The causal relation between top executive compensation and
firm performance: A European investigation
Master Thesis
University of Groningen – Faculty of Economics and Business
Master of Science - International Financial Management
Abstract
This study tries to create a causal pay-performance relation by using as sample the top 80 listed non-financial firms of the European Union. After conducting three different fixed-effects panel analyses (firm, industry, and group) and adding up to two lates (t+1 and t+2),
industry and group fixed-effects reveal that the direction of causality is stronger from performance to pay. This effect is robust to adding both an accounting-based (return on assets) and a market-based (annual stock returns) firm performance measure. The study shows that only using contemporaneous relations underestimates the pay-performance link and the major role of corporate governance.
Keywords: pay-performance causal relation, fixed effects, contemporaneous and late relations, corporate governance
1. Introduction
The relation between firm performance and executive compensation is a topic that is investigated numerous times in the literature. From the statement of Jensen and Murphy (1990) that CEOs are “paid like bureaucrats” to Mishel and Davis (2014) who show that from 1978 to 2000 the CEO pay had a growth of 1,279%, it is obvious that the pay-performance relation is a long-debated topic fuelled by contrasting outcomes.
Those contrasts have been the driving force for new theories trying to explain the abnormal CEO pay rise after the 1970s, for an ample literature with a main focus on the United States, and for different analytical techniques to be implemented.
In contrast, Duffhues and Kabir (2008) use data from Dutch firms from 1998 to 2001 and do not find any positive link even when they use lagged firm performance measures.
One reason for those different outcomes is the issue of causality. The well known conclusion of Gabaix and Landier (2008) that the six-fold increase in U.S. CEO pay between 1980 and 2003 is explained by the six-fold growth in market capitalization was criticized by Frydman and Saks (2010) on grounds of failing to create a true causal relation. The Generalized Method of Moments (GMM) (Lilling, 2006) and the two stages least squares (Lee and Chen, 2011) are examples of regression techniques applied to create a causal pay-performance relation.
The contrasting results, the widespread literature on US firms, and its lack of considering the issue of causality have motivated me to focus on the top 80 listed non-financial firms of the European Union and try to create a causal pay-performance relation.
In order to do that, I conduct three different fixed-effects panel analyses and test not only the contemporaneous pay-performance relation (both compensation and performance are tested in year t0) but also I add up to two lates to examine the causal pay-performance relation
in the future.
Considering all the above, my research question is:
Is there a causal pay-performance relation for the top 80 listed non-financial firms of the European Union?
The structure of the study is the following: Section 2 will present the progress of top executive pay through time, present several theories trying to explain this progress and ends up with the empirical evidence on the pay-performance relation and the proposed hypotheses. Section 3 provides details on the sample, the model and the variables used. Section 4 shows the results of this study and Section 5 explains what are the interpretations of those results. Section 6, the conclusion part, sums up the study and reveals its limitations and suggestions for future research.
2. Literature Review
2.1 The progress of top executive pay
Executive compensation is a topic that is fuelled by controversy (Bereskin and Cicero, 2013). The 1970s are the starting point of the surge of executive compensation and the beginning of an ample literature mainly focused on the United States (Frydman and Jenter, 2010; Hayes and Schaefer, 2009; Bereskin and Cicero, 2013; Gabaix and Landier, 2008).
On the one hand, Murphy (1986) after gathering compensation policies data of almost 1,200 large U.S. corporations for the period 1975-1984, concludes that “top executives are worth every single nickel they get” by showing a strong and positive relation between executive pay and firm performance. On the other hand, Mishel and Davis (2014) report that the CEO compensation, inflation-adjusted, increased by a staggering 937% from 1978 to 2013, a number that is more than double the stock market growth during this period. Therefore, the authors suggest policies to reduce this abnormally and non-justifiable escalating executive pay and put a limit on the “rent-seeking” appetite of the CEOs.
Fig. 2.1 Median compensation of CEOs and other top officers from 1936 to 2005. This figure shows the
median level of compensation in a sample of the three highest-paid officers in the largest 50 firms in 1940, 1960, and 1990. Total compensation is composed of salary, bonuses, long-term bonus payments (including grants of restricted stock), and stock option grants (valued using Black-Scholes). Source: Frydman and Jenter (2010).
Fig. 2.2 CEO compensation and the S&P Index (in 2013 dollars), 1965-2013. This figure shows how CEO
pay fluctuates in tandem with the stock market as measured by the S&P 500 Index. CEO annual compensation is computed using the “options realized” compensation series, which includes salary, bonus, restricted stock grants, options exercised, and long-term incentive payouts for CEOs at the top 350 U.S. firms ranked by sales. Source: Mishel and Davis (2014).
However, not only the CEOs show this upward trend in their pay. Studies such as Mishel and Kimball (2015) and Gould (2016) have found that wage inequality has risen since 1979 and the gap between top and low earners is still large. More specifically, Mishel and Kimball (2015) conducted an analysis with data from 1979 to 2014 by separating households according to their level of income. They find out that annual wages of the top 0.1% increased from 1979 to 2014 by 324.4% while the bottom 90%’s growth in pay was only 16.7%. Gould (2016) with a smaller and more recent data sample (2000-2015) reports similar results with the ones mentioned above. By dividing American workers into wage percentiles, Gould (2016) illustrates that through this 16-year period the gap between the middle and the bottom percentiles remains relatively stable, while the gap between the two extremes has grown.
Bob Budley, raised his total compensation by more than 20% to 12.74 million U.S. dollars in 2014 (Shell CEO’s total pay soars to $25.7 million, 2015). The paradox in both cases is that the oil prices during this year were going down. The report of the High Pay Centre, an independent non-party think tank, called “no change there, then” (2016) analyzes the UK’s largest listed companies. It highlights the CEO of Reckitt Benckiser, whose pay in 2015 grew over 10 million pounds more than 2014, which is another indication that “top pay is still rising fast”.
2.2 Theories and other explanations concerning the rise of top executive pay
There is no definitive explanation as to why the top executive pay surged after the mid-1970s. Several theories and explanations have been introduced, mainly to justify the ascending CEO compensation. Four theories have been widely used in the literature: managerial rent extraction, optimal contracting, the scale of firms, and increasing returns to general rather than specific skills (Frydman and Saks, 2010; Bruce and Skovoroda, 2015). Alternative explanations are the “Lake Wobegon Effect” (Hayes and Schaefer, 2009) and the existence of an exogenous shock that leads to the rise of the CEO remuneration through “contagion” (Bereskin and Cicero, 2013). Table 2.1 describes those six theoretical explanations concerning the rise of top executive pay.
governance got weaker through time (Shue and Townsend, 2015) nor were the executives able to “steal” such huge amounts as “80% of their compensation” as Gabaix and Landier (2008) state in their study, as a way to downsize the importance of the theory as being the sole reason for the CEO pay growth.
Optimal contracting is the theory that tries to justify the top executive compensation by using the concept of unobservable factors, such as CEO talent (Bruce and Skovoroda, 2015) and risk generated through a greater use of incentive pay (Frydman and Saks, 2010). Bruce and Skovoroda (2015) present a detailed description of the theory. Optimal contracting is built on the assumptions of a fully competitive market and fully observable CEO talent and follows the arms-length concept, which awards their CEOs exactly in accordance with their talent. However, apart from the extremity of those assumptions, Gabaix and Landier (2008) reveal that “if we rank CEOs by talent, and replace the CEO number 250 by the number one CEO, the value of his firm will increase by only 0.016%” and Frydman and Saks (2010) do not find a strong relation between risk and the rise in executive pay.
Gabaix and Landier (2008) try to give an answer to the puzzle of the escalating executive pay by using the firm size as the main determinant of the rising CEO pay. They conclude that the six-fold increase in U.S. CEO pay between 1980 and 2003 is explained by the six-fold growth in market capitalization over the same period. A criticism of their academic paper is made by the work of Frydman and Saks (2010). They find that the link between compensation and market value of the firm is not so strong from the late 1940s to the 1970s. Especially during the 1950s and 1960s, the considerable increase of market capitalization is not followed by an equal increase in the executive pay. Therefore, the scale of firms cannot solely be used as an explanation for the upward trend in the executive remuneration.
Alternative explanations have also been suggested as a potential cause for the booming CEO pay. Hayes and Schaefer (2009) create a game-theoretic model and show that the “Lake Wobegon Effect” can occur. The Lake Wobegon Effect materializes when one firm distorts upward the CEO pay in an attempt to make “the company look strong”. Overpaying a CEO has a determining effect on the value of the firm, as having a CEO that is paid “below average” can seriously tarnish the public’s perception about the firm.
Bereskin and Cicero (2013) find that a governance shock created by legal developments can provide a justification for the ever increasing CEO pay since the late 1990s. Building on the Delaware case law that gave executives “greater ability to defend against hostile takeover bids”, the authors follow the “contagion hypothesis” proposed by Gabaix and Landier (2008); the Delaware-incorporated firms increased their CEO compensation and the non-Delaware ones, when the legal changes took place, did the same as they realized that this trend was prevalent among other firms in their industry.
However, even though the theories are not “mutually exclusive” (Bruce and Skovoroda, 2015) and “combining these explanations might have greater explanatory power” (Frydman and Saks, 2010) the empirical data are not fully consistent with any theory proposed (Frydman and Jenter, 2010) as none of them can completely explain the surge of top executive pay.
Table 2.1
Theoretical explanations concerning the rise of top executive pay
The table shows the most widely used theories in literature regarding the growth in CEO pay. Every theory is accompanied by a short explanation. Supporting and contrasting evidence are shown for each theory.
Theory Supporting evidence Contrasting evidence
1. Managerial rent extraction Frydman and Saks (2010) Gabaix and Landier (2008) Bruce and Skovoroda (2015) Shue and Townsend(2015)
2. Optimal contracting Bruce and Skovoroda (2015) Gabaix and Landier (2008)
Frydman and Saks (2010)
3. Scale of firms Gabaix and Landier (2008) Frydman and Saks (2010)
4. Change in managerial skills Frydman and Jenter (2010) Frydman (2005)
5. Lake Wobegon effect Hayes and Schaefer (2009)
6. Contagion hypothesis Bereskin and Cicero (2013) Gabaix and Landier (2008)
2.3 Empirical evidence on the pay-performance relation
I will present studies that employ the contemporaneous and late-lag relations of CEO pay and firm performance from several countries in an attempt to expand this mixed literature. A tie is contemporaneous when both the dependent and independent variables are checked in year t0.
But, authors also try to explore the same relation in the future (late) or in the past (lag) by adding lates or lags to the dependent and/or independent variables.
Most of the literature has focused on the United States. Cooper, Gulen, and Rau (2014) with a sample over 1994 to 2011, try to show the relation between excess incentive pay and future abnormal returns. They identify incentive compensation as the difference between total compensation and total cash compensation and sort the companies into “industry and size benchmark adjusted CEO compensation deciles”. Their results report that there is a strong negative correlation between incentive pay and stock returns for the top three deciles and an insignificant one for the lower seven deciles. By measuring overconfidence by the number of unexercised options, the authors conclude that CEOs who are highly compensated, overconfident and with long tenure earn “three-year abnormal returns of -22.4%”. The latter conclusion is consistent with the “skimming” view, as discussed above. Stock returns are also used as a firm performance measure in the study of Shaw and Zhang (2010). They present a timeline where in year t-1 the CEO signs his payment contract, at the
end of the year t, the year’s earnings and returns are realized and afterwards (at year t+1) the
compensation committee decides on the final remuneration awarded to the CEO according to the performance of the year t. By using the CEO cash compensation as their dependent variable, they show that no matter how low the stock returns are, the CEO cash remuneration is unaffected. Therefore it is supported that CEO cash compensation is not punished for poor firm performance. However, Yang, Dolar, and Mo (2014) by using both an accounting-based firm performance measure (return on assets) and a stock-based one (stock returns) before and after the financial crisis of 2007-2008, draw a different conclusion from the above studies. After checking for total compensation, cash-based compensation, and stock-based compensation, they underline that the return on assets is strongly positively related to all the previous forms of remuneration, whereas stock returns is not statistically significant.
36 companies from New Zealand and use lagged and simultaneous stock returns and Tobin’s Q in order to “represent the past, present and future performance of the firm” . They find that all the previously mentioned performance measures have a strong and significantly positive influence only on total compensation and not on cash compensation. With a sample of 390 UK non-financial firms for the period 1999-2005, Ozkan (2011) measures the firm performance by the shareholder return and finds that both cash compensation (the sum of salary and bonus) and total direct compensation (the sum of salary, bonus, value of stock options, and long term incentive plans granted during the year) are positively correlated with this performance metric. On the contrary, Duffhues and Kabir (2008) do not find a positive sign on the pay-performance relation for the Netherlands. Their sample consists of Dutch listed companies from 1998 to 2001 and they employ accounting-based (return on assets, return on sales), capital market-based (annual stock return) and a hybrid firm performance measure (Tobin’s Q) to test if the executive compensation is a reflection of the corporate performance. However, after testing for both contemporaneous and lagged firm performance relations, neither of their dependent variables (cash compensation and total compensation) has a positive sign and this is accounted to the “rent-seeking” behaviour of the top executives. Table 2.2 includes a summary of all the previously discussed studies.
Table 2.2
Summary of previous studies using compensation as dependent variable.
The table shows a summary of all the studies mentioned in Section 2.3 of the literature review which have as dependent variable the top executive compensation.
Compensation as dependent variable
Sample Time Dependent Independent
Authors characteristics period Method variables variables Findings
Cooper et al. listed companies, US 1994-2001 panel incentive future stock - (2014) regression compensation returns
Shaw and N=14,632, US 1992-2005 pooled OLS CEO cash stock returns 0 Zhang (2010) compensation
Yang et al. N=3,286, US 1992-2011 fixed effects total, cash, stock returns 0 (2014) and stock ROA + compensation
Gunasekaragea N=36, New Zealand 1998-2000 OLS cash and total lagged and 0 and Wilkinson compensation simultaneous +
(2002) stock returns
Tobin’s Q
Ozkan N=390, UK 1999-2005 fixed effects total and cash shareholder + (2011) non-financial firms compensation return +
Duffhues and N=145, Netherlands 1998-2001 pooled OLS total and cash return on assets - Kabir (2008) compensation return on sales -
stock returns
Tobin’s Q
equity-based pay is inefficient and it is something that compensation committee members should take into consideration. Table 2.3 shows a summary of those two academic papers.
Table 2.3
Summary of previous studies using performance as dependent variable.
The table shows a summary of all the studies mentioned in Section 2.3 of the literature review which have as dependent variable the firm performance.
Performance as dependent variable
Sample Time Dependent Independent
Authors characteristics period Method variables variables Findings
Li et al. N=308, US 1993-2005 quantile ROE CEO stock + (2015) non-financial firms regression Tobin’s Q compensation +
model
Kuo et al. N=216, US 1994-2008 threshold ROE stock-based 0 (2013) non-financial firms regression to total
model compensation
1995-2004, Lee and Chen (2011) show that Tobin’s Q (as a performance measure) and CEO compensation (salary and bonus) are positively related.
2.4 Hypotheses
Section 2.3 was about what the literature has offered so far to the pay-performance relation. However, the different approaches adopted by each author lead to contradictory outcomes in the literature. A combination of some of the techniques applied from previous works can provide a more refined result.
The use of lagged relations is widespread in the literature. As presented in Section 2.3, authors such as Gunasekaragea and Wilkinson (2002), Ozkan (2011), and Duffhues and Kabir (2008) add lags to test the influence of the past on the pay-performance relation. However, performance can also affect the future compensation. The much quoted work of Jensen and Murphy (1990) finds that for every $1,000 increase in firm value, the CEO compensation goes up by just $3.25. Therefore, CEOs are characterized to be “paid like bureaucrats”. Pay can be considered as a reward for earlier performance, which leads to the first hypothesis:
Hypothesis 1: Performance in year t0 determines future CEO pay.
indication of future performance (Banker, Darrough, Huang, and Plehn-Dujowich, 2013). Therefore, the following is suggested:
Hypothesis 2: CEO pay in year t0 determines future performance.
The literature has also revealed other patterns that do not follow the causal pay-performance link. Theories such as the “Lake Wobegon Effect” of Hayes and Schaefer (2009) and the Delaware case of Bereskin and Cicero (2013) point out that firms do not consider corporate performance to be a essential determinant for a higher CEO pay. So, the last hypothesis is:
Hypothesis 3: There is no causal pay-performance relation.
3. Data and methodology
3.1. Sample
However, due to missing data for the rest top executives, and, in some cases, because of CEO being the sole executive, this study is focusing only on CEOs. Furthermore, I include remuneration data for the current CEO and his/her predecessor, so that a comparison between “new” and “previous” CEO can be made. The final sample, after excluding companies with missing data, consists of an unbalanced panel of the top 80 listed non-financial firms of European Union for the period 2005-2014, a time period that includes data before and after the financial crisis of 2008, thus giving me the opportunity to extract useful outcomes.
For the remaining variables, two databases are used. For the firm performance measure (return on equity), firm-level data (firm size and firm age) and control variables (leverage and dividend payout ratio) the Worldscope database is used. Data concerning CEO characteristics (CEO age and CEO tenure) and corporate governance variables (board size and board independence) is extracted by BoardEx. Because of the presence of outliers, all the variables are winsorized at the 1% and 99% level.
The full sample consists of ten countries: Denmark, Finland, France, Germany, Great Britain, Ireland, Italy, the Netherlands, Spain, and Sweden. Although all those countries are part of the European Union, their backgrounds differ significantly from each other. Therefore, I split my sample into two categories: companies that are part of the United Kingdom (33 from Great Britain plus one from Ireland) and companies that are part of the Continental Europe (all the rest countries, which equals to a subtotal of 46 companies). The rest of this study will reveal the significance of this differentiation.
3.2. Dependent variables
The natural logarithm of CEO remuneration is the dependent variable of the first regression model. There are five different dependent variables that represent the natural logarithm of CEO pay: total compensation, annual base salary, cash-based variable remuneration, equity-linked variable remuneration, and other forms of remuneration. Although total and cash compensation (Duffhues and Kabir, 2008; Ozkan, 2011; Gunasekaragea and Wilkinson, 2002) and equity compensation (Yang, Dolar, and Mo, 2014) have been widely used in literature, annual base salary and other forms of remuneration are two pay components that are rarely used as dependent variables. Even though empirical results lead to eventually drop those two variables, this study shows why total, cash, and equity compensation are the only compensation parts to be included (see Section 4.5).
Return on equity is the metric chosen to represent firm performance and the dependent variable of the second regression model. Calculated as the ratio of net income to the shareholders’ equity, I follow previous literature and use this particular accounting-based measure because “top executive payment is based mainly on summary accounting performance measures” (Kuo, Li, and Yu, 2013) and it is “a true bottom-line measure of performance” (Li, Yang, and Yu, 2015).
Table 3.1 reports all the variables used in the study divided into three categories (dependent, independent, and control variables) and accompanied by a brief description.
3.3. Independent variables
Papers written by Palia (2001) and Lee and Chen (2011) report regression models with several variables that have an effect on both CEO compensation and firm performance. I follow those studies and the “common” independent variables used are: firm size, board size, CEO tenure (Lee and Chen, 2011), and board independence (Ozkan, 2011).
Corporate governance variables are connected with the effectiveness of the board of directors in terms of monitoring. Lee and Chen (2011) find out that board size has a different impact on CEO compensation and firm performance accordingly. A large board size is weak in terms of effectiveness, which increases the influence of the CEO on the board enabling him/her to extract rents as supported by the managerial rent extraction view. The “rent seeking” appetite of the CEO creates agency problems that lead to lower corporate value.
Board independence is defined as the ratio of non-executive to executive directors that serve on the board. Although the presence of independent directors is associated with better monitoring and higher firm performance, Coles, Lemmon, and Wang (2009) report a negative relation between Tobin’s Q and board independence. In terms of CEO compensation, the results are mixed. Core, Holthausen, and Larcker (1999) argue that there is a negative correlation between CEO pay and the number of independent directors on the board. However, the exact opposite is proven by Fernandes, Ferreira, Matos, and Murphy (2012). Finally, CEO tenure can be regarded as an accumulation of “experience” (Lee and Chen, 2011) or as a means of greater power for “entrenched” executives (Ozkan, 2011). Higher levels of experience equal to a better handling of crucial situations which can be translated to higher firm performance. However, this superior managerial experience can also be used as a tool to manipulate the board of directors and eventually result to an abnormally high CEO compensation without an equally high performance.
3.4. Control variables
Table 3.1
Variable names and definitions.
The table shows all the variables used in the study accompanied by a short definition. The variables’ operationalization is given in the parentheses. CEO compensation data and characteristics (CEO age and CEO tenure) and corporate governance variables (board size and board independence) are provided by BoardEx database. Data on return on equity, firm size, firm age, leverage, and dividend payout ratio is extracted by Worldscope database. The horizontal lines make a distinction between the variables used in each of the two regression models of the study (see Section 3.5).
Variable Definition
Dependent variables
1. Annual base salary Natural logarithm of the fixed remuneration awarded to a CEO
(Salary) annually
2. Cash-based variable compensation Natural logarithm of the variable remuneration settled in cash
(Cash_C) payments (bonus)
3. Equity-linked variable compensation Natural logarithm of the sum of shares, share options and long (Equity_C) term incentive plans (LTIPs)
4. Other forms of compensation Natural logarithm of the sum of defined contribution pension (Other_C) plans and other benefits
Total compensation (Total_C) Natural logarithm of the sum of all four forms of compensation Return on equity (RoE) Ratio of net income to shareholders’ equity
Independent variables
Board size Sum of executive and non-executive directors serving on the
(BoardSize) board
Board independence Ratio of non-executive to executive directors serving on the
(BoardIndie) board
Firm size (FirmSize) Natural logarithm of total assets
CEO tenure (CEOtenure) Number of years an executive has worked in the firm as a CEO
Control variables
CEO age (CEOage) CEO’s age in years Firm age (FirmAge) Firm’s age in years
Leverage (Leverage) Ratio of total liabilities to total assets
system and incentivize their CEO by awarding him/her with equity-linked rather than cash-based compensation in an attempt to boost firm performance.
The control variables for the second regression are leverage and dividend payout ratio. Leverage has a positive and dividend payout ratio a negative association with firm performance (Lee and Chen, 2011). A high leverage means that the company has the needed funds to make investments which will lead to an improved firm value. On the other hand, dividend payout ratio’s adverse relation is explained on the basis of higher “debt agency costs”.
3.5. Model specification
The goal of the study is to create the causal relation between CEO pay and performance. Therefore, I use both the CEO compensation and firm performance as dependent variable and estimate the following empirical models:
Ln(COMP)it = α + β1RoEit + β2BoardSizeit + β3BoardIndieit
+ β4FirmSizeit + β5CEOtenureit + β6CEOageit + β7FirmAgeit + ε1it (1)
RoEit = γ+ δ1ln(COMP)it + δ2BoardSizeit + δ3BoardIndieit
+ δ4FirmSizeit + δ5CEOtenureit + δ6Leverageit + δ7DIVit + ε2it (2)
where ln(COMP)is the natural logarithm of the total compensation (Total_C) and any other CEO pay component (annual base salary (Salary), cash-based variable remuneration (Cash_C), equity-linked variable remuneration (Equity_C), and other forms of remuneration (Other_C)) and RoE is the return of equity which stands for the firm performance metric for the ith firm at time t.
form) and the size of the board and a negative one in regards with firm performance. BoardIndie, as discussed above, shows mixed results according to the literature, and therefore it is added in order to examine its effect on the top 80 non-financial listed companies of the European Union. I follow the example of Li, Yang, and Yu (2015) and estimate FirmSize as the natural logarithm of the total assets and expect a positive association for both of the dependent variables. Finally, the effect of CEOtenure on the causal pay-performance tie is checked, in order to test whether experience is translated into higher firm value or into a way for the CEO to extract more rents from the firm.
The control variables for Eq. (1) are CEOage and FirmAge and for Eq. (2) Leverage and DIV for the ith firm at time t. ε1it and α and ε2it and γ are the error terms and regression
intercepts respectively for Eq. (1) and Eq. (2).
Joskow and Rose (1994) state in their study that assuming the pay-performance relation is only contemporaneous underestimates its true tie. Contemporaneous relations are observed in the Eq. (1) and Eq. (2). Both the dependent and independent variable are tested in year t0. But, performance (CEO compensation) in year t0 can be positively or negatively
related to CEO compensation (performance) in earlier years (t-1, t-2) or in later years (t+1, t+2).
I take advantage of the unbalanced panel within a ten-year period, and examine the causal pay-performance relation both in year t0 and by using up to two lates (t+1, t+2) as suggested by
Hypotheses 1 and 2.
The statistical technique followed in both regressions is fixed-effects. Previous studies (Palia, 2001; Li, Yang, and Yu, 2015) support that when OLS is used as a regression model it “misidentifies” the pay-performance relation and is weak when it comes to panel data.
4. Results
4.1. Descriptive statistics
The descriptive statistics of the full sample are presented in Table 4.1. I use only the total CEO compensation (Total_C) as it is the sum of all four forms of compensation. The mean of total compensation is 15.32151. Converting the natural log into a number reveals that the average CEO pay is equal to 4,508,658 Euros. In terms of firm performance, the mean of return on equity is almost 16%. The corporate governance variables, board size and board independence, show that the average board consists of 15 directors and the ratio of independent to dependent directors is approximately 69%. An average executive serves as a CEO for at least 5 years and his/her age is over 53. The top 80 listed non-financial companies of Europe have assets worth 30,494.15 million Euros in average, a mean age of 74 years, an amount of debt equal to 61% and give back to the shareholders 47% of the money they make in dividends.
Table 4.2 shows the correlation matrix of the full sample. By following the work of Cohen (1988), it is concluded that there is not a large correlation between the dependent and the independent and control variables. Cohen (1988) reports that “a correlation of 0.5 is large, 0.3 is moderate, and 0.1 is small”. By looking at Table 4.2, it is clear that only in a few instances a moderate correlation is observed. The lack of a strong correlation between the variables assures that no spurious results will be reported and their interpretations will reveal their true impact on the pay-performance relation.
4.2. The differences between United Kingdom and Continental Europe
Table 4.1
Descriptive statistics of the full sample.
The table shows the descriptive statistics of the top 80 listed non-financial companies of Europe for the period of 2005-2014. Total_C is the natural logarithm of the sum of the four CEO compensation components (Salary, Cash_C, Equity_C, and Other_C). RoE is the ratio of net income to shareholders’ equity. BoardSize is the sum of executive and non-executive directors serving on the board. Boardindie is the ratio of non-executive to executive directors serving on the board. FirmSize is the natural logarithm of total assets. CEOtenure is the number of years an executive has worked in the firm as a CEO. CEOage is the CEO’s age in years. FirmAge is the firm’s age in years. Leverage is the ratio of total liabilities to total assets. DIV is the ratio of annual total dividend divided by earnings per share.
Obs Mean Std. Dev. Min Max
Total_C 710 15.32151 0.6398242 14.27 16.26 RoE 710 0.158169 0.096121 0.03 0.34 BoardSize 710 14.53521 4.548225 9 24 BoardIndie 710 0.6884085 0.1824473 0.33 0.92 FirmSize 710 10.32529 0.9578154 8.82 11.8 CEOtenure 710 5.471831 3.589071 1 12 CEOage 710 53.62394 4.959724 46 62 FirmAge 710 73.6831 51.41191 11 150 Leverage 710 0.6108028 0.1301623 0.4 0.8 DIV 671 0.470313 0.2603346 0.04 0.95
To begin with, all the mean differences are statistically significant apart from the CEOtenure variable. The rest of the variables are highly significant (at the 1% significance level) except for return on equity and dividend payout ratio, which are significant at the 5% and 10% level accordingly.
Table 4.2
Correlation matrix of the full sample.
The table shows the correlation matrix of the top 80 listed non-financial companies of Europe for the period of 2005-2014. Total_C is the natural logarithm of the sum of the four CEO compensation components (Salary, Cash_C, Equity_C, and Other_C). RoE is the ratio of net income to shareholders’ equity. BoardSize is the sum of executive and non-executive directors serving on the board. Boardindie is the ratio of non-executive to executive directors serving on the board. FirmSize is the natural logarithm of total assets. CEOtenure is the number of years an executive has worked in the firm as a CEO. CEOage is the CEO’s age in years. FirmAge is the firm’s age in years. Leverage is the ratio of total liabilities to total assets. DIV is the ratio of annual total dividend divided by earnings per share.
Dependent variables Independent and control variables
Total_C RoE Board Board Firm CEO CEO Firm DIV Leverage Size Indie Size tenure age Age
Total_C 1.0000 RoE 0.1859 1.0000 BoardSize 0.0488 -0.1767 1.0000 BoardIndie -0.0365 -0.1246 -0.0623 1.0000 FirmSize 0.0405 -0.1971 0.3998 0.3078 1.0000 CEOtenure 0.0894 0.0614 -0.0306 -0.1378 -0.1747 1.0000 CEOage 0.0769 0.0263 0.1610 0.0213 0.1881 0.3332 1.0000 FirmAge -0.1135 -0.1500 0.2627 0.1160 0.0242 0.0054 0.1611 1.0000 DIV 0.1555 -0.0356 -0.0820 0.0066 0.1620 -0.0867 -0.0467 0.0231 1.0000 Leverage 0.0524 0.1475 0.1671 -0.1115 0.1955 -0.0036 0.0581 -0.1651 0.1194 1.0000
The means of board size and board independence are higher for companies of Continental Europe. The average Continental European board consists of at least 4 more executives than the United Kingdom’s board and its independence is approximately 10% higher. Regarding firm size, the mean differences are of minor importance.
their firms hold an amount of debt of approximately 60% and almost half of the earnings are returned to the shareholders in dividends.
4.3. The evolution of the composition of CEO compensation in the European Union
The descriptive statistics revealed that the mean of the United Kingdom, concerning the CEO total compensation, is higher than the one of the Continental Europe. However, the contribution of each CEO pay component through the 2005 to 2014 period shows different patterns in those two sub samples. I convert the natural logarithms to numbers, break down total compensation in its four parts and divide each one of them by the total compensation. Fig. 4.1 and Fig. 4.2 depict the share of the remuneration components of the United Kingdom and of the Continental Europe respectively.
The evolution of CEO pay components (as a percentage of total compensation) can be characterized as stable for the top companies of the United Kingdom. Apart from the equity-linked variable remuneration, all the rest components show little fluctuations throughout the ten-year period. Both annual base salary and cash-based variable compensation report their highest scores in 2005, fall right before the financial crisis of 2008 and keep those lower numbers until 2014. Equity-linked variable compensation, on the contrary, played a bigger role after 2008. The lowest contribution percentage is in 2005 and the highest one in 2009, just one year after the crisis. Other forms of remuneration go up slightly after 2012.
United Kingdom 0% 10% 20% 30% 40% 50% 60% 70% C on tr ibu ti on of pa y com po ne nt Salary 20% 18% 16% 18% 15% 15% 16% 18% 16% 16% Cash_C 20% 17% 17% 19% 16% 18% 17% 17% 15% 16% Equity_C 56% 61% 63% 59% 64% 63% 62% 59% 63% 61% Other_C 5% 4% 4% 4% 4% 4% 5% 6% 6% 8% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Fig. 4.1 The share of the remuneration components of the companies that are part of the United Kingdom (as a percentage of total compensation). This figure shows how the contribution of the four pay components
has evolved from 2005 to 2014 for the top companies of the United Kingdom. Salary is the fixed remuneration awarded to a CEO annually. Cash_C is the variable remuneration settled in cash payments (bonus). Equity_C is the sum of shares, share options and long term incentive plans (LTIPs). Other_C is the sum of defined contribution pension plans and other benefits.
In both regions, equity-linked variable remuneration has the biggest share out of the total compensation, which means that mainly this component forms the final compensation awarded to a CEO, especially in the United Kingdom. Annual base salary and cash-based variable compensation follow similar patterns before and after the crisis for both sub samples and their importance becomes almost equal in the post-crisis era. Finally, other forms of remuneration account for a small amount of the total compensation.
However, are those differences in share statistically significant? In order to answer that, I conduct two-sample t tests using as group variable the two different regions of my sample. The results are described in Table A3 of Appendix A.
Continental Europe 0% 10% 20% 30% 40% 50% 60% C on tr ibu ti on of pa y c om po ne nt Salary 23% 24% 25% 30% 29% 27% 24% 25% 25% 24% Cash_C 23% 29% 33% 31% 31% 30% 29% 29% 27% 25% Equity_C 49% 39% 34% 30% 31% 33% 37% 38% 40% 42% Other_C 5% 8% 8% 10% 9% 10% 10% 8% 9% 9% 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
Fig. 4.2 The share of the remuneration components of the companies that are part of the Continental Europe (as a percentage of total compensation). This figure shows how the contribution of the four pay
components has evolved from 2005 to 2014 for the top companies of the Continental Europe. Salary is the fixed remuneration awarded to a CEO annually. Cash_C is the variable remuneration settled in cash payments (bonus). Equity_C is the sum of shares, share options and long term incentive plans (LTIPs). Other_C is the sum of defined contribution pension plans and other benefits.
4.4. Univariate results
Is CEO pay affected by shocks such as the financial crisis of 2008 or does it keep on growing “irrespective of the firm performance and economy”? Is the constant media and public criticism over the ever increasing CEO compensation justified? Vemala, Nguyen, Nguyen, and Kommasani (2014), by using a sample of Fortune 500 firms, conclude that CEOs are still highly paid even after the occurrence of the financial distress.
By focusing on the top 80 listed non-financial firms of European Union, I try to check how much of a difference the financial crisis and the years after this event had on firm performance (measured by return on equity) and on total CEO compensation and its components (annual base salary, cash-based, equity-linked, other forms of remuneration).
In order to do that, I divide my sample into three periods, the pre-crisis (2005-2007), the post-crisis (2009-2011) and the new era (2012-2014). Recent studies label 2012 as the beginning of a new “wave” of abnormally high executive compensation. Sabadish and Mishel (2013) identify 2012 as a year with extraordinarily high CEO pay. Mishel and Davis (2014) show in an individual graph the real growth in the CEO compensation during the 2012-2013 and conclude that CEOs realized an increase in their remuneration up to 11.1% since 2012. Following previous literature (Vemala, Nguyen, Nguyen, and Kommasani, 2014), I exclude 2008 from my sample and conduct paired t-tests. I use the average of 2005-2007 for the pre-crisis, the average of 2009-2011 for the post-crisis and the average of 2012-2014 for the new era. A paired t-test is used to determine whether the mean difference between two groups is statistically significantly different to zero. In my study, those “two groups” are the combinations of the pre-crisis, post-crisis, and new era averages of every CEO pay component and of the firm performance measure (ROE).
Table 4.3
Paired t-tests of the full sample.
This table shows all the paired t-tests conducted for the full sample during the period 2005-2014. The sample is divided into three periods, the pre-crisis (2005-2007), the post-crisis (2009-2011) and the new era (2012-2014). RoE is the ratio of net income to shareholders’ equity. Total_C is the natural logarithm of the sum of all four forms of compensation (Salary, Cash_C Equity_C, and Other_C). Salary is the natural logarithm of the fixed remuneration awarded to a CEO annually. Cash_C is the natural logarithm of the variable remuneration settled in cash payments (bonus). Equity_C is the natural logarithm of the sum of shares, share options and long term incentive plans (LTIPs). Other_C is the natural logarithm of the sum of defined contribution pension plans and other benefits. The *, **, and *** represent statistical significance at 10%, 5%, and 1% levels respectively.
However, the CEO compensation has not decreased through the time periods. On the contrary, the mean of total compensation and of every CEO pay component increased during the after transitioning to the next period and the highest mean difference is observed at the “pre-crisis to new era” time line. The only exception is the cash-based remuneration during the “post-crisis to new era” period, but its mean difference is statistically insignificant.
It is observed that from 2009-2011 to 2012-2014 firm performance’s result is not statistically significant and every CEO remuneration part (apart from cash-based compensation) is significant at 1% and 5% level, a contradictory outcome to the “pre-crisis to post-crisis” era. Finally, the two extremes of my sample confirm all the above conclusions: total compensation and most of the pay components are statistically significant reporting a higher mean, and return on equity is also significant but with a mean decreased by 0.0338064.
Tables A4-A6 of Appendix A examine the three time periods individually after splitting my sample into two categories: United Kingdom and Continental Europe.
4.5. Regression results
As already discussed, the ultimate goal of this study is to find the causal pay-performance relation. To begin with, a firm fixed-effects panel regression is conducted to test the hypotheses. Next, the influence of the industry is tested by adding industry-specific effects. Finally, due to the separation of the sample into companies from United Kingdom and Continental Europe, “group” fixed-effects are tested to check how this specification affects the pay-performance relation.
4.5.1. Firm fixed-effects
Table 4.4 describes the firm fixed-effects regression model with CEO compensation as dependent variable and Table 4.5 the same technique with firm performance as dependent variable. In both models, time-effects are added and the pay-performance relation is tested in year t0 with robust standard errors.
The outcomes show that in any direction taken, annual base salary and other forms of remuneration are not statistically significant. The same occurs even when adding lags or lates (not shown). This leads me to drop them and continue this work with total compensation and the two types of variable remuneration (cash and equity). In this way, I underline the importance of separating cash compensation into salary and bonus as proposed by Banker, Darrough, Huang, and Plehn-Dujowich (2013).
Table 4.4 reports a positive and highly statistically significant (at the 1% level) correlation between CEO pay and return on equity for every compensation part. The same applies for CEO tenure. Board size has no impact on the non-fixed compensation variables (cash-based and equity-linked); there is a positive but insignificant relation. Regarding the total remuneration, however, the relation remains positive but it is now significant at the 5% level.
Table 4.4
Firm fixed-effects with CEO compensation as dependent variable.
This table shows the firm fixed-effects panel analysis with robust standard errors by using CEO compensation as dependent variable. Column (1) is the Total_C and Columns (2)-(5) are Salary, Cash_C, Equity_C, and Other_C accordingly. The definition of each variable is reported in Table 3.1. Year dummies are added. The *, **, and *** represent statistical significance at 10%, 5%, and 1% levels respectively.
Variables (1) (2) (3) (4) (5) RoE 1.073*** 0.125 0.940*** 1.389*** 0.656 (0.263) (0.106) (0.276) (0.361) (0.585) BoardSize 0.0346*** 0.0196*** 0.0132 0.00103 -0.0574 (0.0141) (0.00509) (0.0160) (0.0201) (0.0359) BoardIndie 0.248 -0.0891 0.124 -0.0244 0.346 (0.225) (0.0752) (0.220) (0.399) (0.553) FirmSize 0.0757 -0.0362 -0.0291 0.0172 0.408* (0.0840) (0.0303) (0.0789) (0.130) (0.215) CEOtenure 0.0404*** 0.0223*** 0.0333*** 0.0456*** -0.00821 (0.00899) (0.00352) (0.0111) (0.0125) (0.0237) CEOage -0.00563 -0.000396 0.00213 -0.0115 0.0286* (0.00704) (0.00282) (0.00747) (0.0116) (0.0166) FirmAge 0.0126 0.00615 0.0190 0.0203 0.190*** (0.0163) (0.00710) (0.0179) (0.0262) (0.0538) Obs 710 710 631 520 666 R2 0.687 0.718 0.704 0.761 0.695 Adj. R2 0.638 0.675 0.652 0.711 0.645
Table 4.5
Firm fixed-effects with firm performance as dependent variable.
This table shows the firm fixed-effects panel analysis with robust standard errors by using firm performance as dependent variable. Columns (1)-(5) represent the RoE. The definition of each variable is reported in Table 3.1. Year dummies are added. The *, **, and *** represent statistical significance at 10%, 5%, and 1% levels respectively. Variables (1) (2) (3) (4) (5) Total_C 0.0249*** - - - - (0.00652) Salary - 0.0180 - - - (0.0172) Cash_C - - 0.0246*** - - (0.00653) Equity_C - - - 0.0206*** - (0.00556) Other_C - - - - 0.00211 (0.00240) BoardSize 0.000691 0.00108 0.00221 0.000944 0.000767 (0.00230) (0.00236) (0.00235) (0.00296) (0.00258) BoardIndie -0.101*** -0.0972*** -0.103*** -0.112** -0.0971*** (0.0359) (0.00365) (0.0383) (0.0436) (0.0384) FirmSize -0.0647*** -0.0629*** -0.0579*** -0.0818*** -0.0667*** (0.0138) (0.0139) (0.0136) (0.0155) (0.0148) CEOtenure -0.000400 0.000126 -0.000458 -0.000979 0.000204 (0.000906) (0.000942) (0.000970) (0.00113) (0.000947) Leverage 0.0908 0.0706 0.136* 0.0594 0.0629 (0.0656) (0.0682) (0.0701) (0.0730) (0.0716) DIV -0.0758*** -0.0772*** -0.0848*** -0.0758*** -0.0751*** (0.0178) (0.0181) (0.0199) (0.0208) (0.0184) Obs 671 671 598 496 628 R2 0.689 0.681 0.727 0.705 0.691 Adj. R2 0.637 0.628 0.675 0.640 0.636
Table 4.6
Testing Hypothesis 1 with firm fixed-effects.
This table shows the firm fixed-effects panel analysis with robust standard errors for testing Hypothesis 1, which is: performance in year t0 determines future CEO pay. Columns (1) and (2) represent the Total_C one and two
years ahead accordingly. Columns (3) and (4) represent the Cash_C one and two years ahead accordingly. Columns (5) and (6) represent the Equity_C one and two years ahead accordingly. The definition of each variable is reported in Table 3.1. Year dummies are added. The *, **, and *** represent statistical significance at 10%, 5%, and 1% levels respectively.
Variables (1) (2) (3) (4) (5) (6) RoE 0.178 0.0125 0.250 0.236 0.180 -0.169 (0.307) (0.297) (0.312) (0.334) (0.433) (0.465) BoardSize 0.0216 0.0210 0.0150 0.0137 -0.0117 -0.00489 (0.0158) (0.0182) (0.0167) (0.0183) (0.0268) (0.0273) BoardIndie 0.475** 0.588** 0.200 0.0329 0.406 0.297 (0.239) (0.276) (0.224) (0.252) (0.410) (0.412) FirmSize 0.0209 -0.127 -0.0216 -0.00456 -0.158 -0.641*** (0.0823) (0.0889) (0.0901) (0.0962) (0.171) (0.155) CEOtenure 0.0238*** 0.0247** -0.0112 -0.0140 0.0297** 0.0270* (0.00863) (0.0103) (0.0108) (0.0124) (0.0128) (0.0156) CEOage -0.0102 -0.0120 0.00838 0.00830 -0.0147 -0.0168 (0.00707) (0.00846) (0.00841) (0.00976) (0.0109) (0.0135) FirmAge 0.0347* 0.0432** 0.0339 0.0111 0.0517 0.0492 (0.0196) (0.0209) (0.0212) (0.0265) (0.0329) (0.0364) Obs 630 550 563 491 461 402 R2 0.683 0.701 0.686 0.682 0.750 0.776 Adj. R2 0.627 0.640 0.625 0.611 0.691 0.715
future CEO pay. Cash-based variable remuneration is unrelated with all the variables of the regression model. Concerning the second non-fixed compensation part, firm size is negatively associated with it only in year t+2. CEO tenure matters for up to two years ahead
Table 4.7
Testing Hypothesis 2 with firm fixed-effects.
This table shows the firm fixed-effects panel analysis with robust standard errors for testing Hypothesis 2, which is: CEO pay in year t0 determines future performance. Columns (1), (3), and (5) represent the RoE one year
ahead. Columns (2), (4), and (6) represent the RoE two years ahead. The definition of each variable is reported in Table 3.1. Year dummies are added. The *, **, and *** represent statistical significance at 10%, 5%, and 1% levels respectively. Variables (1) (2) (3) (4) (5) (6) Total_C 0.00755 -0.00211 - - - - (0.00731) (0.00837) Cash_C - - 0.00926 0.00470 - - (0.00844) (0.0100) Equity_C - - - - 0.00412 0.00655 (0.00627) (0.00703) BoardSize 0.00477* 0.00127 0.00465* 0.00256 0.00888*** 0.00529 (0.00257) (0.00300) (0.00256) (0.00315) (0.00316) (0.00384) BoardIndie -0.00689 -0.0187 0.00234 -0.00537 0.0178 0.0162 (0.0415) (0.0498) (0.0419) (0.0520) (0.0496) (0.0597) FirmSize -0.0631*** -0.0461** -0.0601*** -0.0428** -0.0823*** -0.0459** (0.0161) (0.0188) (0.0173) (0.0196) (0.0203) (0.0231) CEOtenure -0.000457 0.000124 -0.000487 0.000562 -0.00181 -0.000972 (0.00110) (0.00135) (0.00123) (0.00144) (0.00133) (0.00162) Leverage 0.174*** 0.0523 0.195*** 0.0580 0.250*** 0.172* (0.0608) (0.0770) (0.0621) (0.0794) (0.0764) (0.0946) DIV 0.0146 0.00733 0.0302* 0.0222 0.0151 0.0103 (0.0155) (0.0178) (0.0175) (0.0194) (0.0182) (0.0226) Obs 591 511 530 459 433 379 R2 0.687 0.683 0.718 0.701 0.710 0.687 Adj. R2 0.628 0.612 0.657 0.626 0.637 0.594
a strong and positive connection with future return on equity for both total and variable compensation as independent variables. The positive impact of the board size is not mentioned because of its small coefficient (less than 0.01 for any specification).
Summing up, firm fixed-effects with time effects added report a positive causal pay-performance relation when it comes to year t0 for both of the variables. But, after checking
for future influences, Hypotheses 1 and 2 are not confirmed, therefore leading to Hypothesis 3.
4.5.2. Industry fixed-effects
Table 1B of Appendix B reports the results of the fixed-effects regression analysis after adding industry dummies. Columns (1)-(3) represent the three compensation variables (Total_C, Cash_C, and Equity_C) and columns (4)-(6) represent the return on equity. Time effects are also added and all the regressions are with robust standard errors.
The contemporaneous pay-performance relation remains positive in this analysis, too. In both directions, all variables are positive and highly statistically significant. Board size and firm age are positively and negatively correlated accordingly with total compensation and cash-based variable remuneration. Equity pay, on the other hand, has a negative relation with board size but a positive one with board independence and CEO tenure.
In terms of firm performance, the two corporate governance variables have a strong negative impact on return on equity. Firm size is relevant only for the pay-equity compensation relation and is accompanied by a negative sign. Leverage, in contrast with the firm fixed-effects, is now positive and statistically significant with any performance-pay combination.
with Hypothesis 1 is the equity compensation. Current performance determines future equity pay and the correlation is positive and significant at least at the 5% level even in year t+2.
The results of Table 3B of Appendix B are consistent with Hypothesis 2, but again only for the equity-linked compensation. The latter in year t0 has a positive effect for up to
two years later and it is significant at the 1% level. Leverage’s effect remains strong and positive not only for the present but also for the future performance. The other control variable, dividend payout ratio, becomes positive and statistically significant after adding the lates in the relation. However, its impact is not as strong as the impact of leverage.
Checking for industry fixed-effects yields the same outcome as with firm fixed-effects for the synchronous pay-performance relation. However, when it comes to testing the hypotheses, the industry specification shows a different pattern. Hypotheses 1 and 2 are consistent with the equity pay-performance link. Moreover, Hypotheses 1 can also be confirmed by total compensation at the 10% level of significance. So, I find partial support for both hypotheses and therefore reject Hypothesis 3.
4.5.3. Group fixed-effects
Because of the different backgrounds of the countries that compile my sample, I divided it, as already mentioned, into companies from the United Kingdom and from the Continental Europe. I apply “group” fixed-effects in order to test how those two different categories affect the pay-performance relation.
Tables 5B and 6B of Appendix B reveal an even stronger association when it comes to future pay and performance. The data support fully the Hypothesis 1. Current performance has a positive and statistically significant impact on every future CEO pay component. In contrast with the previous specifications, cash-based variable remuneration is significant at the 10% level and total compensation at the 1% level up to two years ahead. Out of the rest variables, only board independence is correlated with future CEO compensation, whereas board size remains to have a positive and statistically significant tie with total and cash-based remuneration.
Hypothesis 2, on the other hand, is still partially supported. The relation between performance in year t+1 and total compensation is positive and statistically significant at the
1% level, in contrast with the 5% significance with industry-specific effects. Even though cash-based variable remuneration is still not a determinant for future performance, group-effects provide the strongest evidence for not supporting Hypothesis 3.
4.6 Robustness tests
In order to check the validity of my results, I conduct two robustness tests. The first one is with another accounting-based firm performance measure, which is the return on assets (see Tables 7B and 8B of the Appendix B). The second one is with a market-based firm performance measure, which is the annual stock returns (results available upon request).
As long as the contemporaneous relation is the same for all the above specifications, I conduct my robustness tests only on the late relations and I focus on the group fixed-effects, which report the strongest support for both of my Hypotheses.
By using return on assets as firm performance measure. the results are identical with the ones reported by return on equity in terms of performance at year t0 and future total and
concerning the cash pay is insignificant, which explains those spurious results. In general, the data support the Hypotheses 1 and 2.
However, the same does not apply for annual stock returns. Although in both directions there is a partial support for the Hypotheses, the signs of the coefficients are not consistent with all the tests that were conducted using an accounting-based firm performance metric. Annual stock returns in year t+1 are negatively related to total and equity-linked
compensation at the 5% level of significance. Furthermore, cash-based remuneration in year t+1 is positively associated with the present annual stock returns. Both results, in combination
with the highly significant and positive sign of board size, point out to the managerial rent extraction theory and to the presence of “overconfident” CEOs as proposed by Cooper, Gulen, and Rau (2014). The rent-seeking behaviour of the CEOs is also observed when using both of the accounting-based measures; therefore the two robustness tests end up with the same conclusion. More details on those theories will be discussed in the next section.
5. Discussion
5.1. Interpretation of differences between United Kingdom and Continental Europe
As mentioned in Section 4.2, CEOs in the United Kingdom get paid more than the ones in the Continental Europe. This is in line with Rodionova (2016) who states that half of the ten highest paid CEOs in Europe are working in the United Kingdom. This higher pay is connected with the greater usage of equity-linked variable remuneration, as shown in Section 4.3, which is still going up in contrast with the Continental Europe.
Finally, a surprising outcome is that the mean of the leverage of the Continental Europe is smaller than the equivalent of the United Kingdom. Continental Europe is traditionally labeled as a bank-based, whereas the United Kingdom a market-based system. This would normally translate to higher leverage for the Continental Europe. Incentives associated with tax avoidance for the United Kingdom could be the answer to this “paradox”.
5.2. Interpretation of univariate results
Yang, Dolar, and Mo (2014), by using data gathered by ExecuComp database, find that the accounting-based firm performance measures that they used were positive and statistically significant with total, cash- and stock-based compensation both before and after the financial crisis of 2007-2008.
However, my sample shows a different pattern than this. After dividing the data into three time periods, it is revealed that firm performance is constantly dropping, whereas CEO compensation is on the rise. This outcome can be consistent with the managerial rent extraction theory or with any alternative explanation, such as the “Lake Wobegon Effect” where CEO pay is not determined by the firm’s performance.
Although the univariate analysis offers only preliminary results, this adverse pay-performance relation is revealed in the interpretation of the regression results.
5.3. Interpretation of regression results
Table 4.4 shows that all the compensation components are positively linked with the number of years an executive has served as CEO and only total compensation is associated with the size of the board. However, CEO tenure can be linked either with an “experienced” top executive (Lee and Chen, 2011) or with an “entrenched” one (Ozkan, 2011). Regarding the incentive pay (cash and equity compensation), CEO tenure can be considered a positive aspect for the executive. But, the positive correlation of the total compensation with the size of the board can be translated into more power for his/her selfish goals.
The same conclusion is extracted when the test is conducted using the return on equity as dependent variable. The pay-performance is positive, but the statistically significant negative relations of board independence and dividend payout ratio with firm performance do not support this positive outcome. Consistent with the work of Coles, Lemmon, and Wang (2009), board independence does not promote higher standards in monitoring and eventually leads to a lower firm value. Moreover, Lee and Chen (2011) show in their work the negative impact of dividend payout ratio on firm performance.
Those arguments are confirmed when late relations are added. Not surprisingly, there is not any relation on the future pay-performance tie. So, on the firm-level, it can be hypothesized that companies increase the top executive compensation regardless of the firm performance. It would be tempting to guess that this conclusion is consistent with the “Lake Wobegon Effect”, but additional assumptions should be made for the effect to be applied (Hayes and Schaefer, 2009).
Finally, the group fixed-effects show the strongest future pay-performance relation and again the direction of causality is stronger from performance to pay. Although, in year t0
for both pay and performance, board size is positively related to the total compensation, so is the board independence. Furthermore, all the variables except for the firm age have a strong association with the CEO compensation, something that did not occur on the previous specifications. Because of the group-effects, each corporate governance system (1-tier or 2-tier) is now “connected” with the appropriate group which could be the reason for the strong contemporaneous and future pay-performance relation.
Taking into consideration the performance-to-pay direction of causality, that implies that the compensation committees of the top 80 European companies should stabilize the pay of their CEO. By doing so, the future CEO compensation will be determined by the current performance and if the top executives want to earn more, they should act in a way that improves firm performance, which will eventually raise the firm’s value.
6. Conclusion
support and prove the claim of Joskow and Rose (1994) that the pay-performance relation is underestimated when tested only on a synchronous level. Furthermore, by testing the total CEO compensation and its components (annual base salary, cash-based variable pay, equity-linked variable remuneration, and other forms of remuneration) individually, I underline the importance of separating cash compensation to salary and bonus. The latter was suggested by Banker, Darrough, Huang, and Plehn-Dujowich (2013). The major impact of the strength of corporate governance is analyzed. Finally, the need for using both accounting-based and market-based firm performance measures is described in the robustness tests. .
However, there are also some limitations on this research. The restriction to only companies that are part of the European Union could be a potential selection bias. Including countries that are not members of the Union could provide a more fruitful outcome. Furthermore, by focusing on the top 80 firms, any small- or medium sized corporations are not included. Lastly, the selection according to the list of the Financial Times could also be considered a selection bias. Redoing the same study with another list or by using a different method of categorization would provide support or not on the results of this study.
References
Banker, R., Darrough, M., Huang, R., Plehn-Dujowich, J., 2013. The relation between CEO compensation and past performance. Acounting Review 88, 1-30.
Bereskin, F., Cicero, D., 2013. CEO compensation contagion: evidence from an exogenous shock. Journal of Financial Economics 107, 477-493.
Bruce, A., Skovoroda, R., 2015. The empirical literature on executive pay: context, the pay-performance issue and future directions. High Pay Centre, London.
Cohen, S., 1998. Psychological models of social support in the etiology of physical disease. Health Psychology 7, 269-297.
Coles, J., Lemmon, M., Wang, Y., 2011. The joint determinants of managerial ownership, board independence, and firm performance. Unpublished working paper. Arizona State University, University of Utah, and Chinese University of Hong Kong.
Cooper, M.J., Gulen, H., Rau, P.R., 2014. Performance for pay? The relation between CEO incentive compensation and future stock price performance. Unpublished working paper. University of Utah, Purdue University and Purdue University.
Core, J., Holthausen, R., Larcker, D., 1999. Corporate governance, chief executive compensation, and firm performance. Journal of Financial Economics 51, 371-406.
Croci, E., Gonenc, H., Ozkan, N., 2012. CEO compensation, family control, and institutional investors in Continental Europe. Journal of Banking & Finance 36, 3318-3335.
