accounting, Organizations and Society, 31(4-5), 343-372.
Donaldson, L., & David, J. H. (1991). Stewardship Theory or Agency Theory: CEO Governance and Shareholder Returns. Australian Journal of Management, 16(1), 49 - 56.
Drobetz, W., & Zimmerman, H. (2004). Corporate Governance and Expected Stock Returns: Evidence from Germany. European Financial Management, 10(2), 267-293.
Finkelstein, S., & Hambrick, D. C. (1989). Chief Executive Compensation: A Study of the Intersection of Markets and Political Processes. Strategic Management Journal, 10(2), 121-134.
Frydman, C., & Jenter, D. (2010). CEO Compensation. Annual Review of Financial Economics, 2, 75-102.
Galema, R., Plantinga, A., & Scholtens, B. (2008). The stocks at stake: Return and risk in socially responsible investment. Journal of Banking & Finance, 32(12), 2646-2654.
Gompers, P., Ishii, J., & Metrick, A. (2003). Corporate Governance and Equity Prices. The
Quarterly Journal of Economics, 118(1), 107-155.
Guthrie, J., & Parker, L. D. (1989). Corporate Social Reporting: A Rebuttal of Legitimacy Theory. Accounting & Business Research, 19(76), 343-352.
Hartzell, J. C., & Starks, L. T. (2003). Institutional Investors and Executive Compensation. The
Journal of Finance, 58(6), 2351-2374.
Hayes, R. M., & Schaefer, S. (2009). CEO pay and the Lake Wobegon Effect. Journal of Financial
Economics, 94(2), 280-290.
Hill, C. W., & Phan, P. (1991). Ceo Tenure As a Determinant of Ceo Pay. The Academy of
Management Journal, 34(3), 707-717.
Hong, B., Li, Z., & Minor, D. (2016). Corporate Governance and Executive Compensation for Corporate Social Responsibility. Journal of Business Ethics, 136(1), 199-213.
Houston, J. F., & James, C. (1995). CEO compensation and bank risk Is compensation in banking structured to promote risk taking? Journal of Monetary Economics, 36(2), 405-431. Ittner, C. D., Larcker, D. F., & Rajan, M. V. (1997). The Choice of Performance Measures in
Annual Bonus Contracts. The Accounting Review, 72(2), 231-255.
Konar, S., & Cohen, M. (2001). Does the Market Value Environmental Performance? The Review
