1
The CEO Pay-Performance relationship: a robust
investigation of Europe compared with the US
By Jelco L. Paalhaar, 2015
University of Groningen
Faculty Economics and Business
Abstract
In this study, we investigate the CEO pay-performance relationship for two samples of 828 and 6123 firm-year observations where we compare respectively Europe with the US for the years 2000-2013. With the alignment of company performance and CEO compensation, the agency problem is assumed to decline. In our two-way fixed effects regressions, we only find a moderate significant relationship between CEO compensation and shareholder return for Europe which is significant at the 10%-level, but a highly significant one for the US. In this larger sample, the relationship is much more visible. When we look at the first differences CEO compensation, the resulting performance elasticities are not significantly positive for both samples, but the sales elasticities are. When we take the crisis into consideration, we observe a significant decline during this period in the direct relationship and the elasticity for the US only.
Keywords: CEO, Compensation, Pay, Company Performance, Elasticities JEL classification: G30, G35, J33
Name: Jelco Leon Paalhaar Study: Finance University of Groningen Student number: S1889419 Email: jelcopaalhaar@hotmail.com 1st Supervisor: B. van Oostveen1
2nd Supervisor: R. van Dalen
1 Acknowledgement: I would like to thank my supervisor Mr. Boris van Oostveen for his help and feedback
2
1. Introduction
For decades, researchers studied the existence of an agency problem between shareholders and management. Conflicts of interests arise when managers make decisions on behalf of the financing party, the shareholders. A famous paper from Jensen and Meckling (1976) describes that solving this agency problem is one of the key challenges of effective corporate governance. ―CEO compensation packages have been viewed as important in mitigating the conflict of interest between managers and shareholders in corporations‖ (Ozkan, 2011). A widely researched topic is the executive pay for performance, which is also the aim of this study. In this paper, we investigate the relationship between company performance and CEO compensation in the period 2000-2013. We compare Europe with the US and give robust evidence on a positive CEO pay-performance relationship.
Much literature exists with regard to CEO compensation and company performance. Many researchers found a significant positive relation between these variables (Jensen and Murphy, 1990a; Ozkan, 2011; and Hall and Liebman, 1998). However, most literature is based on the United States. In this study, we broaden the pay-performance investigation with respect to the European market and compare the results of several European countries with the US, which is not done in existing literature. Conyon and Schwalbach (2000) already compared the UK and Germany in a study. They explained the different economic systems with corresponding governance variables which could lead to different outcomes.
CEO compensation is a controversial topic with a lot of debate, especially nowadays in the current financial crisis. ―During the financial crisis, there has been increasing media attention on compensation, with criticism of the payment of large bonuses to top executives. If the companies have solid financial results, the public does not show too much interest in executive compensation. When a crisis occurs, however, compensation is once again a subject for debate‖ (Bührer, 2010). Gilson and Vetsuypens (1993) already showed that the pay-performance relationship can decrease when a company witnesses financial struggles. This study takes the crisis into consideration to investigate whether different economic times affect the CEO pay-performance relationship.
3
We find, next to company size, a strong positive and highly significant relationship between the absolute amounts of shareholder return and CEO compensation (salary and bonus only) for US and a positive relationship for Europe which is marginally significant. However, the difference in coefficients is not statistically significant. When we look at first differences to calculate the performance elasticities, they are not significant for none of the samples. Sales elasticities do have a significant impact. Hence, when the sales increase in some year, this will lead to an increase in the CEO compensation in the corresponding year. As a robustness check, we show how important it is to describe your sample, since sample splits can give varying results on the strength of the relationship. Nevertheless, it seems that bigger companies have a stronger relationship in our research. Lastly, in line with our expectations, the crisis had a significant negative impact on the relationship for the US. For Europe, the relation did not decrease significantly.
This paper will be organized as follows. In section 2, we will discuss the research already done on this topic. Here we introduce the agency theory and show the mainly positive pay-performance relationship, with lacking literature for Europe. At the end of this section, hypotheses are formulated. After that, in the Data and methods section, the obtaining of the samples and the regression formulas are explained. Next, the Results section start with the descriptive statistics and present the outcomes from the regression analysis which answers the hypotheses. Finally, a conclusion will follow.
2. Literature review
2.1 Agency theoryAgency theory describes the issue that many companies are dealing with the management who makes decisions on behalf of the shareholders. ―The problem of inducing an ―agent‖ to behave as if he were maximizing the ―principal’s‖ welfare is quite general‖ (Jensen and Meckling, 1976). When managers do not bear the total risk of their decisions, the separation of ownership and control can cause conflicts of interest. Jensen and Murphy (1990a) argue that this conflict of interest between the Chief Executive Officer and the shareholders is a classic example of the principal-agent problem, where incomplete information and lack of transparency exists. ―CEOs have goals that often conflict with the interests of shareholders. The most direct solution to this agency problem is to align the incentives of executives with the interests of shareholders by granting (or selling) stock and stock options to the CEOs‖ (Hall and Liebman, 1988).
4
from Jensen and Murphy (1990a). According to them, when conflicts of interest appear, agency theory predicts that compensation policy will be designed to give the manager incentives to select and implement actions that increase shareholder wealth. Conyon and Schwalbach (2000) explain that executives are encouraged to promote shareholder interest by making a part of their compensation contingent upon shareholder wealth. Many compensation mechanisms can align interests, to provide value-increasing incentives for the executives. Hall and Liebman (1998) state that it is doubtful that the trillions of dollars of assets in public corporations are being managed efficiently if there is no meaningful link between CEO pay and company performance. In addition, Ozkan (2011) argues: ―It has been widely recognized that compensation packages could potentially play an important role in motivating top managers. Therefore it is important to understand how corporations set the CEO compensation packages and whether there is a link between compensation and performance‖.
2.2 The pay-performance research
During the 80’s, research on the compensation of CEOs already started. Murphy (1985) concludes that corporate performance is strongly and positively related to managerial remuneration. The relation between shareholder wealth and company performance is also investigated by Jensen and Murphy (1990a) for a large US sample in the period 1974-1986. They present the so called pay-performance sensitivity. According to them, the total change in CEO wealth (including a high share for stock ownership) is $3,25 for every $1000 change in shareholder wealth, so a sensitivity of 0,325%. In terms of agency theory, the relation seems small but it is statistically significant. Jensen and Murphy (1990a) did a lot of robustness checks. When measuring the performance with accounting profits, CEO pay appears to be equally sensitive.
From then, the literature about this topic was extending rapidly. Hall and Liebman (1998) studied the relationship for the largest US companies in a 15-year panel dataset. They found a strong positive relationship. ―This relationship is generated almost entirely by changes in the value of CEO holdings of stock and stock options. In addition, we show that both the level of CEO compensation and the sensitivity of compensation to firm performance have risen dramatically since 1980, largely because of increases in stock option grants‖ (Hall and Liebman, 1998). The mean percentage of salaries and bonuses rose by 97% and stock option grants by 683%. This generates an interest in the change of the relationship.
2.3 Non-US research
5
level amounts of CEO compensation and firm performance and found a significant positive relation. Second, he focused on the changes in the variables. These results show a significant positive pay-performance relationship as well, although the elasticity was lower compared with the US market. He found an elasticity of 0,075 ―…indicating that a ten percentage increase in shareholder return corresponds to an increase of 0,75% (0,95%) in cash (total direct) compensation‖ (Ozkan, 2011). The cash compensation consists of salary and bonus and the ―total direct compensation‖ includes the equity-based components of compensation. Since he used the natural logs of the variables, it is not possible to give the label ―sensitivity‖ anymore. Most recent research therefore explains results in terms of ―elasticity‖.
Another study from Conyon and Murphy (2000) also found a lower elasticity in the UK (0,12) compared to US (0,27). They also document that American CEOs earn 45% more in cash pay and 190% more in total pay, compared with British colleagues. One explaining factor they state is the culture: ―The United States, as a society, has historically been more tolerant of income inequality, especially if the inequality is driven by differences in effort, talent, or entrepreneurial risk taking‖.
2.4 Corporate governance systems
Besides the apparent cultural differences, Ozkan (2011) shows the importance of different corporate governance systems, which he added as control variables. These differences between countries can partially explain variation in the sensitivities. He argues that changes over time in corporate governance mechanisms can influence the relationship over different periods. According to his research, institutional ownership has a negative impact on the level of CEO compensation. Hüttenbrink, Oehmichen, Rapp, and Wolff (2014) found that weak shareholder protection and high disclosure requirements are encouraging to implement higher pay-for-performance. Harvey and Shrieves (2001) also showed the importance of corporate governance variables on CEO incentive pay. Outside directors and large block holders have a positive effect on incentive compensation in their study. Lastly, Van Essen, Heugens, Otten and Van Oosterhout (2012) trace the differences in the strength of the relationship in the level of development of formal and informal institutions protecting investors against managerial overcompensation and underperformance.
6
control). Type I economies also have outside or non-executive directors who sit on the company board and act as representatives of shareholders and monitor the performance of the management team (internal control). Type II economies are characterized by dual boards where financial institutions have an important role in representing the owners’ interest‖ (Conyon and Schwalbach, 2000). These systems can be linked with the researched variables by Ozkan (2011); Hüttenbrink, Oehmichen, Rapp, and Wolff (2014); Harvey and Shrieves (2001) and Van Essen, Heugens, Otten and Van Oosterhout (2012), which conclude to a stronger relationship in the US (and UK) compared with most European countries.
2.5 Changes in the relationship
―There is conflicting evidence as to whether the pay-performance relationship has weakened or strengthened over time‖ (Gregg, Jewell and Tonks, 2012). Jensen and Murphy (1990a) discuss the change of the sensitivity. They explain the implicit regulation hypothesis where strong political forces operate in both the private sector (i.e. board meetings, annual stockholder meetings, and internal corporate processes) and in the public sector that affect the pay of executives. Managerial compensation contracts are open to public information and open to public scrutiny and criticism. Furthermore, compensation committees have the authority over compensation decisions which are composed of outside members of the boards of directors who are elected by, but are not perfect agents for, shareholders. ―Fueled by the public disclosure of executive pay required by the Securities and Exchange Commission, parties such as employees, labor unions, consumer groups, Congress and the media create forces in the political environment that constrain the type of contracts written between management and shareholders―. Jensen and Murphy (1990a) conclude that the pay-performance relation for CEOs in the top quartile of NYSE firms has fallen by a factor of 10 over the past 50 years. This is in line with their implicit regulation hypothesis since the overall regulation of corporate America increased substantially.
Crawford, Ezzel and Miles (1995) showed that the average pay-performance sensitivity for salary and bonus of 124 banks on a Forbes list increased significantly after the deregulation of the sector around 1980. An explanation could be the increased attention towards performance bonus plans by firms, not only in the banking sector. The sensitivity for salary and bonus compensation increased from 0,35 to 2,22 after the deregulation.
7
regard to total wealth are even more dramatic. An excessive part of the total relation comes from stock option grants. According to Hall and Liebman (1998), their finding of increased salary and bonus sensitivity is consistent with the view that boards are attempting to increase pay to performance.
2.6 Crisis times
Gilson and Vetsuypens (1993) researched CEO compensation for 77 publicly traded companies which filed for bankruptcy or privately restructured their debt during 1981 to 1987. They argue that compensation policy is often an important part of firms’ overall strategy to deal with financial distress. The CEOs who stay during a year of default see their salaries reduce with substantial amounts. ―Newly appointed CEOs with ties to previous management are typically paid 35% less than the CEOs they replace. In contrast, outside replacement CEOs are typically paid 36% more than their predecessors, and are often compensated with stock options‖ (Gilson and Vetsuypens, 1993). New CEO insiders are punished for the prior poor performance or being perceived to be less productive. Outsiders are paid more because of unique skills, a higher demanded salary for the risk of losing their job, and less political constraint and public pressure on salaries when an outsider is hired. A big part of the firms in their sample took actions to tie senior management wealth more closely to the value of the firm’s stock.
The findings of Gilson and Vetsuypens (1993) with regard to cash compensation show the presence of a positive relationship between pay and performance which is shifting to an absence of any relationship when a company goes into default. This corresponds with their descriptive data of declining bonus plans in crisis years. The total wealth regressions show a significantly positive association in all periods and sharply increase after the bankruptcy period, so when a company is not in default anymore. They conclude that a possible explanation for this higher sensitivity is the better alignment of managers and owners. Lastly, Gregg, Jewell and Tonks (2012) also found that executive pay is less sensitive to performance when stock returns are low, which is the case in crisis times. ―There appears to be a pattern that elasticities are increasing when the stock market is rising, but a weaker relationship between pay and performance when stock prices are falling‖ (Gregg, Jewell and Tonks, 2012).
2.7 Hypotheses
8
Based on the described literature, we expect a positive relationship between company performance and CEO cash compensation over the whole panel in Europe and in the United States. Therefore, we formulate the following hypothesis:
H1: There exists a positive relationship between CEO cash compensation and company performance in Europe and the US during 2000-2013.
However, we expect the relationship to be smaller for Europe and since the already mentioned cultural and governance factors are expected to play a role in the relation. Furthermore, European compensation contracts differ from the US contracts. Salaries and bonuses are lower and stock (option) rewarding is less apparent, which will be further explained in the Data & Methods section. H2: The pay-performance relationship in the US is stronger than in Europe.
Balancing the existing literature, we expect a decreasing relationship between CEO cash compensation and company performance during the economic crisis time. Salaries and bonuses declined in this period and consequently alter the relationship. In addition, the already mentioned implicit regulation hypothesis, which goes hand in hand with increases in regulation in crisis times are factors which can impact the relationship downwards.
In line with this statement, the following hypothesis results:
H3: The pay-performance relationship changed and decreased during the crisis for Europe and the US.
3. Data & Method
3.1 Obtaining CEO compensation data
9
Dollars. For the US sample, the corresponding variable is ―Total Current Compensation‖, which is easily available through CompuStat ExecuComp. All the tickers of the S&P500 companies are collected and inserted in the query. This resulted in an extended list of CEOs with corresponding compensation for the same 14-year period. The ExecuComp output was already structured in US American Dollars.
For the sake of simplicity and the large differences in stock- and option payments in the CEO compensation of European and US companies, only direct/current cash compensation regarding salary and bonuses will be taken into account. Supporting evidence for the fact that stock (option) rewarding is not very usual in European countries is present. Conyon and Murphy (2000) found that share options are much more important in the US and comprise 42% of the CEOs total pay. According to the study of Fernandes, Ferreira, Matos and Murphy (2009), this is on average more than double the amount than in other worldwide 26 countries used in their research. In a recent study of Hüttenbrink, Oehmichen, Rapp, and Wolff (2014), stock-based incentives are still less important in European countries. The use of share options is even declining in the UK, probably due to discouraging and controversial situations. Furthermore, Ozkan (2011) discusses the removal of an accounting advantage for this form of compensation. In addition, Gilson and Vetsuypens (1993) also discusses problems with the inclusion and calculation of options which will lead to biased results. Lastly, BoardEx has a lot of missing data on equity compensation. In sum, there are enough reasons to exclusively study the ordinary salary and bonus.
3.2 Company performance and controls
The main explanatory variable used for this relationship is Shareholder Return. This performance measure is calculated by the continuously compounded rate on the ―Total Return Index‖, where the dividends are reinvested. The Total Return Index is retrieved from Datastream (Code: RI). For converting the index in a return measure, we use the following formula, in accordance with Ozkan (2011):
( )
( )
( )
We calculate the natural log variable which results in more normalized data and reduces the effect of outliers, consistent with most studies in the field of finance.
10
same results. Lastly, we perform the analysis with the accounting measure Return on Assets (ROA) in order to explore any possible deviations with our market performance measures (Datastream code: WC08326).
We also check the effect of the company size on the CEO compensation, since it is likely that company performance is not the only variable which is influencing compensation. Therefore, in accordance with other studies (Ozkan, 2011; Conyon and Murphy, 2001; Conyon and Schwalbach, 2000; Gregg, Jewell and Tonks, 2012), firm size is proved to be an important control variable. Most literature is defining the firm size with Total Sales (Datastream code: WC01001).
Datastream output is in national currency. To be consistent, we conduct the research for Europe in Euros and for the US in US Dollars. However, in this study, we include companies from European countries which do not have the Euro as their national currency (Sweden, Denmark and Switzerland) and some companies for which the Datastream output is in British Pounds. Therefore, all the financial amounts for these companies are converted to Euros, based on their exchange rate data at the last available day of the corresponding year. The BoardEx data for European countries is also converted to Euros. Appendix A shows the used exchange rates for all the currencies in each corresponding years.
All the compensation data, together with the independent variable data, is merged into one database by the unique ISIN-number of each company. The resulting samples consists of 502 US companies from the S&P500 and 75 available European companies from France (27), Germany (17), The Netherlands (9), Switzerland (9), Sweden (6), Spain (5), Denmark (1) and Luxembourg (1). We retrieved 828 firm-year observations for Europe and 6123 firm-year observations for the US, which results in two unbalanced panels.
3.2 Regression formulas
We start with investigating the direct level relationship between the absolute amounts of CEO compensation and the Market Value and Sales to explore whether CEOs of bigger companies are paid more. This results in the following regression which will be performed for both the European and the US sample:
11
Where ―lnCEOCOMPENSATION‖ is the natural log of the CEO compensation with the firm index ―i‖ and time index ―t‖ since we have panel data for many companies over a 14-year period. α is a constant, β is our pay-size coefficient and ɛ is the error term.
Next, we investigate the relationship with our main explanatory variable Shareholder Return. In addition to this measure, the same regression will be performed with the yearly differences in Market Values. Lastly, we will check what the effect of our control variable (Sales) is. Following Ozkan (2011), the regression formula becomes:
( )
( )
( )
In addition to the absolute amounts, we investigate the relative relation by measuring the sensitivity and elasticity in accordance with for example Jensen and Murphy (1990a) and Ozkan (2011) respectively, for further evidence on this relation. They look at the effect of changes in Market Value (also called ―Wealth of Shareholders‖ or ―Shareholder Value‖) on the changes in CEO compensation, where β is the sensitivity or elasticity. We take our usual measure Shareholder Return and also look at the differences in Market Value as a robustness check. In addition, the control variable for firm size (Sales) is again included to check its influence in the increase of CEO compensation next to Shareholder Return.
( )
( )
( )
It is not only current company performance which may affect compensation. Last year’s stock price changes may play a role in for example current bonuses. Therefore, we include one lagged year in the shareholder return, consistent with Jensen and Murphy (1990a), Hall and Liebman (1998) and Ozkan (2011). Including lags also reduces the problem of endogeneity. A further second- and third-year lag did not made any significant changes in Jensen and Murphy (1990a). The term
12
When investigating the crisis effect, a dummy is used to control for the later period. We expect the returns were in a free fall since 2008. Therefore, the years 2008-2013 are seen as crisis years in this study. The regression formula includes a created dummy variable which is valued one for the sample where the year is 2008-2013 and zero for 2000-2007. ―A slope dummy changes the slope of the regression line, leaving the intercept unchanged‖ (Brooks, 2008). Brooks (2008) furthermore explains that when the dummy value is zero, the slope will be β, while for periods when the dummy is one, the slope will be β + γ. The γ will represent the deviation of the slope from the original slope. In the regression formula, the dummy is interacting with the Shareholder Return. When the result of the dummy term is significant, this presents a change in the relationship in this period. The ordinary regression formula becomes:
( )
( )
( )
Again, we look at the pay-performance elasticity by taking the first differences in CEO compensation. This results in the following regression formula:
( )
( )
( )
3.3 Tests
13
inconclusive results for European sample but rejection of the null hypothesis for the US sample. Therefore we will also stick with the fixed effects panel regressions.
4. Results
4.1 Descriptive data
14 Figure 1A
Average & median CEO compensation in the years 2000-2013 for Europe and the US
Figure 1B
Average Shareholder Return in the years 2000-2013 for Europe and the US $0 $500.000 $1.000.000 $1.500.000 $2.000.000 $2.500.000 $3.000.000 $3.500.000 €0 €500.000 €1.000.000 €1.500.000 €2.000.000 €2.500.000 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
CEO
Com
p
en
sation
E
u
rop
e
Year
Average EU (€) Median EU (€) Average US ($) Median US ($)
CEO
com
p
en
sation
US
-60% -50% -40% -30% -20% -10% 0% 10% 20% 30% 40% 50% 60% 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013Re
turn (LN)
Year
15
Figure 1B shows the average Shareholder Returns in all the years over the whole sample. These lines present almost equal results between Europe and the US. 2003, 2008 and 2009 were declining years. In most of the years before the crisis, together with 2010, returns were positive.
Table 1
Average CEO compensation and the yearly differences in % for Europe and the US for the period 2000-2013
Year Avg. Comp. EU % difference EU Avg. Comp. US % difference US
2000 € 1.503.243 $ 2.181.173 2001 € 1.572.557 4,61% $ 2.215.511 1,57% 2002 € 1.611.119 2,45% $ 2.120.599 -4,28% 2003 € 1.913.801 18,79% $ 2.517.733 18,73% 2004 € 2.086.869 9,04% $ 2.761.973 9,70% 2005 € 2.276.893 9,11% $ 2.948.073 6,74% 2006 € 2.160.802 -5,10% $ 1.792.305 -39,20% 2007 € 2.297.297 6,32% $ 1.587.221 -11,44% 2008 € 1.931.029 -15,94% $ 1.547.645 -2,49% 2009 € 2.008.590 4,02% $ 1.390.995 -10,12% 2010 € 2.267.866 12,91% $ 1.533.174 10,22% 2011 € 2.270.896 0,13% $ 1.512.687 -1,34% 2012 € 2.347.702 3,38% $ 1.540.271 1,82% 2013 € 2.222.926 -5,31% $ 1.513.318 -1,75%
Table 2 shows the descriptive statistics of all our used variables in this study for the two samples of Europe and the US. The averages of the direct compensation are almost equal. However, the average size of the company is higher in the European sample, according to the variables sales and market value. This is reasonable because the European sample is smaller and data from BoardEx is scarcer. Therefore, this could be caused by the fact that data is only available for the bigger companies. For the US, the sample contains the S&P500, with a wider range of companies. Maximum compensation and size are much higher in the US sample, as well as the corresponding dispersion, measured by standard deviation. From these statistics one can see that it seems that US CEOs are paid more related to the company size. The maximum direct cash compensation in the US sample in one year is almost 78 million dollar.
16
US trend in compensation over the years is practically the same. The absolute amount, however, is much higher. Until 2006, the dollar amounts were around $1.000.000 higher every year. After the large decrease, the gap declined somewhat. On average, the compensations were $300.000 higher each year. This is supported by Fernandes, Ferreira, Matos and Murphy (2009), who explained that CEOs in the US are overpaid by 170% in their study compared with foreign counterparts, 40% after controlling for size and industry and corporate governance variables and 12% when deleting the incentive compensation. They furthermore explained the large decrease of the premium during the years 2000-2006 and the converging of the compensation structures and pay levels in the world economy.
Due to the fact that the shareholder returns are continuously compounded, the minimums go beyond -100%. Appendix B shows the correlation diagrams of Europe and the US for all our variables. We do not observe too high correlations. Normality Jarque-Bera tests are rejected for the data because the kurtosis is high for most variables. Nevertheless, Brooks (2008) argues that it is not problematic since the sample is large enough.
Table 2
Descriptives Europe and the US for the period 2000-2013
The first panel shows the descriptive statistics for the complete 14-year sample for Europe in Euros. The second panel presents the numbers for the US in US dollars.
Europe Mean Median Maximum Minimum Std. Dev. Obs.
CEOcomp. (€) 2.099.709 1.874.059 9.284.344 34.745 1.273.959 838 CEOcomp. (ln) (€) 14,365 14,443 16,044 10,456 0,693 838 ΔCEOcomp. (ln) (€) 0,025 0,033 3,966 -3,761 0,571 759 Return (%) 4,3 9,8 190,3 -209,8 39,8 998 Return T-1 (%) 3,5 8,9 190,3 -209,8 40,7 923 Sales x000.000 (€) 27.948 18.919 132.000 125 25.261 1036 Sales (ln) x000 (€) 16,665 16,756 18,670 11,733 1,1020 1022 MV x000.000 (€) 26.082 17.986 187.505 93 25.289 1011 ROA (%) 6,0 5,0 46,7 -35,9 7,1 1007
US Mean Median Maximum Minimum Std. Dev. Obs.
17 4.2 Absolute pay-performance relationship
In our first panel regression analysis, we study the relationship between the CEO compensation and the size of the company (Market Value and Sales), to see whether bigger companies have higher paid CEOs. Table 3 shows the outcomes for the absolute amounts of both size variables for Europe and the US, as well as the natural logs of each variable. Since the absolute compensation data lacks normality, logs are more reliable measures which reduce the impact of outliers and normalize the data. This is also consistent with related finance literature. With absolute data, the Market Value and Sales do have a positive impact on compensation in the European sample. When both are included, one extra million in Market Value and Sales causes an increase of €10,86 and €14,81 in CEO compensation respectively. For the US, Market Value does not have any impact, in contrast to the Sales variable, which has a negative influence on the CEO compensation. Since these are absolute amounts, we do not trust these results. When the regression is pooled, the results turn in to positive significant coefficients.
18 Table 3
Panel regression results for the relationship of CEO compensation and firm size
In the upper panel, the dependent variable is CEO compensation. Independent variables are Market Value and Sales, which are both in millions (€ for Europe in the upper left panel and $ for US in the upper right panel). In the lower panel, all variables are in natural logs. The first lines are the coefficients, and the second lines are the corresponding T-values. For all tests, we include cross section and period fixed effects. *=significant at the 10%-level, **=significant at the 5%-level, ***=significant at the 1%-level.
Europe Dep. Variable = CEOCOMPENSATION US Dep. Variable = CEOCOMPENSATION
Intercept 1736072*** 1506287*** 1380152*** Intercept 1928734*** 2046658*** 2035644***
(19,36) (11,3) (9,62) (48,64) (48,25) (42,77)
Market Value 14,10*** 10,86*** Market Value -0,878 1,316
(4,42) (3,22) (-0,62) (0,87)
Sales 20,17*** 14,81*** Sales -8,919*** -9,672***
(4,55) (3,11) (-4,02) (-4,06)
Cross section fixed Yes Yes Yes Cross section fixed Yes Yes Yes
Period fixed Yes Yes Yes Period fixed Yes Yes Yes
R2 0,505 0,505 0,510 R2 0,519 0,519 0,521
Obs 835 824 821 Obs 5984 6089 5983
Europe Dep. Variable = lnCEOCOMPENSATION US Dep. Variable = lnCEOCOMPENSATION
Intercept 12,259*** 6,847*** 7,782*** Intercept 12,885*** 11,765*** 11,671*** (26,13) (4,85) (5,39) (106,16) (72,05) (69,69) MarketValue (ln) 0,216*** 0,127** MarketValue (ln) 0,137*** 0,047*** (4,50) (2,30) (10,45) (3,00) Sales (ln) 0,449*** 0,318*** Sales (ln) 0,27*** 0,232*** (5,32) (3,27) (14,60) (10,39)
Cross section fixed Yes Yes Yes Cross section fixed Yes Yes Yes
Period fixed Yes Yes Yes Period fixed Yes Yes Yes
R2 0,502 0,505 0,507 R2 0,653 0,657 0,66
19
In order to answer our first hypothesis, we performed a panel regression analysis with the Shareholder Return as independent variable. In Table 4, the results are presented for Europe and the US. Shareholder Return is marginally significant for Europe at the 10%-level. The coefficient denotes that companies with a 10% higher Shareholder Return have CEO compensation which are on average 1,18% higher. In the second and third regression, one can conclude that the lag of the Return has a small positive influence which is not significant. In the fourth regression, we include the control variable Sales, which is significant at the 1% level. The explaining power of the Shareholder Return declines a bit. Consistent with other studies, the size variable has more influence on CEO compensation than Shareholder Return. For the US, the results are somewhat different. Shareholder Return is significant in all performed regression analyses. The lag on its own is not significant, but appears to have a positive influence when it is included together with the ordinary Shareholder Return. Again, when the size control is included next to the return and the lagged return, it is highly significant. In the last regression, the coefficient of the return is 0,154 and 0,043 for the lagged return. Companies with a 10% higher Shareholder Return, pay their CEOs on average 1,54% more this year and 0,43% in the next year.
Therefore, we can conclude that our first hypothesis H1 regarding the positive pay-performance relationship is accepted. There exists a positive relationship between shareholder return and CEO compensation. In addition, the relationship is marginally significant for Europe but more present for the US, which is consistent with our expectations. In order to compare the different Betas from the separate samples, we use the following formula to retrieve the T-statistic for differences in coefficients from separate samples:
√
( )
20 Table 4
Panel regression results for the relationship of lnCEOCOMPENSATION and shareholder return
The dependent variable is the natural logarithm of CEO compensation, in the upper panel for Europe, and in the lower panel for the US. Independent variables are Return, lagged Return and the control variable Sales which is in millions (€ for Europe and $ for US). The first lines are the coefficients, and the second lines are the corresponding T-values. For all tests, we include cross section and period fixed effects. *=significant at the 10%-level, **=significant at the 5%-level, ***=significant at the 1%-level.
Europe Dependent Variable = lnCEOCOMPENSATION
Intercept 14,37*** 14,386*** 14,381*** 7,148*** (788,08) (780,73) (771,41) (4,98) Return 0,118* 0,134* 0,111 (1,66) (1,80) (1,56) ReturnT-1 0,056 0,067 (0,82) (0,98) Sales (ln) 0,431*** (5,03)
Cross section fixed Yes Yes Yes Yes
Period fixed Yes Yes Yes Yes
R2 0,495 0,497 0,499 0,51
Obs 829 802 802 815
US Dependent Variable = lnCEOCOMPENSATION
Intercept 14,142*** 14,146*** 14,131*** 11,848*** (2305,87) (2282,82) (2168,17) (64,96) Return 0,122*** 0,157*** 0,154*** (6,31) (7,28) (7,26) ReturnT-1 0,018 0,046** 0,043** (0,94) (2,34) (2,22) Sales (ln) 0,257*** (12,52)
Cross section fixed Yes Yes Yes Yes
Period fixed Yes Yes Yes Yes
R2 0,652 0,665 0,668 0,679
Obs 5812 5306 5306 5306
As a robustness check, we performed the same regression with ΔlnMarketValue as a market based return measure. Lastly, we looked at the effect of the Return on Assets2. For Europe, the first differences in market value had a coefficient of 0,138 which is pretty close at our ordinary return
2
21
measure and significant at the 5%-level. When both ΔlnMarketValue and Shareholder Return are inserted, the differences in Market Value had more influence but none of the variables were significant. Finally, the ROA was not significant. For the US, ROA do have positive influence on CEO compensation and is more conclusive than shareholder return. Furthermore, the ΔlnMarketValue is again comparable with our measure of Shareholder Return for the US. The coefficient of 0,127 is significant at the 1%-level. Again, when both measures are inserted, the ΔlnMarketValue is positively influencing CEO compensation at the 5%-level. Interestingly, our ordinary Shareholder Return measure turns into an insignificant coefficient of 0,039. Therefore, the ΔlnMarketValue is explaining the CEO compensation better than the Shareholder Return. As already explained, the Shareholder Return includes the reinvested dividends. This could lead to a small upward bias in the return. Therefore, CEO compensation tracks the Market Value apparently more closely. This is comprehensible since the real share price is equal to the market value divided by the amount of shares outstanding.
4.3 Pay-performance elasticities
For further investigation in the pay-performance relationship, we look at the elasticities of the relationship. The dependent variable in the next regression is the difference (Δ) of lnCEOCOMPENSATION from year t to t-1. In this regression, the coefficient of the Shareholder Return variable is the partial increase/decrease in CEO compensation, which is caused by the increase/decrease of return. Table 5 shows these results, together with the sales, increase in sales, increase in market value and return on assets.
22 Table 5
Panel regression outcome for the relationship of ΔlnCEOCOMPENSATION and return. The dependent variable is Δ (the first differences) of the natural logarithm of CEO compensation, in the upper panel for Europe, and in the lower panel for the US. Independent variables are Return, the natural logarithm of Sales, Δ (the first differences) of the natural logarithm of Sales, Δ (the first differences) of the natural logarithm of Market Value and ROA. Sales and Market value are in millions (€ for Europe and $ for US). The first lines are the coefficients, and the second lines are the corresponding T-values. For all tests, we include cross section and period fixed effects. *=significant at the 10%-level, **=significant at the 5%-level, ***=significant at the 1%-level.
Europe Dependent Variable = ΔlnCEOCOMPENSATION
Intercept 0,023 -2,887 0,008 0,04 (1,03) (-1,60) (0,37) (1,09) Return 0,027 0,027 -0,012 (0,33) (0,32) (-0,07) Sales (ln) 0,174 (0,16) ΔSales (ln) 0,731*** (4,07) ΔMarketValue (ln) -0,036 (-0,21) Return on Assets -0,003 (-0,55)
Cross section fixed Yes Yes Yes Yes
Period fixed Yes Yes Yes Yes
R2 0,058 0,061 0,076 0,058
Obs 756 742 723 740
US Dependent Variable = ΔlnCEOCOMPENSATION
Intercept -0,016** 0,265 -0,044*** -0,048** (-2,38) (1,42) (-5,92) (-4,92) Return 0,022 0,021 0,064 (1,03) (0,99) (1,14) Sales (ln) -0,032 (-1,51) ΔSales (ln) 0,324*** (8,04) ΔMarketValue (ln) -0,078 (-1,52) Return on Assets 0,005*** (4,61)
Cross section fixed Yes Yes Yes Yes
Period fixed Yes Yes Yes Yes
R2 0,134 0,134 0,15 0,136
23
A lot of literature (see, e.g. Gilson and Vetsuypens, 1993; Jensen and Murphy, 1990a) discusses the ―sensitivity‖ of the relationship, were they regress the first differences in absolute amounts of CEO compensation against the first differences in market value of common stock. Jensen and Murphy (1990a), found that an increase in Market Value of $1000 leads to an increase in CEO cash compensation of only $0,30. As a robustness check in addition to the log variables performed above, we performed the same absolute regressions. The results are not surprising. For Europe, an extra €1000,- in market value only causes an increase of €0,004 or 0,4 eurocents in CEO compensation, which is not significant. For the US, the coefficient is the same. An extra $1000,- in market value leads to an increase of $0,004 in CEO compensation.
Jensen and Murphy (1990a) discuss the effect of firm size on the pay-performance sensitivity. They showed that in smaller firms, sensitivities are higher. This could be the result of the amount of influence a CEO has in a smaller company or the fact that smaller firms are less visible and therefore less influenced by the implicit regulation hypothesis. They divide the sample on the median of the Market Value and conclude that the smaller firms have a sensitivity which is 4 times higher as the larger firms. We also check the effect of firm size by dividing the sample at the median of our size variable Sales. For Europe, results did not change into significant relationships for the smaller or bigger companies. The direct pay-performance relationship and elasticity is somewhat bigger for the upper half of the sample compared with the lower half3. For the US, we also see a smaller direct relationship at the lower half of the sample in comparison with the upper half. The coefficients are respectively 0,09 and 0,21, and the difference is also statistically significant. This is contradicting with Jensen and Murphy (1990a), but consistent with Gregg, Jewell and Tonks (2012) who also found a stronger pay-performance relationship for larger firms than smaller firms. The results of the elasticity of the US are not that consistent. When we split the sample in half, the elasticity for the upper half is bigger and significant. However, a cut-off point with the smallest 20%, the elasticity coefficient is higher than our lower 50%. So these results are not always consistent and hence varying sometimes. Therefore, we are cautious with our conclusions with respect to the effect of firm size on the pay-performance relationship, although it seems that bigger companies have stronger relationships in our sample. Apparently, large companies attempt to link pay for performance more than small companies.
4.4 The crisis effect
We now turn to discussing the changes of the pay-performance relationship during the crisis. Since return and compensation are not only decreasing in 2008, the years 2006, 2007 and 2009 are also taken into consideration. Table 6 shows the results of the regression with the 2008 dummy included.
24 Table 6
Panel regression outcome for the pay-performance relationship with the dummy included The dependent variable is the natural logarithm of CEO compensation in the left panel and Δ (the first differences) of the natural logarithm of CEO compensation in the right panel, where the upper panels are for Europe and the lower panels for the US. Independent variables are Shareholder Return and the interaction effect of the 2008 year dummy with the Shareholder Return. The first lines are the coefficients, and the second lines are the corresponding T-values. For all tests, we include cross section and period fixed effects. *=significant at the 10%-level, **=significant at the 5%-level, ***=significant at the 1%-level. Europe Dep. Variable = lnCEOCOMPENSATION Europe Dep. Variable = ΔlnCEOCOMPENSATION Intercept 14,375*** Intercept 0,026 (776,6) (1,17) Return 0,025 Return -0,042 (0,26) (-0,35) DUMMY2008 *Return 0,195 DUMMY2008 *Return 0,138 (1,39) (0,83) Cross section fixed Yes Cross section fixed Yes
Period fixed Yes Period fixed Yes
R2 0,496 R2 0,059 Obs 829 Obs 756 US Dep. Variable = lnCEOCOMPENSATION US Dep. Variable = ΔlnCEOCOMPENSATION Intercept 14,142*** Intercept -0,017** (2305,34) (-2,52) Return 0,171*** Return 0,076*** (6,85) (2,80) DUMMY2008 *Return -0,121*** DUMMY2008 *Return -0,135*** (-3,08) (-3,16) Cross section fixed Yes Cross section fixed Yes
Period fixed Yes Period fixed Yes
R2 0,653 R2 0,136
Obs 5812 Obs 5812
25
Therefore, the results are not conclusive. We also replaced the dummy year with new dummies which split the sample in the year 2006, 2007 and 2009. The results were comparable, and none of the dummies have a significant influence. In addition, we performed the regression with the first differences of CEO compensation (Δ). The ordinary return coefficient is even negative, but again inconclusive, so is the dummy effect. As a final check, we performed the tests with the Sales variable. The size-performance relationship did not change either. To conclude, we cannot say anything about the change of the pay- or size-relationship in Europe.
For the US we find quite different results. The relationship is positive and the interaction term with the dummy is significantly negative at the 1%-level. The coefficient of -0,121 represents the change in the relationship. So from 2008 to 2013, the slope is only 0,05 (0,171-0,121). To be concrete, when the CEO compensation was 1,71% more for every firm with a 10% higher shareholder return at the start, this reduced with 1,21 percent point from 2008 onwards. This result is consistent with slightly different coefficients when we replace the year dummy 2008 with 2006, 2007 and 2009. When return is replaced by the Sales variable, this relationship also heavily decreased at all our year dummies. The panel of the first differences also shows a decrease in the elasticity. The coefficient of 0,076 means that a 10% rise in Shareholder Return will increase CEO compensation by 0,76%, which is reduced by 1,35% from the year 2008 onwards, so the summed coefficient (-0,59) is even negative. This effect is also consistent with the 2006, 2007 and 2009 year dummies, only producing slightly different coefficients. When we check the effect of the Sales elasticity, we observe negative dummies, but only significant for the 2006 and 2009 dummies. In sum, our third hypothesis is accepted for the US. The pay-performance relationship did decrease during the crisis in the US. In Europe, the results were not conclusive and so the third hypothesis is rejected for this sample.
5. Conclusion
In this study, we investigated the CEO pay-performance relationship in a robust way for Europe and the US in the period 2000-2013. For the US, literature is available on the relationship but for Europe, consistent research is lacking. In terms of agency theory, it is assumed that the conflicts of interest between managers and shareholders will reduce when there is an alignment of CEO compensation and company performance. Two samples are created from BoardEx and CompuStat where we study many regressions in a two-way fixed effects panel analysis. We only use the salary and bonus compensation due to the mentioned reasons.
26
increased consolidation requirements. However, when the samples are like-for-like, the US CEOs always earn more. Total average company returns calculated by the Total Return Index were almost equal for Europe and the US. Not surprisingly, the company size, measured by Sales and Market Value, are both highly significant factors for the amount of CEO compensation. Therefore, CEOs from bigger companies earn more.
When we perform the regression analysis with Shareholder Return as the independent variable, we only see a moderately significant relation for Europe at the 10%-level, which is reduced a bit when the control variable Sales is included. For the US, this relationship is highly significant at the 1%-level. Therefore, the first hypothesis regarding the positive pay-performance relationship is accepted. The pay-performance coefficient of the US is a bit stronger in absolute terms but more conclusive in the separate sample due to the lower standard deviation. After a T-test for coefficients for different samples, we find that the coefficients from Europe and the US are not significantly different. Therefore, we can only marginally accept the second hypothesis regarding the presence of a stronger relationship in the US. Without doubt, the control for size (Sales) has a high influence on CEO compensation. We also observe a higher explanatory power for ΔMarket Values for both samples and Return on Assets for the US compared with our ordinary Shareholder Return.
In addition to the absolute level regression performed, we also regress the first differences to obtain reliable evidence of the existence of a pay-performance relationship. The resulting elasticities are positive but not conclusive, except for the Sales elasticities in both samples. When we split the samples by our Sales measure to look at the effect of firm size on the strength of the relationship, it seems that bigger companies have a stronger pay-performance relationship. For the US, this difference was significant. However, results are varying with different cut-off points and therefore different samples. Therefore, we have to interpret these conclusions with caution. The crisis had a significant declining effect on the strength of the relationship in the US, measured by our dummy of 2008, but is in addition consistent with a year dummy of 2006, 2007 or 2009. For Europe, the crisis did not give any concluding results. Our third hypothesis regarding the change in the strength of the pay-performance relationship is therefore only accepted for the US.
27 Limitations and further research
In this study, we did not include equity compensation for the already widely discussed reasons. Without doubt, equity compensation is present in Europe. Many studies (e.g., Harvey and Shrieves, 2001) present an increasing part of equity or incentive pay on total compensation. With the correct inclusion of these amounts, results can be different, less biased and comparable with previous US literature. Nevertheless, the precise equity compensation is difficult to obtain or to value and will therefore take a lot of effort.
We did not take some variables into consideration in this study. First, CEO turnover can be correlated with CEO pay. In our database, there are changes in the CEO position for companies when looking at the whole 14-year investigation period. With the study of Gilson and Vetsuypens (1993) in mind, it is imaginable that new CEOs can break a trend in earnings and get a higher or lower compensation compared with previous CEOs. This turnover of CEOs can be controlled for in extending future research. Second, Ozkan (2011), Conyon and Schwalbach (2000), Van Essen, Heugens, Otten and Van Oosterhout (2012) and many more showed the importance of corporate governance variables, which can also have a significant impact on compensation. Lastly, the regulation effect with respect to disclosure requirements, which is possibly apparent in our US sample, can be an influencing factor to include in the research.
28
References
Baker, G. P., Jensen, M. C., Murphy, K. J., 1988. Compensation and Incentives: Practice vs. Theory. The Journal of Finance, Vol, XLIII, NO. 3, 593-616.
Brooks, C., 2008. Introductory econometrics for finance. second edition, Cambridge University Press. Bührer, M. S., 2010. CEO and Chairperson Compensation: The Impact of the Financial Crisis. Dissertation of the University of St. Gallen.
Conyon, M., Schwalbach, J., 2000. Executive Compensation: Evidence from the UK and Germany. Long Range Planning 33, 504-526.
Conyon, M., Murphy, K., 2000. The prince and the pauper? CEO pay in the United States and United Kingdom. The Economic Journal 110, F640-F671.
Crawford, A., Ezzel, J., Miles, J., 1995. Bank CEO pay-performance relations and the effects of deregulation. Journal of Business 68, no. 2, 231-256.
Fernandes, N., Ferreira, M. A., Matos, P. P., Murphy, K. J., 2009. The Pay Divide: (Why) Are U.S. Top Executives Paid More? Finance Working Paper No. 255.
Fernandes, N., Ferreira, M. A., Matos, P. P., Murphy, K. J., 2012. Are US CEOs Paid More? New International Evidence, Review of Financial Studies.
Gilson, S., Vetsuypens, M., 1993. CEO compensation in financially distressed firms: an empirical analysis. Journal of Finance No.2, 425-458.
Gregg, P., Jewell, S., Tonks, I., 2012. Executive pay and performance: did bankers’ bonuses cause the crisis? International Review of Finance, 12:1, 89-122.
Hall, B., Liebman, J., 1998. Are CEOs really paid like bureaucrats? Quarterly Journal of Economics, Vol. CXIII, Issue 3, 653-691.
Harvey, K. D., Shrieves, R. E., 2001. Executive Compensation Structure and Corporate Governance Choices. The Journal of Financial Research, Vol. XXIV, NO. 4, 495-512.
29
Jensen, M., Murphy, K., 1990a. Performance pay and top-management incentives. Journal of Political Economy 98, 225-264.
Jensen, M., Murphy, K., 1990b. CEO incentives—it’s not bow much you pay, but how. Harvard Business Review 68, 138-153.
Jensen, M., Meckling, W., 1976. Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, October, 1976, V. 3, No. 4, 305-360.
Murphy, K. J., 1985. Corporate performance and managerial numeration. Journal of Accounting and Economics, No. 7, 11-42.
Ozkan, N., 2011. CEO Compensation and Firm Performance: an Empirical Investigation of UK Panel Data. European Financial Management, Vol. 17, No. 2, 260–285.
Van Essen, M., Heugens, P., Otten, J., Van Oosterhout, H., 2012. An institution-based view of executive compensation: A multilevel meta-analytic test. Journal of International Business Studies, 43, 396-423.
30
Appendix A: Exchange rates
31
Appendix B: Correlation diagrams
32
Appendix C: Redundancy test
Redundant Fixed Effects Tests Europe
Effects Test Statistic d.f. Prob.
Cross-section F 9.134795 (74,740) 0.0000 Cross-section Chi-square 537.957426 74 0.0000 Period F 3.432565 (13,740) 0.0000 Period Chi-square 48.540940 13 0.0000 Cross-Section/Period F 8.262964 (87,740) 0.0000 Cross-Section/Period Chi-square 562.702538 87 0.0000
Redundant Fixed Effects Tests US
Effects Test Statistic d.f. Prob.
