The effect of relative performance evaluation on long term
compensation for a CEO
Name: Daan van Esch Student Number: 10580751 Bachelor: Economie en Bedrijfskunde Track: Economie en Financiering Supervisor: Evgenia Zhivotova Date: 25-06-2018
Statement of Originality
This document is written by Student Daan van Esch who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract This paper analyses the compensation levels of CEO’s within the 400 largest companies in Europe. The explicit usage of relative performance evaluation (RPE) will be researched and its effects on total compensation, together with long term compensation, will be evaluated. Taking the data from the Institutional Shareholders Services database, all the characteristics of the CEO’s and their corresponding compensation levels were acquired, together with the companies’ financials, industries and country specifications. This research finds evidence for a positive effect of RPE on total compensation for executives, however the effect of RPE on long term compensation was hard to interpret, because of the lack of a good measurement device for long term compensation.
Introduction On the 13th of march this year the Dutch financial paper or “het Financiele Dagblad” posted an article about Paul Polman, the current CEO of Unilever. Headline in this article was “CEO Paul Polman verdient 292 keer zoveel als gemiddelde Unilevermedewerker”, referring to the fact that Paul Polman earns 292 times as much as the average worker from Unilever. After this article on the compensation scheme of Paul Polman, numerous amounts of articles have been posted about the dissatisfaction of shareholders from Unilever regarding the high amounts of compensation of their CEO. This is not a new subject in the daily papers, but an ongoing point of discussion within the largest companies in the world. Especially after the depression that started in 2007, which led to a global downfall of many companies and banks, high amounts of compensation including bonuses are difficult to explain and defend. One example of an article which states this anger amongst shareholders is “Veel weerstand van beleggers tegen beloningsbeleid Unilver” which states the resistance of investors in Unilever against the total compensation policy for their CEO. Is this resistance justifiable? Or are different companies in the same country, industry or from the same magnitude working with similar compensation schemes? These compensation schemes are built up out of different fees, which are either long- or short term compensation and consist of salary, bonuses and for example shares in the company. However, research in the U.S. have proven that the rise in CEO payment can be fully contributed to the rise in market value of the company (Gabaix & Landier, 2008). They studied the top 500 and 1000 companies with regard to size of the firm from 1980 until 2003 in the United States. Their conclusion was that all of the rise over those years in compensation is related to the rise in market capitalization of these firms. To evaluate what types of payment are mostly used and how this contributes to the performance of the firm, the 400 biggest companies in Europe are taken into account. These companies will be evaluated according to size, industry and fiscal year. Using Stata and the data form Institutional Shareholder Services, regressions can be made including different variables over a certain period of time, concerning the compensation for CEO’s and the performance of their company. To evaluate the CEO of a company in order to compensate him or her on their performance, relative performance evaluation or RPE is used by some these companies (Gong et al., 2011). They concluded that the peer group for the RPE and
detectable influence of implicit approach of RPE, in comparison to the explicit approach (Gong et al., 2011). Implicit meaning the company doesn’t have an obligation to implement RPE but only uses it to compare compensation to other companies, instead of having an explicit RPE contract (Gong et al., 2011). This RPE results in a performance based payment and its becoming a larger part of the compensation packages for CEO’s. It is therefore important in evaluating a firm’s performance, to also take into account the sustainability of a firm and their relative and absolute performance goals, as set by the company. This research will try to answer the question which characteristics of a firm determine the type and size of the executive compensation package and how is this compensation package built up. Also how does this compare to the different peer groups the company is in. This will hopefully answer the hypotheses about Relative Performance Evaluation and the the effect of RPE on pay-duration, which will be a measurement device for long term compensation. This pay duration will be quantified as Radhakrishnan et al. (2014) did in their research, looking at the difference between short term and long term compensation. Putting these two together and testing those on the European companies defined in the Institutional Shareholders Services database, will give insight in the possible correlation between the two. If there is a positive effect of RPE on compensation levels of CEO’s, this implies that comparing company performance to establish the height of the compensation works in favour of the CEO’s total pay-out. If this is accompanied by higher pay duration, or in other words this is done through long term compensation, it could mean that RPE result in longer compensation. If this is the case, the use of RPE could mean there is a less short termism in these companies, which works in favour of the company and shareholders (Radhakrishnan et al., 2014). Therefore, having a higher compensation for the CEO of a company could lead to less short termism in this company. To do so I will perform a quantitative research concerning compensation levels of CEO’s across companies in Europe, the usage of RPE and the amount of long term compensation they are rewarded.
Literature Review Relative performance evaluation Workers are rewarded on the basis of their performance, which applies for employees all the way up to the CEO of a corporation. In order to measure the performance of an employee or employer, their performance is not only determined individually, but also in comparison to a co-worker or peer (Gibbons & Murphy, 1990). Relative performance evaluation protects the worker from uncertainties which also affects the workers they are compared to, therefore excluding the worker from common risk (Gibbons & Murphy, 1990). To do so in the compensation packages of chief executive officers, firms of the same industry or by measurement of the same macroeconomic variable are chosen. This will separate the effects of the economy or industry from the actions of the CEO (Dye, 1992). The downside of relative performance evaluation is when it becomes very expensive to measure the performance of a worker. Though when it comes to CEO performance evaluation this measurement is not expensive, given the fact that the stock rates of a company are available to anyone on a daily basis (Gibbons & Murphy, 1990). The second problem with evaluating a workers’ performance is the interaction with its co-workers, which could lead to sabotage or shirking. Again looking at CEO’s level, there is little or no interaction between CEO’s of different companies, therefore sabotaging is impossible (Gibbons & Murphy, 1990). They found in their research using the data of about 1700 CEO’s from about 1100 companies in the US, that RPE strongly influences the way the compensation is composed for CEO’s, between the years of 1974 and 1986. In the end they concluded that their hypotheses concerning RPE was supported by the results, stating that the executives’ compensation will be more related to the market movements than the industry movements. This translates into the following: RPE will be used more with respect to movements in the market than movements in the industry. This industry performance was measured using 11000 firms from Compustat with the same one-, two-, three-, or four digit SIC codes, using the New York Stock Exchange returns. They didn’t find explicit evidence for this industry effect however, when they controlled for the market movements (Gibbons & Murphy, 1990). Hypotheses 1: Companies using RPE will have a higher compensation scheme on average in comparison to companies who don’t use RPE.
Relative performance evaluation in large companies uses peer groups, in order to measure performance relative to companies in the same industry or from the same size. When using different peer groups for a company, that are either set by the company itself or are outsourced to another company, will have different effects on the CEO’s compensation level. The data concerning CEO compensation was taken from the ISS database and contains roughly 400 of the largest companies is Europe. Each company has the same variables, from the compensation package, to the different peer groups on the basis of relative performance and peer groups containing the benchmark compensation comparisons. Also the holding periods for the awarded stock options are given and the personal information for the CEO of the company. This research focuses on the data from 2011 until 2016. The expectation for the total compensation is that it will be higher for companies using RPE, relative to companies who don’t use RPE. The reason for expecting this higher compensation scheme is the self-serving bias. This states that companies drive to selecting peers for RPE is dependent of the performance of the company itself. Companies who perform better could trough selecting different performance targets for RPE influence this output and therefore executive compensation (Gong et al., 2011). Compensation Duration In the past decades there has been a change in the way executive compensation is structured and the magnitude this compensation takes on. Throughout the 1940’s until the 1970’s the level of compensation for CEO’s was almost flat, followed by thirty years of exponential growth and a change in the way the CEO is rewarded (Frydman & Saks, 2010). They found that from the 1950’s until the present, compensation has changed and shifted to more incentive payment methods including pay-out in stock options. Resulting from this shift in compensation set-up is the ongoing discussion in compensation literature about short-termism, which is the result form short term compensation (Radhakrishnan et al, 2014). The question therefore rises what the difference is between short term and long term compensation. This is researched (by Radhakrishnan et al.) in order to state the difference between long- and short term, creating a new variable to measure different types of compensation called “pay duration”. They composed this measure from a few components
in the compensation packages of CEO’s, like stock options (restricted or unrestricted), salary and bonuses. Their research concluded the presence of “short termism”, or the motivation to enhance short-term performance, when the duration of compensation is shorter. Also they found that the larger the firm, the higher the growth opportunities and the better the firm performed in the past, the longer the pay duration in the CEO’s compensation package. Hypotheses 2: RPE will have a positive correlation with pay duration, therefore resulting in higher long term compensation when RPE is used. Short termism Certain compensation schemes can encourage the CEO of a company to increase the value of the company in the short run, which goes at the expense of the long term value. Laverty (1996) suggested that short termism can be the consequence of organizational characteristics and individual factors. Avoiding short termism can therefore be done through changing individuals themselves, or the organisational values and norms (Marginson & McAulay, 2008). One of these values from the company is its compensation package and the way performance is measured. Looking at how both these systems are set up can give an insight in the way to avoid short termism. The duration of compensation quantifying the extent to which pay is long term or short term, is measured by the vesting and holding period of stock options (Radhakrishnan et al, 2014). They constructed a measurement device for obtaining the variable duration, by looking at the stock awards that the CEO receives as compensation. Conclusively they also found a correlation between firm size and the executives’ pay duration. This correlation was positive, meaning the pay duration was longer in in firms with a higher market value. In order to test these hypotheses, the next four models have been created. All of them are using the natural logarithmic form for the dependent variable, creating the elasticity between compensation and companies market value.
Models for testing the effect of RPE on total compensation. Model 1: 𝐿𝑛(𝑇𝑜𝑡𝑎𝑙𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/) = 𝛽4+ 𝛽6∗ 𝐿𝑛(𝑀𝑎𝑟𝑘𝑒𝑡𝑉𝑎𝑙𝑢𝑒/) + 𝛽=∗ 𝑅𝑃𝐸 + 𝜀/ Controlling for panel data in stata and the clusters in industry for all the companies Model 2, when adding the different general industries as dummy variables: 𝐿𝑛(𝑇𝑜𝑡𝑎𝑙𝑐𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/) = 𝛽4 + 𝛽6∗ 𝐿𝑛(𝑀𝑎𝑟𝑘𝑒𝑡𝑉𝑎𝑙𝑢𝑒/) + 𝛽=∗ 𝑅𝑃𝐸 + 𝛽B∗ 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 + 𝛽E ∗ 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡𝑎𝑡𝑖𝑜𝑛 + 𝛽F ∗ 𝐵𝑎𝑛𝑘 + 𝛽H∗ 𝐼𝑛𝑠𝑢𝑟𝑎𝑛𝑐𝑒 + 𝛽J ∗ 𝑂𝑡ℎ𝑒𝑟𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 + 𝜀/ Models for testing the effect of RPE on long term compensation. Model 3: 𝐿𝑛(𝑇𝑜𝑡𝑎𝑙𝐿𝑜𝑛𝑔𝑇𝑒𝑟𝑚𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/) = 𝛽4+ 𝛽6∗ 𝐿𝑛(𝑀𝑎𝑟𝑘𝑒𝑡𝑉𝑎𝑙𝑢𝑒/) + 𝛽=∗ 𝑅𝑃𝐸 + 𝜀/ Model 4: 𝐿𝑛(𝑇𝑜𝑡𝑎𝑙𝐿𝑜𝑛𝑔𝑇𝑒𝑟𝑚𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/) = 𝛽4+ 𝛽6∗ 𝐿𝑛(𝑀𝑎𝑟𝑘𝑒𝑡𝑉𝑎𝑙𝑢𝑒/) + 𝛽= ∗ 𝑅𝑃𝐸 + 𝛽B∗ 𝑈𝑡𝑖𝑙𝑖𝑡𝑦 + 𝛽E ∗ 𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡𝑎𝑡𝑖𝑜𝑛 + 𝛽F∗ 𝐵𝑎𝑛𝑘 + 𝛽H∗ 𝐼𝑛𝑠𝑢𝑟𝑎𝑛𝑐𝑒 + 𝛽J∗ 𝑂𝑡ℎ𝑒𝑟𝐹𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 + 𝜀/ Model 3 and 4 contain the dependent variable long term compensation, which is constructed in two different ways using two assumptions that will be elaborated further in the results section. Data Collection The Institutional Shareholders Services database contains data of about 400 European companies regarding their executive compensation, relative and absolute performance goals and peer groups for compensation and performance. This compensation consists of the pay mix, long- and short term compensation and the holding requirements for stock- and option awards. This database contains observations from the year 2009 until 2017 from 400 of the largest companies in Europe. All the companies can be found in the appendix.
The Datastream database gave an insight in the financials for the companies stated in the ISS database. To control for different variables in the regression, such as market value and industry the company is in, these financials were added to the information about CEO, company data, peer data and compensation schemes. Below are a few statistic tables given concerning the acquired compensation data from the European Companies. The countries that are used are specified in table 1, with the corresponding frequencies alongside them. From these frequencies is becomes clear that of all the companies examined in this research, most companies are located in the United Kingdom, France and Germany. These values are not stated by fiscal year, but taken on the entire time window from the dataset. This way companies who are missing observations in one or more fiscal years are included in the dataset, as they will still contribute to the analysis. Table 1
Nation Frequency Percent Cumulative
Austria 18 0.84 0.84 Belgium 51 2.38 3.22 Denmark 49 2.28 5.50 Finland 67 3.12 8.62 France 312 14.55 23.17 Germany 271 12.63 35.80 Ireland 46 2.14 37.95 Italy 72 3.36 41.31 Luxembourg 23 1.07 42.38 NA 6 0.28 42.66 Netherlands 107 4.99 47.65 Norway 42 1.96 49.60 Portugal 16 0.75 50.35 Russian Federation 1 0.05 50.40 Spain 65 3.03 53.43 Sweden 127 5.92 59.35 Switzerland 191 8.90 68.25 United Kingdom 681 31.75 100.00 Total 2,145 100.00
Table 2 shows all the different components from the total compensation of the CEO and the number of observation in all fiscal years combined. All values for compensation below zero were excluded from the dataset, assuming these were measurement errors or faults in the database. Also because for the fiscal years 2009,2010 and 2017 there were significant less values than for the other fiscal years, these were also excluded from further research. This leaves six fiscal years in the database. Table 2 Statistics of all the compensation components for chief executive officers in the dataset.
Variable Mean N Max Min S.D.
Salary 1279219 2129 6.87e+07 0 1671950 Bonus 39755.07 2145 8271353 0 374055 Stock Awards 1294229 2145 1.14e+08 0 4340612 Option Awards 114141.5 2145 1.55e+07 0 774300.3 Non-Equity Compensation 1289677 2091 3.14e+07 0 1623446 Change in Pension Value and NQDC 502440.9 1306 2.25e+07 0 1096887 Other Compensation 290178.9 967 2.16e+07 0 1222146 Total Compensation 4632865 2142 1.17e+08 0 6106303 The total compensation is build up out of different components as is shown above in table 2. The most important ones for this research are total compensation, stock awards and option awards. Total compensation is used in order to investigate if relative performance evaluation results in a higher average compensation for the companies that use it, against companies who don’t. Stock awards and stock options can be seen as long term compensation, depending on which kind of condition their corresponding holding requirements are set. As is defined by Radhakrishnan et al., this can be seen as the variable pay duration and will be compared to the level of RPE that is used within these companies. To indicate which types of companies are used, table 3 states the average market value of all the companies combined. This value is obtained through the Datastream database and is formulated in millions of dollars. These values were obtained in local currency and have all been converted to Euro’s, as most of the countries and corresponding observations used are inside the EMU. Looking at the maximum and minimum market values of all the companies, you can see the smallest company in each fiscal year is fluctuating between a hundred and
two hundred and fifty million-dollar range. The largest company value increases every year and reached its max in 2016 of around 220 billion. The average lies around 17 billion taken over all the fiscal years. Looking at the FT 500 list from the Financial Times in 2015, with the biggest 500 companies in Europe, you can see this average lies around the 20 billion dollars. Bearing in mind the fact these values are in euros, this is roughly the same as the average for 2015 in this dataset. In fact, almost all of the companies in the top 400 of that list are examined in this research. Table 3 Statistics from the Market Value of the companies in each fiscal year. Fiscal
Year Mean N Max Min S.D.
2011 15967.22 346 166233.1 2.832.379 21991.45 2012 13926.46 353 154735.3 1.412.606 20486.82 2013 15759.18 357 166890 168.293 22223.81 2014 19046.14 353 182851.1 2.726.342 26136.87 2015 19037.26 363 217008.8 2.710.869 26669.63 2016 19329.01 373 206396.5 6.380.858 26847.91 Total 17207.58 2145 217008.8 6.380.858 24296.72 When observing the data, it becomes clear that not all values will be usable in the regression. Some of them are outliers and some of them could be measurement errors. The total compensation for the CEO of a company, of which he is also the founder and owner, could be disproportionally high in comparison to the other CEO’s. However, because this turns out to be inconsistent when looking at the database, in this case the total compensation will be transformed to its logarithmic form. The database showed that also some CEO’s who aren’t founder or owner of the company, still make ten times more than the average of the total compensation.
Table 4
Statistics from the Total Compensation in each fiscal year. The first statistic shows result of the whole dataset, whereas the second only uses the values of companies that use RPE. Again all the fiscal years are taken into consideration.
Variable Observations Mean S.D. Min Max
Total Compensation 2,142 4632865 6106303 0 1.17e+08 Total Compensation of RPE using Companies 354 5866189 4491024 16872.12 3.62e+07 The most important variable of the compensation table is total compensation. This variable will be used to determine whether RPE will have a positive influence on the amount of yearly compensation for the CEO. Presented in the statistics table for the total compensation is the mean, maximum, minimum, observations and standard deviation of the total compensation. Another reason for taking the logarithmic values of this total compensation in the regressions, is the value of standard deviation. The average of the total compensation taken over all the fiscal years is lower than the corresponding standard deviation. This implies that there are large differences between executives’ compensation, which also becomes clear from the difference between the maximum and minimum values of compensation. Table 5 Summary of statistics concerning companies that are using RPE and companies who are not. RPE equal 1
when used Frequency Percent Cumulative
0 1,817 83.69 83.69 1 354 16.31 100.00 Total 2,171 100.00
Results
Table 6
Dependent variable is the natural logarithm of the total compensation (in dollars).
(1) (2) (3)
VARIABLES OLS Fixed Effects Fixed Effects
logmarketvalue 0.287*** 0.322*** 0.332*** (0.0179) (0.0203) (0.0384) RPE 0.234*** 0.208** 0.211** (0.0544) (0.0762) (0.103) Constant 12.32*** 12.00*** 11.91*** (0.163) (0.186) (0.349) Observations 2,133 2,133 2,133 R-squared 0.127 0.145 0.116 Number of IND 7
Year FE Yes Yes
Industry FE Yes Yes
Number of industry 88
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 To test the effect of relative performance evaluation, indicated with the variable RPE, on the total compensation, a few regressions were performed. In all of these regressions the corresponding companies’ market value is included as control variable. The first regression in this output is an ordinary least squares regression, not taking into account there are different years of data and industry effects. Regression (1) shows positive coefficients for the market value variables. These are statistically significant different from zero within 1 percent. Considering the log-log form that is used in all three regressions, these coefficients need to be interpreted like the elasticity of total compensation with respect to the companies’ market value. This translates to the following relationship between the two variables: a 1 percent change in market value will mean a 𝛽PQRSTUVWXYTPZW percent change in
total compensation. In regression (1) the beta for RPE is positive and statistically significant. When a company uses RPE this dummy variable will be equal to 1. Its log-linear form means there is a change 𝛽[\]*100 percent when RPE goes from 0 to 1. In (1) this indicates a 23.4
percent rise in total compensation if the corresponding company uses RPE.
The regressions (2) and (3) were performed in Stata with respect to the fixed effects, using the xtreg command. This will take into account the panel data that is used in this research, consisting of the years 2011 until 2016, and controls for the year fixed effects in the regression. After the regression is adjusted for time, is also needs to be adjusted for different industries the firms are in. The compensation level of the chief executive officer is also dependent of industry the firm operates in. Firstly, this could be due to the difference in managing the company. Some companies are run by an organization such that the capability of the executive officer isn’t of great importance. Secondly, different firm require higher levels of labour and capital to achieve the same amount of sales, therefore having different amounts of rewards. Thirdly, because of a change in managers due to underperforming of the firm, the compensation ratio for the new CEO could be higher (Kostiuk, 1990). As is shown in the regression (3) the standard errors have been adjusted for the 88 different industries the firm are in. This is defined as clusters of industry in the regression. The second regression (2) is the same xtreg regression in Stata, still taking the panel data into account. The only difference now is the variable for industry. All the firms are now labelled according to their general industries, instead of their specific industries. This will bring down the total amount of industries from 88 to 7. These are: Industrial, Utility, Transportation, Bank, Insurance, Other Financial and not announced. When you compare the regressions of (2) and (3) you can see the same positive relation between market value and total compensation. However, the regression (2) when clustering for only 7 industries, will give lower standard errors for the estimated coefficients. This raises the question whether controlling for too many different industries benefits the regression results. If you compare the R-squared for both regressions it becomes clear that controlling for these 7 industries will give a higher value. However, when looking at all the different industries, you will notice that for instance brewers aren’t in the same industry as food products. The same holds for footwear companies that aren’t in the clothing & accessory industry. When comparing these on the managerial level of the different executives, it may be unnecessary to split these into two industries, as they are most likely very similar. Therefore, in the regressions performed in this research, both industry appointments types are evaluated by letting Stata take them into account. The research from Kostiuk (1990) also concluded that the firm size has a positive and time consistent influence on the compensation of the CEO. He conducted research in
different time periods concerning U.S. companies. His regressions from companies in the 80’s, having almost the same dependent variable, namely the natural logarithm of compensation (Salary+Bonus), concluded a positive and statistically significant result between sales and compensation. He observed an estimate for log(sales) of around 0.25, meaning a 1 percent change in sales resulted in a 0.25 percent change in compensation. The corresponding coefficients in Table 6 conclude the same thing in a certain way, a positive correlation between company value and total compensation. The difference in this case is the market value for the company is chosen as controlling variable, but this will be closely related to sales of a company throughout the year. Here the dummy variables for industry are also added, leaving out Industrial in order to avoid collinearity and NA for only having four observations. Also, because I’m working with multiple countries, regression (2) has been conducted using a fixed effects regression for Nation. Table 7 (1) (2)
VARIABLES OLS Fixed Effects
logmarketvalue 0.322*** 0.340*** (0.0184) (0.0201) RPE 0.207*** 0.288** (0.0541) (0.110) Utility -0.379*** -0.357*** (0.0650) (0.102) Transportation -0.0228 -0.0289 (0.128) (0.136) Bank -0.455*** -0.436** (0.0821) (0.177) Insurance -0.0485 -0.146 (0.0861) (0.149) OtherFinancial 0.136 0.000751 (0.0913) (0.156) Constant 12.06*** 11.90*** (0.166) (0.194) Observations 2,133 2,133 R-squared 0.152 0.166 Number of Nation 17 Year FE Yes Nation FE Yes
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Instead of clustering for industries in table 6 regression (2) and (3), these general industries were added to the regressions in table 7. Here, in regression (2), the different Nations are clustered and the fact that panel data is used is taken into consideration. Still having the dependent variable of total compensation, it becomes clear that RPE and market value estimates are staying positive when the different industries are added. These dummy variables are relative to the Industrial industry, as this is the largest industry in the dataset. The only statistically significant variable for industry is Bank, in which the total compensation is around 43 percent lower compared to Industrial. This effect can possibly be explained by negative attention in the media regarding public banks after the recession, which makes companies move to less controversial types of payment and therefore having a lower overall performance based payment (Kuhnen and Niessen, 2012). I think this negative attention is less in companies that are in an industrial industry, because they have a much lower form of social responsibility like public banks have. Table 8 Dependent variable Is natural logarithm of 𝐿𝑜𝑛𝑔𝑇𝑒𝑟𝑚𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/ as defined in the formula and follows assumption 1. (1) (2) (3)
VARIABLES OLS Fixed Effects Fixed Effects
logmarketvalue 0.859*** 1.004*** 1.537*** (0.134) (0.208) (0.326) RPE 3.014*** 2.819*** 2.777*** (0.410) (0.306) (0.763) Constant -1.087 -2.375 -7.223** (1.221) (1.879) (2.939) Observations 2,145 2,145 2,145 R-squared 0.051 0.054 0.067 Number of IND 7
Year FE Yes Yes
Industry FE Yes Yes
Number of industry 88
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
To see if RPE will have an influence on the amount of long term compensation that is awarded to the CEO, a few assumptions had to be made. In the research from
compensation with the variable “pay duration”. The following formula was used to create this variable:
Duration= ^TPTU_`aQbZc `^TPTU_`aQbZc`hfij[WcXU/dXWe ^XQdV[WcXU/dXWe ^XQdVf g Xf` hfijklX/Qbm g Xm f` h fij hfijklX/Qbm
The salary and bonus for each fiscal year for the CEO are found in the database, together with the stock awards and the option awards. In this formula, i and j denote the different grants of either stocks or options, and t denote the corresponding holding or vesting period (Radhakrishnan et al., 2014). In the dataset that is used for this research, all these variables for compensation are given. The difference however with their research, is that the holding or vesting period isn’t always stated for the corresponding stock- or option awards. There are three different ways the holding requirements are formulated: Holding requirements stated as a multiple of base salary, as a number of shares or as a specific dollar value. Where the first and second statement sometimes have corresponding holding time denoted, not always though, in the last statement where the holding requirement is a specific dollar value there is no corresponding holding period. Missing these values have created a problem concerning the allocation of compensation to long- and short term. Only a small number of observations have a number of months given for the holding period. The vast majority of observations is missing these values, which makes using this formula not possible and unreliable if still done so. The rest of the holding requirements are given in multiples of salaries and as actual dollar amounts, which are not possible to appoint a certain time value to. In order to still be able to measure the amount of long term remuneration for the executives, I have made a few assumptions concerning stock- and option awards. In the first regression the total amount of stock- and option awards is taken as a measurement for long term compensation. The idea behind this is that executives are given these awards at the end of the year and have no intention to sell them directly. If they remain with the company the next year and perform their managerial task as they are expected to do, not taking into account the market and industrial effects that have an impact on the value of the company, they could enhance the value of their stocks. This way the executive would like to keep their interest in the company and will hopefully look at the long term performance abilities of their company. This will therefore avoid short termism the same as awards with an actual holding period given. In the second case only the observations containing a certain time
used in the regression, leaving out all the left over observations for stock- and option awards. When an observation contained a time value of over 12 months the stock- and option awards were labelled as long term compensation. The same holds if the observation had a dollar value, a multiple of base value over 1 (which refers to a holding requirement of at least 1 time the base salary value) or when it is formulated as a certain dollar value. All these requirements can be seen as a certain holding period for the rewarded stocks and options. This way not all the stocks- and options are taken into account, but only the ones with actual requirements given. To interpret these requirements as actual time variables is very hard and would be guessing, therefore they are all seen as long term compensation requirements. The regressions for these models, with both assumptions stated above, have been done again in three different ways. The first one is an ordinary least squares regression, with no fixed effects for the panel data or the different industries taken into account. The second and third regression do take these fixed effects into account, but again differ in the amount of industries the firms are appointed to. The second regression looks at the industries provided by Worldscope and consist of 7 general industries. The last regression takes the specific industries into account, consisting of a total of 88 different industries. Table 8 provides the results for the regression with the first assumption, all stock- and option awards are seen as long term compensation. So the formula for long term compensation in this case is: 𝐿𝑜𝑛𝑔𝑇𝑒𝑟𝑚𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/ = 𝑆𝑡𝑜𝑐𝑘𝑎𝑤𝑎𝑟𝑑𝑠/ + 𝑂𝑝𝑡𝑖𝑜𝑛𝑎𝑤𝑎𝑟𝑑𝑠q After calculating this variable, the logarithmic form was taken to perform the regressions. The dependent variable therefore is the natural logarithm of long term compensation, whereas the independent variables are the natural logarithm of the market value for the companies and RPE, which equals one if the company uses relative performance evaluation. This results in a rise in the variable long term compensation as stated above for a higher market value and for the use of RPE. Still using the log-log model states that a 1% increase in market value results in a 0.859, 1.004 or 1.537 percent change in long term compensation. RPE for being a dummy variable results in a 3014, 2819 or 2777 percent rise in long term compensation when it equals one. All these estimations are significantly different from zero at a 5 percent level (except for the contants in (1) and (2)) and the interpretation follows the same explanation as the regressions in table 6. The values for RPE seem high but this is due
to the fact the dependent variable is the total long term compensation, which in half of the observations equals zero. Still I have chosen for the actual values of compensation rather than a dummy variable for if there was any long term compensation, due to the possibility that in lather the effect would be distorted. This would give low values of long term compensation the same weight as high values, because they would in both cases equal one with a dummy variable. Table 9 Dependent variable Is natural logarithm of 𝐿𝑜𝑛𝑔𝑇𝑒𝑟𝑚𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛/ as defined in the formula and follows assumption 1. (1) (2)
VARIABLES OLS Fixed Effects
logmarketvalue 1.004*** 1.156*** (0.139) (0.191) RPE 2.817*** 1.780** (0.410) (0.793) Utility -2.330*** -0.920 (0.491) (0.533) Transportation 1.552 0.762 (0.963) (1.006) Bank -0.973 0.274 (0.621) (1.255) Insurance -0.593 -1.035 (0.653) (0.757) OtherFinancial 1.301* 0.766 (0.690) (0.982) Constant -2.132* -3.503* (1.249) (1.728) Observations 2,145 2,145 R-squared 0.066 0.049 Number of Nation 18 Year FE Yes Nation FE Yes
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Just like the regressions for testing the effect of RPE on the total compensation, here the last regressions were executed using the general industry as dummy variables and the specific countries as clusters. This will make it possible to compare the model with and without nations included. Again this doesn’t change the signs of the estimates for RPE and market
be due to the fact that adding the industries as variables, instead of adjusting the standard errors for clusters of industry, don’t add any explanation of variance to the long term compensation value. If this is because of the assumption made about long term compensation is not clear. Table 10 Dependent variable is the natural logarithm of long term compensation as stated in the second assumption. (1) (2) (3)
VARIABLES OLS Fixed Effects Fixed Effects
logmarketvalue 0.571*** 0.688*** 1.056*** (0.116) (0.139) (0.279) RPE 2.142*** 1.920*** 2.161*** (0.355) (0.240) (0.761) Constant -2.252** -3.282** -6.676** (1.057) (1.287) (2.565) Observations 2,145 2,145 2,145 R-squared 0.033 0.035 0.048 Number of IND 7
Year FE Yes Yes
Industry FE Yes Yes
Number of industry 88
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 The last regressions table contains the three regressions regarding the long term compensation as formulated by the second assumption. These regressions test the value of long term compensation as stated in the assumption against the companies’ market value and the usage of RPE. Again controlling for the year fixed effects and the industry effects in regression (2) and (3), all of the three regressions are stated in table 10. After observing the standard errors for the estimates, all of the estimates are statistically different from zero at least the five percent level. The usage of RPE positively effects the long term compensation in all three cases. The same holds for the market value. Comparing this to the all of the stock and options awards, which follows from assumptions 1, the R-squared values drops in the case of long term compensation as dependent variable as described by assumption 2. Changing the regressions by adding the dummy variables for general industry, plus adjusting for different countries, will enhance this value. Stated in table 11 are these new regressions,
for which the OLS has the most significant values. The Fixed effect model raises the R-squared, however it does make the estimation for RPE inaccurate. Table 11 Dependent variable is the natural logarithm of long term compensation as stated in the second assumption. (1) (2)
VARIABLES OLS Fixed Effects
logmarketvalue 0.689*** 1.251*** (0.120) (0.331) RPE 1.913*** 1.344 (0.354) (1.011) Utility -0.885** -0.207 (0.424) (0.354) Transportation -1.547* -1.633* (0.831) (0.916) Bank -1.097** -0.784 (0.536) (0.951) Insurance 0.598 0.300 (0.564) (0.879) OtherFinancial 3.094*** 1.904*** (0.595) (0.615) Constant -3.271*** -8.318** (1.079) (2.936) Observations 2,145 2,145 R-squared 0.052 0.075 Number of Nation 18 Year FE Yes Nation FE Yes
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 The OLS regression (1) in table 11 also shows positive estimates for RPE and market value. This contributes to the hypothesis that RPE has a positive influence on long term compensation, stated as pay duration by Radhakrishnan et al. (2014), which is tried to be simulated in this research by following the second assumption. Observing the different industries in this regression concludes a negative relationship between Transportation, Bank and Utility compared to Industrial. The segment of Other Financial industry shows a positive influence when it equals 1.
Conclusion This research is looking to find a connection between RPE and pay duration or long term compensation, in order to give an explanation for high compensation with the prospect of less short termism. As was stated in the hypotheses, RPE is expected to increase the total compensation of CEO’s. Looking at regressions (1), (2), and (3) in table 6, the hypotheses is not rejected at the 5 percent confidence level. The effect of RPE on total compensation is positive and around 20 percent in all three cases. The same result is shown in the regressions of table 7. Added to these regressions were industry effects, which are researched by Kostiuk (1990) and are found to be a determining factor for the compensation level of executives. Table 4 Total Compensation table with RPE=1, meaning only the companies who use RPE are taken into consideration.
Variable Observations Mean S.D. Min Max
Total Compensation 2,142 4632865 6106303 0 1.17e+08 Total Compensation of RPE using Companies 354 5866189 4491024 16872.12 3.62e+07 To illustrate this difference in compensation across firm who do and don’t use RPE, table 4 provides some statistics about total compensation. Looking at table 4, the average total compensation lies around 4.6 million dollars a year for all the companies. Taking only the companies who do use RPE in table 8, this average climbs to about 5.8 million, which is in conformity with the regression result and the hypotheses stating that RPE will result in a higher total compensation. The number of companies that use RPE in each year in comparison to the all the companies is around 16,3 percent. To avoid multicollinearity the only financial controlling variable added is the companies’ market value. In further research the model could be expanded with more variables concerning the companies’ or executives’ characteristics. Multiple extended regressions have been done, but couldn’t provide any additional value to the already performed outputs. One of the reason this is hard to do is because of multicollinearity between variables of company financials. For instance, the variables earnings per share (EPS) and return on invested capital (ROC) were added to the model, but didn’t come with significant estimates for their beta’s. This is probably due to the fact that they are highly correlated with the companies’ market value. The same has been tried with the market values for the RPE peer groups for benchmark compensation. Adding
the average peer group market values for each fiscal year to the model as control variable, didn’t provide any additional information due to collinearity with the variable RPE. This dummy variable equalled zero when RPE was absent in a company. Changing this to actual performance values of the companies meant including the value zero for these absent companies and therefore creating a highly correlated with RPE variable MVpeer. To accomplish the objective by finding a correlation between the usage of RPE and the long term rewards for chief executive officers in large European companies, certain assumptions had to be made in order to quantify this long term compensation. Despite having access to all the forms of compensation executives obtain by their companies, actually quantifying this long term compensation couldn’t be successfully done. The pay- duration variable that was formed in the research from Radhakrishnan et al. (2014) in order to quantify this long term variable, could not be implemented in this research due to the lack of data. Therefore, composing the assumptions to still obtain certain values for long term compensation, could imply a biased research. The obtained values for effects of RPE, industries, countries and market values are all in favour of the second hypotheses. However, because there was missing a precise and accurate measurement device for long term compensation, not much value can be derived from these results. Linking the avoidance of short termism trough long term compensation to the usage of RPE, which results in higher average compensation, has therefore only been proven by taking into consideration the assumptions as stated in the results. In the future, having more determents for the total compensation of CEO’s and a more accurate measurement for long term compensation, will enhance the results of the research.
References Datastream. (2018) Thomson Reuters Datastream. [Online]. Available at: Subscription Service (Accessed: June 2018) Dye, R. (1992). Relative Performance Evaluation and Project Selection. Journal of Accounting Research, 30(1), 27-52. Frydman, C., & Saks, R. (2010). Executive Compensation: A New View from a Long-Term Perspective, 1936-2005. The Review of Financial Studies, 23(5), 2099-2138. Kostiuk, P. (1990). Firm Size and Executive Compensation. The Journal of Human Resources, 25(1), 90-105. Gabaix, X., & Landier, A. (2008). Why Has CEO Pay Increased so Much? The Quarterly Journal of Economics, 123(1), 49-100. Gibbons, R., & Murphy, K. (1990). Relative Performance Evaluation for Chief Executive Officers. Industrial and Labor Relations Review, 43(3), 30S-51S. Gong, G., Li, L., & Shin, J. (2011). Relative Performance Evaluation and Related Peer Groups in Executive Compensation Contracts. The Accounting Review, 86(3), 1007-1043. GOPALAN, R., MILBOURN, T., SONG, F., & THAKOR, A. (2014). Duration of Executive Compensation. The Journal of Finance,69(6), 2777-2817. Incentive Lab - Europe (2018) Institutional Shareholders Services Database (Online). Available at: Wharton Research Data Services (Accessed: June 2018) Kuhnen, C., & Niessen, A. (2012). Public Opinion and Executive Compensation. Management Science, 58(7), 1249-1272. Marginson, D., & McAulay, L. (2008). Exploring the Debate on Short-Termism: A Theoretical and Empirical Analysis. Strategic Management Journal, 29(3), 273-292. Rajgopal, S., Shevlin, T., & Zamora, V. (2006). CEOs' outside Employment Opportunities and the Lack of Relative Performance Evaluation in Compensation Contracts. The Journal of Finance,61(4), 1813-1844. Stock, J. H., & Watson, M. W. (2012). Introduction to econometrics. Third Edition. London: Pearson Education Limited
Appendix I Companies: AALBERTS INDUSTRIES ABB LTD N ACCOR ACTELION (OTC) ADECCO 'R' ADIDAS AEGON AGEAS (EX-FORTIS) AGGREKO AIR FRANCE-KLM AIRBUS AKER AKZO NOBEL ALFA LAVAL ALLIANZ ALLIED IRISH BANKS (OTC) ALSTOM AMADEUS IT GROUP HEINEKEN ANHEUSER-BUSCH INBEV ANTOFAGASTA ARCELORMITTAL ARKEMA ASHTEAD GROUP ASML HOLDING ASSA ABLOY 'B' ASSICURAZIONI GENERALI ASSOCIATED BRIT.FOODS ASTRAZENECA ATLAS COPCO 'A' ATOS AVIVA AXA ACS ACTIV.CONSTR.Y SERV. ADMIRAL GROUP ADP ARYZTA ASOS BAE SYSTEMS BBV.ARGENTARIA BANCO SANTANDER BARCLAYS BAYER BEIERSDORF PROXIMUS BERENDSEN DEAD - 13/09/17 BERKELEY GROUP HDG. BHP BILLITON BMW BNP PARIBAS BODYCOTE BOUYGUES BP BRENNTAG BRITISH AMERICAN TOBACCO BRITISH LAND SKY BT GROUP BUNZL BUREAU VERITAS INTL. BABCOCK INTERTIOL BARRATT DEVELOPMENTS BURBERRY GROUP CAIXABANK CAPITA CAPITAL & CNTS.PROPS. CARREFOUR CENTRICA RICHEMONT N MICHELIN CLARIANT CNP ASSURANCES COCA-COLA HBC (CDI) COLOPLAST 'B' COLRUYT COMPASS GROUP CONTINENTAL CREDIT AGRICOLE CREDIT SUISSE GROUP N CRH CRODA INTERTIOL CAP GEMINI (SWX) CARNIVAL CELESIO (OTC) COBHAM COCA COLA EUROPEAN PTNS. SAINT GOBAIN DAILY MAIL 'A' DAIMLER DANONE DANSKE BANK DASSAULT AVIATION DASSAULT SYSTEMES DAVIDE CAMPARI MILANO DCC DSV 'B' DEUTSCHE BANK DEUTSCHE BOERSE DEUTSCHE LUFTHANSA DEUTSCHE POST DEUTSCHE TELEKOM DEUTSCHE WOHNEN BR.SHS. DIAGEO DKSH HOLDING DNB DRAX GROUP DUFRY 'R' E ON N EASYJET ELECTRICITE STRASBOURG EDP ENERGIAS DE PORTUGAL ELECTROCOMP. ELECTROLUX 'B' ENBW ENGE.BADEN-WURTG. ENDESA ENEL ENI ERICSSON 'B' ESSILOR INTL. EUROMONEY INSTL.INVESTOR EVONIK INDUSTRIES EXPERIAN ENGIE FERROVIAL FIAT CHRYSLER AUTOS. LEORDO FLSMIDTH & CO.'B' FORTUM FRAPORT FRESENIUS FUGRO G4S GALP ENERGIA SGPS GAS TURAL SDG GECI REIT GEMALTO GENEL ENERGY GETINGE GIVAUDAN 'N' GKN GLAXOSMITHKLINE GLENCORE HENNES & MAURITZ 'B' HALMA HAMMERSON HANNOVER RUCK. HARGREAVES LANSDOWN HAYS HSBC HOLDINGS HENKEL HERMES INTL. HEXAGON 'B' LAFARGEHOLCIM UPM-KYMMENE BOSS (HUGO) HEIDELBERGCEMENT IBERDROLA NEX GROUP ILIAD IMI IMPERIAL BRANDS INFINEON TECHNOLOGIES INFORMA ING GROEP INTL.CONS.AIRL.GP.(CDI) INTERTEK GROUP INTESA SANPAOLO INTU PROPERTIES INVESTEC ITV IWG INMARSAT ICTL.HTLS.GP. SAINSBURY J WOOD GROUP (JOHN) JOHNSON MATTHEY JULIUS BAER GRUPPE TALANX AKTGSF. KBC GROUP KERRY GROUP 'A' KINGSPAN GROUP KLEPIERRE KONE 'B' KONINKLIJKE AHOLD DELHAIZE KPN KON PHILIPS ELTN.KONINKLIJKE KUEHNE+GEL INTL. BASF KINGFISHER DSM KONINKLIJKE VOPAK AIR LIQUIDE L'OREAL LAGARDERE GROUPE LAND SECURITIES GROUP LANXESS LEGRAND LINDE CHOC.LINDT & SPRUENGLI LLOYDS BANKING GROUP LONDON STOCK EX.GROUP LONZA GROUP
LVMH LEGAL & GENERAL MAN GROUP MAN MAPFRE MARKS & SPENCER GROUP MEGGITT MELROSE INDUSTRIES MERCK KGAA METSO MILLICOM INTL.CELU. MONDI MODERN TIMES GP.MTG 'B' MTU AERO ENGINES HLDG. MUENCHENER RUCK. MAILRU GROUP GDR (REG S) MERLIN ENTERTAINMENTS TIOL GRID NESTE NESTLE 'R' NEXANS NEXT NOKIA NOKIAN RENKAAT NORDEA BANK NORSK HYDRO NOVARTIS 'R' NOVO NORDISK 'B' NOVOZYMES OCADO GROUP OLD MUTUAL OMV ORANGE ORKLA PANDORA PARTNERS GROUP HOLDING PEARSON PENNON GROUP PERNOD-RICARD PETROLEUM GEO SERVICES PHAROL SGPS PORSCHE AML.HLDG.PREF. PRUDENTIAL PRYSMIAN PUBLICIS GROUPE PADDY POWER BETFAIR(DUB) PERSIMMON PEUGEOT PROVIDENT FINCIAL RANDSTAD RED ELECTRICA RELX REULT RENTOKIL INITIAL REPSOL YPF REXEL RHOEN-KLINIKUM RIGHTMOVE RIO TINTO ROBERT WALTERS ROCHE HOLDING ROLLS-ROYCE HOLDINGS ROTORK ROYAL BANK OF SCTL.GP. ROYAL DUTCH SHELL A RSA INSURANCE GROUP RTL GROUP RWE RYAIR HOLDINGS ST.JAMES'S PLACE ORD SAFRAN SAGE GROUP SAIPEM SAMPO 'A' SANDVIK SANOFI SAP ANGLO AMERICAN SCHINDLER 'R' SCHNEIDER ELECTRIC SE SCHRODERS SEB SECURITAS 'B' SEGRO SERCO GROUP SES FDR SGS 'N' SIEMENS SEB 'A' SKANSKA 'B' SKF 'B' SMITH & NEPHEW SMITHS GROUP SMURFIT KAPPA GROUP SM SOCIETE GENERALE SODEXO SOLVAY SONOVA N SPECTRIS SPIRAX-SARCO ENGR. SSE STANDARD CHARTERED STANDARD LIFE ABERDEEN STHREE STMICROELECTRONICS (PAR) STORA ENSO 'R' SUEZ SULZER 'R' SVENSKA HANDBKN.'A' THE SWATCH GROUP 'B' SWEDBANK 'A' SWEDISH MATCH SWISS RE SWISSCOM 'R' SYMRISE SYNGENTA (OTC) SAAB 'B' SALZGITTER SEVERN TRENT SOFTWARE N SPORTS DIRECT INTL. SUEDZUCKER SVENSKA CELLULOSA B(SWX) TELECOM ITALIA TELEFONICA TELENET GROUP HOLDING TELENOR TELIA COMPANY TERIS TER RETE ELETTRICA Z THOMAS COOK GROUP THYSSENKRUPP TOTAL TRAVIS PERKINS TUI TULLOW OIL TALKTALK TELECOM GROUP TATE & LYLE TELE2 'B' TESCO THALES WEIR GROUP UBM UBS GROUP UCB UMICORE WFD UNIBAIL RODAMCO STAPLED UNITS UNICREDIT UNILEVER (UK) UNITED UTILITIES GROUP VALLOUREC VEDANTA RESOURCES VEOLIA ENVIRONNEMENT VERBUND VINCI VIVENDI VODAFONE GROUP VOESTALPINE VOLKSWAGEN VOLVO 'B' VONOVIA WARTSILA WILLIAM DEMANT HLDG. WOLTERS KLUWER WPP WILLIAM HILL YARA INTERTIOL ZODIAC AEROSPACE (OTC) ZURICH INSURANCE GROUP
